Special Issue Reprint

Advanced Power
Electronics Technology

Edited by
Wenzhong Ma, Xingtian Feng and Shuguang Song

mdpi.com/journal/energies
Advanced Power Electronics
Technology
Advanced Power Electronics
Technology

Guest Editors
Wenzhong Ma
Xingtian Feng
Shuguang Song

Basel • Beijing • Wuhan • Barcelona • Belgrade • Novi Sad • Cluj • Manchester

Guest Editors
Wenzhong Ma Xingtian Feng Shuguang Song
School of Shida Shanneng School of Shida Shanneng School of Shida Shanneng
Renewable Energy Renewable Energy Renewable Energy
China University of China University of China University of
Petroleum (East China) Petroleum (East China) Petroleum (East China)
Qingdao Qingdao Qingdao
China China China

Editorial Office
MDPI AG
Grosspeteranlage 5
4052 Basel, Switzerland

This is a reprint of the Special Issue, published open access by the journal Energies (ISSN 1996-1073),
freely accessible at: https://www.mdpi.com/journal/energies/special issues/10IT914P94.

For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:

Lastname, A.A.; Lastname, B.B. Article Title. Journal Name Year, Volume Number, Page Range.

ISBN 978-3-7258-4789-1 (Hbk)

ISBN 978-3-7258-4790-7 (PDF)
https://doi.org/10.3390/books978-3-7258-4790-7

© 2025 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license. The book as a whole is distributed by MDPI under the terms
and conditions of the Creative Commons Attribution-NonCommercial-NoDerivs (CC BY-NC-ND)
license (https://creativecommons.org/licenses/by-nc-nd/4.0/).
Contents

About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Wenzhong Ma, Xiao Wang, Yusheng Wang, Wenyan Zhang, Hengshuo Li and Yaheng Zhu
Adaptive Quasi-Super-Twisting Sliding Mode Control for Flexible Multistate Switch
Reprinted from: Energies 2024, 17, 2643, https://doi.org/10.3390/en17112643 . . . . . . . . . . . 1

Yupeng Cai, Lujie Yu, Meng Wu, Shengyang Lv, Ziyu Fu, Wenhao Tong, et al.
Grid-Forming Control for Solar Generation System with Battery Energy Storage
Reprinted from: Energies 2024, 17, 3642, https://doi.org/10.3390/en17153642 . . . . . . . . . . . 24

Vaishali H. Kamble, Manisha Dale, R. B. Dhumale and Aziz Nanthaamornphong

Optimization of PID Controllers Using Groupers and Moray Eels Optimization with
Dual-Stream Multi-Dependency Graph Neural Networks for Enhanced Dynamic Performance
Reprinted from: Energies 2025, 18, 2034, https://doi.org/10.3390/en18082034 . . . . . . . . . . . 40

Xingyu Liu, Shuguang Song, Wenzhong Ma and Yusheng Wang

Stability Analysis and Controller Optimization of MMC in Standalone Mode
Reprinted from: Energies 2024, 17, 5474, https://doi.org/10.3390/en17215474 . . . . . . . . . . . 63

Shu Niu, Shuai Li, Jizhong Liang, Guodong Li, Fan Hu, Hai Zhang, et al.
Simulation and Characterization of Micro-Discharge Phenomena Induced by Glitch
Micro-Defects on an Insulated Pull Rod Surface
Reprinted from: Energies 2024, 17, 2594, https://doi.org/10.3390/en17112594 . . . . . . . . . . . 84

Zhanqing Zhou, Lingyue Xue, Chen Li and Qiang Geng

A Spread-Spectrum Modulation Scheme for a 3 × 6 Indirect Matrix Converter Based on a
Current Ripple Model
Reprinted from: Energies 2024, 17, 2546, https://doi.org/10.3390/en17112546 . . . . . . . . . . . 94

Minseung Kim, Donghee Choi and Soo Hyoung Lee

A DCM-Based Non-Isolated Step-Down DC Transformer
Reprinted from: Energies 2024, 17, 940, https://doi.org/10.3390/en17040940 . . . . . . . . . . . . 112

Jixiang Tan and Zhongfu Zhou

An Optimized Switching Strategy Based on Gate Drivers with Variable Voltage to Improve the
Switching Performance of SiC MOSFET Modules
Reprinted from: Energies 2023, 16, 5984, https://doi.org/10.3390/en16165984 . . . . . . . . . . . 125

Fangwei Liang, Hanhua Luo, Xianhao Fan, Xuetong Li and Xu Wang

Review of Surface Charge Accumulation on Insulators in DC Gas-Insulated Power
Transmission Lines: Measurement and Suppression Measures
Reprinted from: Energies 2023, 16, 6027, https://doi.org/10.3390/en16166027 . . . . . . . . . . . 141

v
About the Editors
Wenzhong Ma
Wenzhong Ma received the B.S. and M.S. degrees from the Harbin Institute of Technology,
Harbin, China, in 1993 and 1995, respectively, and the Ph.D. degree from the Institute of Electrical
Engineering, Chinese Academy of Sciences, Beijing, China, in 2006, all in electrical engineering. He
was a visiting professor at the University of Alberta, Edmonton, AB, Canada, from 2013 to 2014. Since
1995, he has been a faculty member at the China University of Petroleum, Qingdao, China, where
he is currently a professor in the Department of Electrical Engineering. He has conducted extensive
research in electrical engineering and automation, including power electronic systems, converters,
renewable energy, distributed microgrid, HVdc, motor design, and motor drives. From 2002 to 2006,
he was involved in a National Key Project for Shanghai High-Speed Maglev Train Systems, which
is the first commercial high-speed maglev train. He took charge of the commissioning and testing
work of the long stator linear motor, propulsion system, and power distribution systems. He holds
six patents in his areas of interest. He has authored six books and authored or coauthored more than
30 technical papers. Dr. Ma is a Committee Member of the China Electrotechnical Society.

Xingtian Feng
Xingtian Feng received B.S. and M.S. degrees in electrical engineering from the China University
of Petroleum (East China), Qingdao, China, in 2001 and 2004, respectively, and the Ph.D. degree
in electrical engineering from the Institute of Electrical Engineering, Chinese Academy of Sciences,
Beijing, China, in 2012. Since 2004, he has been with the China University of Petroleum (East
China), Qingdao, China. His research interests include power electronics, power quality, distributed
generation, and energy storage technology.

Shuguang Song
Shuguang Song received his B.S. degree and Ph.D. degree from Xi’an Jiaotong University, Xi’an,
China, in 2014 and 2021, respectively, both in electrical engineering. He is currently a lecturer with the
College of New Energy, China University of Petroleum (East China). His current research interests
include modeling and control of multilevel converters, HVDC, and power quality.

vii
 energies
Article
Adaptive Quasi-Super-Twisting Sliding Mode Control for
Flexible Multistate Switch
Wenzhong Ma 1 , Xiao Wang 1, *, Yusheng Wang 2 , Wenyan Zhang 1 , Hengshuo Li 1 and Yaheng Zhu 1

1 College of New Energy, China University of Petroleum (East China), Qingdao 266580, China;
mawenzhong@126.com (W.M.); 19960009@upc.edu.cn (W.Z.); 19930065@upc.edu.cn (H.L.);
mawenzhg@163.com (Y.Z.)
2 PetroChina Planning and Engineering Institute, Beijing 100083, China; wenning99@tom.com
* Correspondence: wxiao200068@163.com

Abstract: The mathematical model of a flexible multistate switch (FMSS) exhibits nonlinear and strong
coupling characteristics, whereas traditional power decoupling control makes it difficult to completely
decouple the output power. The traditional proportional–integral control parameters are difficult
to adjust, and their robustness and dynamic performance are poor, which affects the stability of the
voltage of the power distribution network and feeder power. To address these problems, this study
first converted the original system into a linear system via coordinate transformation using feedback-
accurate linearization to decouple active and reactive currents. Thereafter, a super-twisting sliding
mode control (ST-SMC) algorithm was introduced, and an adaptive quasi-super-twisting sliding
mode control (AQST-SMC) algorithm comprising the quasi-sliding mode function and adaptive
proportional term was proposed. An FMSS double closed-loop controller was designed to achieve
improved vibration suppression and convergence speed. A three-port FMSS simulation model was
developed using MATLAB/Simulink, and the simulation results show that the proposed control
strategy enhances the robustness and dynamic performance of the system.

Keywords: flexible distribution network; flexible multistate switch; exact feedback linearization;
adaptive quasi-super-twisting sliding mode control

1. Introduction
With the generalization of distributed energy sources and new types of load access
to power systems, traditional distribution networks are experiencing problems such as
voltage overruns, bidirectional tidal currents, and feeder power imbalances [1–3]. Flexible
multistate switches (FMSSs), widely used in medium-voltage distribution grids, can realize
continuous and precise power control, improve feeder voltage distribution, and increase
the reliability of grid power supply. Furthermore, they can realize a flexible closed-loop
operation of the distribution system and increase the ability to consume distributed en-
ergy resources [4,5]. Recently, many related studies on FMSSs have been conducted. The
proportional–integral (PI) control strategy has been applied [6]; however, this presents
challenges, such as difficult parameter adjustment and weak anti-interference ability of
multiport FMSSs. In a previous study [7,8], instead of PI control, finite-set model predic-
tive control (MPC) was proposed, which showed excellent dynamic characteristics and
anti-interference in a steady state along with a high current response speed. However,
such a model presents challenges, such as unstable switching frequency, high control
system computation, and model mismatch. Another study [9] proposed an FMSS-based
fault detection method for distribution networks under grid-side AC faults, in which the
traditional fault management devices were replaced, with the method achieving better
dynamic performance during faults. However, the double closed loop still uses PI control,
which is parameter-sensitive and insufficiently robust.

Energies 2024, 17, 2643. https://doi.org/10.3390/en17112643 1 https://www.mdpi.com/journal/energies

Energies 2024, 17, 2643

FMSSs belong to nonlinear systems exhibiting a strong coupling degree, and it is

challenging to design FMSS controllers by applying traditional control theory. When
feedback linearization is employed, a nonlinear system can be linearized by selecting an
appropriate coordinate mapping method, which can completely decouple the state variables
of the system, thus considerably improving the control performance of the system [10].
Sliding mode control (SMC) is a nonlinear control method that requires a model exhibiting
high accuracy but considerably short response time and high robustness; it has been
widely used in the design of FMSS controllers and observers. In a previous study, second-
order SMC was combined with a nonlinear disturbance observer to design a strategy for
suppressing direct current (DC)-side voltage fluctuations caused by the voltage imbalance
of FMSS ports [11], which resulted in accurate tracking of the reference voltage and current.
However, the nonlinear disturbance observer was sensitive to the system model error
and parameter perturbation, and exhibited poor robustness. A soft open point (SOP)
strategy based on adaptive voltage drop outer-loop control and feedback-linearized sliding
mode inner-loop control was proposed to fully harness the regulation capability of the
converter in the SOP system to avoid reaching the voltage limit [12]. Furthermore, the
strategy was proposed to balance the tidal current in the distribution network. However,
the inner loop uses a conventional first-order linear sliding mode, which may not be
sufficiently robust to external perturbations and uncertainties, leading to a decline in the
control system’s performance. A full-order terminal SMC method was proposed to enhance
system immunity to disturbances by establishing a mathematical FMSS model considering
parameter uptake [13]. However, the full-order terminal sliding mode response was slower
in the initial moments and cases of large nonlinear disturbances. The super-twisting SMC
(ST-SMC) algorithm is a second-order sliding mode algorithm. This algorithm only uses
information about the system state and its first-order derivatives in designing the system
control law, and can be directly used to design controllers and observers for relative-degree-
one systems, which makes it widely applicable in second-order sliding mode control [14].
Moreover, it is more accurate and robust than first-order SMC. Furthermore, extensive
research has been conducted on the mathematical computation of ST-SMC algorithms.
Xu et al. proposed a variable-step closed-loop angle compensation method using the
tangent reaching law based on SMC and the convergence law. Results revealed that the
algorithm exhibited good dynamic compensation performance and disturbance resistance
while maintaining good steady-state compensation accuracy [15]. Maged et al. proposed a
new application of ST-SMC based on an artificial hummingbird optimization technique,
in order to enhance the transient performance of islanded microgrids [16]. Ghazi et al.
proposed an ST algorithm based on the circle search algorithm to improve the efficiency of
grid-connected photovoltaic systems; its parameters were optimized in combination with
the gray wolf optimizer to improve the robustness of the system, tracking speed, and the
optimal convergence to the minimum error value [17].
In recent years, ST-SMC has been widely used in power electronics and power drives;
moreover, considerable research has focused on improving the traditional ST-SMC by
combining it with other nonlinear control algorithms to optimize the dynamic and steady-
state characteristics of the system. Saadaoui et al. proposed an ST-SMC method for ultrafast
chargers of pure electric vehicles for grid imbalance conditions, which can always achieve
the lowest performance index compared with other nonlinear methods, demonstrating the
fast response speed, strong robustness, and low steady-state error of ST-SMC [18]. Celik
et al. designed a Kalman filter using an ultra-twisted sliding-mode strategy, which ensures
a shallow DC-link voltage ripple and achieves a fast dynamic response for a battery electric
vehicle charger based on a three-phase two-level rectifier and a conventional DC/DC buck-
boost converter [19]. Tiwary et al. designed an ST-SMC controller based on an isolated
DC–DC bidirectional dual active bridge (DAB) power converter, in order to achieve DC
bus voltage stability and high robustness for accurate reference voltage tracking [20].
To enhance the performance of the control system in complex environments, some
improved ST-SMCs have been created. Wang et al. proposed a model-free predictive current

2
Energies 2024, 17, 2643

control strategy based on an adaptive super-twisting sliding-mode observer, in order to

design an adaptive gain matrix and dynamically adjust the observer parameters to suppress
the effect of boundary uncertainty in the dynamic part of the rectifier system, and to ensure
the accuracy of the prediction model [21]. Pati et al. proposed adaptive ST-SMC and dual-
loop control for a three-phase differential boost inverter in a grid-connected photovoltaic
system to regulate voltage across a DC bus capacitor, and to compensate for harmonics
under nonlinear loads [22]. However, for FMSS systems, ST-SMC exhibiting improved
efficiency should be developed to withstand various complex operating conditions, and
exhibit enhanced dynamic performance in complex environments. Composite second-order
SMC was proposed to apply ST-SMC in the outer loop of an FMSS [23]. Nevertheless,
the full-order fast terminal SMC used in the inner loop requires high model accuracy,
is sensitive to measurement errors, and is prone to cause instability in the controller’s
performance. Another paper [24] proposed an ST sliding-mode observer to improve the
immunity of an FMSS system, and applied the conventional ST-SMC to design a voltage
outer-loop controller to enhance its dynamic performance and robustness. However, the
inner loop was poorly controlled using three-vector MPC, which requires high system data
accuracy and complex computation.
In this study, a linear radiation model of a three-terminal FMSS was first obtained by
applying the exact feedback linearization method, which decouples active and reactive
currents. Subsequently, an adaptive QST-SMC (AQST-SMC) algorithm was proposed to
design the voltage outer-loop and current inner-loop controllers of the three-terminal FMSS.
The algorithm adopted the sigmoid function in the quasi-sliding mode to discretize the
switching function and reduce the jitter. An adaptive scaling term was introduced to
adjust the response speed of the system, which considerably shortens its response time and
enhances immunity to interference. Finally, a three-terminal FMSS simulation model was
constructed to verify the effectiveness of the proposed AQST-SMC strategy.
The rest of the paper is organized as follows: Section 2 describes the establishment of
an affine nonlinear model of the FMSS system, which provides the basis for the subsequent
construction of an AQST-SMC controller. Section 3 explains the construction of the AQST-
SMC controller and compares it with conventional control strategies, such as first-order
SMC. Section 4 describes the construction of the voltage outer-loop and current inner-
loop controllers based on a radiative nonlinear model and mathematical equations of the
FMSS system. Section 5 presents the establishment of a simulation of the three-terminal
FMSS system, and the verification of its performance under different operating conditions.
Section 6 presents the conclusions of this study.

2. FMSS Nonlinear Control Design

2.1. FMSS System Architecture
In medium-voltage distribution networks, feeders of different voltage levels are con-
nected through an FMSS. However, with the upgrade and transformation of these distri-
bution networks and the rapid development of loads, double-ended flexible switchgear
cannot fulfill the requirements of the interconnection of multi-distribution areas in complex
environments. In addition, the switchgear cannot regulate the power between the feeders if
one of the ports fails. Therefore, it is important to study three-port and multiport FMSSs. A
three-port FMSS can connect multiple feeders and reduce the construction cost. In addition,
if one end is out of operation owing to failure, the rest of the converter can flexibly switch
the working mode to support the normal operation of the power grid, which is more
reliable. Figure 1 shows the structure of the three-terminal FMSS system mentioned earlier.
For each feeder, the head of the transmission line is typically connected to the substation.
Thus, the AC system connected to the FMSS can be equated to a Davignon equivalent
circuit comprising a fixed alternating current (AC) voltage source and a series impedance
at the point of common coupling.

3
Energies 2024, 17, 2643

power flow

AC system 1 Feeder 1 DFIG AC

DC AC
DC Load FMSS
wind solar Load
power power
Distributed power
power flow

AC system 2 Feeder 2 AC
DFIG DC AC
DC Load MMC
wind solar Load
power power
Distributed power

power flow
AC system 3 Feeder 3 AC
DC Load AC
DC Load

Load Load

Figure 1. Structure diagram of three-terminal FMSS system.

Modular multilevel converters (MMCs) are widely used in power electronics, owing
to their small loss and good quality of voltage and current waveforms generated by
power conversion in the working state [25]. Thus, this study used MMCs as the main
circuit topology.

2.2. FMSS Radiation Modeling

Assuming the AC system is balanced in three phases, the MMC equation of state was
obtained from the circuit law as follows:
      
d isd − R ωL isd usd −md Udc
L = + , (1)
dt isq −ωL − R isq usq −mq Udc

where ω represents the AC voltage angular frequency. In the rotating coordinate system, isd
isq and usd usq represent the output current and voltage of any phase of the port, respectively.
Moreover, md mq represents the modulation switching function components of any phase
of the port.
The transformation of (1) yielded the following nonlinear radiation model of the
FMSS system:  .
x = f(x) + g(x)u
, (2)
y = h(x)
   R 
f1 ( x ) − L x1 + ωx2 + uLsd
f(x) = = u , (3)
f2 ( x ) − RL x2 − ωx1 + Lsq
 T  Udc 
g1 (x) − L 0
g(x) = = , (4)
g2 (x) 0 − ULdc

where x = [x1 , x2 ]T = [isd , isq ]T , u = [u1 , u2 ]T = [md , mq ]T , and y = [y1 , y2 ]T . h(x) satisfies
h(x) = [h1 (x), h2 (x)]T = [isdref− x1 , isq ref− x2 ]T , and isdref , isq ref denotes the inner-loop current
reference.
The Lie derivative of function h(x) along the vector field f (x) is defined as follows:

∂hi ( x )
L f ( x ) hi ( x ) = f( x ). (5)
∂x
From (5), we obtain the following:

∂L f ( x) hi ( x )
L gi ( x ) L f ( x ) h i ( x ) = gi ( x ) . (6)
∂x

4
Energies 2024, 17, 2643

According to the differential geometry theory, the nonlinear radiation model satisfies
the following constraints, and the matrix, B(x), is nonsingular:
⎧
⎪
⎪ Lkf ( x) hi ( x ) = L f ( x) Lkf − 1
h (x)
⎪
⎪ (x) i
⎪
⎨ L
gi ( x ) L f ( x ) h j ( x ) = 0
k
k = 1, 2, · · · i, j = 1, 2. (7)
⎪ L f ( x ) hi ( x ) = hi ( x )
⎪ 0
⎪
⎪
⎪
⎩ L r j −1
gi ( x ) L f ( x ) h j ( x )  = 0

 r −1 r −1   
L g1 ( x) L f1( x) h1 ( x ) L g2 ( x) L f1( x) h1 ( x ) Udc
0
B(x) = = L , (8)
L g1 ( x) Lrf2(− 1
L g2 ( x) Lrf2(− 1 Udc
x) 2
h (x) x) 2
h (x) 0 L

where r = [r1 , r2 ] corresponds to the subrelativity of function h(x).

When the FMSS satisfies the normal operation of the aforementioned equation and
r1 + r2 = 2 is equal to the order of the system, the choice of coordinate transformation can
fully linearize the nonlinear model of the FMSS.

2.3. Coordinate Mapping and Control Law Solving

The following coordinate transformation matrix was selected:
     
z1 h1 (x) isdref − x1
z= = = . (9)
z2 h2 (x) isqref − x2

The control variable after the exact feedback linearization takes the following form:

R
ν = A(x) + B(x)u + z. (10)
L
 r −1   
L f1( x) h1 ( x ) R
− ωx2 − usd
A(x) = = L x1 L . (11)
Lrf2(−
usq
1
x) 2
h (x) R
L x2 + ωx1 − L

The reintegrated linear system was as follows:

 
−1 R
u = B (x) ν − A(x) − z . (12)
L
   usd  
md
L
Udc ν1 + ωx2 + L − Udc isdref
R
u= = u . (13)
mq L
Udc ν2 − ωx1 + Lsq − UR isqref
dc

3. Construction of Adaptive Quasi-Super-Twisting Sliding Mode Controller

Second-order SMC uses the differentiation of the control input as a new virtual control
quantity, and applies it to the higher-order derivatives of the sliding mode surface to ensure
that the sliding variable, s, converges to zero in a restricted time [26]. This weakens the
jitter, eliminates relative-order limitations, and improves the accuracy and reliability of
the SMC.
The super-twisting algorithm (STA) is a second-order SMC algorithm designed for
systems of relative order 1, which ensures that the system reaches the ideal state (s). Its
first-order derivative is zero in a finite time. Conversely, conventional first-order SMC can
only render the system state (s) zero in a finite time. The general expression of the STA is
as follows: 
u = −α|s|1/2 sign(s) + u1
. . , (14)
u1 = − βsign(s) + ϕ
where s represents a sliding variable; u1 represents an auxiliary variable; ϕ represents a
perturbation term; and α and β represent sliding mode gain coefficients.

5
Energies 2024, 17, 2643

.
If (14) satisfies  ϕ ≤ δ and δ > 0, then α and β simultaneously satisfying the following
conditions can cause (14) to converge in a finite time [27].

α>2
α3 +(4α−8)δ2 . (15)
β> α(4α−8)

Figure 2 shows the phase plane trajectory of the STA, and point S1−A is the intersection
.  .2
of the system-state trajectory curve with the axis, s = 0 and 2 K1 − A β − L s1− A = s0 , where
 
.
K1 − A and L are positive constants. The calculation shows that s1− A = − α2 KL + β s1/2 1− A ,
.  .  .  .   .   . 1− A
which affords s1− A /s0  < 1. Considering  . that
s1  ≤ s1− A , s1 /s0  < 1 holds;
.  .  . 
repeating the process ensures the inequality sk+1 /sk  = sk+1 /sk  < 1 holds. Similarly, it
can be shown that |sk+1 /sk | = |sk+1 |/|sk | < 1 holds, indicating that the closed-loop system
can converge to stability in a finite time [28].

s 0, s1 B

s0
0, s2  A
System status
s2  A , s2  A system operating domain
s2 traces

s1 A , 0 s
s2  A , 0 s3
s1
0, s2  B STA control rhythm
s1 A , s1 A
0, s1 A

Figure 2. STA phase plane trajectory.

Based on (14), the introduction of the integral term can effectively attenuate the jitter
and eliminate the static error of the system. However, there is still a sign function with
poor smoothing performance near the sliding mode surface in the control law. This can
cause high-frequency jitter with rapid and frequent switching on the sliding mode surface,
and reduce the control accuracy of the system. A study [29] added a sinusoidal saturation
function, sat (s), in the super-twisting control law to reduce the system jitter. However,
the saturation function presents the problem of selecting the appropriate thickness of the
boundary layer, which is extremely large to lead to the insufficient control accuracy of the
system, and excessively small to cause oscillations. Thus, the following smoother sigmoid
function was designed:
2
sd(s) = − 1. (16)
1 + e−ns
A constant (n > 0) was used to adjust the rate of the curve rise, which determines
the function boundary layer thickness. Figure 3 shows that the sign function changes
stepwise at t = 0, and that the sigmoid function continuously and smoothly changes with
the input. The quasi-smooth mode changes more gradually than the switching function,
which can alleviate the shock of the smooth mode switching and reduce the appearance of
high-frequency jitter to a certain extent. This can improve the response speed and control
accuracy of the FMSS system.

6
Energies 2024, 17, 2643

Function(signǃsigmoid)
1

sign(s)
0.5
n=0.5
n=1
F(s)

0 n=2
n=4
sigmoid
0.5

1
10 8 6 4 2 0 2 4 6 8 10
time(s)

Figure 3. Symbolic and sigmoid functions for different values of n.

Equation (14) shows that the proportional term, −α|s|1/2 sign(s), improves the con-
vergence speed of the algorithm. However, the exponent of the state vector is always 1/2,
which prevents the system state from converging to a smaller region. The immunity of the
algorithm to interference and the arrival speed are directly affected by its gain. A previous
study [18] introduced the proportional term in a traditional STA to improve the conver-
gence trajectory of the system. However, when s < 1, i.e., ks is zero; the proportional term is
only left with the original square root term, and the problem of low arrival speed remains.
To solve the aforementioned problem, this study introduced an adaptive linear term to
realize the adaptive regulation of the convergence speed with the following expression:

q(s) = k|s|c·sign(|s|−1) s, (17)

where k > 0, 0 < c < 1. When the system state tends to the sliding mode surface, i.e.,
s < 1, the adaptive term becomes q(s) = k |s|−c s, α|s|1/2 sign(s) + k |s|−c s  α|s|1/2 sign(s),
which speeds up the convergence. Furthermore, when the system state tends to the sliding
mode surface, i.e., s > 1, the adaptive term becomes q(s) = k|s|c s, α|s|1/2 sign(s) + k|s|c s 
α|s|1/2 sign(s), which speeds up the convergence.
Using the sigmoid function and adaptive scaling term, the AQST-SMC mathematical
expression can be written as follows:
⎧
⎪ 1/2
⎨ u. = −α|s| sd(s.) − q(s) + ω
ω = − βsd(s) + ϕ . (18)
⎪
⎩ q(s) = k |s|c·sign(|s|−1) s

.
If (18) satisfies 2 ϕ ≤ δ and σ > 0, then α, β, and k simultaneously satisfying
condition (19) can cause (18) to converge to the origin in a finite time. The stability of the
AQST-SMC algorithm was confirmed, as described in Appendix A. It was discovered that
the stability of the system will not be affected by the introduction of the adaptive linear
term, q(s), into the super-twisting algorithm, as long as it is guaranteed that k > 0.
⎧
⎪
⎪ α>2
⎪
⎨ 2
α+k| x1 |1/2
δ2
β> + . (19)
⎪
⎪ 4 α + k | x1 | 1/2 2
−8
4α
⎪
⎩
k>0

To verify the effectiveness of the proposed AQST-SMC, it was compared with the
conventional first-order SMC, conventional STA, and STA without the adaptive acceleration
term. Figures 4 and 5 show the comparison results. First-order SMCs can usually reach
the sliding mold surface in finite time. Nevertheless, they cannot maintain stability and

7
Energies 2024, 17, 2643

equilibrium, and most conventional first-order SMCs are less efficient than second-order
SMCs in terms of control performance and the tracking error, as shown in Figure 4a,b;
the tracking accuracy and convergence speed of conventional first-order SMCs are worse
than those of the conventional ST-SMC and the ST-SMC that contains only the sigmoid
function. Using the designed adaptive convergence term equation q(s), the AQST-SMC
can approach the surface of the sliding mode quickly, and the system’s state variables can
converge smoothly. From Figure 5, the introduction of the continuous smooth sigmoid
function attenuates the system’s jitter and enhances the system’s perturbation resistance
compared to the sign function, confirming the effectiveness of the AQST-SMC.




)LUVWRUGHU60&
 
6LQJOH67$
 −
67$VLJPRLG − )LUVWRUGHU60&


s
s

$46760& 6LQJOH67$
−
        67$VLJPRLG
 VOLGHPROGVXUIDFH
− $46760&
VOLGHPROGVXUIDFH

−
                 
WLPH V WLPH V

(a) s > 0 (b) s < 0


k  k  k  k 


 
 
 
− YLEUDW LRQ  QRWUHDFKHGs 
  −
s

  − 
 YLEUDWLRQ
YLEUDW LRQ
− QRWUHDFKHGs 
− − −
     −    
       
 k  k  k  k  k 
−
                 
WLPH V WLPH V

(c) s > 0 (d) s < 0

Figure 4. Comparison of the effect of different sliding mode control laws to converge to s = 0.

30
First-order SMC
20 Single STA
STA+sigmoid
Significantly reduced jitter
10
u

 10

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
time(s)

Figure 5. Comparison of vibration suppression effects of different sliding mode control laws.

Owing to the limitation of the stability condition Equation (19), control gain k must
be selected within a certain range. When k is excessively small, the adaptive term q(s)
tends to zero, thus preventing the achievement of the desired acceleration effect and
simultaneously departing from the limitations of the stability condition. This causes
the system trajectory to deviate from the sliding mode surface; however, although the
convergence speed increases when k becomes larger, it may cause new problems such as

8
Energies 2024, 17, 2643

considerable overshooting or system oscillations due to the integral term. Therefore, k

must be designed in a reasonable range to ensure that the system state variables converge
smoothly in a stable range. Figure 4c,d show simulations that were conducted to test the
effect of different gain coefficients k on the convergence of the system state to the surface of
the sliding mode. When a higher value of k is considered (e.g., k = 200), the convergence
is fast; thus, oscillations may occur. Conversely, when a small value of k is considered
(e.g., k = 2 or 5), the system state does not reach the surface of the sliding mode completely,
but the convergence speed is still faster than that of the conventional ST-SMC. Notably,
in Figure 4, the three ST-SMCs use the same gain coefficients. Conversely, traditional
first-order SMCs use the integral sliding mode surface and the exponential convergence
law. Their gain parameters are selected to ensure that the system reaches the surface of the
sliding mode in a finite amount of time and remains stable.

4. FMSS AQST-SMC Controller Design

4.1. AQST-SMC Voltage Outer Ring Design
During normal operation, the FMSS needs to operate in the Udc Q mode at one end
to maintain a constant voltage on the DC side. The remaining ports operate in the PQ
mode for power exchange. However, in the traditional outer-loop PI controller, when
the system parameters are changed, the dynamic performance and response time of the
system will increase because of the constant parameters of the PI. In this research, the
AQST-SMC was designed to effectively suppress the jitter and speed up the response; it
exhibits good dynamic performance and response time. The steady-state inverse model [7]
was used to improve the traditional power outer-loop controller, which improves the outer-
loop regulation rate and eliminates the static error of the system. For example, this study
analyzed MMC1 running in the Udc Q operation mode, and MMC2 and MMC3 running in
the PQ mode.
First, the three-port FMSS DC side satisfied (20), as follows:
⎧ 3
⎪
⎨ C dUdtdc = idc1 + idc2 + idc3 = ∑ idck
k =1 , (20)
⎪
⎩ idck = m ak i ak + mbk ibk + mck ick = ∑ m jk i jk
j= a,b,c

where idck represents the DC side current of port k, and mjk , ijk denotes the one-phase
switching function and output current of port k.
Combining (20) with coordinate mapping transformation shows the following:

dUdc
C = m ak i ak + mbk ibk + mck ick
dt . (21)
3 3
= ∑ −mdk isdk + mqk isqk
k =1 2

In the rotated coordinate system, mdk and mqk denote the port k switching function compo-
nents. Moreover, usdk and isdk denote the voltage and current of port k, respectively.

isqk = 0
disdk . (22)
dt = 0

Equation (21) can be transformed into the following:

dUdc 3 3 3
C = − md1 isd1 − md2 isd2 − md3 isd3 , (23)
dt 2 2 2
where mdk satisfies mdk = (usdk − Rk isdk )/Udc .

9
Energies 2024, 17, 2643

The sliding mode surface s0 was selected as the DC voltage tracking error.

s0 = Udcref − Udc . (24)

Equation (24) was differentiated as follows:

. .
s0 = −U dc . (25)

The uncertain disturbance ϕ was unknown; however, it was assumed to be bounded

and satisfies |ϕ| ≤ m, where m represents a known positive constant. To accurately track
the reference DC voltage, it was obtained according to the AQST-SMC theory:
. . 
s 0 = −U dc = α0 |s0 |1/2 sd(s0 ) + k0 |s0 |c0 ·sign(|s0 |−1) s0 + β 0 sd(s0 )dt 
3 usd1 − R1 isd1 u − R2 isd2 u − R3 isd3 . (26)
= · isd1 + sd2 · isd2 + sd3 · isd3
2C Udc Udc Udc

When the FMSS system is three-phase-balanced and in a steady state, the d-axis current
reference value is obtained via simplification as follows:

2C 
Udc α0 |s0 |1/2 sd(s0 ) + k0 |s0 |c0 ·sign(|s0 |−1) s0 + β 0 sd(s0 )dt
isdref = 3
usd1 − R1 id1 . (27)
−isd2 (usd2 − R2 isd2 ) − isd3 (usd3 − R3 isd3 )
+
usd1 − R1 isd1

Figure 6 shows the AQST-SMC-based FMSS voltage outer-loop control.

AQST-SMCControl outer loop control

sigmoid( s0 ) E0 1s
U dcref s0 isd 3 usd 3  R2isd 3
sigmoid( s0 ) D0

U dc 2CU dc 3 isdref
abs( s0 ) sqrt

s0 isd 2 u sd 2  R2isd 2
1 X
Y
Y X
1 usd 1  R1isd 1
c0
sign(.) k0
adaptive term

Figure 6. Voltage outer loop AQST-SMC block diagram.

4.2. AQST-SMC Current Inner Loop Design

The current inner-loop sliding mode surfaces s1 and s2 of the FMSS were expressed
as follows:    
s1 z1
s= = . (28)
s2 z2
When the system reaches a steady state, the derivatives of s1 and s2 are zero, and the
system control law is as follows:
   R 1/2 c1 ·sign(|s1 |−1)

ν1 z 1 − α 1 | s 1 | sd ( s 1 ) − k 1 | s 1 | s 1 − β 1 sd ( s 1 )
ν= = RL . (29)
ν2
L z2 − α2 | s2 |
1/2
sd(s2 ) − k2 |s2 |c2 ·sign(|s2 |−1) s2 − β 2 sd(s2 )
Q
Furthermore, the original system control volume is obtained as the following:
   
L ν1 1 − Risdref + ωLx2 + usd
u= + . (30)
Udc ν2 Udc − Risqref − ωLx1 + usq

Figure 7 shows a schematic of the final FMSS AQST-SMC integrated control strategy.

10
Energies 2024, 17, 2643

MMC1 voltage outer loop

Udcref control(Active class)

U dc isdref
AQST-SMC Accurate linearization of AQST-SMC control with current inner
loop feedback
Build
MMC2, MMC3 Power outer loop SMC
s 0 Q1 Feedback
PPref control(Active class)
ref
+isdid z1 Surface s1 s1
precise
u
P
P
PI
+ isdref
isd s2
linearization
1 generate
s 0 T
modulated
isq waveform
¦ Pref u
Transient
isd z2 Build s2 Q2 Feedback
2
SMC AQST-SMC precise
Qref Power Theory Comtrol
isq Surface s s 0 linearization
Q isqref
PI +isq
Qref
MMC1,MMC2, MMC3 outer loop
control(Reactive class)

Figure 7. Block diagram of three-terminal FMSS AQST-SMC control strategy.

5. Analysis of System Simulation Examples

According to the demand of a distribution network under different working conditions,
the FMSS mainly operates in the Udc Q, PQ, and Uac f modes. To enable the FMSS to precisely
control the current and fast power supply restoration, the following collaborative control
methods are adopted [30]:
(1) During normal operation, the MMC of one port typically operates in the Udc Q
mode to stabilize the DC bus voltage and regulate the reactive power of that port. The
MMCs of the rest of the ports operate in the PQ mode to separately regulate the active and
reactive power of the ports.
(2) The MMC running in the Udc Q mode is the more vulnerable link in the entire
system. When the MMC running in this mode fails, it is necessary to switch the MMC in
the PQ mode to the Udc Q mode to maintain the DC side voltage.
(3) In the event of a port feeder failure, the FMSS can rapidly isolate the fault and switch
to the Uac f mode to provide continuous power to critical loads in the area experiencing loss.
The three-port FMSS model (Figure 1) was built using MATLAB 2023a/Simulink
software. Table 1 lists the system parameters of the three ports.

Table 1. Simulation parameters of the FMSS system.

System Parameter Symbol Value

AC voltage rating us 10 kV
AC side-rated frequency fs 50 Hz
Rated port capacity S 10 MVA
DC voltage reference Udc 40 kV
Submodule point capacitance C 5000 μF
Bridge arm inductance L0 2 mH
Number of sub-modules N 20

5.1. AQST-SMC Performance Analysis

The simulations of the conventional first-order SMC, conventional STA, and improved
AQST-SMC were compared to study the optimization performance of the algorithms and
validate the theoretical analyses presented in Section 3. The traditional first-order SMC
uses an integral sliding mode surface and the exponential convergence law. In addition,
the sign function in the original exponential convergence law is replaced by a saturation
function, which is more effective in suppressing jitter (Appendix A). The Pref and Qref of
port 2 were 0.2 p.u. and 0.3 p.u. At t = 0.02 s, the active and reactive powers of port 2 were
assumed to suddenly increase and decrease by 0.2 p.u., respectively. Figure 8 shows the
simulation results.

11
Energies 2024, 17, 2643

0.5

0.4
First-order SMC
0.3 Traditional STA
P/p.u.

Reference Value
0.2 AQST-SMC
0.215
0.1 0.21
0.205
0.2
0 0.195
0.19
0.1 0.011 0.012 0.013 0.014 0.015
0 0.01 0.02 0.03 0.04 0.05
time(s)

(a)
First-order SMC
0.3
Traditional STA
0.25 Reference Value
AQST-SMC
Q/ p. u.

0.2

0.15 0.11
0.105
0.1 0.1
0.095
0.05 0.09
0.034 0.035 0.036
0
0 0.01 0.02 0.03 0.04 0.05
time(s)

(b)
Figure 8. Power response characteristics of port 2. (a) Active and (b) reactive power characteris-
tic curves.

The power response curves presented in Figure 8a,b show that the proposed AQST-
SMC method had a faster response speed at the startup moment of the system. Furthermore,
it reached a steady state in a shorter time because of the rapid adjustment of convergence
when the power changed. Based on the calculations, the power oscillation of the AQST-
SMC algorithm using the sigmoid function was considerably smaller in the steady-state
operation than in other SMCs. It was approximately 66.71% and 60.13% of the active and
reactive oscillations, respectively, under the proposed STA, compared with that under the
traditional STA, effectively reducing the power oscillation in the steady-state operation and
improving the transient performance of the system.
Table 2 lists the output power response time at the system startup moment and the
regulation time to return to a steady state when disturbed. Table 2 shows that the active
power response times at the startup moment using the proposed AQST-SMC method were
1.96 and 8.43 times those of the traditional STA and first-order SMC algorithm, respectively.
The reactive power response times were 2.61 and 11.70 times better than those of the
traditional STA and first-order SMC algorithm, respectively. The active power regulation
time to return to a steady state when disturbed under the AQST-SMC algorithm improved
compared with those of the traditional STA and first-order SMC algorithm. The active
power regulation times of the AQST-SMC algorithm were 3.75 and 12.34 times better than
those of the traditional STA and first-order SMC algorithm, respectively. Finally, the reactive
power regulation times were 3.33 and 14.67 times better than those of the traditional STA
and first-order SMC algorithm, respectively. Thus, the proposed AQST-SMC method
effectively improved the speed of the system to reach a steady state, and considerably
reduced the degree of fluctuation of the system under transient operation. In addition, the
robustness and anti-interference capability of the system improved.

12
Energies 2024, 17, 2643

Table 2. Performance characteristics of FMSSs.

Steady-State
Control Modes Response Time (ms) Adjustment Time (ms) Oscillation
Amplitude
P: 4.30 P: 3.95 P: 9.87%
First-order SMC
Q: 5.50 Q: 4.40 Q: 8.91%
P: 1.00 P: 1.20 P: 6.96%
Traditional STA
Q: 1.23 Q: 1.00 Q: 8.36%
P: 0.51 P: 0.32 P: 4.64%
AQST-SMC
Q: 0.47 Q: 0.30 Q: 5.03%

5.2. Dynamic Simulation Verification of FMSS System

To further investigate the dynamic performance and anti-interference ability of the
control system, this study compared the simulations of several typical working conditions
prone to occur under new energy access. The initial operating state of the three-port FMSS
is as follows: the output power value of port 1 is Q1ref = 0 p.u.; the output power values of
port 2 are P2ref = 0.1 p.u. and Q2ref = 0.3 p.u.; and the output power values of port 3 are
P3ref = 0.2 p.u. and Q3ref = 0.5 p.u. Table 3 lists the parameters of different controllers. The
first-order traditional SMC features an additional control action based on the exponential
convergence law. Notably, the present study only applied the AQST-SMC in port MMC1
(traditional Udc Q outer loop) and MMC2, and 3 (traditional PQ control outer loop) used PI
control. Furthermore, the controller parameters were calculated and then carefully selected
for the optimal steady-state operation of the system, as well as to achieve the desired
response speed and convergence.

Table 3. Parameters of different FMSS controllers.

Parameters of Outer Loop Parameters of Inner Loop

Control Method
Controller Controller
Traditional PI (MMC 1) kp1 = 0.5, ki1 = 100 kp1 = 22, ki1 = 3.46
Traditional PI (MMC 2) kp2 = 0.00005, ki2 = 70 kp2 = 22, ki2 = 3.46
Traditional PI (MMC 3) kp3 = 0.000028, ki3 = 80 kp3 = 22, ki3 = 3.46
First-Order SMC ε = 1, q = 40, c1 = 30, Δ = 0.05 ε = 3, q = 4000, c1 = 30, Δ = 0.05
AQST-SMC α1 = 2, β1 = 30, k1 = 10 α1 = 110,000, β1 = 20,000, k1 = 10

5.2.1. Comparison of Simulations with Disturbed System Output Power

The active power variation affects the stability of the DC-side voltage. The simulation
of port 3 was subjected to power fluctuation after accessing distributed power supply and
a new type of load. The traditional PI and proposed AQST-SMC algorithm were simulated
and compared; Figure 9 shows the output feedback curves of the three-terminal FMSS.
Figure 9 shows that the system startup moment, using the traditional PI control under
the output power response, was slow. It took approximately 0.1 s to stabilize. Based on
the active power reference of MMC3 at t = 0.2 s, the voltage jumped from 0.2 (standard) to
0.5 to simulate the power fluctuation of the FMSS system caused by the distributed power
supply and new load access. Under PI control, the system required approximately 0.12 s to
reach a steady state after being perturbed. The dynamic mathematical model of the DC
side of the three-port FMSS and the active power balance relationship showed that the
active power perturbation affected the stability of the voltage on the DC side [9]; moreover,
owing to the inability of the PI controller to adapt to the dynamics of the system under the
sudden change of power, the DC voltage fluctuated considerably at 0.2 s and 0.45 s, and
the sizes of the oscillations were approximately 112.05 V and 105.85 V, respectively. When
the reactive power of MMC1 fluctuated at 0.35 and 0.55 s, its active power suffered from a
certain degree of fluctuation, verifying the proposed control strategy for the effectiveness
of the power decoupling. Under the proposed AQST-SMC, the system stabilized within

13
Energies 2024, 17, 2643

0.001 s at the startup moment and after being disturbed. The dynamic response time was
reduced by 100 times, and the DC voltage remained stable. Thus, the feasibility of the
AQST-SMC was verified, the anti-disturbance performance of the system considerably
improved, the decoupling between the output powers of the system was realized, and the
stability of the additive system improved.

1 P1 PI P1 (AQST-SMC)
P2 PI P2 (AQST-SMC)
P3 PI P3 (AQST-SMC)
0.5
P/p.u.

0.11s 0.55s
0.5
0.35s

1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
time(s)

(a)
1

0.5
Q/p.u.

Q1 PI Q1 (AQST-SMC)
Q2 PI Q2 (AQST-SMC)
Q3 PI Q3 (AQST-SMC)
0.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
time(s)

(b)
40.2

PI
AQST-SMC
40.1
U dc kV

40.0

39.9

39.8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
time(s)

(c)
Figure 9. Output feedback of FMSS system under power perturbations. (a) Active power, (b) reactive
power, and (c) DC voltage response curves.

5.2.2. System AC Measured Voltage Amplitude Dips

To validate the proposed method for AC fault measurements, the MMC3 AC voltage
was simulated at an amplitude drop of −0.4 (units) in 0.2–0.3 s. Figure 10a shows the AC
voltage and current responses; the current response was rapid and accurate. Figure 10b,c
show that during voltage dips, the proposed AQST-SMC exhibited good transient charac-

14
Energies 2024, 17, 2643

teristics at instances of fault occurrences and good steady-state characteristics during the
fault period, compared with the conventional first-order SMC and conventional PI control.
The active and reactive powers were rapidly maintained, with fast response and robustness.
Figure 11a,b show that during voltage dips at the MMC3 side, MMC2 can function normally
without being affected, and its active and reactive powers can remain stable. This indicated
that the dynamic response of the system was fast, with good anti-interference properties.

usa usb usc

1

0.5 isa isb isc

U(p.u.) I(p.u.)

0.5

1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

time(s)
(a)
0.4
PI
First-order SMC
AQST-SMC
0.3 Reference value
P3 (p.u.)

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time(s)
(b)
0.15
PI
0.1 First-order SMC
AQST-SMC
Reference value
0.05
Q3 p.u.

0.05

0.1

0.15
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time(s)
(c)

Figure 10. MMC3 output feedback under voltage drop. (a) MMC3 voltage and current. (b) MMC3
active power. (c) MMC3 reactive power.

15
Energies 2024, 17, 2643

usausb usc
1

0.5
I/p.u.

isaisb isc
0
U/p.u.

0.5

1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

time(s)

(a)
0.4
Q2 P2

0.3
P2 Q2 p.u.

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
time(s)
(b)
Figure 11. MMC2 output feedback under voltage drop. (a) MMC2 voltage and current. (b) MMC2
active and reactive power.

5.2.3. Disturbance of System Electrical Parameters

During practical operation, the AC-side equivalent impedance of the system generated
parameter perturbations in response to changes in grid conditions, environmental factors,
and load conditions. To analyze the immunity of the system, it was assumed that the AC
inductance on the MMC1 side increased to thrice its original value and reduced to 40% of
the original value at t = 0.2 s. Figures 12 and 13 show the response curves of the d-axis
currents and DC side voltages using the three methods. Under PI control, the system
exhibited a slow response at the initial moment, and produced fluctuations in the d-axis
currents of approximately 93 A when the inductance parameters were changed. However,
the STA and proposed AQST-SMC overcame the problem of slow response at the initial
startup moment of the system. The ability of the proposed AQST-SMC to track the reference
value of the d-axis current in the steady state and after parameter changes was superior
to that of the conventional STA, and the jitter considerably reduced. The DC voltage of
the proposed AQST-SMC responded rapidly and fluctuated to a lesser extent at the time
of parameter ingestion. Furthermore, the voltage was stabilized at the reference value of
40 kV, confirming the superiority of the AQST-SMC strategy.

16
Energies 2024, 17, 2643

100
PI
0 Tradit ional STA
AQST-SM C
Reference value
100
153A
isd A

200

300
idref 245A
339A
400

500
0 0.1 0.2 0.3 0.4 0.5 0.6
time(s)

(a)
100
PI
Traditional STA
0 AQST-SMC
Reference value
100

200
isd A

300
idref 245A
400

500
0 0.1 0.2 0.3 0.4 0.5 0.6
time(s)

(b)

Figure 12. Current response of phase a of MMC1 under parameter perturbation. (a) L tripled. (b) L
reduced to 40% of the original value.

40.4
PI
40.3 Traditional STA
AQST-SMC
40.2 Reference value

40.1
U dc kV

40.0

39.9 40.1

39.8 40.0

39.7
39.9

39.6 0.2 0.22 0.24 0.26 0.28 0.3

0 0.1 0.2 0.3 0.4 0.5 0.6
time(s)

(a)
40.4
PI
Traditional STA
AQST-SMC
40.2 Reference value
U dc kV

40.0

40.1
40.05
39.8 40.0
39.95
39.9
0.2 0.22 0.24 0.26 0.28 0.3
39.6
0 0.1 0.2 0.3 0.4 0.5 0.6
time(s)

(b)

Figure 13. DC voltage response under parameter perturbation. (a) L tripled. (b) L reduced to 40% of
the original value.

17
Energies 2024, 17, 2643

5.3. Validation of Uac f Mode of Operation

When the feeder connected to the FMSS fails, to guarantee the operation of important
loads, a stable AC voltage and grid frequency must be provided to the power loss area;
this is called the Uac f mode of operation. To verify the effectiveness of the AQST-SMC
method during power loss, port 2 was set to fail and the load connected at the end was
purely resistive (value = 120). Figure 14 shows the three-port AC measurement output
response. Figure 14a shows the output voltage and current waveforms of the failed port
2. The output voltage and current were smooth and stable, and the faulty power supply
function was realized. Figure 14b,c show that the voltage and current waveforms of ports
1 and 3 remained stable and unaffected by port 2. This confirmed the feasibility of the
proposed control strategy, and ensured continuous power supply in the faulty area.

usa usb usc

1
isj (p.u.)

0.5
isa isb isc
0
usj (p.u.)

0.5

1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

time(s)

(a)
usausb usc
1

0.5
usj (p.u.) isj (p.u.)

isa isb isc

0.5

1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

time(s)

(b)
usausb usc
1

0.5 isa isb isc

usj (p.u.) isj (p.u.)

0.5

1

0.1 0.15 0.2 0.25 0.3 0.35 0.4

time(s)

(c)

Figure 14. AC side response of ports in the Uac f mode. Voltage and current of ports (a) 2, (b) 1, and
(c) 3.

18
Energies 2024, 17, 2643

6. Conclusions
This study proposes adaptive super-twisting SMC based on feedback-accurate lin-
earization decoupling for the nonlinear and strongly coupled mathematical model of an
FMSS in the distribution network and the difficulty in the traditional PI double-closed-loop
control parameter tuning. The main conclusions are as follows:
(1) The three-terminal FMSS nonlinear radiation model was transformed into a linear
model via exact feedback linearization and applied to the FMSS inner-loop control design.
This completely decoupled the output power, achieved a simple controller design, and
improved the operating characteristics of the system.
(2) To address the large jitter vibration and low convergence speed of the first-order
SMC, STA control was introduced. AQST-SMC was designed using the sigmoid function
and adaptive proportional term to realize better jitter vibration suppression and higher
convergence speed, respectively. The inner- and outer-loop composite control system of
the FMSS was designed to effectively improve the dynamic response speed, robustness,
and steady-state characteristics of the system.
(3) The proposed method improved the transient and steady-state characteristics of
the system, with good anti-interference ability under several working conditions with high
incidence rates. This can provide insights into innovative solutions for the application of
FMSSs in smart distribution networks and new power systems.
Although this study implemented the AQST-SMC algorithm in the Udc Q and PQ oper-
ating modes with the corresponding validation analyses, the dynamic process of switching
between different operating modes has not yet been considered. In the future, the different
operating conditions of FMSSs should be investigated from the perspective of mode switch-
ing. In addition, the proposed AQST-SMC should be applied to the virtual synchronous
control of new grid-type energy sources to improve its robustness and response speed.

Author Contributions: W.M. and X.W.: methodology, software, validation, formal analysis,
writing—original draft preparation, supervision, and writing—review and editing; Y.W.: project
administration and funding acquisition; W.Z., H.L. and Y.Z.: investigation, resources, and data
curation. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China, grant
number 52277208.
Data Availability Statement: Data are contained within the article.
Acknowledgments: The authors would like to express their gratitude to all those who helped them
during the writing of this paper. The authors would like to thank the reviewers for their valuable
comments and suggestions.
Conflicts of Interest: The authors declare no conflicts of interest.

Appendix A
For the actual situation of the system, the constant-value disturbance can be a special
case of the time-varying disturbance [31]. Simultaneously, the adaptive term evolved from
the proportional term, ks; thus, the quadratic-like Lyapunov function was used to perform
the STA stability proofs in terms of the time-varying perturbation.
First, (17) was replaced using the following variables:
⎧
⎪
⎪
x1 =s
⎨ .
x1 = u = −α| x1 |1/2 sd( x1 ) − kx1 + x2
. (A1)
⎪
⎪ x =ω
⎩ .2 .
x2 = − βsd( x1 ) + ϕ

The following class of quadratic Lyapunov functions [32] was selected:

V ( x1 , x2 ) = ζ T Pζ, (A2)

19
Energies 2024, 17, 2643

   
where ζ T = ζ 1 ζ 2 = | x1 |1/2 sd( x1 x2 ; real symmetric positive definite matrices P
and ζ satisfy the following:
⎡ 2 ⎤
4β + α + k | x1 |1/2 − α + k| x1 |1/2
P=⎣ ⎦, (A3)
− α + k| x1 |1/2 2
 
−2
1 1
. 1
−α| x1 | 2 sd( x1 ) − kx1 + x2
2 | x1 |
ζ = .
− βsd( x1 ) + ϕ , (A4)
= 2|ζ | ( Aζ + Bφ)
1
1
   
− α + k| x1 |1/2 1 0 .
where A = ,B= , φ = 2|ζ 1 | ϕ.
−2β 0 1
According to (17), V(x1 , x2 ) is a radially unbounded continuous positive definite
function, and V(x1 , x2 ) is differentiable to any point except the set {x1 = x2 = 0}, for which
the following derivation was obtained:
. .T .
V ( x1 , x2 ) = ζ Pζ + ζ T Pζ 
= 2|1ζ | ( Aζ )T + ( Bφ)T Pζ + ζ T P 2|1ζ | ( Aζ + Bφ)
1  1 
= 2|1ζ | ζ T AT Pζ + φT BT Pζ + ζ T PAζ + ζ T PBφ . (A5)
1
 T  T  
ζ A P + PA PB ζ
= 2|1ζ |
1 φ BT P 0 φ

φ is a scalar, and an expansion of BT Pζ reveals that it is also a scalar; therefore, we

obtain the following:
T T
BT Pζ = BT Pζ = ζ T PT BT = ζ T PB. (A6)

Taking n = BT Pζ = ζ T PB and n2 = ζ T PBBT Pζ, this follows from the inequality

shown below:
(n − φ)2 = n2 − 2nφ + φ2 ≥ 0, ∀n, φ ∈ R. (A7)
Through (A7), the following results:

ζ T PBBT Pζ ≥ φT BT Pζ + ζ T PBφ − φ2 . (A8)

.
By (19), the perturbation term was assumed to be Lipschitz continuous 2 ϕ ≤ δ and
δ > 0.
.
Combined with φ = 2|ζ 1 | ϕ, this can be obtained by simplifying the following inequal-
ity operation:
δ2 ζ 12 − φ2 ≥ 0. (A9)
At C = [1 0], the following can be obtained:

δ2 ζ 12 = δ2 ζ T CT Cζ. (A10)

20
Energies 2024, 17, 2643

Combining (A5) through (A10), the following conclusions were drawn:

 T   
. ζ AT P + PA PB ζ
V ( x1 , x2 ) = 1
2| ζ 1 |
φ BT P
T  T
0 φ
  
ζ A P + PA PB ζ
≤ 2|1ζ | + δ2 ζ 12 − φ2 . (A11)
1 φ BT P 0 φ
 
≤ 2|1ζ | ζ T AT P + PA + δ2 CT C + PBBT P ζ
1
= − 2|1ζ | ζ T Qζ
1

 
Let AT P + PA + δ2 CT C + PBBT P = − Q < 0; therefore, according to Equation (A11),
we obtain the following:
. 1 T
V ( x1 , x2 ) = − ζ Qζ < 0. (A12)
2| ζ 1 |
Equation (A12) shows that the Lyapunov function of the system satisfies the stability
conditions in the Lyapunov stability theory, which is radially unbounded, and a < b.
However, when a time-varying perturbation exists, it is not guaranteed that a symmetric
positive definite matrix P can be found to satisfy the algebraic equations for any positive
definite matrix, Q. However, the idea of (15) can be used to find the range of values of the
control parameter that ensures the finite time convergence of the system, such that Q is
positive definite.
By solving the calculation for the Q matrix, the following expansion was obtained:
 T 
Q =−
⎡ A P + PA + δ C C + PBB P
2 T T
⎤
3 2 2
1/2 1/2 1/2 1/2 1/2
⎢ 2 α + k | x1 | − α + k | x1 | + 4β α + k| x1 | − δ2 2 α + k | x1 | − 2 α + k | x1 | ⎥ . (A13)
=⎣ 2 ⎦
1/2
2 α + k | x1 | − 2 α + k| x1 |1/2 2 α + k | x1 | 1/2
−4
.
If Q is nonnegative, then V < 0; it can be deduced that a sufficient condition for Q to
be a symmetric positive definite matrix is the following:
⎧ 1/2
⎨ α + k | x1 |
⎪ >2
3
α+k| x1 |1/2 +δ2 α+k| x1 |1/2 −2δ2 . (A14)
⎪
⎩ β>  
4 α+k| x1 |1/2 α+k| x1 |1/2 −2

Considering that k > 0, and there must be | x1 |1/2 ≥ 0 in the FMSS system, (A14) can
be transformed as follows:
⎧
⎪
⎪ α>0
⎪
⎨ 2
α+k | x1 |1/2 δ2
β> + . (A15)
⎪
⎪ 4 α+k| x1 |1/2 −8 4α
⎪
⎩
k>0

According to Lyapunov’s stability theorem, the system is stable when the conditions
.
in (A15) are satisfied, and there is a symmetric positive definite matrix Q in that V < 0.
The traditional first-order SMC used in this paper is the exponential convergence law
and the integral slip mode surface, the expression of which is shown below:
⎧
⎨ ν1 = Udcref −
 Udc
s = ν1 + c1 ν1 dt . (A16)
⎩ .1
s = −εsat(s1 ) − qs1 ε, q > 0

where s1 represents the sliding-mode surface, c1 is the sliding-mode surface control param-
eter, ν1 is the DC voltage error signal of the FMSS system, and ε and q are the exponential

21
Energies 2024, 17, 2643

convergence law control parameters. Saturation function sat(s1 ) was selected for reducing
the system jitter.

References
1. Chen, Y.; Tang, Z.; Weng, X.; He, M.; Zhang, G.; Yuan, D.; Jin, T. A novel approach for evaluating power quality in distributed
power distribution networks using AHP and S-transform. Energies 2024, 17, 411. [CrossRef]
2. Liu, Y.; Hu, J.; Shi, J.; Yang, B. Adaptive integrated control strategy for MMC-MTDC transmission system considering dynamic
frequency response and power sharing. J. Electr. Eng. Technol. 2023, 147, 108858. [CrossRef]
3. Ivic, D.R.; Stefanov, P.C. An extended control strategy for weakly meshed distribution networks with soft open points and
distributed generation. IEEE Access 2021, 9, 137886–137901. [CrossRef]
4. Mardanimajd, K.; Karimi, S.; Anvari-Moghaddam, A. Voltage stability improvement in distribution networks by using soft open
points. J. Electr. Eng. Technol. 2024, 155, 109582. [CrossRef]
5. Li, Z.; Ye, Y.; Wang, Z.; Wu, Y.; Xu, H. Integrated planning and operation evaluation of micro-distribution network based on
flexible multistate switch interconnection. CES Trans. Electr. Mach. Syst. 2021, 36, 10. [CrossRef]
6. Hou, Y.; Xu, Y.; Wang, Z.; Chen, X.; Cui, H. Research on application of three-port SNOP based on dual closed-loop control in
the distribution network. In Proceedings of the IEEE International Conference on Smart Technologies and Management for
Computing, Chennai, India, 2–4 August 2017; pp. 389–395. [CrossRef]
7. Wang, Z.; Sheng, L.; Huo, Q.; Hao, S. An improved model predictive control method for three-port soft open point. Math. Probl.
Eng. 2021, 2021, 9910451. [CrossRef]
8. Peng, B.; Zhang, G. Coordination control strategy for three-port SNOP based on FCS-MPC. J. Eng. 2019, 2019, 1005–1010.
[CrossRef]
9. Aithal, A.; Li, G.; Wu, J.; Yu, J. Performance of an electrical distribution network with soft open point during a grid side AC fault.
Appl. Energy 2018, 227, 262–272. [CrossRef]
10. Li, Z.; Hao, Q.; Gao, F.; Wu, L.; Guan, M. Nonlinear decoupling control of two-terminal MMC-HVDC based on feedback
linearization. IEEE Trans. Power Deliv. 2019, 34, 376–386. [CrossRef]
11. Zhang, G.; Tang, B.; Shen, C.; Wang, T.; Xu, C.; Xia, Z. A strategy for suppression of voltage fluctuation on DC side when SOP
port is unbalanced. Electr. Meas. Instrum. 2021, 61, 95–101. [CrossRef]
12. Li, B.; Liang, Y.; Wang, G.; Li, H.; Ding, J. A control strategy for soft open points based on adaptive voltage droop outer-loop
control and sliding mode inner-loop control with feedback linearization. Int. J. Electr. Power Energy Syst. 2020, 122, 106205.
[CrossRef]
13. Zhou, M.; Su, H.; Zhou, H.; Feng, Y.; Wang, D.; Cheng, J. Full-order sliding mode control strategy for soft open point. Proc. CSEE
2023, 43, 8622–8635. [CrossRef]
14. Liu, Y.C.; Laghrouche, S.; Depernet, D.; N’Diaye, A.; Djerdir, A.; Cirrincione, M. Disturbance-observer-based speed control
for SPMSM drives using modified super-twisting algorithm and extended state observer. Asian J. Control 2024, 26, 1089–1102.
[CrossRef]
15. Xu, Q.; Jiang, D.; Wang, Y.; Zhang, X.; Liu, J.; Chen, Y. Variable-step close-loop angle compensation method of PMSM rotor
position estimation based on super-twisting sliding-mode observer using tangent reaching law. Energy Rep. 2023, 9, 356–361.
[CrossRef]
16. Maged, N.A.; Hasanien, H.M.; Ebrahim, E.A.; Tostado-Véliz, M.; Jurado, F. Optimal super twisting sliding mode control strategy
for performance improvement of islanded microgrids: Validation and real-time study. Int. J. Electr. Power Energy Syst. 2024, 157,
109849. [CrossRef]
17. Ghazi, G.A.; Al-Ammar, E.A.; Hasanien, H.M.; Ko, W.; Lee, S.M.; Turky, R.A.; Tostado-Véliz, M.; Jurado, F. Circle search
algorithm-based super twisting sliding mode control for MPPT of different commercial PV modules. IEEE Access 2024, 12,
33109–33128. [CrossRef]
18. Saadaoui, A.; Ouassaid, M. Super-twisting sliding mode control approach for battery electric vehicles ultra-fast charger based on
Vienna rectifier and three-phase interleaved DC/DC buck converter. J. Energy Storage 2024, 84, 110854. [CrossRef]
19. Çelik, D.; Ahmed, H.; Meral, M.E. Kalman filter-based super-twisting sliding mode control of shunt active power filter for electric
vehicle charging station applications. IEEE Trans. Power Deliv. 2023, 38, 1097–1107. [CrossRef]
20. Tiwary, N.; Naik, N.V.; Panda, A.K.; Narendra, A.; Lenka, R.K. A robust voltage control of DAB converter with super-twisting
sliding mode approach. IEEE J. Emerg. Sel. Top. Ind. Electron. 2023, 4, 288–298. [CrossRef]
21. Wang, F.; Yang, A.; Yu, X.; Zhang, Z.; Wang, G. Model-free predictive current control for three-phase Vienna rectifier based on
adaptive super-twisting sliding mode observer. Trans. China Electrotech. Soc. 2024, 39, 1859–1870. [CrossRef]
22. Pati, A.K.; Sahoo, N.C. Adaptive super-twisting sliding mode control for a three-phase single-stage grid-connected differential
boost inverter based photovoltaic system. ISA Trans. 2017, 69, 296–306. [CrossRef]
23. Hou, J.; Wang, Z.; Zhou, H.; Xu, H.; Wei, W. Soft open point control strategy based on super-twisting-full-order fast terminal
composite sliding mode. Autom. Instrum. 2023, 3, 143–147. [CrossRef]
24. Wang, Z.; Zhou, H.; Su, H. Disturbance observer-based model predictive super-twisting control for soft open point. Energies 2022,
15, 3657. [CrossRef]

22
Energies 2024, 17, 2643

25. Ma, W.; Lü, Q.; Zhang, Y.; Zhang, K.; Han, J.; Li, M. Suppression strategy for fault current of flexible multi-state switch in
distribution network fault state. Power Syst. Technol. 2021, 45, 4251–4258. [CrossRef]
26. Zhang, K.; Wang, L.; Fang, X.; Zhang, K.; Han, J.; Li, M. High-order fast nonsingular terminal sliding mode control of permanent
magnet linear motor based on double disturbance observer. IEEE Trans. Ind. Appl. 2022, 58, 3696–3705. [CrossRef]
27. Li, P. Convergence of super-twisting algorithm based on quadratic-like Lyapunov function. Control Decis. 2011, 26, 949–952.
[CrossRef]
28. Levant, A. Sliding order and sliding accuracy in sliding mode control. Int. J. Control 1993, 58, 1247–1263. [CrossRef]
29. Yang, H.; Bai, Y.; Chien, Y.R. Generalized super-twisting sliding mode control of permanent magnet synchronous motor based on
sinusoidal saturation function. IEICE Electron. Express 2022, 19, 20220066. [CrossRef]
30. Fu, K.; Yi, H.; Zhuo, F.; Li, S. Self-healing control strategy based on snop under severe fault condition. In Proceedings of the 2021
IEEE Sustainable Power and Energy Conference (iSPEC), Nanjing, China, 23–25 December 2021. [CrossRef]
31. Li, X.; Liu, J.; Yin, Y.; Zhao, K. Improved super-twisting non-singular fast terminal sliding mode control of interior permanent
magnet synchronous motor considering time-varying disturbance of the system. IEEE Access 2023, 11, 17485–17496. [CrossRef]
32. Tripathi, V.K.; Kamath, A.K.; Behera, L.; Verma, N.K.; Nahavandi, S. Finite-time super twisting sliding mode controller based on
higher-order sliding mode observer for real-time trajectory tracking of a quadrotor. IET Control Theor. Appl. 2020, 14, 2359–2371.
[CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

23
 energies
Article
Grid-Forming Control for Solar Generation System with Battery
Energy Storage
Yupeng Cai 1 , Lujie Yu 2, *, Meng Wu 1 , Shengyang Lv 1 , Ziyu Fu 2 , Wenhao Tong 2 , Wei Li 1 and Songjie Shi 1

1 State Grid Liaoning Electric Power Co., Ltd., Electric Power Science Research Institute,
Shenyang 110006, China; cai8872@126.com (Y.C.); wumeng_ldk@163.com (M.W.); sgdlck@126.com (S.L.);
yzylw2007@126.com (W.L.); shisongjie2009@126.com (S.S.)
2 School of Electrical and Information Engineering, Tianjin University, Tianjin 300072, China;
fuziyu2021@163.com (Z.F.); 15958966510@163.com (W.T.)
* Correspondence: lujie.yu@outlook.com

Abstract: Solar generation systems with battery energy storage have become a research hotspot in
recent years. This paper proposes a grid-forming control for such a system. The inverter control
consists of the inner dq-axis current control, the dq-axis voltage control, the phase-locked loop
(PLL) based frequency control, and the DC voltage control. The proposed control embeds the PLL
into the grid-forming inverter control, offering the advantages of better synchronization and fault
current-limiting capability. With the proposed control, the battery energy storage is able to provide
inertial and primary frequency support during the grid frequency disturbance. Simulation models
are established in PSCAD/EMTDC, and the results during the active power variation and AC voltage
variation, the grid frequency disturbance, grid fault, and mode switch validate the effectiveness of
the proposed control.

Keywords: battery energy storage; frequency control; grid-forming control; inverter control; solar
generation

1. Introduction
The use of fossil energy sources and air pollution issues drive the fast development of
renewable energy. Among all the renewable energy sources, wind power generation and
solar generation are the most attractive ones. In the US, wind generation accounts for 10%
of the total electrical generation while solar generation will take up 5% of the total electrical
generation by the end of 2023. China has set the goal that before 2030 carbon emissions
will reach their peak value and before 2060 carbon neutrality will be achieved. By the first
quarter of 2023, solar generation in China will exceed 228 GW while wind generation in
China will exceed 310 GW. It has been reported that China is now the leader in renewable
energy installation capacity worldwide and holds around 50% of the total wind and solar
capacity.
However, the high penetration of renewable energy sources raises some concerns
about power system stability. For the wind and solar energy sources, they are all connected
to the power grid through power-electronic converters. These converters generally cannot
provide the same inertial or damping as the traditional synchronous generators. In [1,2], the
grid-connected converters for renewable generation sources are classified as grid-following
converters and grid-forming converters.
Among the renewable generation sources in commission, they generally operate on
the grid-following mode [3]. However, with the increased integration of the grid-following
renewable generation sources and the decrease in the grid strength, the converter-driven
oscillation becomes a serious and annoying issue. In [4], a small signal state space model
is established for VSC-HVDC, and eigenvalue results show that the maximum power
generated is significantly affected by the PLL and the system finds it difficult to maintain

Energies 2024, 17, 3642. https://doi.org/10.3390/en17153642 24 https://www.mdpi.com/journal/energies

Energies 2024, 17, 3642

stability when the short circuit ratio is smaller than 1.3. Based on the work in [4], the authors
of [5,6] establish the small signal model for doubly fed induction generator (DFIG)-based
wind turbines and fully rated converter-based wind turbines, finding that the PLL and rotor-
side converter power control has a noticeable effect on the DFIG connected system stability,
while the weak grid, high wind power export, low voltage, and low PLL bandwidth are
the main causes for the stability issues in the latter system. In [7], the harmonic state–space
model considering the PLL under the asymmetrical grid is established and it is observed
that the stability region of the system is reduced during the increase in the impedance
asymmetrical index. In [8,9], an impedance model is built, and a generalized Nyquist plot is
used to determine the system stability. It is observed that the negative resistor characteristic
of qq impedance, which is mainly caused by the PLL, reduces the system stability margin.
In [10], an improved feedforward control method considering PLL dynamics is proposed
to improve the stability of the grid-following converter connected to a weak grid. In [11], a
current error-based compensation control is proposed, where a magnitude compensation
is implemented with a proportional controller whose input is the reactive current, and
an angle compensation is realized with a proportional and integral (PI) controller whose
input is the active current. Both eigenvalues and simulation results show the above control
increases the system stability, especially when the short circuit ratio of the system is small.
In [12], another advanced vector control is proposed for the grid-following converter, where
the error between the active power reference and the measured one is not only the input of
the active power PI controller but also the input of the reactive power PI controller. Such a
control improves the system stability but suffers from the multi-parameter tuning.
An alternative way to increase the system’s small signal stability is to change the
grid-following control to the grid-forming control. In [13,14], a power synchronization
control is proposed using the internal synchronization mechanism in AC systems and tries
to emulate a synchronous machine with the converter and the results show that the control
enables the converter to provide a strong voltage support to the weak grid. In [15,16],
the virtual synchronous generator and droop control-based grid-forming converter are
compared, and the results show the virtual synchronous generator has larger inertia than
the droop control, but when a proper lead–lag component is embedded in the active
power droop control, the same small signal model has been derived with the virtual
synchronous generator model. In [17,18], the sequence impedance model of the virtual
synchronous generator is provided, and the stability analysis of the voltage-controlled
virtual synchronous generator is compared to the current-controlled virtual synchronous
generator. It is concluded that the voltage-controlled virtual synchronous generator is more
similar to the traditional synchronous generator and is more stable under a weak grid
when compared with the current-controlled virtual synchronous generator. In [19,20], the
inertial support and primary frequency support from different types of the grid-forming
converter are compared, and the simplified virtual synchronous compensator shows the
advantage of providing frequency support and damping during the active power change
and frequency variation.
In all, current research focuses on the control and stability analysis of the grid-forming
converters without the PLL, and the synchronization is implemented through the power
control during the power generation mode. But the reality is that the PLL still plays a key
role for the grid-forming converter. The main reasons are as follows: (1) when the offline
renewable generation sources try to connect to the grid, the phase of the grid needs to be
known for smooth connection; (2) when a fault happens, the current limiting control needs
a PLL. Thus, in this paper, a grid-forming control with a reserved PLL is proposed for the
converters for solar generation.
The rest of this paper is organized as follows. Section 2 describes the topology of
the solar generation. Section 3 presents the proposed grid-forming control for the solar
generation converter. Section 4 presents the simulation under the power and voltage
variation, during the grid frequency change, AC fault, and the mode switch from the
grid-connected mode to the islanded mode. Finally, Section 5 draws the conclusion.

25
Energies 2024, 17, 3642

2. Topology of the Studied System

Figure 1 shows the studied grid-connected solar generation, which mainly consists of
the following parts: (1) a photovoltaic (PV) array; (2) a front boost converter; (3) a battery
energy storage system (BESS); and (4) an inverter. The PV array transforms the solar
energy to the electricity, the DC power is injected to the front boost converter, the battery is
connected to the DC bus through a bidirectional DC/DC converter, the DC power collected
from both the PV and the BESS is then converted to AC power through the grid-connected
inverter. The AC voltage at the inverter terminal is filtered by the LC circuit and is increased
to a collection level through the transformer.

39 )URQWERRVWFRQYHUWHU ,QYHUWHU
$&
JULG
iSY iSY/ iRDEF iSDEF
uSY
CSY uGF
SER CGF LI
CI uSDEF3&&

%(66

RBESS iBESS
'&'&
LBESS FRQYHUWHU
UBESS

Figure 1. Topology of the studied system.

3. Proposed Grid-Forming Control for Solar Generation

This section presents the control of the solar generation, mainly includes the control of
the front boost converter, control of the BESS, and the proposed grid-forming control of
the inverter.

3.1. Control of the Front Boost Converter

The dynamics of the capacitor and the inductance of the PV front boost converter can
be expressed as  dupv
Cpv dt = ipv − ipvL
dipvL (1)
Lpv dt = upv − (1 − dpv )udc
where Cpv and Lpv are the capacitor and inductance value, upv and ipv are the voltage
and current of the PV array, ipvL is the inductance current, udc is the DC voltage of the
grid-connected inverter, and dpv is the duty cycle of the front boost converter. Based on the
dynamics of the capacitor and the inductance of the PV front boost converter, the control
of the front boost converter for the maximum power point tracking (MPPT) is shown in
Figure 2 and can be expressed as
 
kib ∗
dpv = kpb + upv − upv (2)
s

where upv∗ is the voltage reference for the u , which is obtained from the MPPT, and k
pv pb
and kib are the proportional and integral parameters of the controller.

26
Energies 2024, 17, 3642

)URQWERRVWFRQYHUWHU

uSY iSY iSY/

uGF
uSY uSY d SER uSY
iSY 0337 3, 3:0 CSY
SER

Figure 2. Control of the front boost converter.

3.2. Control of the BESS

For the BESS, the control structure consists of the two layers, as shown in Figure 3. The
upper-level control takes the Deblock signal, state of charge (SOC), and discharge/charge
mode as the input and outputs the signal Enboost, Enbuck, and PrefBESS . Figure 4 shows
the flow chart of the control logic. “Enboost” works under the following conditions: (1) The
Deblock signal is equal to 1, which indicates the DC/DC converter is not blocked. (2) The
SOC is larger than 5%, which indicates the battery has enough stored energy to support
the AC grid. (3) The discharge signal is equal to 1, which indicates the power needs to be
transmitted from the battery to the grid and the DC/DC converter should operate on the
boost control mode. On the other hand, “Enbuck” works under the following conditions:
(1) The Deblock signal is equal to 1, which indicates the DC/DC converter is not blocked.
(2) The SOC is less than 100%, which indicates the battery has the capacity to absorb
the energy. (3) The charge signal is equal to 1, which indicates the power needs to be
transmitted from the grid or solar generation to the battery and the DC/DC converter
should operate on the buck control mode.

'HEORFN 62& 'LVFKDUJH

&KDUJHPRGH

'HEORFN  'HEORFN 
(QERRVW (QEXFN
62&! 62&
'LVFKDUJH $1' &KDUJH $1'
PRGH  PRGH 
f
'
P%UHI
f P
+ 8SSHUOHYHO
df/dt FRQWURO

(QERRVW P%UHI (QEXFN

(QERRVW %RRVWFRQWURO (QEXFN %XFNFRQWURO

P%(66 P%(66
uUHI SERRVW P uUHI SEXFN
P%UHI %UHI
3, 3:0 3, 3:0

/RZHUOHYHO S
FRQWURO
%(66

R%(66 i%(66
'&'&
L%(66 FRQYHUWHU
U%(66

Figure 3. Control of the BESS.

27
Energies 2024, 17, 3642

6WDUW

,QSXW62&'HEORFN
'LVFKDUJH&KDUJHPRGHVLJQDOV

1
'HEORFN ˛
<

'LVFKDUJH 1
PRGH ˛
<
&KDUJHPRGH ˛
1
62&!˛ <
< 1
62&˛
7KHERRVWFRQWURORI%(66LV
<
DFWLYDWHG
7KHEXFNFRQWURORI%(66LV
DFWLYDWHG

%(66EORFNHG

Figure 4. Flow chart of the control logic.

The dynamic of the resistor and the inductance can be expressed as

diBESS
LBESS = uBESS − ut + RBESS iBESS (3)
dt
where LBESS and RBESS are the inductance and resistance of the BESS, uBESS is the voltage of
the battery, and iBESS is the output current of the battery. The control for the bidirectional
DC/DC converter in the BESS is expressed as

uref1 = ( PBref − PBESS )(kpdc1 + kidc1
s ) Boost/discharging mode
kidc2 (4)
uref2 = ( PBref + PBESS )(kpdc2 + s ) Buck/charging mode

where PBESS is the active power of the BESS, PBref is the active power reference of the BESS,
kpdc1 and kidc1 are the proportional and integral parameters for the power control in the
boost/discharging mode, kpdc2 and kidc2 are the proportional and integral parameters for
the power control in the buck/charging mode, and uref1 and uref2 are the voltage references
for the pulse modulation width (PMW) in the two different modes, respectively.
When the grid frequency changes, in order to support the system inertia and primary
frequency, the BESS active power reference PBref is determined by the upper-level control as

df
PBref = H + D ( f − f 0 ) + P0 (5)
dt
where H is the inertial time constant, D is the parameter coefficient for the primary fre-
quency support, f is the grid frequency, f 0 is the nominal frequency, and P0 is the active
power feedforward term.

3.3. Grid-Forming Control of the Inverter

The inverter control mainly consists of the inner current control, inner voltage control,
PLL-based frequency control and the DC voltage control, as shown in Figure 5.

28
Energies 2024, 17, 3642

)URQWERRVW
39 FRQYHUWHU ,QYHUWHU QZW $&
iGF7 iGF
PZW JULG
iSY iSY/ iZDEF u iVDEF
uSY IDEF iZDEF iVDEF
uGF șI șI
CSY DEF dq DEF dq
SER CGF LI
CI uSDEF iZd iZq iVd iVq
uIDEF
%(66 DEF dq
Ȧ
RBESS iBESS '&'& uId uIq +
abc dq  șI
LBESS FRQYHUWH șI 3, + Ȧ
UBESS U s
f
uId ±
*
uFG*   iwd     uGF
*

B 3, B 3, B kI 3, B
uIT* B f* 
ȦLIiZT iZG uIT uGF
ȦLIiZG f f
*
  iwq 
3, B 3, B
uFT*  iZT uIG*
uIq uIG

Figure 5. Proposed grid-forming control of the inverter.

For the grid-connected inverter, the dynamic of the inductance in the dq frame is
expressed as 
Lf didtwd = ucd − ufd + ωLf iwq − Rf iwd
di (6)
Lf dtwq = ucq − ufq − ωLf iwd − Rf iwq
where Lf and Rf are the inductance and resistance of the grid-connected converter, ucd and
ucq are the dq-axis voltage at the inverter terminal, ufd and ufq are the dq-axis voltage at the
inverter filter, iwd and iwq are the dq-axis converter current, and ω is grid angular frequency.
Thus, the inner current control is designed as
 
∗ = (k + kii ∗ −i
ucd pi s ) iwd wd − ωLc iwq + ufd
∗ = (k + kii ∗ −i (7)
ucq pi s ) iwq wq + ωLc iwd + ufq

where ucd * and ucq * are the dq-axis inverter voltage references, iwd * and iwq * are the active
and reactive current references, iwd and iwq are the dq-axis inverter current, kpi and kii are
the parameters of the inner current control.
On the other hand, when the inverter behaves like an AC voltage, the active power
and the reactive power Pinv and Qinv can be expressed as
 Uf E sin(θf −θe )
Pinv = Xac
(U − E cos(θf −θe ))Uf (8)
Qinv = f Xac

where Uf is the RMS value of the line-to-line voltage of the inverter, E is the RMS value
of the line-to-line voltage of the grid, θ f and θ f are the phase angles of the voltages at the
inverter filter and the AC grid, respectively, and Xac is the inductance between the inverter
and the grid.
Considering that the relationship between the active power and reactive power and
the dq-axis inverter current is

Pinv = 1.5ufd iwd + 1.5ufq iwq
(9)
Qinv = −1.5ufd iwq + 1.5ufq iwd

and the θ f is mainly determined by the ufq , as shown in Figure 6, the d-axis current can
be used to adjust the ufq , while the q-axis current can be used to adjust the ufd , thus the
voltage control of the inverter is designed as

∗ = (k kiv1 ∗ −u
iwd pv1 + s ) ufq fq
∗ = (k kiv2 ∗
 (10)
iwq pv2 + s ) ufd − ufd

29
Energies 2024, 17, 3642

where ufd * and ufq * are the dq-axis inverter voltage references, kpv1 and kiv1 are the pa-
rameters of the q-axis voltage control parameters, and kpv2 and kiv2 are the parameters
of the d-axis voltage control parameters. Generally, the values of kpv1 , kiv1 , kpv2 , and kiv2
are chosen to ensure that the bandwidth of the voltage control loop is smaller than that
of the current control loop for the sake of better system stability. In this paper, the inner
current control bandwidth is designed as 300 Hz, which is smaller than f s /10, where f s is
the sampling frequency of the inverter and the voltage control bandwidth is designed as
one forth of the inner current control bandwidth.

qD[LV
vIT vI ,QYHUWHU
YROWDJH
vIG
ș dD[LV
E
*ULGYROWDJH
Figure 6. Voltage vector of the inverter and grid.

For the PLL-based frequency control, it is designed as

∗
ufq = kf ( f ∗ − f ) (11)

where f * is the frequency reference, kf is the proportional parameter for the frequency
control. The frequency f obtained from the PLL is expressed as

k
ωpll = kpp ufq + sip ufq
ωpll (12)
f = 2π

where kpp and kip are the parameters of the PLL and ω pll is the frequency of the PLL. When
the inverter frequency reference is larger than the measured one, the ufq * is larger than 0.
With the combined action of the ufq control loop and the PLL, the frequency and the phase
angle will be increased. Thus, more active power can be exported. In contrast, when the
inverter frequency reference is less than the measured one, the ufq * is smaller than 0. With
the combined action of the ufq control loop and the PLL, the frequency and the phase angle
will be decreased. Thus, less active power will be exported.
For the inverter DC capacitor, the dynamic can be expressed as

dudc
Cdc = idcT − idc (13)
dt
where Cdc is the DC capacitor value of the inverter, idcT is the DC current from the PV and
BESS, and idc is the DC current of the inverter. As seen, the DC voltage is highly related to
the DC current, which is related to the active power. Considering the connection between
the frequency and the active power, the DC voltage control for the inverter is expressed as
 
∗ 1 ∗
f = kpdc + kidc (udc − udc ) + f 0 (14)
s
∗ is the inverter DC voltage reference and k
where udc pdc and kidc are the parameters of the
DC voltage control.
When a fault happens, the inverter switches from the grid-forming mode to the
current-limiting mode automatically. The dq-axis current references of the inverter are
based on the look-up table. The surplus power is absorbed by the BESS and when the
DC voltage of the inverter is higher than the threshold value, the chopper inside the
inverter will be triggered and the power will also be consumed partly by the resistor in
the chopper. When the studied system is disconnected with the AC grid, the inverter is

30
Energies 2024, 17, 3642

able to work on the islanded mode, as the control provides the frequency and AC voltage
establishment capability.
Compared with the existing grid-forming control without the PLL, the proposed
control strategy reserves the PLL. The main advantages of the proposed control are (1) when
the offline renewable generation sources try to connect to the grid, the phase of the grid
can be easily obtained for smooth connection and (2) when a fault happens, the current
limiting control can be achieved smoothly through the automatically switching from the
grid-forming control to the inner current control loop combined with PLL.

4. Simulation Result
In order to validate the effectiveness of the proposed control, a simulation model is
established in PSCAD/EMTDC, as shown in Figure 7. The solar generation is modelled as
aggregated model, which is rated at 300 MW. The battery energy storage (BES) is connected
to the DC side of the solar generation through a buck/boost converter. The DC voltage of
the battery is at 0.5 kV and the inverter DC voltage is at 1.25 kV. The rated capacity of the
BES is 16.7 (kA·h). The other detailed parameters are presented in Table 1.

Figure 7. Simulation model in PSCAD/EMTDC of the studied system.

Table 1. Parameters of the system.

Parameters Values
DC voltage 0.5 kV
Rated Capacity 16.7 (kA·h)
Resistive Drop 0.001 pu
BESS Voltage at Exponential Point 1.03 pu
0.2 μH
Inductance and capacitor in DC/DC 3000 mF
0.5 μH
Rating 300 MVA
DC voltage 1.25 kV
AC voltage 0.69 kV
DC capacitor 30,000 mF
Inverter
Transformer ratio 0.69/33 kV
Transformer inductance and resistor 0.08, 0.002 pu
Converter reactance and resistor 0.15 pu, 0.0015 pu
Filter capacitor and resistor 0.10 pu, 0.001 pu

4.1. Performance Evaluation of the Grid-Forming Control during Power Variation and AC
Voltage Variation
The performance of the grid-forming control is evaluated during the solar power
generation variation and AC voltage variation. The timeline of the simulation is presented
in Table 2.

31
Energies 2024, 17, 3642

Table 2. Sequence of simulation.

Time Events
0–3 s Solar power generation at 0.1 pu, BES operation off
3–3.5 s Solar power generation increases from 0.1 pu to 1 pu
4.5–4.6 s Solar power generation decreases from 1 pu to 0.2 pu
6–6.2 s Inverter AC voltage decreases from 1 pu to 0.98 pu
8–8.2 s Inverter AC voltage increases from 0.98 pu to 1.02 pu

Figure 8 shows the results of the system during such variations. As can be seen from
Figure 8a, the solar power is initially generated at 30 MW before 3 s, and from 3 s to 3.5 s,
the power is increased from 30 MW to 300 MW. During the active power increase, the
inverter DC voltage and AC voltage can both be maintained at 1 pu, as shown in Figure 8c,d.
The reactive power from the inverter sees a small increase during the power variation
and is maintained at 8 MVar after 3.5 s. The magnitude of the AC current is increased
from 0.035 MA to 0.35 MA, as presented in Figure 8e,f, which shows the DC current of the
inverter, and the DC current is increased from 0.024 MA to 0.24 MA during 3 s to 3.5 s.
From 4.5 s to 4.6 s, the active power generation is decreased from 300 MW to 60 MW,
as can be seen in Figure 8a. The DC current of the inverter is correspondingly decreased
from 0.24 MA to 0.048 MA, as can be seen in Figure 8f. Figure 8c,d shows the DC voltage
and AC voltage can be maintained at 1 pu after such an active power variation, although a
small voltage oscillation is observed during the active power variation. Figure 8b shows
the inverter reactive power. As can be seen, the reactive power is decreased from 8 MVAr
to 0 MVAr from 4.5 s to 4.6 s. The magnitude of the inverter AC current decreases from
0.24 MA to 0.07 MA, as seen in Figure 8e.
From 6 s to 6.2 s, the inverter AC voltage reference decreases from 1 pu to 0.98 pu. As
seen in Figure 8d, the measured AC voltage follows the reference and reaches 0.98 pu at
6.6 s. The measured AC voltage follows the reference decrease and shows the over-damped
dynamic, indicating the strong stability of the voltage control. With the voltage variation,
the inverter reactive power is decreased from 0 MVar to —100 MVar, as shown in Figure 8b.
The inverter AC current magnitude is increased from 0.07 MA to 0.145 MA, as shown in
Figure 8e. From 8 s to 8.2 s. the inverter AC voltage reference is increased from 0.98 pu to
1.02 pu, and the measured AC voltage follows the reference increase and is controlled at
1.02 pu after 8.5 s, as presented in Figure 8d. The inverter reactive power is increased from
−100 MVAr to 105 MVAr, as shown in Figure 8b. The results are matched by the fact that
when the AC voltage magnitude is increased, the inverter will output more reactive power.
During the AC voltage variation, the AC current sees the initial decrease and then increase,
as presented in Figure 8e. Such a performance is caused by the change in the absolute value
of the reactive power. Figure 8a,c,f shows the performance of the inverter’s active power,
DC voltage, and DC current, respectively. As can be seen, all these variables experience
little change during the inverter AC voltage variation.

32
Energies 2024, 17, 3642



D P 0: 





E Q 09$U








F VGF SX







G VDF SX




    
t V



H IDF 0$







I IGF 0$






    
t V

Figure 8. Simulation results during power variation and AC voltage variation: (a) inverter active
power; (b) inverter reactive power; (c) inverter DC voltage; (d) inverter AC voltage (RMS); (e) inverter
instantaneous AC current; (f) inverter DC current.

4.2. Performance Evaluation of BESS during the Grid Frequency Change

The performance of the BESS is evaluated during the grid frequency change and
the corresponding results are presented in Figure 9. It is worth mentioning that the time
sequence of this part is different from that of Section 4.1, so the simulation time in Figure 9
can be different from Figure 8.
From 2 s to 3 s, the grid frequency is decreased from 60 Hz to 59.6 Hz, as shown in
Figure 9a. Due to the inertial and frequency support capability of the BESS, the inertia time
constant of the system is 10 s and the primary frequency coefficient is 50 MW/Hz, and the
active power from the BESS is increased from 0 MW to 20 MW, as presented in Figure 9b.
Figure 9c,d shows the voltage of the BESS and the current of the BESS, respectively. As
can be seen, the voltage is decreased from 556.8 V to 551 V and the current of the BESS is
increased from 0 kA to 39 kA. Due to the discharging of the BESS, the SOC of the battery
is decreased from the initial value of 94.98% at 2 s to 96.96% at 3 s, as shown in Figure 9e.
With the power support from the BESS, the active power of the inverter is increased from
the initial value of 150 MW to 170 MW, as shown in Figure 9f. It should be noted the initial

33
Energies 2024, 17, 3642

power of 150 MW before 3 s is exported by solar generation. During the power increase
from the BESS, the inverter DC voltage is increased from 1.25 kV to 1.272 kV, as shown in
Figure 9g. Figure 9h shows the AC voltage of the inverter. As can be seen, the AC voltage
is maintained at a constant value.



D f +=





E PBES 0:








F VBESS 9







G IBESS N$






H 62& SX






I PLQY 0:








J VGF N9





    
t V


K VLQY SX




    
t V

Figure 9. Simulation results of BESS during the grid frequency change: (a) grid frequency; (b) battery
power; (c) battery voltage; (d) battery current; (e) state of charge; (f) inverter active power; (g) inverter
DC voltage; (h) inverter AC voltage.

After 3 s, the grid frequency is constant at 59.6 Hz, as shown in Figure 9a. At this stage,
the BESS keeps providing the active power for the primary frequency regulation and the
exported active power is at 20 MW, as presented in Figure 9b–d, which demonstrates the
voltage of the BESS and the current of the BESS, respectively. As can be seen, the voltage

34
Energies 2024, 17, 3642

and the current are maintained at a constant value of 551 V and 39 kA. The SOC of the
battery keeps decreasing and is at 94.84% at 5 s, as shown in Figure 9e–h, which shows that
the inverter active power, DC voltage, and AC voltage can all operate stably after 3 s, at
170 MW, 1.25 kV, and 1 pu, respectively.

4.3. Performance Evaluation of System during the Grid Fault

The performance of the system during the grid fault is assessed and the results are
presented in Figure 10. Before 3 s, the system operates at rated values with the active
power at 300 MW, while the reactive power from the inverter is around 0 MVAr, as shown
in Figure 10d,e. At 3 s, a three-phase fault happens and the voltage of the converter is
decreased to 0.13 pu, as shown in Figure 10a,b, which shows the grid side instantaneous
voltage, the magnitude of the voltage decreases from 27 kV to 3.4 kV, and the voltage
is dominated by the fundamental frequency component. After the severe AC fault, the
converter operates on the current-limiting control mode, and the converter AC current
magnitude is controlled at 0.59 MA, as shown in Figure 10c. After 200 ms, the fault is
cleared. As can be seen in Figure 10a,b, the voltage recovers to 1 pu. The inverter is no
longer working on the current-limiting mode and recovers to transmit the active power. As
seen in Figure 10d, the active power generation is quickly restored after the fault clearance.
Figure 10e shows the reactive power recovery performance, which can also be restored to
0 MVAr, although its recovery speed is slower than the active power generation.



D 9LQY SX








E YLQY N9






F LLQY 0$





G 3LQY 0:







H 4LQY 09$U






     
W V

Figure 10. Simulation results of the system during the grid fault: (a) grid PCC voltage (RMS); (b) grid
PCC instantaneous voltage; (c) inverter current; (d) inverter active power; (e) inverter reactive power.

35
Energies 2024, 17, 3642

4.4. Performance of System during the Transition from the Grid-Connected Mode to Islanded Mode
The performance of the system during the transition from the grid-connected mode
to the islanded mode is evaluated and the results are presented in Figure 11. Before 3 s,
the system works on the rated value, with the AC voltage at 1 pu, frequency at 60 Hz,
DC voltage at 1.25 kV, active power at 300 MW, and reactive power at 3 MVar, as shown
in Figure 11a–e. At 3 s, the inverter is disconnected from the grid. As can be seen in
Figure 11a,b, the voltage and frequency experience some oscillation after the operation
mode switch but can stabilize after 0.8 s. Figure 11e shows the reactive power output from
the inverter. As seen, before and after the mode switch, the reactive power is decreased
from 3.3 MVAr to 2 MVAr. Figure 11f shows the inverter current. As can be seen, the
current is not disturbed by the mode switch.

D VLQY SX





    
tV



E f +=






F VGF N9





G 3LQY 0:



噿 



H 4LQY 09$U





I LLQY 0$





    
WV

Figure 11. Simulation results during the transition from the grid-connected mode to islanded mode:
(a) inverter AC voltage; (b) inverter frequency; (c) inverter DC voltage; (d) inverter active power;
(e) inverter reactive power; (f) inverter AC instantaneous current.

36
Energies 2024, 17, 3642

4.5. Performance of System during Phase Jump and Grid Impedance Variation
The performance of the system during phase jump is evaluated and the results are
presented in Figure 12. At 3 s, the phase of the grid experiences a sudden jump, with a
relatively large variation value at 30◦ , as seen in Figure 12a. The active power and reactive
power of the inverter drift away from the initial steady-state value after such a disturbance
but can stabilize after 0.1 s, as shown in Figure 12b–e, which shows the DC voltage and AC
voltage, respectively. As presented, the DC voltage decreases after the AC grid phase jump
while the inverter AC voltage sees a small oscillation, but both can become stable again
after the disturbance. Figure 12f shows the inverter AC instantaneous current. Due to the
disturbance, the three-phase current sees some harmonic components in the first place but
is able to become symmetrical quickly after 40 ms.



D vJLUG N9






E PLQY 0:





F QLQY 09$U






噿



G VGF SX






H VLQY SX






I iLQY 0$




      

Figure 12. Simulation results during phase jump: (a) grid instantaneous AC voltage; (b) inverter
active power; (c) inverter reactive power; (d) inverter DC voltage; (e) inverter AC voltage; (f) inverter
AC instantaneous current.

The performance of the system during grid impedance variation is presented in

Figure 13. At 3 s, the grid impedance is increased from 0.2 pu (SCR = 5) to 0.33 pu
(SCR = 3), and at 4 s, the grid impedance is further increased to 0.5 pu (SCR = 2). As can be

37
Energies 2024, 17, 3642

seen in Figure 13a,b, the proposed control has the capability to control the inverter active
power and inverter AC voltage during different grid impedances.



D PLQY 0:





E VLQY SX





      

Figure 13. Simulation results during grid impedance ratio variation: (a) inverter active power;
(b) inverter AC voltage.

5. Conclusions
This paper proposes a grid-forming control for solar generation systems with battery
energy storage. The inverter control consists of the inner current control, voltage control,
the PLL-based frequency control, and the DC voltage control. With the proposed control,
the battery energy storage is able to provide inertial and primary frequency support during
the grid frequency disturbance. Simulation models are established in PSCAD/EMTDC,
and the results during the active power variation and AC variation, the grid frequency
disturbance, grid fault, and mode switch validate the effectiveness of the proposed control.

Author Contributions: Methodology, Y.C. and M.W.; conceptualization and supervision, L.Y.;
writing—original draft, S.L., Z.F. and W.T.; writing—review and editing W.L.; validation S.S. All
authors have read and agreed to the published version of the manuscript.
Funding: This research is supported by the Science and Technology Project of State Grid Liaoning
Electric Power Co., Ltd. (2023YF-90). The funder had the following involvement with the study: the
study design and the decision to submit it for publication.
Data Availability Statement: The original contribution presented in this study is included in the
article and further inquiries can be directed to the corresponding author.
Conflicts of Interest: Authors Yupeng Cai, Meng Wu, Shengyang Lv, Wei Li, and Songjie Shi were
employed by the State Grid Liaoning Electric Power Co., Ltd. The remaining authors declare that the
research was conducted in the absence of any commercial or financial relationships that could be
construed as potential conflicts of interest.

References
1. Rocabert, J.; Luna, A.; Blaabjerg, F.; Rodríguez, P. Control of Power Converters in AC Microgrids. IEEE Trans. Power Electron.
2012, 27, 4734–4749. [CrossRef]
2. Pogaku, N.; Prodanovic, M.; Green, T.C. Modeling, Analysis and Testing of Autonomous Operation of an Inverter-Based
Microgrid. IEEE Trans. Power Electron. 2007, 22, 613–625. [CrossRef]
3. Harnefors, L.; Wang, X.; Yepes, A.G.; Blaabjerg, F. Passivity-Based Stability Assessment of Grid-Connected VSCs—An Overview.
IEEE J. Emerg. Sel. Top. Power Electron. 2016, 4, 116–125. [CrossRef]
4. Zhou, J.Z.; Ding, H.; Fan, S.; Zhang, Y.; Gole, A.M. Impact of Short-Circuit Ratio and Phase-Locked-Loop Parameters on the
Small-Signal Behavior of a VSC-HVDC Converter. IEEE Trans. Power Deliv. 2014, 29, 2287–2296. [CrossRef]
5. Fan, L. Modeling Type-4 Wind in Weak Grids. IEEE Trans. Sustain. Energy 2019, 10, 853–864. [CrossRef]
6. Hu, J.; Huang, Y.; Wang, D.; Yuan, H.; Yuan, X. Modeling of Grid-Connected DFIG-Based Wind Turbines for DC-Link Voltage
Stability Analysis. IEEE Trans. Sustain. Energy 2015, 6, 1325–1336. [CrossRef]
7. Lin, X.; Liu, Y.; Yu, J.; Yu, R.; Zhang, J.; Wen, H. Stability analysis of Three-phase Grid-Connected inverter under the weak grids
with asymmetrical grid impedance by LTP theory in time domain. Int. J. Electr. Power Energy Syst. 2022, 142, 108244. [CrossRef]
8. Wen, B.; Boroyevich, D.; Burgos, R.; Mattavelli, P.; Shen, Z. Analysis of D-Q Small-Signal Impedance of Grid-Tied Inverters. IEEE
Trans. Power Electron. 2016, 31, 675–687. [CrossRef]

38
Energies 2024, 17, 3642

9. Sun, J. Impedance-Based Stability Criterion for Grid-Connected Inverters. IEEE Trans. Power Electron. 2011, 26, 3075–3078.
[CrossRef]
10. Zhang, X.; Xia, D.; Fu, Z.; Wang, G.; Xu, D. An Improved Feedforward Control Method Considering PLL Dynamics to Improve
Weak Grid Stability of Grid-Connected Inverters. IEEE Trans. Ind. Appl. 2018, 54, 5143–5151. [CrossRef]
11. Givaki, K.; Chen, D.; Xu, L. Current Error Based Compensations for VSC Current Control in Weak Grids for Wind Farm
Applications. IEEE Trans. Sustain. Energy 2019, 10, 26–35. [CrossRef]
12. Egea-Alvarez, A.; Fekriasl, S.; Hassan, F.; Gomis-Bellmunt, O. Advanced Vector Control for Voltage Source Converters Connected
to Weak Grids. IEEE Trans. Power Syst. 2015, 30, 3072–3081. [CrossRef]
13. Zhang, L.; Harnefors, L.; Nee, H. Power-Synchronization Control of Grid-Connected Voltage-Source Converters. IEEE Trans.
Power Syst. 2010, 25, 809–820. [CrossRef]
14. Zhang, L.; Harnefors, L.; Nee, H. Interconnection of Two Very Weak AC Systems by VSC-HVDC Links Using Power-
Synchronization Control. IEEE Trans. Power Syst. 2011, 26, 344–355. [CrossRef]
15. Liu, J.; Miura, Y.; Ise, T. Comparison of Dynamic Characteristics Between Virtual Synchronous Generator and Droop Control in
Inverter-Based Distributed Generators. IEEE Trans. Power Electron. 2016, 31, 3600–3611. [CrossRef]
16. Meng, X.; Liu, J.; Liu, Z. A Generalized Droop Control for Grid-Supporting Inverter Based on Comparison Between Traditional
Droop Control and Virtual Synchronous Generator Control. IEEE Trans. Power Electron. 2019, 34, 5416–5438. [CrossRef]
17. Wu, W.; Chen, Y.; Zhou, L.; Luo, A.; Zhou, X.; He, Z.; Yang, L.; Xie, Z.; Liu, J.; Zhang, M. Sequence Impedance Modeling and
Stability Comparative Analysis of Voltage-Controlled VSGs and Current-Controlled VSGs. IEEE Trans. Ind. Electron. 2019, 66,
6460–6472. [CrossRef]
18. Li, M.; Wang, Y.; Hu, W.; Shu, S.; Yu, P.; Zhang, Z.; Blaabjerg, F. Unified Modeling and Analysis of Dynamic Power Coupling for
Grid-Forming Converters. IEEE Trans. Power Electron. 2022, 37, 2321–2337. [CrossRef]
19. Mallemaci, V.; Mandrile, F.; Rubino, S.; Mazza, A.; Carpaneto, E.; Bojoi, R. A comprehensive comparison of Virtual Synchronous
Generators with focus on virtual inertia and frequency regulation. Electr. Power Syst. Res. 2021, 201, 107516. [CrossRef]
20. Barać, B.; Krpan, M.; Capuder, T.; Kuzle, I. Modeling and Initialization of a Virtual Synchronous Machine for Power System
Fundamental Frequency Simulations. IEEE Access 2021, 9, 160116–160134. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

39
Article

Optimization of PID Controllers Using Groupers and Moray Eels

Optimization with Dual-Stream Multi-Dependency Graph
Neural Networks for Enhanced Dynamic Performance
Vaishali H. Kamble 1 , Manisha Dale 2 , R. B. Dhumale 3 and Aziz Nanthaamornphong 4, *

1 Department of Electronics and Communication Engineering, DES Pune University, Pune 411004, India;
vaishaliraichurkar@gmail.com
2 Department of Electronics and Telecommunication, MES Wadia College of Engineering, Pune 411004, India;
mpdale@mescoepune.org
3 Department of Electronics and Telecommunication Engineering, AISSMS Institute of Information Technology,
Pune 411001, India; rbd.scoe@gmail.com
4 College of Computing, Prince of Songkla University, Phuket 83120, Thailand
* Correspondence: aziz.n@phuket.psu.ac.th

Abstract: Traditional proportional–integral–derivative (PID) controllers are often utilized

in industrial control applications due to their simplicity and ease of implementation. This
study presents a novel control strategy that integrates the Groupers and Moray Eels
Optimization (GMEO) algorithm with a Dual-Stream Multi-Dependency Graph Neural
Network (DMGNN) to optimize PID controller parameters. The approach addresses key
challenges such as system nonlinearity, dynamic adaptation to fluctuating conditions, and
maintaining robust performance. In the proposed framework, the GMEO technique is
employed to optimize the PID gain values, while the DMGNN model forecasts system
behavior and enables localized adjustments to the PID parameters based on feedback.
This dynamic tuning mechanism enables the controller to adapt effectively to changes in
input voltage and load variations, thereby enhancing system accuracy, responsiveness,
and overall performance. The proposed strategy is assessed and contrasted with existing
strategies on the MATLAB platform. The proposed system achieves a significantly reduced
settling time of 100 ms, ensuring rapid response and stability under varying load conditions.
Additionally, it minimizes overshoot to 1.5% and reduces the steady-state error to just
0.005 V, demonstrating superior accuracy and efficiency compared to existing methods.
These improvements demonstrate the system’s ability to deliver optimal performance
while effectively adapting to dynamic environments, showcasing its superiority over
existing techniques.

Keywords: buck-boost converters; proportional integral derivative; Schottky diode; error

signal; control signal; steady-state error; tuning methods

1. Introduction
The simplicity, convenience of use, and typically satisfactory performance of
proportional–integral–derivative (PID) controllers make them popular in industrial appli-
cations [1]. These controllers are often employed in power electronic systems, like DC–DC
converters and inverters, for voltage, current, and power regulation [2]. DC–DC converters
are especially important in modern energy systems due to their role in efficiently man-
aging and converting power from renewable sources like solar panels and batteries [3].
They enable voltage regulation and energy transfer across different subsystems, improving

Energies 2025, 18, 2034 https://doi.org/10.3390/en18082034

40
Energies 2025, 18, 2034

overall energy efficiency and system reliability [4]. These converters are integral in electric
vehicles, smart grids, and portable electronic devices, where consistent and adaptive power
delivery is critical. Their capacity to provide controlled voltage levels makes them essential
for integrating intermittent and variable energy sources into stable usable outputs [5]. The
traditional PID tuning methods like Ziegler–Nichols and Cohen–Koon often yield subopti-
mal results in these nonlinear systems, as they are designed for linear or time-invariant
conditions [6]. Consequently, conventional PID controllers, which are typically tuned for
steady-state performance [7], may fail to provide an adequate transient response or stability
in applications where load or input voltage fluctuates [8]. Moreover, in the presence of
high-frequency switching noise, which is characteristic of buck-boost converters, improper
PID tuning can lead to performance degradation such as overshooting or instability [9].
The control process can become more complicated when the system is subjected to fluc-
tuating load conditions, making it crucial to adjust PID parameters to maintain optimal
performance [10]. Recent advancements have focused on adaptive PID control strategies,
which dynamically adjust the controller parameters to handle the system’s nonlineari-
ties and improve its robustness [11]. One approach is using optimization techniques like
particle swarm optimization or genetic algorithms, which fine-tune PID parameters to
achieve better performance in nonlinear and time-varying environments [12]. An addi-
tional promising strategy involves incorporating artificial intelligence or machine learning
models to dynamically adjust PID gains, enhancing the controller’s adaptability [13]. De-
spite these advancements, traditional PID control still faces limitations in achieving optimal
performance under highly dynamic and nonlinear conditions [14].
In buck-boost converters, the PID controller must maintain a delicate balance between
stability, transient response, and steady-state accuracy, which is often challenging due to the
varying dynamics of the system [15]. A key challenge lies in selecting PID gains that remain
effective across a variety of operational circumstances, including different load levels and
input voltages [16]. In order to improve the performance of PID controllers in buck-boost
converters, current research is still investigating hybrid and adaptive control techniques [17].
The importance of continuous adjustment and fine-tuning of PID parameters becomes
evident as the converter’s dynamic response must be matched with external changes,
requiring constant optimization [18]. High-frequency noise from the switching operation
of the converter can further complicate the controller’s performance, leading to higher
overshoot and instability unless properly addressed [19]. Therefore, more sophisticated
approaches that integrate global optimization and adaptive feedback mechanisms are
crucial for improving the overall stability and efficiency of PID-controlled buck-boost
converters under dynamic conditions [20].
In the literature, various research works are available based on PID controller optimiza-
tion, adaptive control strategies, and performance enhancement in buck-boost converters
using various techniques and aspects. A few of these works are reviewed as follows.

1.1. Optimization Techniques

S Sangeetha et al. [21] suggested a hybrid technique for fractional-order propor-
tional integral derivative (FOPID) controller performance analysis of buck converters. The
Capuchin Search Algorithm (CapSA) and the Golden Jackal Optimization (GJO) were
integrated into this hybrid technique. The Capuchin Search Algorithm was used to im-
prove the Golden Jackal Optimization’s update behavior, resulting in the enhanced GJO
(IGJO) approach. Because power converters were nonlinear, they were difficult to regu-
late, and there was a constant quest for efficient and effective controllers. Recently, it has
been demonstrated that fractional-order controllers are more efficient in power electronic

41
Energies 2025, 18, 2034

systems. The best design for a fractional-order PID controller for the buck converter was
found using the IGJO approach.
P Warrier et al. [22] presented a complex-order PI controller for DC–DC buck and
boost converter control that combines a complex-order integrator. Four parameters in the
intricate PID controller require adjustment. The Metaheuristic Cohort Intelligence method
was used to optimize the design of the complex-order PI controller. The outcomes were
contrasted with those of a PID controller of fractional order. The results showed that the
complex PI controller provided a better response than the FOPID controller and was more
robust to parameter changes.
N F Nanyan et al. [23] presented an improved Sine Cosine Algorithm (ISCA) for the
optimization of a DC–DC buck converter using a PID controller. Through two distinct im-
provements, the limitations of the traditional Sine Cosine Algorithm (SCA) were addressed,
resulting in a synergistic usage of nonlinear equations in the instrumental mechanism to
revise the average location. To address the problem of local optima, the initial revision
included an instrumental function to update the average location. The second change
applied a nonlinear equation to the algorithm’s reducing position-updating mechanism,
coordinating the traditional SCA’s disproportional exploration and exploitation phases.
L.K. Fong et al. [24] proposed the Archimedes Optimization Algorithm (AOA) as
a metaheuristic method for optimizing a PID controller in a closed-loop DC–DC buck
converter. The primary duty of the converter was to control output voltage, ensuring
stability in the face of input voltage variations and load variations. The performance of
the converter heavily depends on the PID controller’s gain settings, which need to be
optimized for robust operation. AOA helps identify the optimal integral, proportional,
and derivative gains, improving the controller’s ability to minimize voltage errors, reduce
overshoot, and enhance response time across varying load conditions. However, AOA may
require significant computational resources for complex systems.

1.2. Controller Structure

S M Ghamari et al. [25] designed a Lyapunov-based model reference PID controller
for a DC/DC buck converter using their approach. For more dependable functioning, the
parameters must be returned, and the PID approach is unsuitable for real-world applica-
tions due to a variety of disruptions. To solve this, the PID approach used an adaptive
mechanism based on the Lyapunov definition, which improved the stability of the system
and resilience to a range of shocks. The system was also handled as a “black-box”, which
removes the requirement for precise mathematical modeling and eases installation and
computing load. As a contemporary adaptive algorithm, the Lyapunov notion may provide
optimum solutions more quickly while maintaining dependable stability and accuracy.
P. Sharma et al. [26] suggested an optimized FOPID controller, designed to generate
optimal switching signals for the converter based on input voltage. The controller’s
performance was evaluated for both buck and boost modes of operation. It was verified that
zero voltage switching (ZVS) and zero current switching (ZCS) are achieved under boost
mode, enhancing efficiency and reducing switching losses. However, the implementation of
the FOPID controller can be complex and requires high computational resources, especially
in systems with varying load conditions.
D M Acevedo et al. [27] have presented an adaptive speed control strategy for DC
motors using a DC/DC converter (in buck and boost modes) based on the inverse opti-
mal control (IOC) method. Their method ensures stability and optimal performance of
nonlinear systems through Lyapunov theory, minimizing a specified cost function. An
integral action enhances performance, ensuring asymptotic stability without affecting
convergence properties. The strategy was implemented on buck and boost converter/DC

42
Energies 2025, 18, 2034

motor systems. However, the IOC approach can be computationally intensive and may not
be ideal for systems with rapidly changing dynamics. Table 1 displays the summary of the
research work.

Table 1. The summary of research work.

Authors Methodology Advantages Limitations

Efficient in power
Improved Golden Jackal Nonlinear nature of power
electronic systems,
S Sangeetha et al. [21] Optimization (IGJO) for converters, complexity in
improved
PID control controller tuning
fractional-order control
Higher parameter
Cohort Intelligence More robust to
P Warrier et al. [22] complexity, challenging for
Algorithm for PID control parameter variations
practical implementation
Improved Sine Cosine Better local optima Limited to specific types of
N F Nanyan et al. [23] Algorithm (ISCA) for handling, enhanced systems, less efficient for
PID control PID tuning highly nonlinear systems
Effective for optimizing Requires significant
Genetic Algorithm-based
L.K. Fong et al. [24] PID parameters in computational effort
PID control
complex systems and time
Improved stability,
Lyapunov-based model May not work efficiently
S M Ghamari et al. [25] robustness
reference PID control for rapid dynamic changes
under disturbances
Global optimization,
Firefly Algorithm for Require fine-tuning
P. Sharma et al. [26] effective for
PID control of parameters
nonlinear systems.
Effective in finding optimal High computational cost,
Ant Colony Optimization
D M Acevedo et al. [27] PID gains, suitable for sensitivity to
for PID control
complex systems initialization parameters

The generic review of recent research highlights various optimization techniques

designed to enhance the control systems’ performance, particularly in buck-boost convert-
ers, by optimizing PID controllers for better stability, efficiency, and dynamic adaptation
under varying load conditions. The existing techniques include Improved Golden Jackal
Optimization (IGJO) optimized PID, Improved Sine Cosine Algorithm (ISCA) optimized
PID, Cohort Intelligence Algorithm (CTA) optimized PID, and Lyapunov-based adaptive
PID (L-Based Adaptive PID). The IGJO-Optimized PID suffers from slow convergence,
high overshoot, higher steady-state error, and sensitivity to initial conditions, making it
less suitable for precision-critical applications. It also struggles to maintain stability in
systems with significant variations in load or input, reducing its reliability in dynamic
environments. The ISCA-Optimized PID faces challenges with response speed and preci-
sion, and its optimization process can be computationally expensive, limiting scalability for
larger or more complex systems. Additionally, it may not handle nonlinearities effectively.
The CTA-Optimized PID has limited adaptation to highly dynamic conditions, resulting
in slower adjustments in unpredictable systems, and reduced robustness under external
disturbances, leading to performance degradation. The L-Based Adaptive PID relies on
Lyapunov functions, limiting its application to specific system types. It also exhibits slower
adaptation in dynamic environments, higher computational complexity, and reduced effec-
tiveness in nonlinear systems, making it less practical for fast-changing or highly nonlinear
applications. Very few approach-based studies are offered in the literature to deal with this
problem; these issues and disadvantages served as the impetus for this study work.

43
Energies 2025, 18, 2034

The proposed GMEO-DMGNN method was chosen due to its unique ability to ad-
dress the complex optimization challenges in PID controllers for buck-boost converters.
This combination overcomes existing drawbacks by offering better adaptability, faster
convergence, and enhanced stability compared to conventional methods. Advantages
include improved stability, reduced overshoot, minimized steady-state error, and faster
settling times under varying loads. The novelty lies in integrating the GMEO with the
DMGNN model for fine-tuning PID parameters, ensuring superior performance. The
method works by leveraging GMEO for global search and DMGNN for localized fine-
tuning of PID parameters, enhancing efficiency in dynamic systems. Compared to existing
techniques, the GMEO-DMGNN method ensures better adaptability, responsiveness, and
optimal performance in buck-boost converters under changing conditions, making it a
more robust and efficient solution.
The following are the paper’s primary contributions:
• This work uses GMEO and DMGNN to create an adaptive tuning mechanism for
PID-controlled buck-boost converters, evaluated across dynamic loads, disturbances,
and diverse signal profile;
• The proposed system utilizes a standard PID controller designed specifically for buck-
boost converters, with enhancements that focus on optimizing its parameters for
improved performance. The proposed method effectively addresses the nonlinear
behavior and feedback noise commonly present in buck-boost converters;
• Through the integration of GMEO and DMGNN, the control system achieves a notable
reduction in settling time of 100 ms, delivering an efficient response even under rapidly
shifting operational conditions;
• The proposed method effectively stabilizes the nonlinear voltage gain characteristics of
the buck-boost converter, which vary with the duty cycle, ensuring consistent output
regulation across a wide range of input conditions and operating modes.
The remainder of the document is structured as follows: Section 2 clarifies the configu-
ration for Optimizing PID Controllers in buck-boost Converters. Section 3 discusses the
Simulation Configurations and Setup. Section 4 discusses the proposed GMEO-DMGNN.
The results and discussion are clarified in Section 5, and Section 6 contains the conclusions.

2. Configuration for Optimizing PID Controllers in Buck-Boost Converters

Figure 1 illustrates the block diagram for optimizing PID controllers in buck-boost
converters. It shows that the reference voltage is compared with the output voltage through
a feedback loop to regulate the system. Subtracting from Vre f creates the error signal, which
is then sent to the PID controller. Based on the error signal, the PID controller modifies the
control signal u(t), which is then used to regulate the buck-boost converter’s output voltage
Vout . The PID controller’s performance is improved by optimizing the PID parameters
through the integration of the proposed GMEO-DMGNN technique. The error signal and
the control signal are dynamically adjusted through the GMEO-DMGNN method, which
fine-tunes the PID parameters for optimal performance under varying load conditions.
The system’s stability and efficiency are increased by the integration of GMEO-DMGNN,
which makes sure that the output voltage is kept at the intended set point while adjusting
to dynamic changes. The GMEO-DMGNN method works by continually fine-tuning the
PID parameters, allowing the system to adapt to these changing conditions. As a result,
the system is able to maintain a steady output voltage (Vout ) that is as close as possible
to the reference voltage (Vre f ), while also improving the entire stability, efficiency, and
performance of the converter. Through this method, the proposed system not only ensures
precise voltage regulation but also reduces overshoot, improves settling time, and enhances
system robustness under various operating scenarios.

44
Energies 2025, 18, 2034

Figure 1. Block diagram of optimizing PID controllers in buck-boost converters.

2.1. Modeling of the Buck-Boost Converter

A buck-boost converter is a type of DC–DC converter that integrates the fundamental
principles of both bucks and boost converters in one circuit as shown in Figure 2. The
output voltage of this converter might be more or lower than the input voltage. The
modeling used in this study is adapted from [28].

Figure 2. Buck-boost converter.

The energy balance method may be used to find the optimal value for the load resistor
R L , assuming the converter runs in boost mode with optimal efficiency.

2
Vmax
RL = (1)
Pmax

The boost converter’s voltage gain equation while it operates in continuous conduction
mode (CCM) may be used to determine the duty cycle, while the energy balance concept is
used to calculate the load resistor.
Vmax
D= (2)
Vmax + VPmax

where Vmax , the input voltage, is given, and the duty cycle D is then calculated. The output
voltage is defined in Equation (3)

Vin · D
Vout = (3)
1−D

45
Energies 2025, 18, 2034

where Vout is the output voltage. Once the duty cycle has been established, the inductor is
made to restrict the input current ripple ΔIL_Pmax .

VPmax × D
L= (4)
ΔIL_Pmax × f s

where the current ripple ΔIL_Pmax and the switching frequency f s are established, the
input inductor’s inductance is then calculated. The subsequent equation may be used to
determine the input capacitor, Cin , as follows:

ΔIL_Pmax
Cin = (5)
ΔVin_Pmax × f s

ΔVin_Pmax is the minimum input voltage. The output capacitor is given in Equation (5)

Imax × D
Cout = (6)
ΔVmax × f s

ΔVmax is the maximum output voltage. This guarantees a thorough and accurate
depiction of the converter’s operation at different load levels. The transfer function from
the control input (duty ratio D) to the output is expressed as follows:

VOS K
Gs = = (7)
DS ( Ls + R) + (CS + 1)

where R represents the load resistance, L is inductance, and C represents the capacitance of
the converter.

2.2. PID Controller Design

The standard PID controller is favored for its simplicity and reliability, making it
suitable for a wide range of industrial applications. Compared to more complex controllers
such as sigmoid-PID, FOPID, and Bandwidth-Enhanced Linear Bias Integral Controller
(BELBIC), the standard PID controller offers easier implementation, reduced computational
burden, and greater compatibility with existing systems. While advanced controllers
may provide improved performance in specific scenarios, their complexity and tuning
challenges often limit their practical use.
A feedback control loop is used by a PID controller to make clear linkages between
the system variables, minimize the impact of disturbances, and guide the system toward
the desired state [29]. The error at time t, represented by et , is sent into the controller. The
discrepancy between the measured and reference values is known as this error. The action
applied to the system or plant is represented by the PID controller’s output, at . The integral
gain Ki multiplied by the error’s integral, the proportional gain K p multiplied by the error
magnitude, and the derivative gain Kd multiplied by the error’s derivative are the three
terms that make up the control signal or actuation. Some of these terms may be set to zero.
 t
det
u ( t ) = K p e t + Ki et dt + Kd (8)
0 dt

The control signal u(t), which is the total of the P, I, and D terms, is returned by the
PID controller. In this case, K p denotes proportional gain, Kd stands for derivative gain,
and Ki denotes integral gain.
The PID controller’s transfer function is described in Equation (9), as follows:

Ki
C (s) = K p + + Kd s (9)
s

46
Energies 2025, 18, 2034

C (s) is the transfer function of the PID controller in the Laplace domain.
The frequency response of the PID controller is given in Equation (10)

Ki jK
C ( jω ) = K p + + Kd ( jω ) = K p − i + jKd ( jω ) (10)
jω jω

C ( jω ) is the PID controller’s frequency response. ω is the input signal’s angular

frequency in radians per second.
Shortly, the controller can be shown with its gain and phase values, as follows:

C ( jω ) = |C ( jω )|e j<C( jω ) (11)

e j<C( jω ) iseuler’s formula representation of a phase shift.

The gain value of the PID controller’s frequency response is given in Equation (12)

2
Ki
|C ( jω )| = K2p + (− + Kd ω ) (12)
ω

The phase value of the controller’s frequency response is given in Equation (13)
! "
− Kωi + Kd ω
< C ( jω ) = arctan (13)
Kp

where arctan is the inverse tangent function that calculates the phase shift. Thus, the
frequency response of the PID controller is obtained. The initial values of the PID parameter
are given in Table 2.

Table 2. The initial values of the PID parameter.

PID Parameter Range

Kp 0.8 to 1.5
Ki 1000 to 3000
Kd 0.0001 to 0.001

3. Simulation Configurations and Setup

The buck-boost converter is developed with suitable properties before incorporating
optimization techniques into a PID controller. The converter can handle the desired input
and output voltages, loads, and switching frequencies, and certain characteristics and factor
values are given in Table 3. An essential part of a buck-boost converter, the inductor (L), is
responsible for both storing and transferring energy.

Table 3. The system specifications.

Parameters Values
Input Voltage Range (Vin ) 5 V to 15 V
Output Voltage (Vout ) 12 V
Maximum Output Current (Iout ) 2A
Switching Frequency (fs ) 100 Khz
Duty Cycle (D) 0.2 to 0.8

47
Energies 2025, 18, 2034

The value of the inductor has significant effects on the converter’s performance and
efficiency. The Equation (14) is used to obtain the inductance value.

Vin (1 − D )
L= (14)
fs ΔL

Here, IL indicates the highest output current that can pass through the inductor,
usually set between 20% and 40% of the total allowable ripple current. The assumption
that (for 40% ripple) ΔIL = 0.4IOut L is given Equation (15).

15V(1 − 0.6)
L= ≈ 150 μH (15)
100 Khz 0.4 × 2A

The selected value of L is 150 μH. An output capacitor is required to lessen voltage
ripple at the buck-boost converter’s output. This helps stabilize the output by reducing
voltage swings and removing high-frequency switching noise. Using the Equation (16),
the capacitance value is determined based on the permitted ripple voltage, inductor ripple
current, and switching frequency.

ΔIL
Cout = ≈ 1.25 μf (16)
8 · f s Δout

The output voltage remains stable with the correct capacitance, ensuring reliable
converter performance. In a low-side switching setup, a MOSFET (e.g., IRF540N, 100 V,
33 A) handles the maximum input voltage and current. A Schottky diode (1N5822) is
selected for its fast recovery time and high current handling with minimal forward voltage
drop. The LM5118 IC controls Pulse Width Modulation (PWM) switching and feedback.
Resistors set the output and reference voltages and feed back to the control loop. After
initial tuning, the GMEO algorithm combined with the DMGNN model adaptively refines
the PID constants to enhance control efficiency and system performance.

VOS Vin (1 − D )
Gs = = (17)
Ds ( L · C · S2 ) + ( R · C · s ) + 1

The initial values for K p , Ki , and Kd are obtained using the Ziegler–Nichols method
and further fine-tuned through simulation to finalize the PID controller design. To ensure
optimal performance across varying operating conditions, machine learning techniques
using the GMEO-DMGNN framework are employed to dynamically adjust these parame-
ters. The proposed controller continuously updates the gains based on system data, such as
fluctuations in input voltage or load changes, effectively minimizing output voltage errors.

4. Proposed GMEO-DMGNN Method for Enhancing PID Control in

Buck-Boost Converters
This section outlines the integration of a Dual-Stream Multi-Dependency Graph Neural
Network (DMGNN) with the Groupers and Moray Eels (GMEO) for optimizing PID
controller parameters. GMEO is used to optimize PID gains (K p , Ki and Kd ), while the Dual-
Stream Multi-Dependency Graph Neural Network (DMGNN) predicts and locally adjusts
these parameters to enhance performance. GMEO explores large solution spaces, maintains
diversity to avoid local optima, and handles nonlinear systems to improve PID controller
performance by enhancing stability, reducing overshoot, and optimizing response time.
DMGNN further refines optimization by capturing complex dependencies and learning
both global and local patterns, which accelerates convergence and improves performance
in dynamic systems like buck-boost converters. The combination of GMEO’s global search

48
Energies 2025, 18, 2034

with DMGNN’s local adjustments optimizes PID parameters more efficiently, improving
system stability, response time, and adaptability while ensuring faster convergence in
complex dynamic systems. Figure 3 depicts the flowchart of the GMEO-DMGNN approach.

Figure 3. Flowchart of GMEO-DMGNN.

4.1. Optimization Using Groupers and Moray Eels (GMEO)

In this section, the GMEO is described [30] and utilized to optimize the controller
parameters of PID gains, such as, Kd , K p , and Ki . The GMEO algorithm offers a robust
global search mechanism, efficiently optimizing PID parameters to enhance performance
in nonlinear dynamic systems, ensuring improved stability, faster response, and better
adaptability in buck-boost converters. In buck-boost converters, it increases overall system
stability and performance by improving the PID controller’s capacity to adjust to changing
system circumstances, lowering overshoot, settling time, and steady-state error. GMEO was
chosen for its ability to effectively handle the complexities of nonlinear dynamic systems
by offering a balanced global search approach, optimizing PID parameters for improved
control and performance in buck-boost converters.
Step 1: Initialization
Set the input variables to first values. In this instance, the input variables are the PID
parameters, which are specified as Kd , K p , and Ki .
Step 2: Random Generation
In matrix form, the input variables were generated at random.
⎡ ⎤
k1,1 k1,2 ... k1,m
⎢ ⎥
q = ⎣k2,1 k2,2 ... k2,m ⎦ (18)
k3,1 k3,2 ... k3,m

49
Energies 2025, 18, 2034

where q indicates the random generation, k indicates the system parameters, and m indicates
the count of decision variables.
Step 3: Fitness Function
The fitness was evaluated, which was described by

F = min( J ) (19)
 T
J= t|et |dt (20)
0

where J refers to an Integral of Time-Weighted Absolute Error (ITAE), t specifies the time
variable, and et specifies the error signal at time t.
Step 4: Primary Search (PS) Phase
GMEO agents explore the search space for optimal PID controller parameters (Kd , K p ,
and Ki ), mimicking the zigzag swimming pattern of groupers hunting prey. This random
exploration ensures thorough coverage of the solution space, aiming to find the optimal
controller gains that minimize performance errors like transient response, steady-state
error, and overshoot.

Xinitial
ij = lowj + rand · (upper j − lowj ), i
(21)
= 1, 2, 3 . . . N,j = 1, 2, 3, . . . D

Here, Xinitial
ij specifies the first location of ith search agent of jth dimension, upperj and lowj
specify the search space’s upper and lower bounds, D specifies the overall count of dimensions,
N specifies the number of search agents, and rand specifies a random vector that follows a
uniform distribution, with values ranging from 0 to 1.
Step 5: Pair Association (PA) Phase
In this phase, the best-performing agents (groupers) collaborate with other high-
quality agents (moray eels) to improve search efficiency. This cooperative interaction
enhances the exploration of promising regions in the solution space. By leveraging the
strengths of both agents, the search process becomes more targeted, accelerating conver-
gence toward the optimal PID parameters. The agents dynamically adjust their positions
based on the most promising solutions, ensuring a balance between global exploration and
local refinement for improved accuracy in optimizing the PID controller.
Step 6: Encircling or Extended Search (ES) Phase
Agents refine their search by adaptively adjusting their positions toward promising
regions. This phase enhances local exploration, allowing agents to dynamically focus
on areas with higher potential for optimal PID parameters. The cooperative movement
mimics the coordinated behavior of groupers and moray eels, ensuring a balance between
exploitation and exploration. This adaptive search approach improves the likelihood of
finding the global optimum and helps avoid premature convergence.

L
cmj = Xgmj + (X − Xgmj ) (22)
dis Emj
Here, cmj denotes the coordinates of the mth prey in each dimension, Xgmj specifies
the location of a grouper, XEmj specifies the location of an eel, L specifies the separation
between the prey and the grouper, and dis specifies the distance between the grouper and
the eel.
Step 7: Attacking and Catching Phase
Agents converge on the best solution by intensifying the search around the optimal
PID gains. This phase improves convergence accuracy by gradually reducing the search
radius, ensuring precise identification of the optimal controller parameters. The shrinking

50
Energies 2025, 18, 2034

mechanism enables a finer search around the most promising solution, refining the PID
gains for better system performance. This stage also helps to reduce the steady-state
error and enhances system stability by continuously updating the solution based on the
best-performing agents.
Ri+1 = (1 − μ) ∗ Ri (23)

where i = 1, 2, 3, . . . AB − 1 and μ specifies the shrinking ratio and R refers to the radius.
Step 8: Termination Criteria
The procedure ends if the answer is ideal; if not, it goes back to step 3 for fitness
assessments and keeps processing the next steps until the best answer is discovered. Thus,
GMEO effectively optimized the controller parameters of PID gains. The flowchart of
GMEO is illustrated in Figure 4.

Figure 4. Flowchart of GMEO.

4.2. Dual-Stream Multi-Dependency Graph Neural Network (DMGNN)

In this section, the prediction using a Dual-Stream Multi-Dependency Graph Neural
Network (DMGNN) is discussed [31]. DMGNN enhances optimization by efficiently
predicting and adjusting PID parameters, capturing complex dependencies in dynamic
systems. It was chosen for its ability to model both global and local patterns through its
dual-stream architecture, making it well-suited for optimizing PID parameters in nonlinear

51
Energies 2025, 18, 2034

and time-sensitive systems like buck-boost converters. This capability ensures faster
convergence and significantly improves performance.

− 12 − 12
GCN ( X, A) = δ( D AD XW ) (24)

where the trainable weight matrix for feature transformation is specified by Dij = ∑ j Aij
and W.
DMGNN captures the complex dependencies and relationships between the PID
parameters and the system’s dynamic behavior. Its dual-stream architecture learns both
global patterns (long-range dependencies) and local patterns (short-range dependencies)
between the system states, improving PID parameter adjustment.

AWs = (sigm(W1 F + b1 ) · tanh(W2 F + b2 ))W3 + b3 (25)

Here, the learnable transformation matrix and bias is specified by W and b, correspond-
ingly, while sigm indicates the sigmoid function. The DMGNN refines the parameters K p ,
Kd , and Ki by adjusting them locally, ensuring that the PID controller performs optimally
under different load conditions and varying operational environments.

MAM = α + (1 − α) · sigm( AWs ) × sigm( AWs ) T (26)

AFM × AWs
F = so f tmax( β √ + (1 − β) AWs ) × [ F1 , F2 ] (27)
dk
where a pre-defined hyper-parameter and reduction to 0 as the training proceeds is specified
by α, β is the pre-defined hyper parameter, the features from the two branches are specified
by F1 = FUB( F ) and F2 = GAB( F ), and the concatenation operation is specified by [, ].
The adjusted PID parameters are used to control the buck-boost converter, ensuring that the
system’s output voltage is maintained at the desired set point while adapting dynamically
to load variations.
yj
f surv (y j , Fj ) = ∏ (1 − f hazard (i, Fj )) (28)
i =0

where F and y j variables related to the system’s performance and state, respectively, while
f hazard represents the computation process of hazard rates. DMGNN dynamically adjusts
the ideal PID parameters Kd , K p , and Ki based on system conditions by learning both global
and local patterns in the system’s behavior.

5. Result and Discussion

The proposed method’s performance is demonstrated in this section using the findings
of the simulation. The GMEO-DMGNN approach for buck-boost converter PID controller
optimization was proposed in this study. Using the MATLAB R2022b environment, the
proposed method is assessed and contrasted with other current methods. The proposed
method’s simulation is shown below. All simulations begin with an initial capacitor
voltage of 12 V to simulate a realistic steady-state start-up scenario rather than zero-
state initialization.
Figure 5 depicts the voltage response over time using the IGJO-Optimized PID con-
troller. The voltage reaches a peak of approximately 13.6 V, overshooting the setpoint by
around 1.4 V. It then takes about 1000 ms to settle within a 0.2 V band around the setpoint.
The larger overshoot and prolonged oscillations highlight a key drawback, leading to a
less stable and slower response. Figure 6 depicts the voltage response over time using the
ISCA-Optimized PID controller. The voltage rapidly increases, reaching a peak of approxi-
mately 13.2 V within the first second, with an overshoot of 1.2 V. The system then settles in

52
Energies 2025, 18, 2034

about 500 ms; although the settling time is relatively high, the overshoot of 1.2 V suggests
that further optimization could improve system stability and reduce peak voltage. Figure 7
depicts the voltage response over time using the CTA-Optimized PID controller. The initial
voltage rise shows a peak of 13.6 V, with a substantial overshoot of 1.6 V above the set point.
After the peak, the voltage begins to oscillate but stabilizes within approximately 200 ms.
However, the significant overshoot and the time taken to stabilize highlight a drawback of
the CTA-Optimized PID system, as it results in slower settling and takes longer to reach
steady-state performance.

Figure 5. Performance evaluation of the IGJO-Optimized PID controller.

Figure 6. Performance evaluation of the ISCA-Optimized PID controller.

Figure 8 illustrates the voltage response over time using the L-Based Adaptive PID.
Initially, the voltage rises quickly, peaking at around 13.3 V, with an overshoot of 1.3 V
above the set point. After the peak, the voltage starts to oscillate, but these oscillations
dampen over a period of about 150 ms, allowing the system to settle around the set point,
with the persistence of oscillations leading to a slower stabilization process that takes
several seconds to fully settle.

53
Energies 2025, 18, 2034

Figure 7. Performance evaluation of CTA-Optimized PID controller.

Figure 8. Performance evaluation of L-Based Adaptive PID.

Figure 9 depicts the performance comparison of different control strategies. It com-

pares the performance of various PID control strategies, including IGJO-Optimized PID,
ISCA-Optimized PID, CTA-Optimized PID, L-Based Adaptive PID, and the proposed
GMEO-DMGNN method. Each strategy shows an initial overshoot followed by oscillations
before stabilizing. The IGJO-Optimized PID controller reaches a peak of approximately
13.6 V, overshooting the set point by around 1.4 V, and takes about 1000 ms to settle within
a 0.2 V band around the set point. ISCA-Optimized PID reaches a peak of 13.2 V, with a
similar overshoot of 1.2 V, and settles in about 500 ms seconds. CTA-Optimized PID shows
a peak of 13.6 V, with a substantial overshoot of 1.6 V, and takes approximately 200 ms to
stabilize. L-Based Adaptive PID peaks at around 13.3 V with an overshoot of 1.3 V, settling
in about 150 ms. The proposed GMEO-DMGNN method shows a peak of 12.9 V, with a
more controlled overshoot of 1.1 V, and settles within approximately 100 ms, demonstrating
faster convergence and less oscillation. Therefore, the GMEO-DMGNN method outper-
forms the other techniques, providing a more stable, faster, and more efficient response for
controlling buck-boost converters.

54
Energies 2025, 18, 2034

Figure 9. Performance comparison of various control strategies.

Figure 10 depicts the dynamic response characteristics of various control algorithms

under different input signals of step and ramp. In subplot (a), under the step input
condition, the proposed GMEO-DMGNN method achieves a settling time of 100 ms,
an overshoot of only 1.5%, and an extremely low steady-state error of 0.005 V. These
values clearly outperform all other compared algorithms. This superior transient response
demonstrates GMEO-DMGNN’s ability to stabilize the system significantly faster, with
minimal overshoot and highly accurate voltage regulation. The improvement is due to the
integration of the GMEO optimization algorithm with the DMGNN model, which allows
the system to identify optimal PID gains and adaptively refine them based on system
behavior. In subplot (b), under the ramp input condition, the GMEO-DMGNN controller
maintains a smooth and accurate tracking response with error oscillations confined within
±20, showing stable behavior and precise adaptation to gradual input changes. The
smoother slope and absence of sudden fluctuations highlight the method’s ability to track
dynamic inputs effectively, confirming its superior adaptability over other algorithms.

Figure 10. Dynamic response characteristics of various control algorithms under different input
signals: (a) step input and (b) ramp input.

Figure 11 depicts a comparative analysis of the anti-interference performance of var-

ious algorithms based on MSE under different iteration counts, indicating convergence
behavior and different SNRs. In subplot Figure 11a, the proposed GMEO-DMGNN method
demonstrates superior convergence behavior with a significantly lower MSE of 0.020 at
100 iterations compared to IGJO’s 0.060, highlighting its faster and more precise opti-

55
Energies 2025, 18, 2034

mization capability. This indicates that GMEO-DMGNN consistently outperforms other

methods in minimizing error across iterations. In subplot Figure 11b, the proposed GMEO-
DMGNN method demonstrates superior anti-interference performance under varying
noise conditions. Even at a low SNR of 0 dB, it maintains a low MSE of 0.050. This indi-
cates its strong noise resilience. Compared to algorithms like ISCA, which records a high
MSE of 0.185 at 0 dB, GMEO-DMGNN shows significantly better robustness and stability,
effectively handling signal disturbances with minimal performance degradation.

Figure 11. Comparative analysis of the anti-interference performance of various algorithms based
on MSE under (a) different iteration counts, indicating convergence behavior, and (b) different
Signal-to-Noise Ratio (SNR).

5.1. Case Studies

Case 1: Trajectory Tracking Analysis
Figure 12 illustrates the trajectory tracking performance of the GMEO-DMGNN under
a complex time-varying voltage. Despite rapid changes in the input, the controller accu-
rately tracks the trajectory, with the voltage stabilizing at approximately 12.5 V by 8 s. The
peak overshoot reaches close to 13.8 V around 1 s, yet the system quickly compensates
and settles without prolonged oscillations. This performance confirms the controller’s
capability to manage dynamic voltage references with minimal error and fast recovery,
thereby validating the effectiveness of the proposed method in complex trajectory tracking
scenarios beyond simple step inputs.

Figure 12. Trajectory tracking performance of the GMEO-DMGNN under a complex time-varying voltage.

Case 2: Disturbance Analysis

56
Energies 2025, 18, 2034

Figure 13 depicts the disturbance rejection response when a sudden disturbance is

introduced at the converter output. The graph shows the output response of the converter
when a disturbance is injected at t = 6 s. Before the disturbance, the system maintains a
steady amplitude of approximately 12.5 V. This indicates stable operation under normal
conditions. After the disturbance, the GMEO-DMGNN-based PID controller quickly recov-
ers the output with minimal overshoot and fast settling, demonstrating strong disturbance
rejection capability.

Figure 13. Disturbance rejection response when a sudden disturbance is introduced at the converter
output.

Case 3: Uncertainty Analysis

Figure 14 presents the output voltage response of the converter under parameter
uncertainty conditions, where some nominal model parameters were deliberately varied.
At around t = 2 s and t = 6 s, noticeable disturbances occur due to the introduced parameter
changes. The system shows voltage deviations confined within ±0.5 V from the nominal
value of 12.5 V when model parameters deviate from their nominal values. Despite the
uncertainty, the controller demonstrates a quick recovery, stabilizing the output back to
steady state within approximately 1 s after each disturbance. These results confirm the
proposed controller’s strong robustness and adaptability in the presence of parameter
uncertainties, ensuring stable and reliable system performance.

Figure 14. Output voltage response of the proposed controller under parameter uncertainty.

Table 4 compares the performance of various PID optimization methods based on

settling time, overshoot, and steady-state error. The IGJO-Optimized PID method has the
longest settling time at 1000 ms, with a significant overshoot of 25% and a steady-state

57
Energies 2025, 18, 2034

error of 0.15 V, indicating slower stabilization and moderate accuracy. With a settling
time of 500 ms, a 15% decrease in overshoot, and a lower steady-state error of 0.05 V, the
ISCA Optimised PID approach performs better in terms of stability and accuracy. The
CTA-Optimized PID method performs even better, with a settling time of 200 ms, a further
reduced overshoot of 5%, and a steady-state error of 0.02 V, showing faster convergence
and enhanced precision. The L-Based Adaptive PID method offers the fastest settling time
at 150 ms, with a minimal overshoot of 3% and an almost negligible steady-state error of
0.01 V, making it highly stable and precise. Finally, the GMEO-DMGNN (Proposed) method
outperforms all others with the fastest settling time of 100 ms, the smallest overshoot of
1.5%, and the lowest steady-state error of 0.005 V, providing superior performance in terms
of speed, accuracy, and stability. Therefore, the GMEO-DMGNN method stands out as the
most efficient and reliable optimization technique for PID controllers.

Table 4. Performance comparison of PID tuning methods.

Methods Settling Time (ms) Overshoot (%) Steady-State Error (V)

IGJO-Optimized PID 1000 ms 25% 0.15 V
ISCA-Optimized PID 500 ms 15% 0.05 V
CTA-Optimized PID 200 ms 5% 0.02 V
L-Based Adaptive PID 150 ms 3% 0.01 V
GMEO-DMGNN (Proposed) 100 ms 1.5% 0.005 V

Table 5 presents a comparison of statistical values between the proposed and existing
methods. The proposed method achieves the highest mean value of 4.9876 × 10−7 and
the highest standard deviation (SD) of 6.9452 × 10−12 , surpassing the performance of the
existing systems. This indicates that the proposed method offers superior variability and
greater adaptability in comparison.

Table 5. The statistical performance comparison of proposed with existing techniques.

Methods Mean Standard Deviation

IGJO-Optimized PID 4.9876 × 10−7 6.9452 × 10−12
ISCA-Optimized PID 3.2344 × 10−7 5.4322 × 10−12
CTA-Optimized PID 3.9087 × 10−7 4.3322 × 10−12
L-Based Adaptive PID 3.2890 × 10−7 5.9880 × 10−12
GMEO-DMGNN (Proposed) 4.0345 × 10−7 5.4333 × 10−12

Table 6 compares ITAE and ISE. The proposed system achieves an ITAE of 2.09998998
× 10−13 and an ISE of 1.19052756 × 10−6 . In comparison, existing methods show
higher ITAE and ISE values, with the IGJO-Optimized PID achieving an ITAE of
2.39850000 × 10−13 and an ISE of 2.73120000 × 10−6 , the ISCA-Optimized PID yielding
an ITAE of 2.36580000 × 10−13 and an ISE of 2.62490000 × 10−6 , the CTA-Optimized PID
showing an ITAE of 2.96550000 × 10−13 and an ISE of 3.52110000 × 10−6 , and the L-Based
Adaptive PID resulting in an ITAE of 2.9 × 10−13 and an ISE of 2.43120000 × 10−6 . These
results highlight the enhanced accuracy and prediction capability of the GMEO-DMGNN
method compared to other systems.

58
Energies 2025, 18, 2034

Table 6. Comparison of Time-Weighted Absolute Error (ITAE) and Integral of Squared Error (ISE).

Methods ITAE ISE

IGJO-Optimized PID 2.39850000 × 10−13 2.73120000 × 10−6
ISCA-Optimized PID 2.36580000 × 10−13 2.62490000 × 10−6
CTA-Optimized PID 2.96550000 × 10−13 3.52110000 × 10−6
L-Based Adaptive PID 2.9 × 10−13 2.43120000 × 10−6
GMEO-DMGNN (Proposed) 2.09998998 × 10−13 1.19052756 × 10−6

A comparison of efficiency between the proposed and existing methods is displayed

in Table 7. The proposed method achieves the highest efficiency at 98.5%, showcasing
its ability to deliver optimal solutions quickly and effectively. The IGJO-Optimized PID
method follows with an efficiency of 82%, indicating slightly lower effectiveness. The
ISCA-Optimized PID method achieves an efficiency of 89%, while the CTA-Optimized PID
method has an efficiency of 89.6%. The L-Based Adaptive PID reaches an efficiency of 91%.
Overall, the proposed method stands out for its superior efficiency, making it the most
effective solution for optimal performance.

Table 7. Comparison of the efficiency of proposed and existing techniques.

Methods Efficiency (%)

IGJO-Optimized PID 82%
ISCA-Optimized PID 89%
CTA-Optimized PID 89.6%
L-Based Adaptive PID 91%
GMEO-DMGNN (Proposed) 98.5%

Table 8 compares the optimal PID controller parameters K p , Ki , Kd for different meth-
ods. The proposed GMEO-DMGNN method has the highest K p at 1.4, Ki at 1001, and
a relatively higher Kd at 0.0001, making it more responsive and adaptive compared to
existing methods. The way each algorithm tunes these parameters varies based on its
internal mechanism. IGJO adjusts gains using Levy flight-based exploration, enabling
wider search jumps and resulting in moderately high K p and Ki but conservative Kd . ISCA
uses sinusoidal chaotic sequences for better diversity, which yields slightly lower K p and
Kd . CTA introduces a threshold-aware mutation to refine tuning, leading to balanced gains
with a focus on higher Kd damping. The L-based adaptive method modifies gains based
on error trends prioritizing Ki for steady-state accuracy. GMEO-DMGNN dynamically
learns and adapts PID gains by combining evolutionary search with graph-based learning
of system behavior over time, which enables the most effective and balanced tuning among
all methods. This makes the GMEO-DMGNN approach more robust and better suited for
controlling nonlinear dynamics in the buck-boost converter.

Table 8. Optimal parameters of the PID controller.

Methods Optimal Kp Optimal Ki Optimal Kd

IGJO-Optimized PID 0.89 1766 0.0004
ISCA-Optimized PID 0.98 1733 0.0007
CTA-Optimized PID 0.67 1266 0.0003

59
Energies 2025, 18, 2034

Table 8. Cont.

Methods Optimal Kp Optimal Ki Optimal Kd

L-Based Adaptive PID 0.88 1233 0.0009
GMEO-DMGNN (Proposed) 1.4 1001 0.0001

5.2. Discussion
The GMEO-DMGNN approach focuses on optimizing PID controller parameters in
buck-boost converters, enhancing their performance under dynamic load conditions. By
combining the global optimization capabilities of GMEO with the localized adjustment
strengths of DMGNN, the approach effectively fine-tunes PID controller parameters for
improved performance in buck-boost converters. The integration of both methods ensures
more accurate optimization of PID parameters, enhancing system stability, responsiveness,
and overall performance under dynamic conditions. This combined framework offers a
more robust and efficient solution for optimizing PID parameters, resulting in improved
adaptability, faster convergence, and better control performance in buck-boost converters.
The proposed method consistently outperformed existing PID optimization techniques,
achieving the fastest settling time of 100 ms, the lowest overshoot of 1.5%, minimal steady-
state error of 0.005 V, and the highest efficiency at 98.5%. It also demonstrated superior
statistical values with a mean of 4.9876 × 10−7 and a standard deviation of 6.9452 × 10−12 ,
along with the lowest error integrals, including an ITAE of 2.09998998 × 10−13 and an ISE
of 1.19052756 × 10−6 . This result demonstrates that the proposed GMEO-DMGNN system
provides superior performance, exhibiting enhanced speed, accuracy, stability, and effi-
ciency compared to existing PID optimization techniques. This finding indicates the GMEO-
DMGNN method’s comprehensive superiority in PID optimization, achieved through a
combination of optimized performance metrics and enhanced hardware utilization, specifi-
cally the highest capacitance. Despite its strong performance, the GMEO-DMGNN method
may face a few practical limitations. While simulation results are promising, the lack
of real-time hardware validation limits immediate deployment insights. The method’s
reliance on simulated conditions may also result in issues when exposed to unforeseen
physical disturbances or sensor noise not captured in the modeling environment. However,
these drawbacks can be addressed with further development and real-world testing.

6. Conclusions
The proposed GMEO-DMGNN approach has been introduced to address the limi-
tations of traditional PID controllers in buck-boost converters. In this study, the GMEO-
DMGNN combines the global search capabilities of GMEO with the local adjustment
strengths of DMGNN to optimize PID controller parameters. Additionally, this method is
highly adaptable to dynamic load conditions, maintaining consistent performance across
varying operational environments. Unlike conventional methods that may struggle with
nonlinearities or varying conditions, the GMEO-DMGNN approach provides enhanced
robustness and adaptability. The proposed approach exhibits better performance, achieving
a settling time of 100 ms, 1.5% overshoot, and a steady-state error of 0.005 V. This approach
outperforms existing methods by offering better adaptability, stability, and efficiency in
dynamic and nonlinear environments, making it a robust solution for optimizing PID
controllers in buck-boost converters. It also achieves a high efficiency of 98.5% ensuring
optimal performance while maintaining system stability under varying conditions. The
proposed system reduces the ITAE to 2.09998998 × 10−13 and ISE to 1.19052756 × 10−6 ,
highlighting the proposed system’s superior efficiency in reducing error and improving
control performance. The proposed GMEO-DMGNN approach significantly enhances

60
Energies 2025, 18, 2034

PID controller performance in buck-boost converters, improving stability, efficiency, and

adaptability under dynamic and nonlinear conditions. Future work should prioritize strate-
gies to enhance real-world implementation and reduce computational demands, ensuring
that the optimization process remains efficient and practical for practical applications in
power electronics.

Author Contributions: Conceptualization, R.B.D. and M.D.; methodology, R.B.D.; software, V.H.K.;
validation, R.B.D., M.D. and A.N.; formal analysis, R.B.D.; investigation, R.B.D.; resources, R.B.D.;
data curation, R.B.D.; writing—original draft preparation, R.B.D.; writing—review and editing, A.N.;
visualization, R.B.D.; supervision, A.N.; project administration, V.H.K.; funding acquisition, M.D. All
authors have read and agreed to the published version of the manuscript.

Funding: This research received no external funding.

Data Availability Statement: The original contributions presented in the study are included in the
article, further inquiries can be directed to the corresponding author.

Conflicts of Interest: The authors declare no conflict of interest.

References
1. Esfandiari, S.; Davari, M.; Gao, W.; Yang, Y.; Al-Haddad, K. A Novel Converter-Based PV Emulator Control Using Lambert W
Method and Fractional-Order Fuzzy Proportional-Integral Controller Trained by Harris Hawks Optimization. IEEE J. Emerg. Sel.
Top. Ind. Electron. 2024, 5, 1493–1507. [CrossRef]
2. Mughees, A.; Mughees, N.; Mughees, A.; Mohsin, S.A.; Ejsmont, K. Enhancing stability and position control of a constrained
magnetic levitation system through optimal fractional-order PID controller. Alex. Eng. J. 2024, 107, 730–746. [CrossRef]
3. Abdolahi, M.; Adabi, J.; Mousavi, S.Y.M. An adaptive extended Kalman filter with passivity-based control for DC-DC converter
in DC microgrids supplying constant power loads. IEEE Trans. Ind. Electron. 2023, 71, 4873–4882. [CrossRef]
4. Kaplan, O.; Bodur, F.; Ozdemir, M.B. Arbitrary Fixed-Time Sliding Mode Control for Buck Converter with Matched and
Mismatched Disturbances Based on Fixed-Time Observer. IEEE Access 2025, 13, 4582–4596. [CrossRef]
5. Gangula, S.D.; Nizami, T.K.; Udumula, R.R.; Chakravarty, A. Self-learning Controller Design for DC–DC Power Converters with
Enhanced Dynamic Performance. J. Control Autom. Electr. Syst. 2024, 35, 532–547. [CrossRef]
6. Kart, S.; Demir, F.; Kocaarslan, İ.; Genc, N. Increasing PEM fuel cell performance via fuzzy-logic controlled cascaded DC-DC
boost converter. Int. J. Hydrogen Energy 2024, 54, 84–95. [CrossRef]
7. Dey, U.; Veerachary, M. Two-Part Controller Design for Switched-Capacitor Based Buck–Boost Converter. IEEE Trans. Ind.
Electron. 2024; in press. [CrossRef]
8. Gaied, H.; Flah, A.; Kraiem, H.; Prokop, L. A comparison between the Quality of Two level and three levels bidirectional
Buck-boost converter using the neural network controller. IEEE Access 2024, 12, 94323–94336. [CrossRef]
9. Lins, A.W.; Krishnakumar, R. Tuning of PID controller for a PV-fed BLDC motor using PSO and TLBO algorithm. Appl. Nanosci.
2023, 13, 2911–2934. [CrossRef]
10. Liu, X. Design of CCM boost converter utilizing fractional-order PID and Lyapunov-based PID techniques for PF correction.
Electr. Eng. 2024, 107, 3451–3462. [CrossRef]
11. Acosta-Rodríguez, R.A.; Martinez-Sarmiento, F.H.; Múñoz-Hernandez, G.A.; Portilla-Flores, E.A.; Salcedo-Parra, O.J. Validation
of Passivity-Based Control and array PID in High-Power Quadratic Buck Converter through Rapid Prototyping. IEEE Access
2024, 12, 58288–58316. [CrossRef]
12. Sahin, A.K.; Cavdar, B.; Ayas, M.S. An adaptive fractional controller design for automatic voltage regulator system: Sigmoid-based
fractional-order PID controller. Neural Comput. Appl. 2024, 36, 14409–14431. [CrossRef]
13. Kanungo, A.; Kumar, P.; Gupta, V.; Saxena, N.K. A design an optimized fuzzy adaptive proportional-integral-derivative controller
for anti-lock braking systems. Eng. Appl. Artif. Intell. 2024, 133, 108556. [CrossRef]
14. Yakut, Y.B. A new control algorithm for increasing efficiency of PEM fuel cells–Based boost converter using PI controller with
PSO method. Int. J. Hydrogen Energy 2024, 75, 1–11. [CrossRef]
15. Ghamari, S.; Hajihosseini, M.; Habibi, D.; Aziz, A. Design of an Adaptive Robust PI Controller for DC/DC Boost Converter Using
Reinforcement-Learning Technique and Snake Optimization Algorithm. IEEE Access 2024, 12, 141814–141829. [CrossRef]
16. Rajamony, R.; Wang, S.; Ming, W. Modelling and optimal design of a multifunctional single-stage buck-boost differential inverter.
IEEE Open J. Power Electron. 2024, 5, 1328–1339. [CrossRef]
17. Ashok, B.; Michael, P.A. Integration of cascaded controllers for super-lift Luo converter with buck converter in solar photovoltaic
and electric vehicle. Analog Integr. Circuits Signal Process. 2024, 118, 449–466. [CrossRef]

61
Energies 2025, 18, 2034

18. Srivastava, A.; Manas, M.; Dubey, R.K. Optimal electrical vehicle-to-grid integration: Three-phase three-level AC/DC converter
with model predictive controller based bidirectional power management scheme. Electr. Eng. 2024, 1–23. [CrossRef]
19. Liu, J.; Wei, T.; Chen, N.; Wu, J.; Xiao, P. Fuzzy logic PID controller with both coefficient and error modifications for digitally-
controlled DC–DC switching converters. J. Electr. Eng. Technol. 2023, 18, 2859–2870. [CrossRef]
20. Liu, W.; Liu, Y.; Wang, A. Design of CCM boost converter using fractional-order PID and neural-network techniques for power
factor correction. J. Eng. Appl. Sci. 2024, 71, 227. [CrossRef]
21. Sangeetha, S.; Revathi, B.S.; Balamurugan, K.; Suresh, G. Performance analysis of buck converter with fractional PID controller
using hybrid technique. Robot. Auton. Syst. 2023, 169, 104515. [CrossRef]
22. Warrier, P.; Shah, P.; Sekhar, R. A Comparative performance evaluation of a complex-order PI controller for DC–DC converters.
Results Control Optim. 2024, 15, 100414. [CrossRef]
23. Nanyan, N.F.; Ahmad, M.A.; Hekimoğlu, B. Optimal pid controller for the dc-dc buck converter using the improved sine cosine
algorithm. Results Control Optim. 2024, 14, 100352. [CrossRef]
24. Fong, L.K.; Islam, M.S.; Ahmad, M.A. Optimized PID Controller of DC-DC Buck Converter based on Archimedes Optimization
Algorithm. Int. J. Robot. Control Syst. 2023, 3, 658–672. [CrossRef]
25. Ghamari, S.M.; Khavari, F.; Mollaee, H. Lyapunov-based adaptive PID controller design for buck converter. Soft Comput. 2023,
27, 5741–5750. [CrossRef]
26. Sharma, P.; Palwalia, D.K.; Sharma, A.K.; Priyadarshi, N.; Padmanaban, S. Coati optimized FOPID controller for non-isolated
DC–DC converters in EV charging application. IET Power Electron. 2024, 17, 2771–2784. [CrossRef]
27. Acevedo, D.M.; Gil-González, W.; Montoya, O.D.; Restrepo, C.; González-Castaño, C. Adaptive Speed Control for a DC Motor
Using DC/DC Converters: An Inverse Optimal Control Approach. IEEE Access 2024, 12, 154503–154513. [CrossRef]
28. Chouay, Y.; Ouassaid, M. An Enhanced Buck-Boost Converter for Photovoltaic Diagnosis application: Accurate MPP Tracker and
IV Tracer. Sci. Afr. 2025, 27, e02561. [CrossRef]
29. Prag, K.; Woolway, M.; Celik, T. Data-driven model predictive control of DC-to-DC buck-boost converter. IEEE Access 2021,
9, 101902–101915. [CrossRef]
30. Mansour, N.A.; Saraya, M.S.; Saleh, A.I. Groupers and moray eels (GME) optimization: A nature-inspired metaheuristic algorithm
for solving complex engineering problems. Neural Comput. Appl. 2025, 37, 63–90. [CrossRef]
31. Wang, Z.; Ma, J.; Gao, Q.; Bain, C.; Imoto, S.; Liò, P.; Cai, H.; Chen, H.; Song, J. Dual-stream multi-dependency graph neural
network enables precise cancer survival analysis. Med. Image Anal. 2024, 97, 103252. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

62
Article
Stability Analysis and Controller Optimization of MMC in
Standalone Mode
Xingyu Liu 1 , Shuguang Song 1, *, Wenzhong Ma 1 and Yusheng Wang 2

1 College of New Energy, China University of Petroleum (East China), Qingdao 266580, China
2 PetroChina Planning & Engineering Institute, Beijing 100080, China
* Correspondence: shuguang.song@upc.edu.cn

Abstract: The modular multilevel converter (MMC) plays an important role in large-scale renewable
energy integration and transmission, and it can also operate in standalone mode, powering AC
passive loads. This paper focuses on the impact of load variation on the stability of the MMC. First,
the impact of load variation on the MMC transfer function is analyzed in detail using the harmonic
state-space (HSS) modeling method. Then, by means of the impedance-based stability analysis
method, it is found that the MMC tends to become unstable with the increase in inductive loads. If
the controller is not well-designed, the system may fail when loads change. Therefore, the worst-case
design is used to guarantee the overall system’s stability under all load conditions. The impact of
traditional proportional resonant (PR) controller parameters on the system’s stability is analyzed,
revealing that the stability margin and control performance of the controller are limited. Thus, an
improved controller structure with an additional series of compensators is proposed. Extensive
simulation results in MATLAB/Simulink R2024a verify the analysis of this work and the effectiveness
of the proposed controller.

Keywords: controller design; load variation; modular multilevel converter; stability analysis

1. Introduction
Nowadays, high-voltage direct current systems based on modular multilevel con-
verters (MMC-HVDC) have been widely adopted owing to their advantages, such as low
switching frequency, high efficiency, and high performance [1–3]. However, stability issues
like wide frequency oscillation have been increasingly emerging. A number of incidents
have been reported in practical MMC-HVDC projects, e.g., 20~30 Hz sub-synchronous os-
cillation (SSO) was reported in the Nan’ao three-terminal MMC-HVDC system in China [4]
and a 1270 Hz resonance occurred in the Luxi back-to-back MMC-HVDC project [5]. There-
fore, the oscillation issues in the MMC-HVDC system have received a lot of attention.
In order to analyze the oscillation phenomenon, many small-signal modeling methods
have been developed, such as the dynamic phasor method [6–8], harmonic state-space (HSS)
method [2,9,10], and multi-harmonic linearization method [11,12]. Guo et al. developed
a comprehensive dynamic phasor-based small-signal model for MMC systems under
unbalanced grid conditions, which considered the internal dynamics of the MMC system as
well as the more complex controllers, and revealed the reasons for unstable oscillations [6].
Sakinci et al. developed a generalized dynamic phasor model to reduce the order of
the conventional dynamic phasor model using the natural frequency separation of the
variations, where the impacts of high-order harmonics on the small-signal stability of the
MMC were obtained [8]. The HSS modeling approach was first introduced to characterize
the multi-harmonic coupling behavior of the MMC in [2], where the accurate small-signal
impedance models were developed. Wu et al. established an HSS-based MMC small-signal
impedance model with the frequency-coupling dynamics considered, which was used for
analyzing the SSO phenomenon [9]. Xu et al. proposed a complete HSS-based MMC small-
signal impedance model considering the impedance of the grid, and the mechanism of

Energies 2024, 17, 5474. https://doi.org/10.3390/en17215474 63 https://www.mdpi.com/journal/energies

Energies 2024, 17, 5474

MMC grid impedance coupling was illustrated in detail based on the proposed model [10].
Zhu et al. presented the sequence impedance model of MMC based on the multi-harmonic
linearization method and analyzed potential resonance modes, where the impact of the
circulating current controller on the MMC impedance and system stability was studied [11].
In ref. [12], a special form of the multi-harmonic linearization method was adopted to
reduce the complexity of the sequence impedance modeling of MMC depending on the
simpler harmonic spectra.
Moreover, many oscillation suppression solutions have also been proposed to enhance
the stability of the MMC, which can be divided into two categories: control parameter
optimization and control structure modification. Lyu et al. analyzed the impacts of the pro-
portional gain of the AC voltage controller on the AC-side impedance of MMC, confirming
that a larger proportional gain can cause a smaller magnitude of the MMC impedance but
hardly affected the resonant frequencies [2]. Huang et al. obtained the optimal controller
parameters by traversing all the values within the derived boundaries of the regulator
parameters, considering the delay generated by digital control [13]. Wang et al. investi-
gated the effects of different control loop parameters on the hybrid passivity property and
proposed a parameter design guideline based on the hybrid passivity criterion [14]. More-
over, many scholars also focus on the controller structure [15–21]. Wang et al. proposed a
full power decoupling strategy with a new closed-loop DC-bus current proportional (P)
controller for the direct-modulation-based MMC, which improved the stability of the MMC
system [15]. Gao et al. presented an improved virtual impedance control method, which
was the addition of the virtual resistance-inductance to ameliorate the virtual impedance
control that can be equivalent to the parallel connection of resistor and inductor, and it
was demonstrated that the proposed method showed better performance compared to the
traditional virtual resistance control and the phase lag compensation control method [17].
Guo et al. proposed a supplementary damping controller to mitigate the low-frequency os-
cillations in the MMC-HVDC system under low inertia conditions by introducing a reactive
power compensation signal into the reactive power outer loop control of the MMC [19]. Zhu
et al. proposed a damping method with virtual arm impedance by introducing common-
mode compensation terms in the insertion index to compensate for the stability defects
caused by the circulating current injection methods, reducing the submodule capacitor
voltage ripples [21].
Generally, MMC-HVDC systems can operate in grid-connected mode [22,23] and
standalone mode [9,16]. In grid-connected mode, the MMC connects directly to a power
grid, which can be either stiff or weak. In standalone mode, the MMC can connect to
renewable power plants and power other different AC loads. Up to now, the grid-connected
mode has been studied extensively. However, the standalone mode has not been examined
in detail.
In this operation scenario, the connected loads can be modeled as resistive load Rs
and inductive load Ls . Based on existing MMC stability analysis and controller design, the
system can achieve stable operation under certain load conditions. However, in practical
operation, the loads may vary according to the requirement, which changes the MMC
operation points. When the load variation is small, the system can remain stable. However,
with the increase in load variation, oscillations may appear, and in severe cases, the MMC
system could even fail. How the load variation impacts the MMC internal and external
small-signal characteristics still remains unknown. Therefore, it is of great importance to
find out the MMC stability evolution trend to assist in designing controller structure and
its parameters.
Moreover, the basic proportional resonant (PR) controller is commonly adopted for
the MMC in standalone mode to achieve AC voltage control. However, the control degrees
of freedom for the PR controller are limited. Given the complex harmonic interactions and
small-signal dynamics of the MMC, the traditional PR controller can hardly stabilize the
MMC system [9]. Additionally, the transient behavior also shows large voltage overshoot
and long settling time, which is unsatisfactory. To improve the stability margin and control

64
Energies 2024, 17, 5474

performance, the existing controller structure needs to be redesigned with consideration of

the MMC internal and external small-signal characteristics.
To address the above issues, in this paper, the time-domain model and frequency-
domain model are first established based on the average topology of the MMC [4,10], which
can effectively reflect the dynamic information of the sub-synchronous frequency range. It
is worth noting that the delay generated by digital control affects the high-frequency range
dynamics of the MMC. But this paper focuses on the sub-synchronous frequency range,
so the delay effect is not considered. Then, the impact of load variation on the transfer
function of internal small-signal perturbation to state variables is analyzed in detail using
HSS. It is found that load variation significantly affects the transfer function of internal
small-signal perturbation to the AC voltage perturbation response. Furthermore, the AC
terminal impedance of the MMC is analyzed. It is shown that AC impedance varies along
with the AC load, and the system tends to be unstable with the increase in inductive load.
Therefore, the MMC controllers need to be well-designed. Otherwise, the system may fail
under certain load conditions. Based on the above analysis, the worst-case design for the
MMC controller considering load variation is discussed in this paper, which can ensure
overall system stability. In existing literature, the proportional resonant (PR) controller is
commonly used in the control loop of the MMC system [2,9,24–26]. However, given that
the traditional PR controller has limited control performance and is hard to further improve
the stability margin, an improved output voltage controller structure is proposed, which
introduces additional series compensators according to the MMC internal small-signal
dynamic nature. It is demonstrated that the proposed controller can suppress the resonant
peak and decrease the phase differences up to ~110◦ .
The research highlights of this paper are concluded as follows:
(1) Based on the HSS method, the impact of load variation on the MMC internal (control
to state variable transfer function) and external (AC terminal impedance) small-signal
characteristics is studied in detail.
(2) The worst-case design for the MMC controller is adopted to ensure the overall system’s
stability. The control performance of the traditional PR controller is analyzed and
proven to be limited.
(3) An improved output voltage controller structure with additional series compensators
is proposed according to the actual MMC internal and external small-signal dynamic
nature, which improves both the stability margin and control performance.
The remainder of this paper is organized as follows: Section 2 reviews the time-domain
mathematical model and HSS-based small-signal model of the MMC. The impact of load
variation on the MMC transfer function is examined in Section 3. Then, MMC stability is
analyzed in detail in Section 4. Section 5 discusses the limitation of the PR controller and
proposes the improved controller structure. Section 6 concludes this paper.

2. Modeling of Modular Multilevel Converter

In this section, the topology and mathematical model of the MMC are reviewed.

2.1. Circuit Configuration

The topology of a three-phase MMC operating in standalone mode is depicted in Figure 1.
In each phase of MMC, there are two arms, namely the upper arm and the lower arm. Each
arm consists of N submodules (SMs) connected in series. and are the arm inductance and
resistance, respectively. The half-bridge SM structure is shown on the right-hand side, which
contains a floating capacitor and two switching devices. The induction motor, which is a
common industrial load, can be expressed as an inductor in series with a resistor [27]. Thus,
the AC load in Figure 1 is represented by an inductor Ls connecting with the resistor Rs .

65
Energies 2024, 17, 5474

LGF
LX$ LX% LX&
ˇ ˇ &60
60  60  60 
XX$
60 1 60 1 60 1
8GF ˉ 5DUPP 5DUPP 5DUP
/DUPP /DUPP /DUP
L$ / V 5V
ˉ
L% /V 5V

L& /V 5V

ˇ /DUPP /DUPP /DUP

5DUPP 5DUPP 5DUP
ˇ
8GF 60  60  60 
XO$
60 1 60 1 60 1
ˉ ˉ LO$ LO% LO&

Figure 1. The topology of a three-phase MMC.

For simplification, Figure 2 gives the average model of one phase leg of the MMC, where
vcu ∑ and vcl ∑ are the sum capacitor voltages for the upper and lower arms, respectively [28].
Ceq represents the arm equivalent capacitance, where Ceq = CSM /N. mu (t) and ml (t) are
the control signals for upper and lower arms, which can be expressed as [2]
⎧
⎪
⎨ m u ( t ) = 1 [1 − m ( t ) − m2 ( t ) + ∑ ∞
n=3 mn ( t )]
2 (1)
⎪ 1
⎩ m l ( t ) = [1 + m ( t ) − m2 ( t ) + ∑ ∞n=3 mn ( t )]
2
where m(t) and m2 (t) denote the fundamental frequency component for output control and
second-order harmonic component for circulating current control, respectively. Moreover,
for n ≥ 3, mn (t) in (1) is normally equal to zero [2].

LX
ˇ PXLX
YFX  &HT YX

8GF
LFP 5DUP
/DUP
ˉ LJ /V 5V
YJ

/DUP
ˇ
5DUP
POLO
8GF
YFO  &HT YO

ˉ LO

Figure 2. The average model of one phase leg of the MMC.

The circulating current icm (t) and the AC output current i g (t) are defined as

iu (t) + il (t)
icm (t) = (2)
2

i g (t) = iu (t) − il (t) (3)

66
Energies 2024, 17, 5474

The equivalent arm output voltages can be expressed as

vu (t) = mu (t)vcu ∑ (t) (4)

vl (t) = ml (t)vcl ∑ (t) (5)

The capacitor charging and discharging dynamics can be expressed as [9]

dvcu ∑
Ceq = mu iu (6)
dt

dvcl ∑
Ceq = ml il (7)
dt
Applying Kirchhoff’s law to the upper arm and the lower arm, two circuit equations
can be obtained as
diu (t) U
vu (t) + R arm iu (t) + L arm + v g (t) = dc (8)
dt 2

dil (t) U
vl (t) + R arm il (t) + L arm − v g (t) = dc (9)
dt 2
where v g (t) = Rs i g (t) + Ls [di g (t)/dt].
Applying the differential-mode (DM) and common-mode (CM) definitions [29], the
sum and difference in the control signals and equivalent SM capacitor voltages for the
lower and upper arms in Figure 2 can be defined as

ml (t) + mu (t) = 2mcm (t) (10)

ml (t) − mu (t) = 2mdm (t) (11)

vcl ∑ (t) + vcu ∑ (t) = 2uCcm (t) (12)

vcl ∑ (t) − vcu ∑ (t) = 2uCdm (t) (13)

Substituting (1) into (10) and (11), CM and DM control signals can be written as

1
mcm (t) = [1 − m2 cos(2ω1 t + θm2 )] (14)
2

1
mdm (t) = m cos(ω1 t + θm1 ) (15)
2
where ω1 represents the fundamental angular frequency.
Substituting (2)–(5) and (10)–(13) into (6)–(9), MMC dynamics in DM and CM repre-
sentation are given as follows:
⎧
⎪ dicm (t) m (t) m (t)
⎪
⎪ = − RLarm
arm
icm (t) − Lcmarm uCcm (t) − Ldmarm uCdm (t) + 2L1arm Udc
⎪
⎪
dt
⎨ di g (t) = − Rarm +2Rs i g (t) + 2mdm (t) u 2mcm (t)
dt L arm +2Ls L arm +2Ls Ccm ( t ) + L arm +2Ls uCdm ( t )
duCcm (t) (16)
⎪
⎪
⎪ dt = mCcmeq(t) icm (t) − m2C
dm ( t )
i g (t)
⎪
⎪
eq
⎩ duCdm (t) = mdm (t) icm (t) − mcm (t) i g (t)
dt Ceq 2Ceq

67
Energies 2024, 17, 5474

2.2. HSS-Based MMC Steady-State Model

Based on (16), choosing the circulating current icm (t), output current i g (t), CM arm
capacitor voltage uCcm (t), and DM arm capacitor voltage uCdm (t) as state variables, the
time-domain state-space model of the MMC can be expressed as
.
x (t) = A(t) x (t) + B(t)u(t) (17)

where
.
x (t) = [icm (t), i g (t), uCcm (t), uCdm (t)] T (18)

⎡ ⎤
− RLarm
arm
0 − mLcmarm(t) − mLdmarm(t)
⎢ 2mdm (t) 2mcm (t) ⎥
⎢ 0 − RLarm +2Rs ⎥
⎢ arm +2Ls L arm +2Ls L arm +2Ls ⎥
A(t) = ⎢ mcm (t) dm ( t )
⎥ (19)
⎢ Ceq − m2C 0 0 ⎥
⎣ eq ⎦
mdm (t) cm ( t )
Ceq − m2C eq
0 0

1
B(t) = diag( , 0, 0, 0) (20)
2L arm

u(t) = [Udc , 0, 0, 0] T (21)

The detailed HSS modeling procedure can be found in [2,9]. Thus, it is not repeated
in this paper. Based on the Fourier series and harmonic balance theory, (17)–(21) can be
transformed into the frequency-domain HSS model, which is expressed as

sX = (A − Q)X + BU (22)

where
X = [ X − h , . . . , X − 1 , X0 , X1 , . . . , X h ] T (23)

⎡ ⎤
A0 A −1 ··· A−h
⎢ .. .. .. .. ⎥
⎢A . . . . ⎥
⎢ 1 ⎥
⎢. .. .. .. ⎥
⎢ .. . . . ⎥
⎢ A0 A −1 ⎥
⎢ ⎥
⎢ .. .. .. ⎥
A = ⎢ Ah . . A0 A −1 . A−h ⎥ (24)
⎢ .. ⎥
⎢ .. .. .. ⎥
⎢ . . A1 A0 . . ⎥
⎢ ⎥
⎢ .. .. .. .. ⎥
⎢ . . . . A −1 ⎥
⎣ ⎦
..
Ah . A1 A0
⎡ ⎤
B0 B−1 ··· B−h
⎢ .. .. .. .. ⎥
⎢B . . . . ⎥
⎢ 1 ⎥
⎢. .. .. .. ⎥
⎢ .. . . . ⎥
⎢ B0 B−1 ⎥
⎢ ⎥
⎢ .. .. .. ⎥
B = ⎢ Bh . . B0 B−1 . B−h ⎥ (25)
⎢ .. ⎥
⎢ .. .. .. ⎥
⎢ . . B1 B0 . . ⎥
⎢ ⎥
⎢ .. .. .. .. ⎥
⎢ . . . . B−1 ⎥
⎣ ⎦
..
Bh . B1 B0

U = [U−h , . . . , U−1 , U0 , U1 , . . . , Uh ] T (26)

Q = diag(− jhω1 I, . . . , − jω1 I, 0, jω1 I, . . . , jhω1 I) (27)

68
Energies 2024, 17, 5474

where A and B are called Toeplitz matrix. The subscripts h of the matrix elements represent
the harmonic order. For instance, Ah , Bh are the corresponding h th Fourier coefficients of
A(t) and B(t), respectively.
Letting the left-hand side of (22) be zero, the steady-state operation point of MMC can
be calculated by
Xss = −(A − Q)−1 (BU) (28)

2.3. HSS-Based MMC Small-Signal Model

To derive the small-signal model of the MMC, the harmonic linearization principle
needs to be adopted. Applying the small perturbation to (17), the MMC small-signal
state-space equation can be obtained as

x p (t) = A(t) x p (t) + A p (t) x (t) + B(t)u p (t) (29)

where subscript p denotes the small-signal perturbation components.

In steady state, MMC contains ±h th (h ∈ Z) steady-state harmonics. According to
the operation principle of the MMC, the injected small-signal perturbation with angular
frequency ω p can be coupled with these steady-state harmonics, which generates a series
of harmonics with angular frequency ω p ± hω1 (h ∈ Z).
Applying the Fourier series and harmonic balance theory, (29) can be transformed into
the frequency-domain model as follows:

sX p = (A − Q p )X p + A p X + BU p (30)

X p = [ X p − h , · · · , X p −1 , X p , X p +1 , · · · , X p + h ] T (31)

U p = [U p − h , · · · , U p −1 , U p , U p +1 , · · · , U p + h ] T (32)

⎡ ⎤
Ap A p −1 ··· A p−h
⎢ .. .. .. .. ⎥
⎢ A p +1 . . . . ⎥
⎢ ⎥
⎢ . .. .. .. ⎥
⎢ .. . . . ⎥
⎢ Ap A p −1 ⎥
⎢ ⎥
⎢ .. .. .. ⎥
A p = ⎢ A p+h . . Ap A p −1 . A p−h ⎥ (33)
⎢ .. ⎥
⎢ .. .. .. ⎥
⎢ . . A p +1 Ap . . ⎥
⎢ ⎥
⎢ .. .. .. .. ⎥
⎢ . . . . A p −1 ⎥
⎣ ⎦
..
A p+h . A p +1 Ap

Q p = diag[ j(ω p − hω1 )I, · · · , j(ω p − ω1 )I, jω p I, j(ω p + ω1 )I, · · · , j(ω p + hω1 )I] (34)

However, in closed-loop control, the small-signal coefficient matrix A p is related to the

small-signal state variables X p . Therefore, (30) cannot be calculated directly, and a variable
separation is needed [30], where (30) can be further reorganized as

sX p = (A − Q p − G)X p + GRp + BU p (35)

where G is the control matrix, which is composed of the product of the controller transfer
function and steady-state state variables, and Rp represents the small-signal perturbation
of the control reference.

69
Energies 2024, 17, 5474

Letting the left-hand side of (35) be zero, the small-signal perturbation of the state
variables X p can be calculated by

X p = −(A − Q p − G)−1 (GRp + BU p ) (36)

Based on the MMC small-signal model, many scholars established the MMC terminal
impedance and transfer function model to investigate the system stability and controller
design issues.

3. Impact of Load Variation

In this section, the traditional controller structure is first introduced. Then, the transfer
functions of internal perturbation to AC voltage and circulating current are analyzed, respec-
tively. To support the analysis, an MMC simulation system is built in MATLAB/Simulink
R2024a. The main parameters of the MMC are given in Table 1. Table 2 shows the different
load conditions studied in this paper, where | Zs | remains the same but Rs /Xs varies.

Table 1. Parameters of the MMC model.

Parameters Symbols Values

Rated output power Po 100 MVA
Grid frequency fg 50 Hz
DC bus voltage Udc ±160 kV
AC L-L voltage (rms) Vgrms 166 kV
Number of the SM per arm N 200
SM capacitance CSM 2.5 mF
Equivalent SM capacitance Ceq 12.5 μF
Arm inductance L arm 0.18 mH
Arm resistance R arm 4.06 Ω

Table 2. Parameters of the AC load.

Load Conditions Rs Ls
Case I 275.56 Ω 0
Case II 194.85 Ω 0.6202 H
Case III 123.23 Ω 0.7845 H
Case IV 87.14 Ω 0.8321 H
Case V 54.04 Ω 0.8601 H
Case VI 45.30 Ω 0.8652 H
Case VII 34.18 Ω 0.8704 H
Case VIII 0 0.8771 H

3.1. Traditional Controller Structure

The MMC always operates in closed-loop mode rather than open-loop mode. There-
fore, the traditional controller structure is introduced. In standalone mode, MMC commonly
uses the voltage-controlled closed-loop form, where the AC voltage controller structure
traditionally adopts the basic proportional resonant (PR) controller [2,9]. Additionally, the
circulating current controller is also added in the closed-loop configuration, where the PR
controller is used to suppress the second-order circulating current.
Figure 3 shows the block diagrams of the PR controllers for AC voltage and circu-
lating current control. The delay generated by digital control affects the high-frequency
range dynamics of MMC [13,31]. However, as this paper focuses on the sub-synchronous
frequency range, the delay effect is not considered. The corresponding transfer functions
are expressed as
2K Rvg ωc s
Gvg (s) = K Pvg + 2 (37)
s + 2ωc s + ω12

70
Energies 2024, 17, 5474

2K Ricm ωc s
Gicm (s) = K Picm + (38)
s2 + 2ωc s + (2ω1 )2
where K Pvg , K Picm , K Rvg , and K Ricm are the proportional and resonant gain for the AC voltage
and circulating current controller, respectively. Moreover, the resonant angular frequencies of
the AC voltage and circulating current controller are tuned at ω1 and 2ω1 , respectively.

YJUHI  P LFPUHI  P
*YJ V *LFP V
 
YJ LFP

(a) (b)
Figure 3. Block diagrams of the MMC controllers (a) AC voltage control; (b) circulating current control.

Based on the control theory, open-loop gain is crucial for analyzing the system stability,
which is composed of controller and control to state variable transfer function. Therefore,
in the following parts, the impact of AC load variation on the control to AC voltage and
circulating current transfer functions is studied in detail.

3.2. Transfer Function of Internal Perturbation to AC Voltage

The AC voltage is controlled by the DM control signal. Thus, the internal DM small-
signal perturbation mdm,p (t) at angular frequency ω p with amplitude Mdm,p in the MMC
control signal should be added, which can be expressed as

mdm,p (t) = Mdm,p cos(ω p t) (39)

The corresponding perturbation matrix consisting of internal DM small-signal per-

turbation can be expressed as Adm,p (t). Then, the state variable small-signal perturbation
response can be expressed by

X p = −(A − Q p )−1 (Adm,p X) (40)

By solving (40), the corresponding AC current small-signal perturbation response

i g,p (t) can be obtained. The value of the injected internal DM small-signal perturbation to
AC current transfer function at angular frequency ω p can be calculated by

Ig,p
TMMC ( jω p ) = (41)
Mdm,p

where Ig,p and Mdm,p represent the phase of the AC current small-signal perturbation and
the injected internal DM small-signal perturbation at angular frequency ω p , respectively.
The value of AC voltage small-signal perturbation at angular frequency ω p can be
obtained by
Ug,p = ( Rs + jω p Ls )Ig,p (42)
As a result, the value of the injected internal DM small-signal perturbation to the AC
voltage transfer function at angular frequency ω p can be calculated by

( Rs + jω p Ls )Ig,p
TMMC1 ( jω p ) = (43)
Mdm,p

And the transfer function measurement process in simulation can be described as

follows: First, a DM small-signal perturbation mdm,p (t) = Mdm,p cos(ω p t) is injected at
angular frequency ω p into the control signal. Then, the fast Fourier transform (FFT) module
is used to obtain the magnitude and phase of the corresponding AC voltage small-signal

71
Energies 2024, 17, 5474

response at angular frequency ω p . Change the perturbation frequency, and repeat the
whole process until the measurement needs are met.
Figure 4 shows the bode diagram of the injected internal DM small-signal perturbation
to AC voltage transfer functions under different AC load conditions. The solid lines and
marks represent the analytical and simulated results of TMMC1 ( jω p ) in (43), respectively.
The analytical results clearly agree well with the simulation results. It can be observed that
the magnitude response is generally flat at lower frequencies. Then, a resonant peak occurs
at ~28 Hz, which is augmented greatly with the increase in inductive load. The magnitude
difference can reach up to 19 dB. Correspondingly, the phase response shows a significant
shift around the frequency where the magnitude peaks.

Figure 4. Model verification of internal DM perturbation to AC voltage transfer functions under

different load conditions.

3.3. Transfer Function of Internal Perturbation to Circulating Current

The circulating current is controlled by the CM control signal. Therefore, a predefined
internal CM small-signal perturbation mcm,p (t) can be injected into the control function,
whose expression can be written as

mcm,p (t) = Mcm,p cos(ω p t) (44)

where Mcm,p and ω p are the amplitude and angular frequency of internal CM small-
signal perturbation, respectively. Similarly, the corresponding state variables small-signal
perturbation can be calculated by

X p = −(A − Q p )−1 (Acm,p X) (45)

Therefore, the transfer function of internal CM small-signal perturbation to circulating

current at angular frequency ω p can be calculated by

Icm,p
TMMC2 ( jω p ) = (46)
Mcm,p

The measurement process is similar except that an internal CM small-signal perturba-

tion is injected in simulation. Figure 5 shows the bode diagram of the injected internal CM
small-signal perturbation to circulating current transfer functions under different load con-
ditions. The solid lines and marks are the analytical and simulation results of TMMC2 ( jω p )
in (46), respectively. The analytical results agree well with the simulation results as well.
Compared with the AC voltage transfer function, both magnitude and phase responses

72
Energies 2024, 17, 5474

show weak correlation with different load cases. Phase differences can only be observed
below 10 Hz, while there is no obvious magnitude difference, indicating that the MMC
internal circulating current small-signal dynamics are hardly affected.

Figure 5. Model verification of internal CM perturbation to circulating current transfer functions

under different load conditions.

Based on the above analysis, the load variation has a larger impact on the transfer
function of internal DM small-signal perturbation to the AC voltage perturbation response. In
other words, it greatly affects the stability of AC voltage control. Therefore, in the next section,
a detailed discussion of the MMC stability under different load conditions is carried out.

4. MMC Stability Analysis Considering Load Variation

The analysis in the previous section verifies that load variation has a great impact on
MMC small-signal dynamics. In this section, MMC stability is discussed in detail based on
the impedance-based stability analysis method.

4.1. Impedance-Based Stability Analysis Method for the MMC System

To analyze the stability of closed-loop systems, the impedance-based stability analysis
method is adopted. Figure 6 shows the Thevenin equivalent circuit of the MMC operating
in standalone mode, where the current flowing from the MMC system to the load can be
written as [32,33]
v MMC (s)
I (s) = (47)
ZLoad (s) + Z MMC (s)
which can be reorganized as

v MMC (s) 1
I (s) = · (48)
ZLoad (s) 1 + Z MMC (s)/ZLoad (s)

=00& ,

Y00& =/RDG

Figure 6. Thevenin equivalent circuit of MMC operating in standalone mode.

73
Energies 2024, 17, 5474

It can be assumed that the MMC voltage is stable when unloaded and the load current
is stable when powered from an ideal source. Therefore, both v MMC (s) and 1/ZLoad (s) are
stable. Then, current stability depends on (49)

1
H (s) = (49)
1 + Z MMC (s)/ZLoad (s)

By linear control theory, H (s) is stable if and only if Z MMC (s)/ZLoad (s) satisfies the
Nyquist stability criterion, where the details are as follows [34]:
“If and only if the total number of counterclockwise encirclements of (−1 + j0) by the
eigen-loci of Z MMC (s)/ZLoad (s) is equal to the total number of right-half-plane (RHP) poles
of Z MMC (s) and 1/ZLoad (s), the closed-loop MMC system shown in Figure 6 is stable”.
Considering that Z MMC (s) and 1/ZLoad (s) have no RHP poles, the above criterion can
be reorganized as follows:
“If and only if the eigen-loci of Z MMC (s)/ZLoad (s) does not encircle (−1 + j0) coun-
terclockwise, the closed-loop MMC system shown in Figure 6 is stable”.

4.2. Stability Analysis of the Closed-Loop MMC System Under Different Load Conditions
Figure 7a shows the bode diagram of Zclose0 (s) and ZLoad (s) of load case II with
controller parameters in Table 3. The analytical results agree well with the simulation
results. It is observed that the phase difference at the intersection point of Zclose0 (s) and
ZLoad (s) is much lower than 180◦ , which satisfies the Nyquist stability criterion. Figure 7b
shows the time-domain waveforms of output voltage and current, indicating that the
system works stably. Figure 7c shows the time-domain waveforms of circulating current in
phase A, which fluctuates around 75 A, indicating stable system operation. And the sum
capacitor voltages for the upper and lower arms remain stable, as shown in Figure 7d. It is
noted that AC voltage controller parameters in Table 3 are designed based on load case II.

(a) (b)

(c) (d)
Figure 7. Bode diagrams and time-domain waveforms of load case II. (a) Zclose0 (s) and ZLoad (s)
(solid lines: analytical results; cross markers: simulation results); (b) output voltage and current;
(c) circulating current; (d) sum capacitor voltages for the upper and lower arms.

74
Energies 2024, 17, 5474

Table 3. Parameters of the traditional AC voltage PR controller.

Controller Parameters KPvg KRvg ωc

Values 0.8 900 π

Figure 8a shows the bode diagram of Zclose0 (s) and ZLoad (s) of load case VI. It can be
seen that the phase difference at the intersection point is higher than 180◦ , which dissatisfies
the Nyquist stability criterion. Figure 8b shows the corresponding time-domain waveforms
of output voltage and current, where harmonic components appear. As shown in Figure 8c,
the circulating current in phase A exhibits distortion. In Figure 8d, the sum capacitor
voltages for the upper and lower arms contain harmonic components.

(a) (b)

(c) (d)
Figure 8. Bode diagrams and time-domain waveforms of load case VI. (a) Zclose0 (s) and ZLoad (s);
(b) output voltage and current; (c) circulating current; (d) sum capacitor voltages for the upper and
lower arms.

Figure 9a shows the bode diagram of Zclose0 (s) and ZLoad (s) of load case VII. It is
observed that the phase difference at the intersection point is higher than 180◦ , which
indicates the system dissatisfies the Nyquist stability criterion. Figure 9b shows the cor-
responding time-domain waveforms of output voltage and current, which contain large
harmonic components, confirming the analysis. The corresponding circulating current
contains harmonic components, as shown in Figure 9c. Similarly, in Figure 9d, harmonic
components also appear in the sum capacitor voltages for the upper and lower arms. The
waveforms in Figure 9c,d have already become distorted. Furthermore, time-domain
waveforms of load case VIII are not shown because the system is severely unstable.

75
Energies 2024, 17, 5474

(a) (b)

(c) (d)

Figure 9. Bode diagrams and time-domain waveforms of load case VII. (a) Zclose0 (s) and ZLoad (s);
(b) output voltage and current; (c) circulating current; (d) sum capacitor voltages for the upper and
lower arms.

Additionally, Figure 10 shows the Fourier analysis results of Figures 8b and 9b.
In Figure 10a, it can be seen that SSO appears in the AC output voltage and current,
where the resonant frequency is 7.5 Hz, which is close to the analytical resonant frequency
of 7.55 Hz obtained in Figure 8a. In Figure 10b, SSO of 7.9 Hz appears in the AC output
voltage and current. The resonant frequency of the simulation is close to the analytical
resonant frequency of 7.99 Hz obtained in Figure 9a.

  +] +]
+] +]
  N9
9$>N9@

N9 N9
9$>N9@

 N9

 +]  +]
 N9  N9
 
 +]  +]
 +] $  +] $
L$>$@

L$>$@

 $  $

 +]  +]
 $  $
 
                     
)UHTXHQF\>+]@ )UHTXHQF\>+]@

(a) (b)

Figure 10. Fourier analysis results of load cases VI and VII. (a) Load case VI; (b) load case VII.

Based on the transfer functions shown in Figure 4 and the analysis above, the stability of
the MMC with the same AC voltage controller under different load conditions is varying. The
system tends to be unstable with the increase in inductive load. Thus, to guarantee the overall
stability, the worst-case design is adopted for the MMC AC voltage controller design.

4.3. Verification of the Worst-Case Design

The worst load condition is case VIII. Therefore, controller parameters in Table 3 need
to be adjusted according to its transfer function feature. The resonant gain K Rvg is changed
to 30. Figure 11 shows the bode diagrams of Zclose0 (s) and ZLoad (s) with all load conditions.
It can be seen that the phase difference at the intersection point of Zclose0 (s) and ZLoad (s)

76
Energies 2024, 17, 5474

in the worst case, VIII, is largest but lower than 180◦ . The phase differences are generally
reduced with the decrease in inductive load, and all load conditions satisfy the Nyquist
stability criterion, which ensures the overall system stability.

Figure 11. Bode diagrams of Zclose0 (s) and ZLoad (s) under different load conditions.

5. Controller Optimization
5.1. Controller Parameters Optimization
By changing controller parameters, the phase differences at the intersection point
of Zclose0 (s) and ZLoad (s) vary as well. Therefore, the impact of controller parameters on
system stability is studied in this part.
When K Rvg is set to 30, the bode diagram of Zclose0 (s) and ZLoad (s) of load case VIII with
the tuned K Pvg is shown as Figure 12a. The proportional gain K Pvg = 0.8 of the AC voltage
controller guarantees that the phase difference is lower than 180◦ . In addition, the increasing
KPvg reduces the phase differences, but its influence on system stability is quite small.

(a) (b)
Figure 12. Bode diagrams of Zclose0 (s) and ZLoad (s) of load case VIII with different controller parame-
ters (Soid lines: Zclose0 (s), dashed green lines: ZLoad (s)). (a) Changing KPvg ; (b) changing KRvg .

As for the resonant gain K Rvg , it significantly impacts the system’s stability, as shown
in Figure 12b, where K Pvg is set to 0.8. By tuning the resonant gain, it can be observed
that the phase differences at the intersection point of Zclose0 (s) and ZLoad (s) are gradually
reduced with the decrease of K Rvg . A small K Rvg = 30 can ensure the system’s stability.

77
Energies 2024, 17, 5474

When the resonant gains further decrease, e.g., 0.01, as shown in Figure 13a, the phase
difference reduces to 172.27◦ but is still close to 180◦ . However, in this case, the AC voltage
cannot track its reference value as shown in Figure 13b.

(a) (b)

Figure 13. Bode diagrams and time-domain waveforms of load case VII (Soid lines: Zclose0 (s), dashed
green lines: ZLoad (s)). (a) Zclose0 (s) and ZLoad (s); (b) output voltage and current.

From the discussion above, it can be found that tuning the parameters of the PR
controller can guarantee the system’s stability. But the phase difference is relatively large
and nearly impossible to further decrease. Therefore, an improved controller structure is
proposed in this paper.

5.2. Proposed Controller Structure

The structure of the proposed controller is shown in Figure 14, which contains
three parts:
(1) The first part is the traditional PR controller, which provides a high gain at the
fundamental frequency, allowing accurate tracking of the reference signal. K Rvg is the
key parameter to guarantee the tracking performance.
(2) A band reject filter (BRF) is introduced, which is the second part. Based on the
obtained internal DM small-signal perturbation to AC voltage transfer function, it is
found that a large resonant peak appears, which affects the control performance and
may threaten the system stability. The center frequency of the introduced BRF is set to
suppress the resonant peak frequency at 28 Hz, and its bandwidth matches the width
of the resonant peak, with a value set to 20 Hz.
(3) The third part is an additional inertia element (IE), which is added in the control loop
to improve the poor stability margin of the existing PR controller. Based on the bode
diagram of Zclose0 (s) and ZLoad (s) under the worst case, the gain K IEvg and cutoff
angular frequency ωie of the IE are designed.

YJUHI  P
35 %5) ,(

YJ

Figure 14. Block diagrams of the proposed controller structure.

The transfer function of the proposed controller can be written as

2K Rvg ωc s K Bvg (s2 + ωb2 ) K IEvg

Gvg (s) = (K Pvg + )· · (50)
s2 + 2ωc s + ω1 2 s2 + 2ωcb s + ωb2 s + ωie

78
Energies 2024, 17, 5474

Based on (50), the proposed controller has more control degrees of freedom, indicating
that the corresponding system has better abilities in stability and control performance
improvement. Figure 15 shows the flowchart of the parameter design of the proposed
controller. And the corresponding parameters are given in Table 4. Figure 16 shows
the bode diagram of the open-loop gain of the MMC system with the PR controller and
proposed controller of the load case VIII. It can be observed that the resonant peak is
suppressed, which verified the effectiveness of BRF. Moreover, the gain at the fundamental
frequency remains relatively high, showing good tracking performance.

2EVHUYHWKH= FORVH V DQG=ORDG V

2EWDLQWKHZRUVWFDVH

'HVLJQWKHWUDGLWLRQDO35FRQWUROOHUSDUDPHWHUV

2EVHUYHWKHFRQWUROORRSJDLQ

'HVLJQWKH%5)SDUDPHWHUV

0HHWUHVRQDQWSHDN
1R VXSSUHVVLRQUHTXLUHPHQW"

<HV
2EVHUYHWKH=FORVH V DQG=ORDG V 

'HVLJQWKH,(SDUDPHWHUV

0HHWVWDELOLW\PDUJLQ
1R UHTXLUHPHQW"

<HV
2EWDLQWKHSDUDPHWHUVRIWKHSURSRVHGFRQWUROOHU

Figure 15. Flowchart of the parameter design of the proposed controller.

Table 4. Parameters of the proposed controller.

Controller Parameters KPvg KRvg ωc KBvg ωb ωcb KIEvg ωie

Values 0.8 30 π 1 56 π 40 π 200 10 π

Figure 16. Bode diagram of the open-loop gain of the MMC with the traditional PR controller and
the proposed controller.

79
Energies 2024, 17, 5474

Figure 17 shows the bode diagram of Zclose0 (s) and ZLoad (s) of all load conditions with
the proposed controller. It can be seen that the phase difference at the intersection point of
Zclose0 (s) and ZLoad (s) of load case VIII is equal to 114.57◦ . The stability margin is greatly
increased, which confirms the validity and effectiveness of the proposed controller structure.

Figure 17. Bode diagrams of MMC of different load conditions with the proposed controller.

Moreover, Figure 18 shows the time-domain simulation results of the MMC with the
PR controller and the proposed controller in the transient state. The AC load switches from
case II to case V at t = 3.1 s. The results indicate that the MMC system reaches a stable
state. However, the settling time with the proposed controller is shorter, which verifies the
effectiveness of the proposed controller. The oscillation of the circulating current with the
proposed controller is also smaller, as shown in Figure 18c,d. Additionally, in Figure 18e,f,
the capacitor voltages also exhibit similar characteristics, further confirming the effectiveness
of the proposed controller structure.

(a) (b)

(c) (d)

Figure 18. Cont.

80
Energies 2024, 17, 5474

(e) (f)

Figure 18. Transient-state time-domain waveforms from case II to case V. Output voltage and current:
(a) traditional PR controller; (b) proposed controller. Circulating current: (c) traditional PR controller;
(d) proposed controller. Sum capacitor voltages for the upper and lower arms: (e) traditional PR
controller; (f) proposed controller.

Furthermore, Figure 19 shows the corresponding time-domain simulation results, where

the AC load switches from case V to case VIII at t = 3.1 s. In Figure 19a, large oscillations
appear in the output voltage and current. Although the system finally reaches a stable state,
the settling time is ~2.6 s, which indicates the poor control performance of the traditional
PR controller. Figure 19b shows the simulation results with the proposed controller. It can
be observed that the voltage and current quickly reach a stable state within only ~0.2 s,
which further verifies the effectiveness of the proposed controller structure. In Figure 19c, the
oscillation of the circulating current gradually decreases, with overshoot amplitude reaching
~100 A. However, for the MMC with the proposed controller, the overshoot amplitude of the
circulating current is only ~50 A, as shown in Figure 19d. Additionally, Figure 19e,f indicate
that the capacitor voltages have better dynamic performance with the proposed controller.
In Figure 19e, the sum capacitor voltages for the upper and lower arms exhibit large over-
shoots, which reach ~73 kV. Figure 19f shows that the overshoot amplitudes of the sum
capacitor voltages for the upper and lower arms are ~65 and ~37 kV, respectively. This further
verifies the effectiveness of the proposed controller structure.

(a) (b)

(c) (d)

Figure 19. Cont.

81
Energies 2024, 17, 5474

(e) (f)

Figure 19. Transient-state time-domain waveforms from case V to case VIII. Output voltage and current:
(a) traditional PR controller; (b) proposed controller. Circulating current: (c) traditional PR controller; (d)
proposed controller. Sum capacitor voltages for the upper and lower arms: (e) traditional PR controller;
(f) proposed controller.

6. Conclusions
This paper investigates the stability variation in MMC in standalone mode. Based
on the HSS, the impacts of load variation on the transfer function of internal small-signal
perturbation to state variables and AC terminal impedance of the MMC are analyzed in
detail. It is found that the system tends to be unstable with the increase in inductive load.
Therefore, a worst-case design needs to be adopted to guarantee the overall stability. Then,
the impact of traditional PR controller parameters on system stability is analyzed. It is
observed that the stability margin and control performance are limited. Thus, this paper
proposes an improved controller structure with an additional series compensator, which can
enhance the stability and suppress the resonant peak. And the transient performance is also
improved. The simulation results align well with the analysis, validating the effectiveness
of the proposed controller. Based on these findings, to guarantee the stability and control
performance, it is recommended that a detailed small-signal analysis of all load conditions
be carried out for the MMC system.

Author Contributions: Methodology, software, and writing, X.L.; formal analysis, validation, and
writing, S.S.; formal analysis, investigation, W.M.; resources, funding acquisition, Y.W. All authors
have read and agreed to the published version of the manuscript.
Funding: This research was funded by the Fundamental Research Funds for the Central Universities,
grant number 21CX06039A.
Data Availability Statement: Data is contained within the article.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Liu, W.; Li, K.J.; Liu, Z.; Wang, M. A Simple and Novel Precharging Control Strategy for Modular Multilevel Converter. IEEE
Access 2019, 7, 170500–170512. [CrossRef]
2. Lyu, J.; Zhang, X.; Cai, X.; Molinas, M. Harmonic State-Space Based Small-Signal Impedance Modeling of a Modular Multilevel
Converter with Consideration of Internal Harmonic Dynamics. IEEE Trans. Power Electron. 2019, 34, 2134–2148. [CrossRef]
3. Perez, M.A.; Ceballos, S.; Konstantinou, G.; Pou, J.; Aguilera, R.P. Modular Multilevel Converters: Recent Achievements and
Challenges. IEEE Open J. Ind. Electron. Soc. 2021, 2, 224–239. [CrossRef]
4. Lyu, J.; Yin, J.; Zhu, H.; Cai, X. Impedance Modeling and Stability Analysis of Energy Controlled Modular Multilevel Converter.
IEEE Trans. Power Deliv. 2023, 38, 1868–1881. [CrossRef]
5. Zou, C.; Rao, H.; Xu, S.; Li, Y.; Li, W.; Chen, J.; Zhao, X.; Yang, Y.; Lei, B. Analysis of Resonance Between a VSC-HVDC Converter
and the AC Grid. IEEE Trans. Power Electron. 2018, 33, 10157–10168. [CrossRef]
6. Guo, C.; Yang, J.; Zhao, C. Investigation of Small-Signal Dynamics of Modular Multilevel Converter Under Unbalanced Grid
Conditions. IEEE Trans. Ind. Electron. 2019, 66, 2269–2279. [CrossRef]
7. Zhu, S.; Liu, K.; Qin, L.; Ran, X.; Li, Y.; Huai, Q.; Liao, X.; Zhang, J. Reduced-Order Dynamic Model of Modular Multilevel
Converter in Long Time Scale and Its Application in Power System Low-Frequency Oscillation Analysis. IEEE Trans. Power Deliv.
2019, 34, 2110–2122. [CrossRef]

82
Energies 2024, 17, 5474

8. Sakinci, Ö.C.; Beerten, J. Generalized Dynamic Phasor Modeling of the MMC for Small-Signal Stability Analysis. IEEE Trans.
Power Deliv. 2019, 34, 991–1000. [CrossRef]
9. Wu, H.; Wang, X.; Kocewiak, Ł.H. Impedance-Based Stability Analysis of Voltage-Controlled MMCs Feeding Linear AC Systems.
IEEE J. Emerg. Sel. Top. Power Electron. 2020, 8, 4060–4074. [CrossRef]
10. Xu, Z.; Li, B.; Han, L.; Hu, J.; Wang, S.; Zhang, S.; Xu, D. A Complete HSS-Based Impedance Model of MMC Considering Grid
Impedance Coupling. IEEE Trans. Power Electron. 2020, 35, 12929–12948. [CrossRef]
11. Zhu, Y.; Pou, J.; Konstantinou, G. Impedance Shaping Effects of Circulating Current Controllers in Modular Multilevel Converters.
In Proceedings of the 2022 International Power Electronics Conference (IPEC-Himeji 2022-ECCE Asia), Himeji, Japan, 15–19 May
2022; pp. 555–561.
12. Zhang, Y.; Chen, X.; Sun, J. Sequence Impedance Modeling and Analysis of MMC in Single-Star Configuration. IEEE Trans. Power
Electron. 2020, 35, 334–346. [CrossRef]
13. Huang, T.; Yang, F.; Zhang, D.; Chen, X. High-Frequency Stability Analysis and Impedance Optimization for an MMC-HVDC
Integrated System Considering Delay Effects. IEEE J. Emerg. Sel. Top. Circuits Syst. 2022, 12, 59–72. [CrossRef]
14. Wang, D.; Zhang, K. Small-Signal Stability Analysis of MMC-HVDC System Based on Hybrid Passivity. IEEE Trans. Power Deliv.
2024, 39, 29–41. [CrossRef]
15. Wang, J.; Wang, P. Decoupled Power Control for Direct-Modulation-Based Modular Multilevel Converter with Improved Stability.
IEEE Trans. Ind. Electron. 2019, 66, 5264–5274. [CrossRef]
16. Xiong, X.; Yang, Y.; Wu, C.; Zhao, C.; Blaabjerg, F. Improving the Stability of Standalone MMCs by Shaping the AC Side Impedance
Using Insertion Index Compensation. IEEE J. Emerg. Sel. Top. Circuits Syst. 2022, 12, 81–89. [CrossRef]
17. Gao, L.; Zhang, H.; Xu, J. An Improved Virtual Impedance Control Based Oscillation Suppression Method for MMC DC System.
CPSS Trans. Power Electron. Appl. 2022, 7, 347–358. [CrossRef]
18. Huang, P.; Vanfretti, L. Multi-Tuned Narrowband Damping for Suppressing MMC High-Frequency Oscillations. IEEE Trans.
Power Deliv. 2023, 38, 3804–3819. [CrossRef]
19. Guo, C.; Xu, L.; Yang, S.; Jiang, W. A Supplementary Damping Control for MMC-HVDC System to Mitigate the Low-Frequency
Oscillation Under Low Inertia Condition. IEEE Trans. Power Deliv. 2023, 38, 287–298. [CrossRef]
20. Liu, W.; Wang, R.; Kerekes, T.; Dragičević, T.; Teodorescu, R. Modulation-Compensated Power Synchronization Control for
Modular Multilevel Converters with Enhanced Performance Under Varying SCR. IEEE J. Emerg. Sel. Top. Power Electron. 2023, 11,
4714–4725. [CrossRef]
21. Zhu, Y.; Jiang, S.; Pou, J.; Konstantinou, G. Virtual Arm Impedance Emulation and Stability Improvement in Modular Multilevel
Converters. IEEE Trans. Power Electron. 2024, 39, 3931–3942. [CrossRef]
22. Liu, Y.; Huang, M.; Tse, C.K.; Iu, H.H.C.; Yan, Z.; Zha, X. Stability and Multiconstraint Operating Region of Grid-Connected
Modular Multilevel Converter Under Grid Phase Disturbance. IEEE Trans. Power Electron. 2021, 36, 12551–12564. [CrossRef]
23. Liu, Y.; Huang, M.; Zha, X.; Iu, H.H.C. Transient current estimation of MMC-HVDC systems under a grid-converter phase
difference. CSEE J. Power Energy Syst. 2021, 7, 730–743.
24. Lyu, J.; Cai, X.; Molinas, M. Frequency Domain Stability Analysis of MMC-Based HVdc for Wind Farm Integration. IEEE J. Emerg.
Sel. Top. Power Electron. 2016, 4, 141–151. [CrossRef]
25. Lyu, J.; Cai, X.; Molinas, M. Optimal Design of Controller Parameters for Improving the Stability of MMC-HVDC for Wind Farm
Integration. IEEE J. Emerg. Sel. Top. Power Electron. 2018, 6, 40–53. [CrossRef]
26. Isik, S.; Alharbi, M.; Bhattacharya, S. An Optimized Circulating Current Control Method Based on PR and PI Controller for MMC
Applications. IEEE Trans. Ind. Appl. 2021, 57, 5074–5085. [CrossRef]
27. Kundur, P. Power System Stability and Control; McGraw-Hill: New York, NY, USA, 1994.
28. Meng, X.; Han, J.; Pfannschmidt, J.; Wang, L.; Li, W.; Zhang, F.; Belanger, J. Combining Detailed Equivalent Model with Switching-
Function-Based Average Value Model for Fast and Accurate Simulation of MMCs. IEEE Trans. Energy Convers. 2020, 35, 484–496.
[CrossRef]
29. Sun, Y.; Teixeira, C.A.; Holmes, D.G.; McGrath, B.P.; Zhao, J. Low-Order Circulating Current Suppression of PWM-Based Modular
Multilevel Converters Using DC-Link Voltage Compensation. IEEE Trans. Power Electron. 2018, 33, 210–225. [CrossRef]
30. Xu, Z.; Li, B.; Wang, S.; Zhang, S.; Xu, D. Generalized Single-Phase Harmonic State Space Modeling of the Modular Multilevel
Converter with Zero-Sequence Voltage Compensation. IEEE Trans. Ind. Electron. 2019, 66, 6416–6426. [CrossRef]
31. Ji, K.; Chen, W.; Wu, X.; Pang, H.; Hu, J.; Liu, S.; Cheng, F.; Tang, G. High Frequency Stability Constraints Based MMC Controller
Design Using NSGA-III Algorithm. CSEE J. Power Energy Syst. 2023, 9, 623–633.
32. Sun, J. Impedance-Based Stability Criterion for Grid-Connected Inverters. IEEE Trans. Power Electron. 2011, 26, 3075–3078. [CrossRef]
33. Middlebrook, R.D. Input filter considerations in design and application of switching regulators. In Proceedings of the PESC’76,
IEEE Power Electronics Specialists, Conference, Cleveland, OH, USA, 8–10 June 1976; pp. 366–382.
34. Hall, S.R.; Wereley, N.M. Generalized Nyquist Stability Criterion for Linear Time Periodic Systems. In Proceedings of the 1990
American Control Conference, San Diego, CA, USA, 23–25 May 1990; pp. 1518–1525.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

83
 energies
Article
Simulation and Characterization of Micro-Discharge
Phenomena Induced by Glitch Micro-Defects on an Insulated
Pull Rod Surface
Shu Niu 1 , Shuai Li 1 , Jizhong Liang 1 , Guodong Li 1 , Fan Hu 1 , Hai Zhang 1 , Yujie Zhu 2, *, Xianhao Fan 2, *
and Chuanyang Li 2

1 State Grid Shanxi Electric Power Company, Taiyuan 030001, China; nius4446sk@163.com (S.N.);
lsgfr4523@163.com (S.L.); chonggen@163.com (J.L.); lgd959278428@126.com (G.L.); hitszhufan@163.com (F.H.);
brighthai@163.com (H.Z.)
2 Department of Electrical Engineering, Tsinghua University, Beijing 100084, China;
chuanyang_li@mail.tsinghua.edu.cn
* Correspondence: zyujiemail@gmail.com (Y.Z.); xianhao_fan@163.com (X.F.)

Abstract: The reliability of GIS (gas-insulated switchgear) circuit breakers significantly depends on
the condition of the insulated pull rods, with micro-defects on their surface posing a potential risk for
micro-discharges and breakdown incidents. Experimentally investigating these micro-discharges
is challenging due to their minute nature. This study introduces a framework to examine the
linkage between micro-defects and micro-discharges, coupled with numerical simulations of the
micro-discharge process in insulated pull rods afflicted by surface infiltration flaws under opera-
tional conditions. Initially, samples containing micro-defects were sectioned via water jet cutting
for microstructural analysis through white light interferometry. Subsequently, a two-dimensional
axisymmetric model simulating positive corona discharge from a needle to a plate electrode was
employed to derive the relationship between charged particle density and the electric field in SF6 and
air. Building on these observations, a micro-discharge model specific to micro-defects was developed.
Comparative analysis of micro-discharge behaviors in SF6 and air for identical defect types was con-
ducted. This research framework elucidates the discharge dynamics of charged particles in SF6 and
air during micro-discharge events, shedding light on the mechanisms underpinning micro-discharges
triggered by insulation rod defects.

Keywords: gas-insulated switchgear; insulated pull rods; corona discharge; simulation analysis

1. Introduction
The insulated pull rod is a crucial insulating component in gas-insulated, metal-
eclosed switchgears (GISs) characterized by its slender, elongated structure and frequent
involvement in disconnection scenarios [1–5]. In operational conditions, it is subjected
to significant mechanical stress and impact vibrations, necessitating superior electrical
insulation and mechanical strength [2,6–9]. The resin-impregnated fiberglass (RIF) process
is instrumental in achieving these properties [1,10–12]. However, inadequacies in the
design of the process or its parameters can lead to micro-defects, such as white spots, burrs,
and poor surface infiltration, compromising the rod’s integrity [2,13,14]. The presence of
these micro-defects can distort the electric field, initiating micro-discharges and potentially
leading to breakdown incidents. While various methods, including X-ray, non-destructive
microwave, and ultrasonic tests, have been proposed for defect detection in insulation
rods, comprehensive studies on the micro-discharge phenomena induced by these defects
remain scarce [2,13,14].
To advance our understanding of micro-discharge phenomena triggered by micro-
defects in insulated pull rods, and given the experimental challenges in direct observation,

Energies 2024, 17, 2594. https://doi.org/10.3390/en17112594 84 https://www.mdpi.com/journal/energies

Energies 2024, 17, 2594

we propose a comprehensive framework. This framework encompasses three main phases:

initially, sections of the insulated pull rod harboring micro-defects are segmented via water
jet cutting, followed by the examination of their surface microstructures through white light
interferometry for physical characterization. Subsequently, we derive a correlation curve
for charged particle density concerning the electric field within SF6 and air environments
using a two-dimensional axisymmetric model of positive corona discharge from a needle
to a plate electrode. The final stage involves the formulation of a micro-discharge model
predicated on the observed microstructural characteristics and the derived correlation
curves. Investigating micro-discharge attributes resulting from insulation rod micro-defects
not only facilitates a theoretical comprehension of insulation failures but also lays the
groundwork for developing theoretical models for the online monitoring of discharge
signals associated with insulation pull rod defects.

2. Characterization of the Surface Defects on Insulated Pull Rods

The efficacy of the resin-impregnated fiberglass (RIF) technique is pivotal to ensuring
the integrity of insulated pull rods [10–12]. The comprehensive process entails several
critical steps: Initially, fiberglass fabric is rolled into a cylindrical form and placed within
a sealed mold. Subsequently, under a controlled vacuum applied at the top of the mold,
liquid epoxy resin is simultaneously injected at a predetermined rate from a feeder at the
bottom. The process concludes with the resin’s thorough impregnation and subsequent
curing phase. A misalignment in the synchronization between the vacuum and resin
injection rates can introduce air into the interface between the flowing resin and the mold’s
interior, leading to areas of poor surface infiltration. This defect may manifest on either the
inner or outer surfaces of the insulated pull rod, compromising its performance.
The predominant method for inspecting the quality of insulated pull rods is artificial
visual inspection. This method, however, faces significant limitations in accurately dis-
tinguishing between inner poor infiltration defects and white spot defects. As shown in
Figure 1a, both defect types manifest as irregular, gray shadows in optical transmission
imaging, differing markedly from the appearance of unaffected areas. Consequently, these
defects are collectively categorized as white spot defects due to their visual similarity.
Notably, a distinguishing factor is the generally higher grayscale intensity observed in the
light transmission images of inner poor infiltration defects compared to white spot defects.
Additionally, areas of poor outer surface infiltration are identifiable by a distinct rough
texture upon tactile examination.

Figure 1. (a) Defects observed in the insulating pull rod through photonic inspection. The dashed
scircles highlight areas with varying levels of light transmittance. They illustrate the location and size
distribution of defects within the rod. (b) Dimensions of the sample sectioned for analysis. (c) Surface
of the sample observed via white light interferometry.

With the cooperation of the factory, we identified that the minor defects in the rods
mainly include five categories: internal microcracks, internal bubbles, internal impurities,
surface scratches, and surface glitches. Surface scratches primarily result from damage
caused by sharp tools during transportation and assembly. Glitches mainly arise from
residues of the demolding material on the rod’s surface during the demolding process. The

85
Energies 2024, 17, 2594

main difference between the two is that scratches are more often inward depressions, while
burrs protrude outward. Our study is mainly focused on surface glitch defects.
To elucidate the nature of poor infiltration defects and facilitate accurate three-dimen
sional reconstruction, a 10 mm × 10 mm × 2 mm square sample from the defective area
of an insulated rod was extracted using water jet cutting. This preparation enabled the
differentiation of internal surface infiltration defects from white spot defects. Figure 1b
showcases a section of the sample with poor surface infiltration, characterized by a macro-
scopically rough surface texture. Notably, these internal defects, sealed within the rod
during assembly, are exposed to air under operational conditions, whereas external defects
are in contact with SF6 gas.
Microscopic examination through white light interferometry revealed that these de-
fects consist of densely packed protrusions (Figure 1c). In the image, the red areas represent
the burrs on the surface, while the surrounding green areas indicate the smooth parts of the
sample. Additionally, there are some interlaced blue sections aside from the green parts.
This is because the sample is a mixture of epoxy resin and glass fibers, which possess a
certain level of translucency. Therefore, during white light interferometry, some light passes
through the sample, causing interference. The blue areas in the image reflect the internal
glass fiber framework. However, since these areas are largely opaque, this does not affect
the results of the white light interferometry. The resulting surface roughness metrics are as
follows: Sa (arithmetic mean surface roughness) = 6.592 μm; Sq (root mean square surface
roughness) = 10.005 μm; and SZ (peak-to-valley height) = 159.921 μm. These measurements
provide a quantitative basis for the textural characterization of poor infiltration defects.

3. Analysis of Defects under Operational Conditions

Extensive literature highlights that a set of three continuity equations, representing
electrons, positive ions (SF6 + /O2 + ), and negative ions (SF6 − /O2 − ), integrated with Pois-
son’s equation, can be used to effectively model and elucidate the dynamics of positive
corona discharge [15–17]. This study considers four charged species: electrons, positive
ions, negative ions, and photoelectrons. Equations (1)–(5) delineate our mathematical
framework, incorporating the drift–diffusion of these species while accounting for their
generation, drift, and dissipation. The source terms within these equations include cru-
cial processes such as ionization, electron attachment, recombination, and photoelectron
absorption. Poisson’s equation facilitates the computation of electric field distribution.
Additionally, the Helmholtz equation is employed to depict the light absorption curve,
approximated as a sum of series, providing a fit to the empirical data.
     
∂ne →  →  →

+ ∇ · −μe E ne − De ∇ne = S ph + αne μe E  − ηne μe E  − k ep ne Np
  (1)
∂t
   
∂np →  →
+ ∇ · μ p E n p − D p ∇n p = S ph + αne μe E  − k ep ne n p − k np nn n p (2)
∂t
   
∂nn →  →
+ ∇ · −μn E nn − Dn ∇nn = ηne μe E  − k np nn n p
 (3)
∂t

e n p − ne − nn
∇ V=−
2
(4)
ε0
⎧
⎨ ∇2 S j − λ P 2 S j = − A P2 I
ph j SF6 ph j SF6 ph
, ( j = 1, 2, 3) (5)
⎩ S ph = ∑ S j
ph

where t (s) is time; ne , nn , np , μe , μn , μp , De , Dn , Dp are the different densities (m−3 ), mobili-

ties (m2 V−1 s−1 ), and diffusion coefficients (m2 s−1 ) of electrons, positive ions, and negative
ions; and α (m−1 ) and η (m−1 ) are the ionization and attachment coefficients, respectively.
kep (m3 s−1 ) and knp (m3 s−1 ) represent the recombination coefficient of positive ions with

86
Energies 2024, 17, 2594

electrons and the recombination of positive ions with negative ions. V, e, and ε0 denote
the applied voltage (V), the electron charge (C), and the vacuum permittivity (F·m−1 ),
→
respectively. E (V·m−1 ) represents the electric field intensity vector. Aj (cm−2 Torr−2 ) and
λj (cm−1 Torr−1 ) represent the fitting parameter, and the photoionization term in equa-
tion j is represented as Sph (mol·m−3 s−1 ). PSF6 (Torr) represents the pressure of SF6 , Iph
(mol·m−3 s−1 ) represents the photoionization radiation coefficient.
Morrow has presented a detailed collection of reliable data on the basic transport
properties of SF6 and the approximate analytical interpolation of the coefficients [18,19].
Tables 1 and 2 show the parameters used in the simulation under air and SF6 .

Table 1. Parameters used in the air condition simulation.

Parameter Air
  
→
Ionization coefficient α (m−1 ) 3.5 × 105 exp −1.65 × 107 / E 
 
  
 →
Attachment coefficient η (m−1 ) 1.5 × 103 exp −2.5 × 106 / E 
 −0.25
Mobility of electrons μe (m2 V−1 s−1 ) →
0.606 ×  E 
Mobility of positive ions μp (m2 V−1 s−1 ) 2.43 × 10−4
Mobility of negative ions μn (m2 V−1 s−1 ) 2.7 × 10−4
Diffusion coefficient of De (m2 s−1 ) 0.18
Diffusion coefficient of positive ions Dp (m2 s−1 ) 2.8 × 10−6
Diffusion coefficient of negative ions Dn (m2 s−1 ) 4.3 × 10−6
Coefficient of recombination of positive ions and
2 × 10−13
electrons kep (m3 s−1 )
Coefficient of recombination of positive ions and
2 × 10−13
negative ions knp (m3 s−1 )

Table 2. Parameters used in the SF6 condition simulation.

Parameter SF6 Value

⎧ −5 E
⎨  6 × 10 , N < 1.2 × 10−19 V·m2
Ionization coefficient α (m−1 ) 1.216 × 10−5 ln NE
+ 5.89× 10−4 , 1.2 × 10−19 ≤ NE < 3.5 × 10−19 V·m2
⎩
−1.897 × 10−5 ln NE − 7.346 × 10−4 , NE ≥ 3.5 × 10−19 V·m2
⎧  2.985 E
⎪
⎨2.0463 × 10−20 − 0.25379 N E
+ 1.4705 × 1018 NE , N < 5 × 10−20 V·m2
E 3
 −
Attachment coefficient η (m−1 ) ⎪ − 3.0078 × 10 36 , 5 × 10 20 ≤ E
 EN < 2 × 10−19 V·m2
⎩ −
N
7 × 10 21 exp −2.25 × 10 N , N ≥ 2 × 10−19 V·m2
18 E


E −0.2576
Mobility of electrons μe (m2 V−1 s−1 ) 1.027 × 1019 N m−1 V−1 s−1
⎧ −5 E
⎨  6 × 10 , N −< 1.2 × 10−19 V·m2
Mobility of positive ions μp (m2 V−1 s−1 ) 1.216 × 10−5 ln NE
+ 5.89× 10 4 , 1.2 × 10−19 ≤ NE < 3.5 × 10−19 V·m2
⎩
−1.897 × 10−5 ln NE − 7.346 × 10−4 , NE ≥ 3.5 × 10−19 V·m2
E 2

Mobility of negative ions μn (m2 V−1 s−1 ) 1.69 × 1032 N + 5.3 × 10−5 m−1 V−1 s−1

E 0.2424
Diffusion coefficient of De (m2 s−1 ) 8.8823 × 1028 N m−1 s−1
⎧
⎨ 2 × 10−13 P0.6336 , 1 < P ≤ 39 kPa
Diffusion coefficient of positive/negative ions Dp,n
2.28 × 10−11 P−0.6559 , 39 < P ≤ 270 kPa
(m2 s−1 ) ⎩
6.87 × 10−10 P−1.279 , 270 < P ≤ 2000 kPa
Coefficient of recombination of positive ions and kT
electrons/negative ions kep,np (m3 s−1 )
D
μ = 39.6 V

This study employs a 2D axisymmetric needle-to-plate electrode model to investigate

positive corona discharge phenomena in SF6 and air environments. Both gases are simu-
lated under specific conditions: SF6 at a pressure of 0.4 MPa and air at 0.1 MPa, filling the
gap between the needle and plate electrodes, with the corresponding temperature being

87
Energies 2024, 17, 2594

room temperature (300 K). The geometric configuration for both gas settings is identical,
featuring a needle tip curvature of 0.1 mm and a 100 mm gap between the needle tip and
the plate electrode. A positive voltage ranging from 1 to 10 kV is applied to catalyze the
positive corona discharge in both gases.
The study quantitatively captures the dynamics of positive corona discharge by plot-
ting the variation in the total numbers of SF6 electrons, negative ions, and positive ions
across a discharge cycle spanning 100 ns to 1000 ns, illustrated in Figure 2. Similarly, for air,
the total counts of electrons, negative ions, and positive ions over a discharge cycle ranging
from 500 ns to 5000 ns are documented, as shown in Figure 3. The data points reflect the
results of the simulation. The lines connecting these data points are fitting curves based on
quadratic interpolation of the data points. These are used to predict missing data values.
It can be observed that the positive corona discharge intensity of SF6 and air gradually
increases with the voltage applied to the needle electrode.
As the discharge progresses to 100 ns for SF6 and 500 ns for air, the rate of change in
the total particle count of both gases stabilizes, eventually reaching a steady concentration.
This stability allows for the calculation of particle density by dividing the steady-state total
particle count by the volume of the integral domain of each gas environment. The flux
of particles generated near the glitch defect is then calculated for the next step analysis.
Here, our statistical method involves using the needle tip as the center to define a circle
with a radius of 2 cm as the statistical area. The electric field strength within this area is
considered the true electric field strength. We measure the electron density, positive ion
density, and negative ion density within this area over a time interval of 0–5000 ns as the
particle yield.

Figure 2. Through high-throughput simulation, the average densities of (a) electrons, (b) positrons,
and (c) negative ions within different electric fields in an SF6 environment were obtained, and (d) the
flux of particles generated in the defect area were quantified.

88
Energies 2024, 17, 2594

Figure 3. High-throughput simulation yielded the average densities of (a) electrons, (b) positrons, and
(c) negative ions within different electric fields in a standard atmospheric pressure air environment,
along with (d) the quantified flux of particles generated in the defect area.

4. Micro-Discharge Analysis in Air and SF6 Environments under

Operational Conditions
We employed the charge generation rates near the detected defects to simulate the
pulling rod under operational conditions. The simulation structure is depicted in Figure 4.
Initially, the discharge process within SF6 was simulated. To further investigate the dynam-
ics and distribution of charged particles during the micro-discharge development process,
an analysis was conducted at four critical time points: t = 25 ns, t = 100 ns, t = 200 ns, and
t = 500 ns.

Figure 4. Schematic diagram of the simulation area.

In the initial stage (up to 100 ns), the charge distribution concentrated near the glitch
defect, allowing only a few electrons to gain enough kinetic energy to initiate ionization
and attachment reactions with neutral molecules. Electrons, positive ions, and negative
ions were distributed evenly around the defect, as depicted in Figure 5a,e,i.

89
Energies 2024, 17, 2594

Figure 5. In an SF6 environment, the process of discharge induced by a glitch defect on the rod
surface. From 25 ns to 200 ns, the evolution of (a–d) electrons, (e–h) negative ions, and (i–l) positive
ions near the glitch defect.

As the micro-discharge developed, electrons predominantly migrated to the left side

of the defect, counter to the electric field’s direction, indicating the progression into later
stages. SF6 ’s strong electronegativity plays a crucial role here, significantly enhancing
electron adsorption and influencing electron density distribution throughout the micro-
discharge process. Electron density peaks were observed at the defect’s left corner and
right wall, creating a distinct electron cloud layer. Despite moving against the initial electric
field, electrons accumulate at these points. Some electrons lose their kinetic energy upon
colliding with the defect’s right wall, adhering to the surface and increasing electron density
in this area.
Conversely, the left wall of the defect, where electron movement originates, witnesses
interactions between electrons and neutral particles, producing additional positive ions and
electrons. This leads to a high concentration of electron density at the left corner. However,
the tendency of electrons to combine with neutral particles to form negative ions results in
a lower electron density on this side.
Here, we did not consider the impact of moisture in the gas environment on the
discharge at defect locations. However, in actual operation, the moisture in the gas envi-
ronment can affect the defects in two main ways: the presence of moisture can influence
discharge parameters such as the discharge ionization coefficient, affecting the actual dis-
charge process; furthermore, the rod’s surface can absorb moisture, increasing its surface
conductivity, thereby facilitating easier migration of the charges produced by discharges
across the surface.
During the SF6 micro-discharge process, the density distribution of negative ions
illustrates that SF- is predominantly formed by electron adsorption in SF6 . Initially, electrons
near the defect are captured, generating negative ions that subsequently move opposite to
the electric field’s direction, diffusing away from the left wall of the defect. This pattern
suggests that negative ions are consistently produced and dispersed outward from the
defect throughout the micro-discharge.
In the case of positive ions in SF6 , they are primarily created when SF6 molecules
collide with electrons, resulting in their detachment. Throughout the micro-discharge, there
is a noticeable concentration of positive ions along both the left and right walls of the defect.

90
Energies 2024, 17, 2594

These ions show a tendency to diffuse slightly outward, aligning with the electric field’s
direction. Unlike electrons and negative ions, however, their diffusion is less extensive.
In an air environment, the intensity of micro-discharge around poorly insulated defects
is higher than in SF6 , due to air’s lower electronegativity. The evolution of micro-discharge
is depicted at six critical times: t = 10 ns, t = 25 ns, t = 50 ns, t = 75 ns, t = 100 ns, and
t = 500 ns, with simulation results illustrated in Figure 6.

Figure 6. In an air environment, the process of discharge induced by a glitch defect on the rod surface.
From 10 ns to 200 ns, the evolution of (a–f) electrons, (g–l) negative ions, and (m–r) positive ions near
the glitch defect.

Initially, electron accumulation at the corner of the defect’s left wall forms a thin cloud
layer. Within 25 ns, there is significant electron buildup on the surface of the epoxy resin.
Driven by the electric field, these electrons ionize neutral particles, leading to the generation
of additional electrons and positive ions. By 50 ns, a charge cluster pattern emerges, which
maintains a consistent electron cloud shape up to 500 ns. By the end of the period, there is
a slight diffusion of low-density electrons observed at 500 ns.
In an air environment, negative ions are primarily produced through the adsorption of
electrons onto N2 molecules. Initially, these negative ions accumulate around the defect and
migrate against the electric field, showing slower movement compared to electrons, which
leads to their initial retention. By 50 ns, there is a significant accumulation of negative ions,
which begins to dissipate outward by 500 ns.
The evolution of positive ions during the micro-discharge process is influenced by
two main factors. Firstly, there is an accumulation of charge-like ions generated by minor
ionization near the defect, distributed to the right of the defect, opposite the electric field.
Secondly, new positive charges are produced by the collision ionization of surface electrons
as they propagate on the left side of the defect, consistently accumulating on the surface of
the insulating rod and forming a distinct charge pattern.
This study on micro-discharge phenomena caused by poor insulation defects in dif-
ferent environments of SF6 and air has enhanced the understanding of the insulating
performance of gas-insulated switchgear equipment. This highlights how the insulating
characteristics of SF6 and air directly influence the intensity and properties of micro-
discharges triggered by such defects. Specifically, SF6 exhibits superior insulating capabili-
ties compared to air, effectively reducing the extent of micro-discharges under identical
operational conditions. The contrasting operational environments of the inner and outer
walls of the insulating rod further illustrate this point: the outer wall is exposed to SF6 gas at
a higher pressure of 4 atm, while the inner wall is surrounded by air at a lower pressure of
1 atm. This difference in gas type and pressure not only underscores the superior insulating
properties of SF6 over air but also amplifies the variation in discharge intensity between
defects located on the inner and outer surfaces.

91
Energies 2024, 17, 2594

Additionally, the study of micro-discharge phenomena also reveals how insulation

defects behave under different pressure and electric field conditions. This is particularly
important in transformers and other high-voltage equipment, where environmental condi-
tions significantly affect micro-discharges. As pressure increases, the insulating properties
of SF6 improve, reducing both the frequency and intensity of micro-discharges. This effect
has been confirmed experimentally; when the pressure of SF6 gas is increased from 1 to
4 atmospheres, micro-discharges from the same defect are noticeably weaker.
In researching micro-discharges, technologies like high-speed cameras and electron
microscopes allow for a detailed observation and analysis of the dynamic processes and
factors influencing micro-discharges. These precise observation methods enable researchers
to accurately track the movements of electrons and ions, leading to a better understanding
of how micro-discharges develop and their specific impacts on insulation performance.
This deep understanding of micro-discharge behavior provides crucial information
for improving and optimizing the design of gas-insulated switchgears. By comparing
how micro-discharges occur in different mediums like SF6 and air, engineers can design
safer and more efficient systems to meet the complex demands of electrical power systems.
Such research not only advances high-voltage insulation technology but also supports the
stability and safety of electrical power systems.

5. Conclusions
This paper introduces a simulation framework for studying micro-discharge phenom-
ena caused by defects in insulation materials. It explores defect characterization and the
processes of micro-discharges through simulations that capture temporal density fluctua-
tions of electrons, negative ions, and positive ions, enhancing our understanding of their
dynamic behavior. The study details how electric field distortion near defects leads to
charged particle generation, influencing micro-discharge activity. It also examines the
impact of environmental conditions, such as gas composition and pressure, on micro-
discharge intensity and highlights the importance of quality control in manufacturing to
improve the reliability and safety of gas-insulated switchgear systems.

Author Contributions: Methodology, S.L.; Software, G.L. and H.Z.; Validation, G.L. and H.Z.; Formal
analysis, S.L. and F.H.; Resources, S.N., J.L. and F.H.; Writing—original draft, S.N., Y.Z. and X.F.;
Writing—review and editing, C.L.; Project administration, J.L. All authors have read and agreed to
the published version of the manuscript.
Funding: This work was supported by the State Grid Corporation Headquarters Technology Project
(under Grant No. 520530230017).
Data Availability Statement: The data supporting the findings of this study are available from the
corresponding author upon reasonable request.
Conflicts of Interest: Authors Shu Niu, Shuai Li, Jizhong Liang, Guodong Li, Fan Hu and Hai Zhang
were employed by the State Grid Shanxi Electric Power Company. The remaining authors declare
that the research was conducted in the absence of any commercial or financial relationships that
could be construed as a potential conflict of interest.

References
1. Jiang, Z.J.; Li, X.; Zhang, H.; Zhang, E.Z. Research Progress and Prospect of Condition Assessment Techniques for Oil–Paper
Insulation Used in Power Systems: A Review. Energies 2024, 17, 2089. [CrossRef]
2. Fan, X.; Niu, S.; Luo, H.; Liang, J.; Liu, F.; Li, W.; Liu, W.; Gao, W.; Huang, Y.; Li, C.; et al. Photon counting technique as a potential
tool in micro-defect detection of epoxy insulation pull rod in gas-insulated switchgears. High Voltage 2024, 9, 267–274. [CrossRef]
3. Fan, X.; Luo, H.H.; Liang, F.W.; Hu, J.; Liu, W.D.; Li, C.Y.; He, J.L. Photon Count Technique as a Potential Tool for Insulation
Micro-Defect Detection: Principles and Primary Results. iEnergy 2024, 2, 258–263. [CrossRef]
4. Deng, Y.; Fan, X.H.; Luo, H.H.; Wang, Y.; Wu, K.Y.; Liang, F.W.; Li, C.Y. Impact of Air Gap Defects on the Electrical and Mechanical
Properties of a 320 kV Direct Current Gas Insulated Transmission Line Spacer. Energies 2023, 16, 4006. [CrossRef]
5. Liang, F.W.; Luo, H.H.; Fan, X.H.; Li, X.T.; Wang, X. Review of Surface Charge Accumulation on Insulators in DC Gas-Insulated
Power Transmission Lines: Measurement and Suppression Measures. Energies 2023, 16, 6027. [CrossRef]

92
Energies 2024, 17, 2594

6. Riechert, U.; Holaus, W. Ultra high-voltage gas-insulated switchgear-a technology milestone. Eur. Trans. Electr. Power 2012, 22,
60–82. [CrossRef]
7. Li, X.; Zhao, H.; Murphy, A.B. SF6-alternative gases for application in gas-insulated switchgear. J. Phys. D Appl. Phys. 2018, 51,
153001. [CrossRef]
8. Li, C.; Lin, C.; Chen, G.; Tu, Y.; Zhou, Y.; Li, Q.; Zhang, B.; He, J. Field-dependent charging phenomenon of HVDC spacers based
on dominant charge behaviors. Appl. Phys. Lett. 2019, 114, 202904. [CrossRef]
9. Li, C.; Hu, J.; Lin, C.; Zhang, B.; Zhang, G.; He, J. Surface charge migration and dc surface flashover of surface-modified
epoxy-based insulators. J. Phys. D Appl. Phys. 2017, 50, 065301. [CrossRef]
10. Liu, Z.; Tu, T.; Euvrard, E. Combining dry type resin impregnated fiberglass paperless transformer bushings with built-in novel
insulation monitoring function. In Proceedings of the 2014 IEEE PES T&D Conference and Exposition, Chicago, IL, USA, 14–17
April 2014.
11. Borodulin, A.S.; Marycheva, A.N.; Malysheva, G.V. Simulation of impregnation kinetics of fabric fillers in the production of
fiberglass articles. Glass Phys. Chem. 2015, 41, 660–664. [CrossRef]
12. Rohatgi, V.; Patel, N.; James Lee, L. Experimental investigation of flow-induced microvoids during impregnation of unidirectional
stitched fiberglass mat. Polym. Compos. 1996, 17, 161–170. [CrossRef]
13. Yan, B.; Yang, W.; Wang, H.; Hou, C.; Kong, X.; Du, B. Identification and Analysis of Micro Defects in Insulation Pull Rod Based on
the Ultrasonic Method. In Proceedings of the 2023 International Symposium on Electrical Insulating Materials (ISEIM), Shimane,
Japan, 24–28 September 2023; pp. 336–339.
14. Luo, H.; Fan, X.; Liang, F.; Hu, J.; Yan, B.; Li, T.; He, J.; Li, C. Surface Defect Detection of GIS Aramid Insulation Pull Rod via
Photon Counting. In IEEE Transactions on Dielectrics and Electrical Insulation; early access; IEEE: Piscataway, NJ, USA, 2024; pp. 1–7.
15. Adamiak, K.; Atten, P. Simulation of corona discharge in point–plane configuration. J. Electrostat. 2004, 61, 85–98. [CrossRef]
16. Weingärtner, E.; Kuster, F.; Wegener, K. Modeling and simulation of electrical discharge machining. Procedia CIRP 2012, 2, 74–78.
[CrossRef]
17. Ming, W.; Zhang, S.; Zhang, G.; Du, J.; Ma, J.; He, W.; Cao, C.; Liu, K. Progress in modeling of electrical discharge machining
process. Int. J. Heat. Mass. Transf. 2022, 187, 122563. [CrossRef]
18. Morrow, R. The theory of positive glow corona. J. Phys. D Appl. Phys. 1997, 30.22, 3099. [CrossRef]
19. Morrow, R. Theory of stepped pulses in negative corona discharges. Phys. Rev. A 1985, 32, 3821. [CrossRef] [PubMed]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

93
 energies
Article
A Spread-Spectrum Modulation Scheme for a 3 × 6 Indirect
Matrix Converter Based on a Current Ripple Model
Zhanqing Zhou 1,2, *, Lingyue Xue 1 , Chen Li 2 and Qiang Geng 1

1 Tianjin Key Laboratory of Intelligent Control of Electrical Equipment, Tiangong University, Tianjin 300387,
China; xuelingyue@tiangong.edu.cn (L.X.); gengqiang@tju.edu.cn (Q.G.)
2 Zhejiang University Advanced Electrical Equipment Innovation Center, Hangzhou 311107, China;
lichen_hz@zju.edu.cn
* Correspondence: zhouzhanqing@tiangong.edu.cn

Abstract: Focused on addressing harmonic suppression in multiphase indirect matrix converters

(IMCs), this study explores spread-spectrum modulation technology through ripple analysis calcu-
lations. We introduce a current ripple spread-spectrum modulation (CR-SSM) method tailored for
multiphase IMC systems. In this approach, a 3 × 6-phase IMC is modeled as a two-port network,
and a small-signal model of the output side is established. The transfer function is utilized to analyze
the two-port network in the complex frequency domain (s-plane). The time-domain expression of
the output current ripple is derived in vector form. Subsequently, the distribution of the output
ripple and locus are determined based on specified constraints. The carrier frequency is dynami-
cally adjusted online according to the specified ripple locus. Compared to classical periodic PWM
methods, this approach offers a broader range of frequency variations and achieves a more uniform
output spectrum. Furthermore, CR-SSM optimizes system efficiency and enhances spread-spectrum
modulation. Experimental results demonstrate that this method effectively enhances the quality of
input and output waveforms in multiphase IMC systems.

Keywords: indirect matrix converter (IMC); current ripple analysis; spread-spectrum modulation;
periodic PWM; electromagnetic interference (EMI)

1. Introduction
The indirect matrix converter (IMC) represents a form of “all-silicon” AC/AC direct
power converter [1]. This converter achieves efficient power conversion without requiring
an energy storage link, owing to its compact structure, ability to deliver sinusoidal input–
output waveforms, adjustable input power factor, low harmonic distortion, flexible phase
adjustment of input current, and other noteworthy features. The IMC has proven highly
effective in applications within the green energy sector [2,3].
In a three-phase system, the conventional configuration of a matrix converter (MC)
is referred to as a direct matrix converter (DMC), featuring a 3 × 3 bidirectional switch
array [4]. A multiphase indirect matrix converter (IMC) comprises a three-phase rectifier
stage and a six-phase inverter stage, each divided into distinct structural and functional
segments. As illustrated in Figure 1, the rectifier stage of the IMC resembles a current
source rectifier (CSR). Conversely, the inverter stage of the IMC is designed to function
as two two-level voltage source inverters (VSIs) [5]. The clamper circuit is specifically
implemented for protective purposes and does not actively facilitate energy transfer. The
inherent design of the IMC allows for direct energy transfer between input and output, but
it also enables the transmission of harmonic distortion from one side of the converter to
the other. The study by [6] enhances the understanding of IMC topology, indicating that
the input filter plays a critical role in the converter system by attenuating high-frequency
harmonics to prevent interference with the normal operation of other equipment at their
common connection point. Additionally, it assists in smoothing out discrete input current

Energies 2024, 17, 2546. https://doi.org/10.3390/en17112546 94 https://www.mdpi.com/journal/energies

Energies 2024, 17, 2546

to generate a continuous sinusoidal current waveform on the source side. One fundamental
and widely used method for improving waveform quality in frequency converters is by
increasing the carrier frequency of the controller. Recent advancements in wide bandgap
semiconductors, such as silicon carbide (SiC), have significantly expanded the feasible
carrier frequency of power converters [7–9]. Various approaches have been proposed
to enhance the effectiveness of power converters using pulse-width modulation (PWM),
including optimization techniques for vector selection among available vectors [10], re-
duction in common-mode components in modulation waves [11,12], refinement of switch
sequences [13–15], and enhancement of carrier properties [16–18].

Figure 1. A 3 × 6 IMC system.

Spread-spectrum modulation (SSM) is an ideal strategy for reducing EMI in multi-

phase IMC systems. Common SSM strategies currently include random PWM, programmed
PWM, and periodic PWM. In comparison to random PWM, periodic PWM uses a predeter-
mined frequency density function to dynamically change the carrier frequency, achieving
a fixed EMI spectrum [19]. By varying the carrier frequency, periodic PWM introduces
additional modulation flexibility. Typical carrier waveforms encompass sinusoidal, expo-
nential, and triangular shapes. Dual Fourier analysis of inverter output voltages reveals
that triangular wave frequency density functions offer smoother spectral performance
compared to sinusoidal and exponential carriers, which exhibit distinct spectral peaks at
specific frequencies [20]. Research on the variation pattern of periodic signal frequency and
waveform under equal maximum frequency deviation is explored in [21], which derives an
optimal segmented modulation signal for superior EMI suppression by optimizing a combi-
nation of three frequency density functions. Furthermore, reference [22] employs a variable
density sawtooth carrier to implement periodic PWM, thus achieving a spread-spectrum ef-
fect while reducing inverter switch frequencies. Reference [23] explores a spread-spectrum
modulation scheme by varying switch frequencies according to uniform distribution,
exponential distribution, and normal distribution and derives corresponding equations.
Experimental results demonstrate that the proposed spread-spectrum modulation scheme
can reduce EMI by 20 dB. Reference [24] proposes an algorithm using sinusoidal-wave
pulse-width modulation (PWM) to suppress conducted-mode (CM) EMI in large-scale
networks. This method minimizes switch frequency overlap between converters, thereby
reducing total EMI levels. The authors of [25,26] focus on using Σ-Δ modulation for voltage
source converters. The proposed modulation techniques significantly reduce switching
losses and reduce lower-order harmonics by varying switch frequencies. Additionally, Y.
Huang utilizes a variable frequency density sawtooth carrier to achieve periodic PWM,
thereby achieving spread-spectrum effects while reducing inverter switch frequencies [27].
Reference [28] presents an optimized periodic PWM scheme that minimizes motor vibra-
tion frequency response peaks by considering the motor’s vibration frequency response
function. Reference [29] details the impact of periodic PWM on the current harmonics of
permanent magnet synchronous motors (PMSMs). Experimental evidence demonstrates
that periodic PWM disperses high-frequency harmonics across wider frequency bands,
thereby suppressing peak harmonics. Reference [30] proposes a spectral analysis method
based on periodic PWM and dual Fourier series to reduce current harmonics in active
power filters. Reference [31] introduces a novel PWM strategy based on uniform distribu-

95
Energies 2024, 17, 2546

tion PWM and experimentally verifies its effectiveness in eliminating EMI peaks, reducing
conducted EMI, and minimizing switch losses. Reference [32] analyzes the suppression
of harmonics and common-mode EMI using periodic PWM, validating the accuracy of
theoretical analysis and the effectiveness of algorithms through experiments.
However, the widespread implementation of current-ripple-based spread-spectrum
modulation (CR-SSM) in multiphase inverter motor control (IMC) systems still encounters
several challenges. Firstly, variations in vector amplitudes of the rectifier and inverter
stages within the IMC, influenced by the DC-link voltage, pose difficulties in establishing a
consistent ripple model. Secondly, the presence of input filters can affect waveform quality
at the grid side and complicate the analysis of the impact of input switching sequences and
transmission ratios on the output current ripple locus. Additionally, the resonant frequency
and bandwidth of filters impose limitations on the range of carrier frequency variations.
To address these challenges, this study investigates the combined impact of the rectifier
and inverter stages in a 3 × 6-phase inverter motor control (IMC) system [33,34]. It involves
an analysis of a small-signal model at the output stage to derive a time-domain expression
for the output current ripple. Finally, based on the constraints related to ripple distribution
and ripple index, the CR-SSM (current ripple spread-spectrum modulation) method is
proposed. This method allows for a broader range of frequency fluctuations while ensuring
safe commutation and achieving higher-quality spread-spectrum modulation. The main
contributions of this paper are outlined as follows:
1. Through the establishment of a small-signal model on the output side, the locus of
the output current ripple can be determined. Furthermore, dynamically adjusting the
carrier frequency based on the ripple locus can effectively broaden the harmonic spec-
trum, improve total harmonic distortion, and enhance output modulation efficiency.
2. The CR-SSM method dynamically adjusts the carrier frequency based on the rip-
ple locus. The CR-SSM method is based on the output current ripple, and thus the
debugging process of traditional periodic PWM is eliminated. It standardizes the con-
struction of carrier frequencies, providing clear numerical indicators and improving
frequency fluctuation efficiency. Furthermore, it reduces harmonic spikes at specific
frequency bands and optimizes system efficiency.
The remainder of this paper is structured as follows: In Section 2, we provide a brief
overview of multiphase IMC basic modulation strategies. Section 3 analyzes the voltage
error locus and establishes small-signal models for multiphase IMC outputs to derive time-
domain expressions for the current ripple. Subsequently, we explore achieving improved
total harmonic distortion (THD) and efficiency by implementing online adjustment of
carrier frequencies along the ripple locus. Section 4 presents experimental results comparing
CR-SSM and classical period PWM. Finally, Section 5 summarizes the key conclusions
drawn from this study.

2. Modulation Method of Multiphase IMC

2.1. Space Vector Modulation
To enhance the utilization of input voltage and facilitate the commutation action
of the switching device, we employed the indirect space vector modulation method for
the multiphase IMC. Specifically, the rectifier stage adopts zero-free vector modulation
based on the expected input current vector Ii * , while the inverter stage utilizes a maximum
four-vector SVPWM modulation according to the anticipated output voltage vector V o * ,
as illustrated in Figure 2. In this paper, the moment represented by the vector position
in Figure 2 is presented as an example. The red color arrows represent the vector of the
rectifier stage in the I sector and the inverter stage in the 1 sector.

96
Energies 2024, 17, 2546

Figure 2. The vectors diagram of 3 × 6 IMC drives: (a) rectifier stage; (b) α-β.

As shown in Figure 2a, the space vector plane of the rectifier is divided into six sectors,
labeled as I, II, . . ., VI. The symbols ix (x = 1, 2, . . ., 6) represent the rectifier active vectors,
and θ#i represents the angle included between Ii * and the starting vector position of the
sector.
2
ix = √ ipn e j(2x−3)π/6 (1)
3
2
Ii ∗ = (ia + ib ej2π/3 + ic ej4π/3 ) (2)
3
where ipn is the DC-link current, which is related to the load current and the inverter
switching state. The rectifier stage is usually modulated by two effective vectors adjacent
to Ii * . The rectifier stage is responsible for synthesizing vpn and regulating the Ii * phase,
and the Ii * amplitude is determined by the output current. The rectifier stage duty cycle is
⎧
⎨ dμ = sin( π − θ#i )/ cos( π − θ#i )
3 6 (3)
⎩ dυ = sin(θ#i )/ cos( π − θ#i )
6
where dμ is the duty cycle of the vector iμ (vector i1 in Figure 2); dν is the duty cycle of
the vector iν (vector i2 in Figure 2). When the effective current vector iμ acts, the DC-link
voltage is vab (corresponding to the input Sab ), and when the effective current vector iν is
acted upon, the DC-link voltage is vac (corresponding to the input Sac ). V im is the amplitude
of the three-phase sinusoidal input voltage so that the average DC-link voltage can be
expressed as follows:

3Vim
vpn = vμ dμ + vν dν = (4)
2 cos(π/6 − θ#i )

In Figure 2a, the switching state Sab corresponds to the DC-link voltage vab . Figure 2b
is a schematic diagram of the inverter stage output voltage space vector. The space vector
plane of the inverter is divided into 12 sectors, labeled as 1 , 2 , . . .., 12 . The symbols vy
(y = 1, 2, . . ., 12) represent the inverter active vectors; v0 and v7 are the zero voltage vectors;
V o * is the desired output voltage vector; θ#o represents the angle included between V o * and
the starting vector position of the sector; and vpn is the DC-link voltage.

2
vy = vpn e j(2y−1)π/12 (5)
3

2
Vo∗ = (va + vb ej2π/3 + vc ej4π/3 + vu ejπ/6 + vv ej5π/6 + vw ej3π/2 ) (6)
3

97
Energies 2024, 17, 2546

The IMC voltage transmission ratio q is defined as follows:

q = Vom /Vim (7)

where Vom and Vim are the desired output and input voltage vector amplitudes.
According to the vector space decoupling (VSD) transformation, each voltage vector of
the inverter stage can be projected to the α-β plane and the harmonic plane x-y. Each plane
contains 60 valid vectors and 4 zero vectors. While synthesizing the α-β plane fundamental
voltage, it is necessary for the x-y plane output voltage to be zero [34]. Thus, a minimum of
four basic vectors are required to fulfill the modulation requirements, as expressed below.
⎡ ⎤ ⎡ ⎤ −1 ⎡ ∗ ⎤
d1 v1α v2α v3α v4α vα
⎢d2 ⎥ ⎢v1β v2β v3β v4β ⎥ ⎢v∗β ⎥
⎢ ⎥=⎢ ⎥ ⎢ ⎥ (8)
⎣d3 ⎦ ⎣ v1x v2x v3x v4x ⎦ ⎣ 0 ⎦
d4 v1y v2y v3y v4y 0

where d1 ~d4 donate the duty cycles of the four basic vectors, namely V 1 (55), V 2 (45), V 3 (44),
and V 4 (64), which are required to synthesize the output reference vector V o * . To ensure
output voltage utilization, V 1 , V 2 , V 3 , and V 4 can be selected as four vectors that are
adjacent to V o * , d0 = 1 − d1 − d2 − d3 − d4 donates the duty cycle of zero vectors. Table 1
provides the applied times of voltage vectors and the switch selection of the inverter sectors
1 , 2 , . . .. 12 within Tc /2. V 0 and V 7 are zero vectors.

Table 1. Voltage vectors and switch selection of the inverter sectors.

Sec 1 Sec 2 Sec 5 Sec 6 Sec 9 Sec 10

V0 00 00 00 00 00 00
V2 45 44 26 22 13 11
V1 55 45 66 26 33 13
V7 77 77 77 77 77 77
V4 64 66 32 33 51 55
V3 44 64 22 32 11 51
V7 77 77 77 77 77 77
Sec 3 Sec 4 Sec 7 Sec 8 Sec 11 Sec 12
V0 00 00 00 00 00 00
V1 44 64 22 32 11 51
V2 64 66 32 33 51 55
V7 77 77 77 77 77 77
V3 66 26 33 13 55 45
V4 26 22 13 11 45 44
V7 77 77 77 77 77 77

Given that the rectifier and inverter stages are directly cascaded, it is necessary to
ensure coordination between the switching sequences of the two stages. In order to reduce
switching losses, each power device is switched only once in each PWM cycle. In this
paper, symmetrical triangular carrier PWM is adopted. The rectifier stage vector sequence
is divided into 3 segments, and the inverter stage sequence is divided into 13 segments.
The vector action sequence is shown in Figure 3. To minimize the switching action, the
rectifier stage vector sequence needs to be switched from iμ -iν -iμ to iν -iμ -iν when Ii * turns
from an odd sector to an even sector. Similarly, the inverter stage follows the principle of
minimum switching action. Tc is a carrier cycle; dμ = Tc dμ and dν = Tc dν is the duration
time of rectifier vector iμ and iν , respectively; d2μ = Tc dμ d2 is the duration time when
rectifier vector iμ and inverter vector v2 are implemented.

98
Energies 2024, 17, 2546

Figure 3. Switching pattern of 3 × 6 IMC.

2.2. Classical Periodic PWM

In the classical periodic PWM modulation, the carrier frequency of the multiphase
IMC inverter stage is no longer constant [35–38] but varies periodically around a specific
center frequency according to certain mathematical rules.

f c (t) = f c0 ± Δ f c vm (t) (9)

where f c0 represents the center frequency of the time-varying carrier frequency, while Δf c
denotes the maximum frequency deviation from the time-varying carrier frequency to
the center frequency. Furthermore, vm (t) stands for a periodic signal with an amplitude
within [−1, 1]. Typically, sine wave, triangular wave, and exponential wave are selected as
modulation signals, which can be expressed as follows:

m(t) = Mm sin(2π f m t) (10)

⎧
⎪ Tm 2
⎪
⎪ 0 < t < ks , m ( t ) = Mm f m t
⎪
⎪ 2 k
⎨ Tm ks
s
Mm
ks < t < (1 − ) Tm , m(t) = (1 − 2 f m t ) (11)
⎪
⎪ 2 2 (1 − k s )
⎪
⎪
⎪ (1 − k s ) T < t < T , m ( t ) = M 2 ( f t − 1)
⎩ m m m m
2 ks
where ks is the symmetry index in the range (0, 1).
⎧
⎪ Tm Mm
⎪
⎪ 0<t< , m(t) = (e pt − 1)
⎪
⎪ 4 p
⎪
⎪
⎪
⎪ ( e 4 f m − 1)
⎪
⎪
⎪
⎪ p
⎪
⎪ Tm Tm Mm
⎪
⎪ <t< , m(t) = 2
(e f m e− pt − 1)
⎪
⎪ p
⎪
⎪ 4 2
⎪
⎪
⎨ ( e 4 f m − 1)
−p (12)
⎪
⎪ T 3T M
⎪
⎪ m < t < m , m(t) = m
(1 − e 2 f m e pt )
⎪
⎪ p
⎪
⎪ 2 4
⎪
⎪
⎪
⎪
⎪
( e 4 f m − 1)
⎪
⎪ p
⎪
⎪
⎪
⎪
3Tm
< t < Tm , m(t) =
Mm f
(1 − e m e− pt )
⎪
⎪ p
⎪
⎪ 4
⎩
( e 4 f m − 1)

where p is the concavity coefficient.

According to the Parseval theorem, in order to ensure constant frequency energy, the
peak harmonic amplitude decreases as the bandwidth of the harmonic increases [39]. By
using periodic PWM modulation, the carrier frequency is no longer fixed, so the high-
frequency harmonics of multiphase IMC will not concentrate near a fixed frequency. This
can greatly reduce the amplitude of high-frequency current harmonics, but there still exists
some harmonic content.

99
Energies 2024, 17, 2546

3. Spread-Spectrum Modulation Based on Current Ripple

3.1. Output Voltage Error Locus
The equilibrium operating points of the vectors vs , is , vi , and ii are the grid voltage
sampling value vector V s , the average grid-side current vector Is , the average input voltage
vector V i , and the input-side current expectation vector Ii * per unit switching cycle. The
#s , #is , and v
disturbance signals are the ripple vectors v #i , and the input current error vector #ii .
Space vector modulation utilizes fundamental voltage vectors and zero vectors to
synthesize the continuously varying reference voltage V o * . When there is a variance
between the actual output voltage of the multiphase indirect matrix converter and the
reference voltage, a vector model can be established to represent the error in voltage
between the reference and actual output voltages [39,40].

# s (t) = vo − Vo∗
V (13)

where V# s (t) represents the voltage fluctuation quantity, vo represents the actual output
voltage vector, and V o * is the reference voltage. Taking V o * located in the first sector as
an example, V 1 , V 2 , V 3 , and V 4 are required to synthesize this vector, corresponding to
inverter stages V 55 , V 45 , V 44 , and V 64 .
Taking V o * located in the first sector as an example, V 1 , V 2 , V 3 , and V 4 required to
synthesize this vector correspond to the inverter levels V 55 , V 45 , V 44 , and V 64 , and the
magnetic chain fluctuations will be presented as five dimensions in the unit switching
period, which are v1μ , v1ν , v2μ , v2ν , v3μ , v3ν , and v4μ , v4ν , along the V 1 ~V 4 . As shown in
the PWM vector sequence, within one switching cycle, the actual input current vector is
composed of five instantaneous vectors: the direction of iμα , iμβ along the direction of i1 ,
and iνα , iνβ along the direction of i2 , as well as the zero vector i0 . Correspondingly, there
are five types of error current vectors, as depicted in Figure 4a.

Figure 4. Error vectors and their locus in the IMC output side in Section :
1 (a) output voltage–error
vectors. (b) output current ripple vectors.

From this, the locus of the output current ripple vector is obtained, as shown in
Figure 4b. At the beginning of the carrier cycle, t = t0 , vo coincides with V o * , and the voltage
ripple vector corresponds to point O in Figure 4b; during the time period of t0 ~t7 , the locus
departs from point O and forms the polygon OABCDEFO along the direction of the arrow;
from the vector sequence layout in Figure 2, it can be seen that the vector locus during the
time period of t0 ~t7 is the same as that during the time period of t7 ~t14 . It needs to be noted
that the ratio of unit length on the Re-Im axis is set to 1:10, which is due to the fact that it
is a qualitative rather than a quantitative calculation of ripples and does not mislead the
understanding of its properties.

3.2. Small-Signal Model

Based on the principle of high-frequency synthesis, the space vector method employs
a finite number of effective and zero vectors to continuously synthesize desired vectors
with varying values. The discrepancy between the actual and desired vectors results in
the input and output of the IMC containing numerous high-frequency voltage harmonics,

100
Energies 2024, 17, 2546

ultimately contributing to the formation of high-frequency ripple components. It should be

noted that Ii * is usually synthesized by the two boundary current vectors in the rectifier
stage. In Figure 3, one section is the action time of the boundary current vector iμ , and
the other sector is the action time of the boundary current vector iν . Thus, the chopper
DC-link voltage of the multiphase IMC will be in a two-level fluctuation state along the
three-phase sine envelope. This implies that the output current ripple of the IMC is not
solely determined by the switching sequence of the inverter stage but is also affected
by the switching sequence of the rectifier stage. In order to mitigate its impact on grid-
side waveform quality, a second-order RLC low-pass filter is implemented at the input
side to provide an independent low-impedance flow path for the ripple current. This
facilitates the analysis of the output current ripple locus in relation to various factors such
as output/input switching sequences and converter regulation system influence [41].
The input current of the IMC is generated by chopping the load current, resulting in a
waveform characterized by a sequence of intermittent current pulses. To ensure the stability
of the input voltage, a parallel filter capacitor Cf is required at the input terminal of the
IMC. Furthermore, a series filter inductor Lf is necessary to attenuate the transmission of
high-frequency harmonic components of the input current between the power supply and
the IMC input. As a result, the high-frequency harmonic current primarily flows through
the low-impedance path provided by Cf , which forms a loop and reduces its impact on
the waveform quality of the source. Collectively, Cf , Lf , along with a damping resistor Rf ,
constitute the input filter. A single-phase equivalent circuit of the IMC system is shown in
Figure 5, where the filter is a typical two-port network.

Figure 5. A double-phase equivalent of the IMC system.

Using transfer functions, the following analysis is performed on the two-port network
in the complex frequency domain of the s-plane: The source voltage vs. and the MC
input-side current ii are selected as the independent variables of the port network; then, the
source current is and the input-side voltage vi can be expressed using the hybrid parametric
equation (h-equation) as follows:
      
vi ( s ) h h12 − ii ( s ) − ii ( s )
= 11 =h (14)
is ( s ) h21 h22 vs ( s ) vs ( s )

The values of the elements of the hybrid parameter matrix h are as follows:
⎧ 
⎪ vi (s)  2 · sRf Lf
⎪
⎪ h11 = −i (s)  = 2
⎪
⎪ s Rf Lf Cf + sLf + Rf
⎪
⎪
i  vs (s)=0
⎪
⎪ v ( s )  sLf + Rf
⎪
⎪ h12 = i  = 2
⎨ vs (s) ii (s)=0
 s R f f Cf + sLf + Rf
L
 (15)
⎪
⎪ i s ( s ) 
⎪
⎪ h21 = = −h12
⎪
⎪ −ii (s) vs (s)=0
⎪
⎪
⎪
⎪
⎪ is (s)  (sLf + Rf )sCf
⎩ h22 = = 2
vs (s) ii (s)=0 s Rf Lf Cf + sLf + Rf

101
Energies 2024, 17, 2546

where Lf , Cf , and Rf are the filter inductance, capacitance, and damping resistance, respectively.
Appropriate filter parameters are a prerequisite for the reliable operation of the IMC.
The input voltages iiA , iiB , and iiC result from the high-frequency modulation of the
output current from the switching matrix. Furthermore, in the multiphase indirect matrix
converter discussed in this paper, a low-pass filter is integrated between the input side and
the grid side. This setup aims to curb the ingress of high-frequency harmonic components
from the input current into the power grid. By doing so, it alleviates interference with
other equipment at the common connection point and effectively filters out undesirable
high-frequency harmonics.
The design of the filter necessitates careful consideration of its limitations and device
selection criteria [30], which are evaluated based on the IMC equivalent circuit model
illustrated in Figure 5. Initially, it is crucial to address the constraints regarding switching
ripple and low-frequency harmonics. Since the fifth and seventh harmonics exhibit the
highest amplitudes in the grid voltage, for simplification purposes, we will exclude the
seventh harmonic from actual calculations. As depicted in Figure 5, the filter amplifies
the harmonic component of the low-frequency voltage on the grid side. Therefore, it is
imperative to restrict the low-frequency voltage gain coefficient of the filter to suppress
low-frequency harmonics in the input voltage.

| G (j7ωb )| ≤ 2 dB (16)

where wb represents the input fundamental angular frequency. Similarly, to effectively

suppress the ripple component of the current carrier frequency on the grid side, it is
necessary to adjust the current gain coefficient of the filter at the carrier angular frequency
wc .
| G (jωc )| ≤ −26 dB (17)
Secondly, it is important to consider the limitations of inductor voltage drop and
capacitor current. The ability to regulate the power factor is dependent on restricting the
fundamental current of the filter capacitor. Therefore, it is crucial to take into account these
limitations when designing the system for optimal performance.

I ≈ ωb Cf vs ≤ 0.2Iirated (18)

where Iirated is the effective value of input current fundamental wave under rated load. To
meet the voltage utilization requirements, it is necessary to limit the maximum fundamental
voltage drop across the filtering inductance, under rated load conditions.
$
2
v ≈ ωb L f I 2 + Iirated ≤ 0.03Vs (19)

where V s is the RMS value of the gride-side voltage. At the same time,
2
Pirated Vorated Ro (85 V)2 × 10 Ω
Iirated = = = ≈ 8.76 A (20)
3Vs 3Vs 3 × 100 V

where Pirated is the rated active power, and Vorated is the RMS of the rated output voltage.
From the above equation, the upper limit of the filter capacitance and inductance is as
follows: 
Cf ≤ 55.8 μF
(21)
Lf ≤ 1.09 mH
Finally, the damping resistance Rf is carefully chosen to prevent oscillation in the CL
filter. It is crucial to limit the maximum power loss of the damping resistor Rf .

3Rf iR 2 ≤ 0.01Pirated (22)

102
Energies 2024, 17, 2546

where iR is the filter resistor current. Based on the above analysis, the main performance
indicators considered in the filter design process are determined according to the required
parameters and comprehensive practical considerations as follows:
1. The attenuation ratio for high-order harmonics should be no less than 26 dB;
2. The gain of the seventh and lower-order harmonics must be limited to 2 dB;
3. The voltage drop in the inductor must not exceed 3% of the fundamental amplitude;
4. Reactive current flowing into the capacitor shall not exceed 20% of the rated current;
5. The loss on the Rf is less than 1% of the rated power.
Similarly, the output voltage vg and input current ii are chosen as the independent
variables of the port network, and the output current ig and input voltage vi can be
expressed using a mixed parameter equation (h-equation).
      
ig ( s ) h h12 ii ( s ) ii ( s )
= 11 =h (23)
vi ( s ) h21 h22 vg ( s ) vg ( s )

The values of the elements of the hybrid parameter matrix h are as follows:
⎧ 
⎪ vi (s)  2 · sRf Lf
⎪
⎪ h =  = 3
⎪
⎪
11
− i ( s ) s R L 2 C + s2 ( R2 L C + L2 ) + 2sL R + R2
⎪
⎪
i  vs (s)=0 f f f f f f f f f f
⎪
⎪ vi (s)  1
⎪
⎨ h12 = v (s) 
⎪ =
sLf + Rf
s ii (s)=0
 (24)
⎪
⎪ i s ( s )  2sRf Lf
⎪ h21 =
⎪  = 2
⎪
⎪ − i ( s )
i  vs (s)=0 s R f f f + sLf + Rf
L C
⎪
⎪
⎪
⎪ i ( s )  1
⎪
⎩ h22 =
s  =
vs ( s )  ii (s)=0 q

3.3. Current Ripple Model

As mentioned above, the SVM used in the IMC uses a limited number of active and
zero vectors to generate a continuously changing target vector. The difference between the
actual and targeted vectors results in the presence of numerous high-frequency harmonics
in the input and output of the IMC. These harmonic components, which are filtered by the
inductor/capacitor filter, manifest as output current ripples.
With the help of the Clarke transformation, the six-phase current quantities can be
transformed into vector forms on the abc plane. The output current vector ig can be
represented as follows:

2   T
ig = 1 1ej2π/3 ej4π/3 ejπ/6 ej5π/6 e j3π/2 · iA iB iC iU iV iW (25)
3
where iA , iB , iC , iU , iV , and iW are the six-phase current on the output side, and the
superscript T denotes matrix transpose. Similarly, the output voltage vector vg , input
current vector ii , and input current vector vi can be obtained. From this, it is easy to see
that by generalizing the variables in (18) to vectors, the equation form remains unchanged.
Using small-signal analysis methods, the vector-expanded form of the nonlinear Equa-
tion (18) can be linearized. Thereafter, by defining small variations #ig (s) = ig (s) − Ig (s),
#g (s) = vg (s) − Vg (s), #ii (s) = ii (s) − Ii (s), and v
v #i (s) = vi (s) − Vi (s) around equilibrium
points Ig (s), V g (s), Ii (s), and V i (s), and assuming that the ripple of input current #ii = 0 when
the power supply is stable, it is possible to derive the small-signal model of the output side
as follows:
#ig (s) = 1
v#g (s) (26)
sLf + Rf

103
Energies 2024, 17, 2546

Taking the inverse Laplace transform of (19), we can obtain the time-domain integral
form of the output current ripple vector expression as follows:
 t
#ig (s) = 1 v#g dτ (27)
Lf 0

Combined with the PWM switching sequence shown in Figure 3, the integration
Equation (27) can be discretized as follows:
 t v#g,k
#ig = 1 v#g dτ = #ig,k + ( t − t k −1 ) (28)
Lf 0 Lf

v#g,k k v#g, k
#ig,k = #ig,k + T = ∑ Tn (29)
Lf k n =1
Lf

where k = 1, 2, . . ., 14, t ∈ (tk−1 , tk ) (indicated as O, A, B. . .in Figure 3), #ig,k−1 is the output
current ripple vector at t = tk−1 and #ig,0 = 0; v#g,k is the output voltage error vector within
(tk−1 , tk ); Tk = tk − tk−1 .
The evaluation of the output current quality is typically conducted through quan-
tification of the total harmonic distortion (THD). Nevertheless, the THD represents a
frequency-domain metric that entails significant computational effort. The ripple consti-
tutes a time-domain variable that incorporates all harmonic components of the output
current. Consequently, the RMS value of the ripple signal can serve as an effective equiv-
alent for characterizing the THD; that is, within a carrier cycle, the THD of the output
current can be defined as follows:
%
1  π #2
i dt
π 0 g,rms
THD | Tc = × 100% (30)
igfund

where igfund refers to the fundamental magnitude of the output current, and #igrms represents
the RMS of output current ripple within unit carrier frequency cycle Tc , which is defined as
follows:
%
2 2 2 2 2 2
#ig,rms = #irms + #irms + #irms + #irms + #irms + #irms (31)
gA gB gC gU gV gW

where #igA
rms , #rms , #
igB rms , #
igC rms , #
igU rms , and #
igV rms denote the RMS of the six-phase source ripple
igW
currents. The maximum value of the ripple is limited to
   
#   
ig,max  = max{|OA|, |OB| . . . |OF|} = max(#ig,k ) (32)
 
 
where max(#ig,k ) is the maximum value representing the output ripple vector, which is
used to analyze the dynamic changes in the output current ripple locus.

3.4. Spread-Spectrum Modulation Based on Current Ripple Vectors

As depicted in Figure 4b, the locus of the output current ripple in the multiphase
IMC is correlated with the amplitude and phase of the input–output voltages, among
other factors. The fluctuation of the DC-link voltage vpn directly determines the output
ripple vector. The carrier frequency based on the dynamic variation in the ripple locus
no longer fluctuates within a specific range and has unique advantages in the spread
spectrum. Matching high carrier frequencies in the high ripple range can expand the
energy concentrated at specific frequencies to a wider spectrum, effectively reducing the
THD. Simultaneously, matching low carrier frequencies in the low ripple range improves
system efficiency while achieving the spread spectrum.

104
Energies 2024, 17, 2546

By combining the two principles mentioned above, the correspondence between the
ripple locus and carrier frequency is further established. From (22), it can be seen that the
length of each segment of the locus is only related to the carrier period Tc . It can be inferred
that the scale of the ripple can be easily manipulated by modulating the carrier frequency.
Therefore, in this section, by dynamically changing the carrier frequency with the ripple
locus, the variation is no longer based on a fixed magnitude. This method aims to optimize
the output spectrum and enhance system operational efficiency.
The variable Tc appears in each term of the ripple expression in (22). Hence, (25) can
be rewritten as follows:
   
   k v#   k v# dn 
#   g,k   g,k 
ig,max  = max( ∑ ) Tn = max( ∑ ) T (33)
 n =1 Lf   n =1 Lf  c

where k = 1, 2, . . ., 14; dn = Tn /Tc . This demonstrates that the output current ripple is
proportional to the carrier period Tc .
Within
 each
 carrier cycle, the carrier
 frequency
∗ is modified to maintain the modulation
#  # 
index ig,max  at its specified value ig,max  . It is evident that the carrier frequency f c,carrier
may be computed uniquely once the amplitude limit of the input voltage ripple vector in
each cycle is known.
 
 k v# dn   ∗
 g,k  # 
f c,carrier = max( ∑ )/ii, max  × 20 kHz (34)
 n =1 Lf 

The RMS of the current ripple can be used in place of the THD as a standard indicator
for assessing the quality of the current on the output side. Given that THD is a physical
quantity that has been established over a long period of time, a suitable method must be
used to assess if it complies with design criteria. In particular, it is sufficient to verify that
the RMS of the output current ripple of any carrier period #ig,rms satisfies the pertinent THD
requirement after averaging over a certain time interval (typically equals to a fundamental
period Tb ).
$
1  Tb #2
T 0 ig,rms dt
THD = b (35)
igrate
The traditional periodic PWM carrier frequencies fluctuate within a specific range.
However, the carrier frequency defined by (19) no longer specifies a range of variation;
instead, it dynamically adjusts with the trajectory of the ripple. This modification leads
to a more standardized usage and eliminates the debugging process inherent in classical
periodic PWM. It also facilitates clearer numerical indexes, thereby enhancing efficiency in
carrier frequency fluctuation. Consequently, this approach can further reduce harmonic
spikes near specific switching frequencies and promote a more uniform energy distribution
within each frequency band.

4. Experimental Results
To assess the impact and effectiveness of the proposed CR-SSM method in terms of
spreading spectrum, enhancing power quality, and improving efficiency for multiphase
IMC systems, a 3 × 6-phase IMC system prototype was utilized for simulation and experi-
mental validation. Furthermore, to further confirm the efficacy of the proposed CR-SSM
method, it was compared with various conventional periodic PWM strategies such as
triangular (TWPWM), sinusoidal (SWPWM), and exponential (EWPWM) periodic PWM.
The experimental prototype of the IMC platform is shown in Figure 6. The controller
was a DSP (TI F28379D) + FPGA (Intel 5CEFA4) architecture, where the SVM modu-
lation algorithm was completed independently by FPGA. And DSP was used for the
online calculation of the carrier frequency. The rectifier circuit consisted of six Infineon
FF200R12KT3 power modules, while the inverter circuit included two two-level IPM mod-

105
Energies 2024, 17, 2546

ules PM25CLA120 from the MITSUBISHI. Measurement equipment such as oscilloscopes

and power analyzers (WT5000) were all from the Yokogawa corporation, Tokyo, Japan. The
specific system parameters are shown in Table 2.

Figure 6. The experimental 3 × 6 IMC platform.

Table 2. Parameter settings.

Parameters Symbol Values

Effective value of grid-side
Vs 100 V
phase voltage
Grid-side voltage frequency fs 50 Hz
Grid-side filter capacitors Cf 35 μF
Grid-side filter inductors Lf 1.37 mH
Grid-side filter resistors Rf 10 Ω
Load resistance R 10 Ω
Load inductance L 30 mH
Output voltage frequency fo 20 Hz
Fixed carrier frequency f c0 10 kHz

4.1. Analysis of Harmonic Suppression

To demonstrate the change in carrier frequency in terms of distribution characteristics,
with respect to the input voltage phase angle θ#i , the scalar value of f c was assigned the
same phase information as θ#i and converted to the vector form, i.e.,

fc = f c ejθi (36)

The waveform of the vectorized carrier frequency f c can be represented using a polar
coordinate system. In this system, the radial distance represents the value of the carrier
frequency f c , and the angular coordinate represents the phase angle θ#i of the input voltage
vector. The orange and purple circles in Figure 7a, respectively, show the carrier frequency
waveform for the CR-SSM method when f c = f c,max = 20 kHz and f c = f c,min = 5 kHz. The
annular region between the two circles represents the feasible range of carrier frequency
modulation. In order to visualize the distribution of the modulated carrier frequency in each
frequency band, Figure 7b illustrates the histogram of the output harmonic distribution
under the CR-SSM method in the 5–20 kHz band. The abscissa represents the harmonic
frequency band, and it is divided into 17 equal segments to show the actual output of 5–20 k.
The ordinate represents the approximate harmonic amplitude distribution in this band.

106
Energies 2024, 17, 2546

(a) (b)

Figure 7. Plot of carrier frequency with CR-SSM of (a) vectorized waveform and (b) distribution
histogram.

A comparison of three types of periodic PWM with the proposed CR-SSM method
was performed for further evaluation of the effectiveness of the proposed method. The
vectorized waveform of the frequency-modulated wave and the harmonic spectrum near
10 k for the other three types of periodic PWM methods are shown in Figure 8. The
following observations can be made:
1. The strategy of periodic PWM to reduce the carrier frequency always results in a lower
range of carrier frequencies compared to the CR-SSM method while maintaining the
same limit on output current ripple. It can be seen that the classical periodic PWM still
has harmonic spikes near the carrier frequency of 10 kHz, the variable band energy is
still more concentrated, and the THD value is higher. The proposed CR-SSM method
has better spectral characteristics, no harmonic spikes at specific frequencies, uniform
high-frequency energy distribution, and better THD values.
2. The carrier frequency waveform displays periodicity along the direction of the output
voltage vector angle θ#o , but its locus is affected by the input frequency θ#i , and it does
fully coincide with the previous cycle after a complete cycle (θ#o = 2π).

Figure 8. Comparison of the frequency variation locus between classic periodic PWM and CR-SSM
switching method.

4.2. Experimental Analysis of the Converter Efficiency

As the carrier frequency of the IMC system decreases, so do the losses. The efficiency
of IMC can be improved by modulating the carrier frequency with CR-SSM, which employs
a wider range of carrier frequency variations compared to the classical periodic PWM
method. Therefore, it is necessary to conduct further experimental verification to determine
the effect of CR-SSM and periodic PWM methods on the efficiency of IMC systems. Figure 9
shows the operational efficiency of the system under different methods. A comprehensive
analysis of the data reveals the following observations:

107
Energies 2024, 17, 2546

1. The efficiency of the CR-SSM method is always lower than that of the periodic PWM
method at different voltage transmission ratios q;
2. As the voltage transfer ratio q increases, the efficiency of the system also increases.

Figure 9. Operating efficiency of various methods under different working conditions.

Figure 9 illustrates the three periodic PWMs (EWPWM, TWPWM, and SWPWM)
and the operational efficiency of the multiphase IMC system under the proposed CS-
SSM strategy. As depicted, the system efficiency of the CS-SSM method escalates from
27.5% to 94.5%; in SWPWM, it increases from 23.1% to 92.1%; in EWPWM, it increases
from 23.3% to 91.9%; and in TWPWM, it increases from 22.8% to 92.2%. Notably, at
modulation 0.8, the CR-SSM method boosts efficiency by 2.4%, 2.3%, and 2.6% compared to
EWPWM, TWPWM, and SWPWM, respectively. Similarly, at modulation 0.6, the CR-SSM
method enhances efficiency by 2.7%, 2.3%, and 2.9% compared to EWPWM, TWPWM, and
SWPWM, respectively. The conjunction of these findings with Figure 9 demonstrates that
the proposed CR-SSM method not only ensures a spread-spectrum effect but also enhances
the operational efficiency of the multiphase IMC system by leveraging the carrier frequency
density function based on the current ripple.

4.3. Experimental Results on Input–Output Waveform Quality

Figure 10 illustrates a comparison of the total harmonic distortion (THD) of load
current under the CR-SSM method and periodic PWM. It can be observed that, under
different voltage transfer ratio q conditions, the THD of the load current under the proposed
CR-SSM method is lower than that of the periodic PWM method. This indicates that
the carrier frequency modulation technique can effectively reduce the harmonic level of
the IMC load current. Furthermore, as the modulation depth increases, the THD of the
proposed CR-SSM is always lower than that of the periodic PWM method.

Figure 10. THD comparison of CR-SSM and three classical periodic PWM methods.

As depicted in Figure 10, the total harmonic distortion (THD) of the output current
diminishes with increasing modulation. Notably, employing the proposed CR-SSM strategy
results in a significant reduction in THD from 41.34% to 7.32%. Similarly, the THD for
the TWPWM method decreases from 41.34% to 7.65%, while for the SWPWM method, it
decreases from 43.09% to 7.89%, and for the EWPWM method, it decreases from 43.75% to

108
Energies 2024, 17, 2546

7.67%. At a modulation index of 0.5, the THD for CR-SSM is 9.86%, while for TWPWM,
SWPWM, and EWPWM, it stands at 11.76%, 11.74%, and 11.17%, respectively. This trend
underscores that, across all modulation methods, the THD of the proposed CR-SSM strategy
remains consistently lower than the other three periodic PWM methods. The findings from
Figures 4–8 corroborate that the proposed CR-SSM strategy enhances the output current of
the multiphase IMC system while ensuring the desired spread-spectrum effect.
The time-domain waveforms of the source current, DC-link voltage, input line voltage,
and load current under various methods are shown in Figure 11. Figure 12 depicts the
dynamic experimental waveforms using four different methods as the output frequency
varies from 20 Hz to 40 Hz. It can be observed from Figures 11 and 12 that the proposed
CR-SSM method ensures a balanced sinusoidal three-phase source current, DC-link voltage,
and load current without significant fluctuations. Moreover, it effectively suppresses
harmonic amplitude at switching frequencies.

(a) (b) (c) (d)

Figure 11. Comparison between the proposed CR-SSM and classic periodic PWM: (a) CR-SSM;
(b) TWPWM; (c) SWPWM; (d) EWPWM.

(a) (b) (c) (d)

Figure 12. Comparison between the proposed CR-SSM and classic periodic PWM dynamics:
(a) CR-SSM; (b) TWPWM; (c) SWPWM; (d) EWPWM.

5. Conclusions
This study analyzes output current ripple to construct a carrier frequency density
function and proposes a CR-SSM method suitable for multiphase IMC systems. The method
selectively modulates carrier frequency without requiring changes to other modules, thus
serving as a complementary tool to optimize PWM modulators like SVM and carrier
modulation. CR-SSM utilizes output current ripple as an indicator to build a switch
frequency density function, addressing the limitations of traditional periodic PWM in center
frequency control. It offers a standardized application, eliminates the debugging process
associated with classic periodic PWM, and defines clear numerical indicators. Additionally,

109
Energies 2024, 17, 2546

it improves switch frequency efficiency, reducing harmoni2c peaks near specific switch
frequencies. This promotes a more even energy distribution across frequency bands and
optimizes system efficiency by correlating switch frequencies with output ripple heights.
By distributing the power spectrum over a broader frequency range, the system’s output
current spectrum is optimized. Experimental results demonstrate that compared to periodic
PWM, the CR-SSM strategy effectively reduces system total harmonic distortion (THD)
while enhancing operational efficiency in achieving spread-spectrum operation modes.

Author Contributions: Conceptualization, Z.Z.; data curation, L.X.; formal analysis, L.X.; funding
acquisition, Z.Z.; investigation, C.L.; methodology, L.X.; project administration, Z.Z.; resources, Z.Z.;
software, Z.Z.; supervision, Z.Z. and Q.G.; validation, L.X.; visualization, L.X.; writing—original
draft preparation, L.X. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the “Zhejiang Provincial Natural Science Foundation of China
under Grant No. LY24E070001”.
Data Availability Statement: The original contributions presented in the study are included in the
article, further inquiries can be directed to the corresponding author.
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Wang, R.; Lin, Z.; Du, J.; Wu, J.; He, X. Direct sequence spread spectrum-based PWM strategy for harmonic reduction and
communication. IEEE Trans. Power Electron. 2016, 32, 4455–4465. [CrossRef]
2. Raghuram, M.; Chauhan, A.K.; Singh, S.K. High-Gain-Integrated Switched Capacitor Indirect Matrix Converter. IEEE J. Emerg.
Sel. Top. Power 2019, 3, 1846–1853. [CrossRef]
3. Shi, T.; Wang, Z.; Ma, C.; Yan, Y.; Song, P. A Variable Switching Frequency PWM Method for Indirect Matrix Converters. In
Proceedings of the 10th International Conference on Power Electronics, Machines and Drives (PEMD 2020), Online Conference,
15–17 December 2020; pp. 943–948.
4. Padhee, V.; Sahoo, A.K.; Mohan, N. Modulation techniques for enhanced reduction in common-mode voltage and output voltage
distortion in indirect matrix converters. IEEE Trans. Power Electron. 2016, 32, 8655–8670. [CrossRef]
5. Nguyen, H.N.; Lee, H.H. An effective SVM method for matrix converters with a superior output perfor-mance. IEEE Trans. Ind.
Electron. 2017, 65, 6948–6958. [CrossRef]
6. Empringham, L.; Kolar, J.W.; Rodriguez, J.; Wheeler, P.W.; Clare, J.C. Technological Issues and Industrial Application of Matrix
Converters: A Review. IEEE Trans. Ind. Electron. 2013, 60, 4260–4271. [CrossRef]
7. Chen, J.; Sha, D.; Zhang, J. Current ripple prediction and DPWM-based variable carrier frequency control for full ZVS range
three-phase inverter. IEEE Trans. Ind. Electron. 2020, 68, 1412–1422. [CrossRef]
8. Casadei, D.; Serra, G.; Tani, A.; Zarri, L. Optimal use of zero vectors for minimizing the output current distortion in matrix
converters. IEEE Trans. Ind. Electron. 2008, 56, 326–336. [CrossRef]
9. Casadei, D.; Serra, G.; Tani, A.; Zarri, L. Theoretical and experimental analysis for the RMS current ripple minimization in
induction motor drives controlled by SVM techniqu. IEEE Trans. Ind. Electron. 2004, 51, 1056–1065. [CrossRef]
10. Attia, H.; Freddy, T.; Che, H.; Hew, W.; Elkhateb, A. Confined band variable carrier frequency pulse width modulation (CB-VSF
PWM) for a single-phase inverter with an LCL filter. IEEE Trans. Power Electron. 2016, 32, 8593–8605. [CrossRef]
11. Qin, X.; Zhou, B.; Wei, J.; Lei, J.; Liu, X.; Han, N. Distortion Analysis and Duty Ratio Correction Algorithm for Asymmetric
Modulation of Two-Stage Matrix Converter. IEEE Trans. Ind. Electron. 2014, 62, 351–362. [CrossRef]
12. Xie, L.; Ruan, X.; Zhu, H.; Lo, Y.-K. Common-Mode Voltage Cancellation for Reducing the Common-Mode Noise in DC–DC
Converters. IEEE Trans. Ind. Electron. 2021, 5, 3887–3897. [CrossRef]
13. Onederra, O.; Kortabarria, I.; De Alegria, I.M.; Andreu, J.; Gárate, J.I. Three-Phase VSI Optimal Switching Loss Reduction Using
Variable Carrier frequency. IEEE Trans. Power Electron. 2016, 32, 6570–6576. [CrossRef]
14. Marouani, K.; Baghli, L.; Hadiouche, D.; Kheloui, A.; Rezzoug, A. A New PWM Strategy Based on a 24-Sector Vector Space
Decomposition for a Six-Phase VSI-Fed Dual Stator Induction Motor. IEEE Trans. Ind. Electron. 2008, 5, 1910–1920. [CrossRef]
15. Gamoudi, R.; Elhak Chariag, D.; Sbita, L. A Review of Spread-Spectrum-Based PWM Techniques—A Novel Fast Digital
Implementation. IEEE Trans. Power Electron. 2018, 12, 10292–10307. [CrossRef]
16. Cano, J.M.; Navarro-Rodriguez, A.; Suarez, A.; Garcia, P. Variable Carrier frequency Control of Distributed Resources for
Improved System Efficiency. IEEE Trans. 2018, 54, 4612–4620.
17. Jamal, H.; Albatran, S.; Smadi, I. Variable carrier frequency algorithm for optimal tradeoff between switching losses and total
demand distortion in grid-tied three-phase voltage-source inverter. In Proceedings of the 2016 IEEE Energy Conversion Congress
and Exposition (ECCE), Milwaukee, WI, USA, 18–22 September 2016; IEEE: Piscataway, NJ, USA, 2016; pp. 1–7.
18. Jiang, D.; Wang, F. Variable carrier frequency PWM for three-phase converters based on current ripple prediction. IEEE Trans.
Power Electron. 2013, 28, 4951–4961. [CrossRef]

110
Energies 2024, 17, 2546

19. Atamaniuk, V.; Kuznyetsov, O.; Lunkova, H. Driving the PWM-Controlled System to Monostable Behavior by Periodic Modulation
of a Carrier Signal. In Proceedings of the 2022 IEEE 4th International Conference on Modern Electrical and Energy System
(MEES), Kremenchuk, Ukraine, 20–22 October 2022; pp. 1–6.
20. Shan, Y.; Pei, X.; Sun, T.; Zhang, M.; Zhou, P.; Jiang, D. Space spread-spectrum strategy for MMC to reduce the conducted EMI.
IEEE Trans. Ind. Electron. 2021, 69, 10807–10818. [CrossRef]
21. Chen, J.; Jiang, D.; Shen, Z.; Sun, W.; Fang, Z. Uniform distribution pulsewidth modulation strategy for three-phase converters to
reduce conducted EMI and switching loss. IEEE Trans. Ind. Electron. 2019, 67, 6215–6226. [CrossRef]
22. Li, Q.; Jiang, D.; Zhang, Y. Analysis and calculation of current ripple considering inductance saturation and its application to
variable carrier frequency PWM. IEEE Trans. Ind. Electron. 2019, 34, 12262–12273.
23. Chen, J.; Jiang, D.; Sun, W.; Shen, Z.; Zhang, Y. A Family of Spread-Spectrum Modulation Schemes Based on Distribution
Characteristics to Reduce Conducted EMI for Power Electronics Converters. IEEE Trans. Ind. Applicat. 2020, 56, 5142–5157.
[CrossRef]
24. Le, D.; Choi, S. Network CM EMI Reduction Using Sinusoidal Frequency Modulated Carrier Wave Indexing. In Proceedings of
the 2023 11th International Conference on Power Electronics and ECCE Asia (ICPE 2023-ECCE Asia), Jeju Island, Republic of
Korea, 22–25 May 2023; Volume 38, pp. 3321–3326.
25. Lumbreras, D.; Zaragoza, J.; Berbel, N.; Mon, J.; Gálvez, E.; Collado, A. Comprehensive Analysis of Hexagonal Sigma-Delta
Modulations for Three-Phase High-Frequency VSC Based on Wide-Bandgap Semiconductors. IEEE Trans. Power Electron. 2021,
36, 7212–7222. [CrossRef]
26. Menon, M.A.; Jacob, B. A New Space Vector Pulse Density Modulation Scheme for Two-Level Five Phase Induction Motor Drive.
IEEE Trans. Ind. Electron. 2022, 69, 12077–12085.
27. Huang, Y.; Xu, Y.; Zhang, W.; Zou, J. Modified Single-Edge SVPWM Technique to Reduce the Switching Losses and Increase
PWM Harmonics Frequency for Three-Phase VSIs. IEEE Trans. Power Electron. 2020, 35, 10643–10653. [CrossRef]
28. Ji, Z.; Cheng, S.; Li, X.; Lv, Y.; Wang, D. An Optimal Periodic Carrier Frequency PWM Scheme for Suppressing High-Frequency
Vibrations of Permanent Magnet Synchronous Motors. IEEE Trans. Power Electron. 2023, 38, 13008–13018. [CrossRef]
29. Ji, Z.; Cheng, S.; Ren, Q.; Li, X.; Lv, Y.; Wang, D. The Effects and Mechanisms of Periodic-Carrier-Frequency PWM on Vibrations
of Multiphase Permanent Magnet Synchronous Motors. IEEE Trans. Power Electron. 2023, 38, 8696–8706. [CrossRef]
30. Chen, N.; Yu, S.; Gao, L.; Huang, Y.; Chen, R. Suppressing Interference Peak in an Active Power Filter via Periodic Carrier
Frequency Modulation Based on a Spectrum Analysis Approach. IEEE Trans. Electromagn. Compat. 2019, 61, 1760–1770. [CrossRef]
31. Xu, J.; Nie, Z.; Zhu, J. Characterization and Selection of Probability Statistical Parameters in Random Slope PWM Based on
Uniform Distribution. IEEE Trans. Power Electron. 2021, 1, 1184–1192. [CrossRef]
32. Ma, Y.; Yang, M.; Lyu, Z.; Li, S.; Xu, D. Application of Periodic Carrier Frequency Modulation for Harmonic and EMI Reduction in
GaN-Based Motor Drive. In Proceedings of the 2020 23rd International Conference on Electrical Machines and Systems (ICEMS),
Hamamatsu, Japan, 24–27 November 2020; pp. 778–782.
33. Nguyen, T.D.; Lee, H.-H. Dual Three-Phase Indirect Matrix Converter With Carrier-Based PWM Method. IEEE Trans. Power
Electron. 2014, 2, 569–581. [CrossRef]
34. Benson, M.; Dornala, A.; Andleeb, M.; Lee, K.; Lee, W. Simple and Robust Carrier-Based PWM Technique for Single-Stage
Three-phase Rectifier Indirect Matrix Converter. In Proceedings of the 2024 IEEE Applied Power Electronics Conference and
Exposition (APEC), Long Beach, CA, USA, 25–29 February 2024; pp. 691–697.
35. Park, H.P.; Kim, M.; Jung, J.H. Spread spectrum technique to reduce EMI emission for an LLC resonant converter using a hybrid
modulation method. IEEE Trans. Power Electron. 2017, 33, 3717–3721. [CrossRef]
36. Dasgupta, A.; Sensarma, P. Filter design of direct matrix converter for synchronous applications. IEEE Trans. Ind. Electron. 2014,
61, 6483–6493. [CrossRef]
37. Qiu, Z.; Chen, Y.; Kang, Y.; Liu, X.; Gu, F. Investigation into periodic signal-based dithering modulations for suppression sideband
vibro-acoustics in PMSM used by electric vehicles. IEEE Trans. Energy Convers. 2020, 36, 1787–1796. [CrossRef]
38. Leoncini, M.; Bertolini, A.; Melillo, P.; Gasparini, A.; Levantino, S.; Ghioni, M. Spread-spectrum frequency modulation in a
DC/DC converter with time-based control. IEEE Trans. Power Electron. 2022, 38, 4207–4211. [CrossRef]
39. Kumar, A.B.; Narayanan, G. Variable-Carrier frequency PWM Technique for Induction Motor Drive to Spread Acoustic Noise
Spectrum with Reduced Current Ripple. IEEE Trans. Ind. Appl. 2016, 52, 3927–3938. [CrossRef]
40. Li, Q.; Jiang, D. Variable Carrier frequency PWM Strategy of Two-Level Rectifier for DC-Link Voltage Ripple Control. IEEE Trans.
Power Electron. 2017, 33, 7193–7202. [CrossRef]
41. Xu, C.; Lu, S. Practical Online Modulation Method for Current Ripple and Switching Losses Reduction in the Three-Phase Voltage
Source Inverters. IEEE Trans. Power Electron. 2021, 2, 1475–1490. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

111
 energies
Article
A DCM-Based Non-Isolated Step-Down DC Transformer
Minseung Kim 1 , Donghee Choi 2 and Soo Hyoung Lee 3, *

1 Department of Electrical Engineering, Mokpo National University, Muan 58554, Republic of Korea
2 Division of Converged Electronic Engineering, Cheongju University, Cheongju 28503, Republic of Korea;
heechoi@cju.ac.kr
3 Division of Electrical, Electronic, and Control Engineering, Kongju National University,
Cheonan 31080, Republic of Korea
* Correspondence: lsh@kongju.ac.kr

Abstract: DC transformers have emerged as essential devices for medium voltage DC (MVDC)-low
voltage DC (LVDC) distribution systems. However, conventional step-down single-level converters
have limits on the voltage level of the MVDC-LVDC distribution system. This paper proposes a
non-isolated step-down (NISD) DC transformer based on discontinuous conduction mode (DCM).
The proposed structure can withstand high voltage levels by sharing voltages between energy storage
modules dividing voltage levels. The proposed NISD DC transformer determines operational modes
based on energy storage modules and performs the voltage conversion process. The effectiveness of
the proposed NISD DC transformer is verified based on a case study using a power system computer-
aided design and electromagnetic transient simulation engine including DC (PSCAD/EMTDC™).

Keywords: DC distribution; DC transformer; modular multilevel converter; non-isolated converter;

step-down converter

1. Introduction
As climate change poses increasingly significant environmental challenges, there is a
worldwide increase in the demand for renewable sources of energy that are sustainable and
environmentally friendly. In accordance with the International Energy Agency, renewable
energy will comprise 29.8% of the world’s overall electricity production by 2022 [1]. The
growth of direct current (DC) power generation is driving the increase in renewable energy
production, along with the rise of DC loads such as electric vehicle charging stations and
data centers [2–4]. This attention is leading to a focus on DC power transmission and
distribution [2–7]. A critical challenge in this area is the design of converters to reduce the
voltage level from medium-voltage DC (MVDC) to low-voltage DC (LVDC) without using
the bulky and expensive AC transformers that are currently in use. A DC transformer,
commonly known as a DC-DC converter, uses switching components and energy storage
implements for voltage transformation [8,9]. However, conventional single-stage step-
down converters require high-rated voltages for each component, which leads to higher
costs, larger sizes, and higher losses. As a consequence, the MVDC to LVDC distribution
voltage levels are limited [10–12].
To overcome the disadvantages of replacing AC transformers, alternative isolated DC
converters using medium- or high-frequency transformers have been proposed instead
of the traditional bulky low-frequency transformers [13–16]. Despite this solution, these
converters still exhibit certain drawbacks, such as losses during the DC-AC-DC conversion
and the continued presence of AC transformers [17].
This paper proposes a non-isolated step-down (NISD) DC transformer based on dis-
continuous conduction mode (DCM). Currently, only DCM operation is considered, and the
proposed structure divides high voltage using an energy storage module (ESM) comprised
of series. The proposed NISD DC transformer determines operational mode based on the
ESMs and performs voltage transformation, providing the following advantages:

Energies 2024, 17, 940. https://doi.org/10.3390/en17040940 112 https://www.mdpi.com/journal/energies

Energies 2024, 17, 940

(1) It is possible to achieve a high voltage ratio with a small number of components and a
simple control method.
(2) It operates in DCM, reducing switching losses and improving efficiency.
(3) There is no requirement for an isolation transformer or galvanic isolation, which
results in reduced cost and size.
The proposed NISD DC transformer uses capacitors to store electric energy in ESMs [18–27],
unlike the conventional DC converter that uses inductors to store magnetic energy and per-
form voltage conversion [3,6,15,16,18,19,28–32]. Also, conventional flying capacitor multilevel
converters have complex control schemes [24–26] in contrast to this structure, enabling flexible
control of the output voltage by adjusting the charge and discharge duty of ESMs. The high
voltage of MVDC is dispersed among ESMs, allowing for a reduction of each circuit element’s
rated voltage to the level of commonly used components. Currently, soft switching is considered
for the only main switches excluding ESMs.
The remaining parts of this paper are structured as follows: Section 2 presents the
topology and configuration of the proposed NISD DC transformer, as well as its operational
modes and control strategy. Section 3 analyzes the IGBT voltage stress and current as well
as the considerations for the inductor, ESM, and output capacitor during the design of the
proposed NISD DC transformer. Section 4 provides the results of simulations conducted
using a power system computer-aided design and electromagnetic transient including DC
(PSCAD/EMTDC) to validate the effectiveness of the proposed NISD DC transformer in
two cases: the steady state and with disturbances. Finally, Section 5 summarizes the key
contributions of this paper and proposes several future research directions.

2. NISD DC Transformer Based on DCM

2.1. NISD DC Transformer Topology
The topology of the non-isolated step-down DC transformer is presented in Figure 1. The
proposed topology operates based on capacitors in contrast to the conventional one that uses
an inductor. It includes n energy storage modules for voltage transformation and operates
according to a six-duty ratio segment mode. Each energy storage module is composed of a
capacitor, marked as (C), and two insulated gate bipolar transistors (IGBTs), labeled as (Sa ,
Sb ), which are charged and discharged by separate groups of IGBTs, marked as (S11 , · · · , S1α
and S21 , · · · , S2α ). Moreover, the addition of the inductor (L) and the output capacitor (Cout )
restrict excessive current and reduce the ripple in output voltage, respectively.

Figure 1. (a) NISD DC transformer topology including (b) ESM.

2.2. Operational Modes

The operational modes of the NISD DC transformer are composed of duty cycles
divided into six segments (M1 − M6 ) based on the charging and discharging of the inductor
and ESM. The operational sequence depicted in Figure 2 is demonstrated to operate in a

113
Energies 2024, 17, 940

steady state. The waveforms of voltage and current are periodic, with identical values at
the beginning (t0 ) and end (t6 ) of each cycle. Figure 3 illustrates the NISD DC transformer’s
operational mode for each segment. And its current path is shown as red highlighted.

ȱ
Figure 2. Operational sequences of the proposed DC transformer.

Figure 3. Operational modes of NISD DC transformer: (a) M1 , (b) M2 , (c) M4 , (d) M5 .

Before analyzing each operational mode, it is assumed that the IGBTs, diodes, and
ESMs’ capacitor balancing are ideal during operation. Moreover, the inductor has low

114
Energies 2024, 17, 940

inductance to make the inductor current periodically zero, and each capacitor is assumed
to have a large enough capacitance to maintain a constant voltage at the output voltage
Vout and the voltage Vcn (n ∈ N) of each ESM.
The voltage reference of the ESM capacitor is defined by the relationship between the
input voltage and the number of ESMs, as shown in (1).

Vin
Vc = (1)
n
The average of each ESM capacitor voltage is expressed as (2), and the operational
mode is determined by the relationship between (1) and (2).

1 n
n i∑
Vc = Vci (2)
=1

2.2.1. First Operational Mode M1 : (t0 − t1 )

During the steady state, the state at t0 is identical to the state at t6 in the prior op-
erational sequence., i.e., both groups of IGBTs are off and the inductor current is zero.
Afterward, IGBT group 1 (S11 , · · · , S1α ) is turned on, which causes the capacitor of the
ESM to charge and the voltage to increase through the path, as shown in Figure 3a. During
this process, the inductor prevents an overcurrent by filling the difference between the
input voltage and the controlled ESM voltage and VM , as shown in (3).

VL = Vin − VM − Vout (3)

2.2.2. Second Operational Mode M2 : (t1 − t2 )

During the first operational mode, when the voltage charged to each ESM is sufficiently
high to satisfy Vc > Vc , the IGBT group 1 (S11 , · · · , S1α ) is turned off at t1 . At this moment,
as illustrated in Figure 3b, current that has been charged to the inductor by circulating
through the antiparallel diodes of IGBT group 2 (S21 , · · · , S2α ), the ESM, and diode group
1 (D11 , · · · , D1β ) delivers energy to ESMs and then discharges to zero. At this time, the
voltage of each ESM capacitor is increased by the ESM capacitor’s voltage ripple ΔVc to the
reference voltage in (1) to Vc + ΔVc . The inductor voltage during discharge in this mode is
determined by (4).
VL = −VM (4)

2.2.3. Third Operational Mode M3 : (t2 − t3 )

As the load on the DC transformer approaches its maximum rated load, which is
M3 = 0 at full load, the duration of the third mode is reduced. However, its changes do not
affect the stability of the output voltage considering the fixed switching frequency.

2.2.4. Fourth Operational Mode M4 : (t3 − t4 )

In the third operational mode, after the inductor current is completely discharged
and IGBT group 2 (S21 , · · · , S2α ) turns on at t3 , the ESM capacitor discharges via the
path shown in Figure 3c and the voltage decreases. As in the first mode, the inductor
prevents overcurrent by covering the gap between the controlled ESM voltage and the
output voltage, as given by (5).
VL = Vout − VM (5)

2.2.5. Fifth Operational Mode M5 : (t4 − t5 )

After the ESM is discharged in the fourth mode and the voltage is sufficient to satisfy
Vc < Vc , IGBT group 2 (S21 , · · · , S2α ) is turned off at t4 . At this point, the inductor
current that has been accumulated is diverted back to the input side and discharged to zero
through the anti-parallel diodes of IGBT group 2 (S21 , · · · , S2α ), the ESM, and diode group
2 (D21 , · · · , D2β ), as seen in Figure 3d. At this time, the ESM capacitors are continuously

115
Energies 2024, 17, 940

being discharged to reduce the voltage to Vc − ΔVc according to the previous mode, and
the voltage of the discharged inductor is equal to (6).

VL = Vin − VM (6)

2.2.6. Sixth Operational Mode M6 : (t5 − t6 )

Similar to the third operational mode, when the DC transformer operates at full
rated load, then M6 = 0. The duration of the sixth operational mode is also longer when
operating at lower load.

2.3. Control Strategy

The control logic of the NISD DC transformer is presented in Figure 4. Thereafter, it
proceeds in sequence through the six operational modes detailed in Section 2.2. At the
outset, the voltage of the ESM begins at zero; therefore, IGBT group 1 (S11 , · · · , S1α ) is
turned on and operates as M1 . Subsequently, IGBT group 1 (S11 , · · · , S1α ) and IGBT group
2 (S21 , · · · , S2α ) are controlled by the voltage of ESMs, and the charging and discharging
of ESMs are determined according to the six operational modes described in Section 2.2.

Figure 4. Control logic for IGBT group.

The total inserted voltage of the ESMs, VM , in Figure 3, is determined by the number of
ESMs inserted in each operational mode, N, and the capacitor voltage of the inserted ESM,
according to (7). Additionally, capacitor balancing causes the ESM charging duty, Vci at Tc , to
be sorted in ascending order while the ESM discharging duty, Vci at Td , is sorted in descending
order to ensure that the capacitor voltage of all ESMs is constantly charged and discharged.

N
VM = ∑ Vci (7)
i =1

At this time, the number of ESMs to be inserted N depends on Nctrl , which is controlled
by the output voltage, and a number of total modules n, which is determined by the
specification of the NISD DC transformer. Therefore, the number of inserted ESMs N
according to the charging and discharging cycle of the ESM is determined as (8). Also, as
the ratio of input voltage to output voltage increases, the change of N, i.e., VM , increases.

n − Nctrl f or Tc
N= (8)
Nctrl f or Td

Depending on the operational mode, the ESM is inserted to charge and discharge,
and the output voltage is determined by M1 and M4 . In order to effectively regulate the
output voltage, the ESM must be inserted variably, and the control variable Nctrl of the
insertion number is controlled flexibly according to the output voltage. Nctrl is calculated
via the control diagram shown in Figure 5, and an additional insertion amount is calculated
through the PI controller for error correction of the output voltage. The insertion amount is
limited to a minimum of 1 and a maximum of n − 1 for the operation described in Figure 3.
Since the number of modules is an integer, the amount of insertion is selected by converting
to UInt (Unsigned Integer) type.

116
Energies 2024, 17, 940

Figure 5. Control diagram for ESM insertion amount.

2.3.1. ESM Charging Duty (Tc )

Using (3) and (4), the voltage output during the ESM charging period at Tc is calculated
from (9). At this time, the total inserted voltage of the ESMs, VM between M1 and M3 is
determined by (7) and (8).
t2 − t0
Vout = Vin − VM (9)
t1 − t0
Meanwhile, the relationship between the output voltage Vout and the number Nctrl
of ESMs inserted while the ESMs are charging is shown in Figure 5. If the output voltage
is higher than the reference voltage, additional modules are inserted to charge the ESMs
and, simultaneously, the high electrical voltage difference between the input and output is
reduced, i.e., if the output voltage is higher than the reference voltage, Nctrl is reduced to
increase the inserted number of ESMs, N. Alternatively, if the output voltage is lower than
the reference voltage, N is reduced to control the output voltage.

2.3.2. ESM Discharging Duty (Td )

The voltage output during the Td discharge period of the ESM is obtained by utilizing
(5) and (6), as demonstrated in (10). Similarly to Tc , (7) and (8) are applied to determine VM
in the M4 to M6 range.
t5 − t3 t5 − t4
Vout = VM − V (10)
t4 − t3 t4 − t3 in
The correlation between the output voltage Vout during the discharge of the ESMs and
the control variable Nctrl of the number of ESMs inserted is shown in Figure 5. The number
of inserted ESMs, N, decreases when the output voltage is higher than the reference voltage,
and, contrary to Tc , the number of inserted ESMs, N, decreases. As a result, the voltage
source of the ESMs is reduced and the output voltage decreases. Conversely, as the output
voltage is lower than the reference voltage, Nctrl increases, which increases the number of
ESMs inserted, and the output voltage is increased by the discharging ESMs.

3. Design Consideration of NISD DC Transformer

In this section, design considerations are provided for the implementation of the NISD
DC transformer based on DCM. The energy conversion process is performed by multiple
ESMs, including capacitors in the NISD DC transformer, and the overcurrent that can be
produced by voltage unbalance during the conversion process is limited by inductors.
Moreover, the capacitor on the output side reduces the ripple in output voltage.
Before analysis, Figure 2 demonstrates that when the NISD DC transformer operates
by segments for t2 − t0 > Tc , i L interrupts the ESM’s capacitor discharge and load current
in the Td duty, which is not discharged in the Tc duty (i.e., i L > 0). Similarly, when
t5 − t3 > Td , i L interrupts the ESM’s capacitor charge and load current in the Tc duty,
which is not discharged in the Td duty (i.e., i L < 0). In both cases, the regular operation
of the NISD DC transformer is interrupted, i.e., the theoretical maximum power transfer
condition is t2 = t3 & t5 = t6 , and when a load with more than the maximum power is
connected, the output is reduced to protect itself from overcurrent. To ensure efficient
operation, it is necessary for the maximum value Imax and minimum value Imin of the
inductor current i L to satisfy the condition Imax = − Imin , regarding the power loss I 2 R in
the conductor.

117
Energies 2024, 17, 940

The values are determined by the switching frequency f S of Group 1 (S11 , · · · , S1α )
and Group 2 (S21 , · · · , S2α ) IGBTs in Figure 1a and the switching frequency f M of the two
IGBTs (Sa , Sb ) of the ESM in Figure 1b, which are influenced by the inductor current i L at
rated load conditions. Thus, in this section, the relationship among f S , f M , the inductor
and capacitors and the quantity of input and output voltages, IGBTs and diodes, and ESMs
in a NISD DC transformer will be discussed.

3.1. Voltage Stress on IGBTs and Diodes

IGBTs and diodes are arranged in groups of α and β (α, β ∈ N), respectively, to
endure high voltages. IGBTs and diodes are controlled to operate, as shown in Figure 3, for
voltage transformation.
IGBT Group 1 (S11 , · · · , S1α ) and Group 2 (S21 , · · · , S2α ), comprising α IGBTs,
alternate in turning on/off to prevent short-circuiting of the NISD DC transformer. The
voltage applied to the component when each IGBT group is operating is expressed as (11)
and (12).
VS1 = Vin + VD2 − VL − VM (11)

VS2 = Vin − VS1 − VL (12)

Diode groups 1 (D11 , · · · , D1β ) and 2 (D21 , · · · , D2β ) are composed of β diodes and
perform opposite operations depending on the charging and discharging duty of the ESM.
The voltage for each diode group is shown as (13), and, according to the equation, the reverse
voltage during the reverse bias of the diode is mainly affected by Vout , and the voltage during
the forward bias is operated to zero by the reverse voltage of the other diode group.

VD1 + VD2 = −Vout (13)

In the design of the NISD DC transformer, the number of IGBTs and diodes is deter-
mined by Vin,re f and Vout,re f , which are the reference of input and output voltages. From
Figure 3, each group of components is organized and operated in series, so the number of
IGBTs and diodes (α, β) for the rated voltage is determined as shown in (14) and (15).
When the rated collector-to-emitter voltage of an IGBT device is VS , then

Vin, re f
α≥ (14)
VS

When the rated breakdown voltage of a diode device is VD , then

Vout, re f
β≥ (15)
|VD |

3.2. Current Calculations of IGBTs and Diodes

During operation, the IGBTs and diodes in each group are organized in series, as
shown in Figure 3; thus, the current in each component within the same group is the same
and is expressed as (16) and (17).

IL f or Tc
IS1 = ID1 = (16)
0 f or Td

0 f or Tc
IS2 = ID2 = (17)
− I L f or Td

3.3. Design on Inductor

The NISD DC transformer’s inductor is installed to reduce the overcurrent that may
occur when switching IGBTs’ operation in accordance with the operating modes in relation to

118
Energies 2024, 17, 940

each duty ratio, and the inductance for the NISD DC transformer to operate can be calculated
by the total voltage of the inserted ESM (7), which is calculated from the inductor voltage in
(6) and the number of ESM inputs in this mode (8), and is calculated as shown in (18).

(n − Nctrl ) Nctrl Vin (nVout − Nctrl Vin )

L= (18)
n3 Imin f S (Vout − Vin )

3.4. Design on Capacitor

The capacitors in the NISD DC transformer can be categorized into two types: the
internal capacitors of the ESM responsible for transforming energy and the capacitors
aimed at reducing the output voltage’s ripple.
The ESM’s internal IGBTs (Sa , Sb ), which are composed of n in series, regulate the
ripples in ESM’s voltage ΔVM by operating at a switching frequency f M . The capacitance
C M of the ESM’s internal capacitor can be calculated using (19).

Imax (n − Nctrl )
CM = (19)
2nΔVM f M

IGBT groups 1 (S11 , · · · , S1α ) and 2 (S21 , · · · , S2α ) operate at a switching frequency
f S , and their respective operational modes are determined. From this, the output capacitor
current Ic and output voltage’s ripple ΔVout are determined, and the capacitance of the
output capacitor Cout configured in parallel with the output terminal is expressed with (20).

Ic Nctrl
Cout = (20)
2nΔVout f S

4. Case Studies
In this section, to verify the proposed NISD DC transformer, a DC transformer with
an input voltage of 35 kV, an output voltage of 1.5 kV, and a rated power of 1 MW is
configured with 35 ESMs. Moreover, the switching frequency is set to 5 kHz, and the ESM
switching frequency is set to 50 kHz. In addition, the overcurrent limit inductor is 40 μH,
the capacitance of each ESM is 20 mF, the output capacitor is 1.45 mF, and the simulation
parameters are configured as shown in Table 1. The simulation is performed as shown in
Figure 1 using PSCAD/EMTDC. Additionally, the performance of the control method for
the capacitor’s voltage is confirmed through seven cases.

Table 1. Parameters of the NISD DC transformer.

Symbol Quantity Value

Vin Input voltage 35 kV
Vout Output voltage 1.5 kV
P Rated power 1 MW
n Number of ESMs 35
fS Switching frequency 5 kHz
fM ESM switching frequency 50 kHz
L Inductance 40 μH
Cout ESM capacitance 20 mF
CM Output capacitance 1.45 mF

Figure 6a demonstrates that the output voltage is verified at 1.5 kV—with the output
power being 1 MW and the output current at 0.667 kA—to achieve the rated power of
1 MW. Furthermore, ESM’s insertion control variable N is regulated based on operational
modes, with a range of 1 to 34, and the total injected voltage VM of the ESM according to N
is properly operated according to (7). The relationship between the current in the IGBTs
and diodes and the inductor’s current is also validated to perform the same as shown in

119
Energies 2024, 17, 940

Figure 2—(16) and (17). The voltage applied to the IGBTs and diodes is also validated using
a simulation, and the voltages in (11)–(13) are verified.

ȱ
Figure 6. NISD DC transformer operation at resistive load of (a) 100%, (b) 50%, and (c) 10%.

4.1. Operations in Steady State

The NISD DC transformer is verified using constant resistive loads of 100%, 50%, and
10% (Figure 6a–c). The output voltage’s ripples decrease by 3.904%, 1.44%, and 0.651%,
respectively, as the load rate is reduced. Correspondingly, the output power’s ripples
decrease by 7.841%, 4.326%, and 1.304%, while the output voltage remains at 1.5 kV. The
operational modes demonstrated consistency with the theoretical expectations, as the
operating times of operational modes M3 and M6 increased with a decrease in load.

120
Energies 2024, 17, 940

4.2. Operations with Disturbances

The simulation results, depicted in Figure 7a, illustrate normal output voltage control
under the weak input voltage condition while operating under a 100% resistive load. in
addition, the analysis of the figure indicates that the inductor current IL is also influenced
by the input at M2 , when compared to Figure 6a, at a weak input voltage.

Figure 7. NISD DC transformer operation with (a) ripples in input voltage, (b) low-voltage side fault
at 4 ms, (c) transition at 4 ms from 100% resistive to inductive load, and (d) transition at 4 ms from
100% resistive to inductive load during ripples in input voltage.

121
Energies 2024, 17, 940

Next, in order to verify the operation and blocking ability in case of the LVDC side
fault, a LVDC side fault was generated at 4 ms during an operation under 100% load,
and the circuit breaker was opened 2 ms later to clear the fault (Figure 7b). Although, in
this case, the fault current and power from the capacitor, Cout , increase to 54.80 kA and
82.58 MW, respectively. Such increases are a common problem for all inverters without
a special protection scheme for an output capacitor. This might be solved by adding
additional circuits, such as a damping resistor. Meanwhile, the rapid fault current reduction
of the proposed converter based on the reduced capacitor size equates to a shorter discharge
time with the same current compared to the other converters. At this time, all IGBTs were
not damaged by immediately being opened. They cause the current and power to drop to
zero and trigger sequential operations of the circuit breaker. Therefore, the proposed NISD
DC transformer should be shortened. Inside the NISD DC transformer, the operation is
similar to the second operational mode M2 , since all IGBTs are opened, and the voltage,
VD2 , of diode group 2 is applied through the anti-parallel diode of IGBT group 2 and the
diode of the ESM.
To evaluate the performance of the transformer during the rapid load changes of the power
system, an inductance of 0.1 H was transitioned at 4 ms while operating at 100% resistive load
rate (Figure 7c). This representatively reflects the aggregated effect caused by independent
changes of various loads located through different feeders, and the system is stable.
Finally, the inductance was transitioned at 4 ms with high-impedance input voltage,
i.e., the conditions in Figure 7a,c were considered simultaneously. Therefore, a weak
input voltage causes the output current to ripple up to 8.077% due to the discharge of the
capacitor, and the output power to ripple up to 15.728%. Although the inductor current is
affected, it operates between 0.798 kA and −2.819 kA. Thus, it does not cause a problem
for IGBT group 1 (S11 , · · · , S1α ) and 2 (S21 , · · · , S2α ) and diode group 1 (D11 , · · · , D1β )
and 2 (D21 , · · · , D2β ). Also, the increase in the ripple of output voltage is negligible. As
shown in Figure 7d, the NISD DC transformer operates reliably, even in harsh conditions.

5. Conclusions
This article proposed a non-isolated step-down (NISD) DC transformer based on DCM
for the MVDC-LVDC distribution system. To verify the performance of the proposed NISD
DC transformer, normal operation was confirmed with loads of 100%, 50%, and 10%. Addi-
tionally, abnormal situations such as inductive load, LVDC side fault, and instability input
voltage were assumed and simulated using a power system computer-aided design and
electromagnetic transient simulation engine including DC (PSCAD/EMTDC). The proposed
NISD DC transformer successfully operated for the purpose of energy transformation.
This study verified the performance of energy transformation from MVDC to LVDC,
i.e., the MVDC-LVDC distribution system. However, further studies are required to apply
the proposed NISD DC transformer to an HVDC-MVDC system or step-up version of the
proposed transformer, even a bidirectional one.

Author Contributions: Conceptualization, S.H.L.; methodology, S.H.L. and D.C.; software, M.K.;
validation, S.H.L. and M.K.; formal analysis, S.H.L. and M.K.; investigation, S.H.L. and M.K.; re-
sources, S.H.L. and M.K.; data curation, S.H.L. and M.K.; writing—original draft preparation, M.K.;
writing—review and editing, S.H.L.; visualization, M.K.; supervision, S.H.L. and D.C.; project ad-
ministration, S.H.L.; funding acquisition, S.H.L. and D.C. All authors have read and agreed to the
published version of the manuscript.
Funding: This work was supported in part by the National Research Foundation of Korea (NRF)
funded by the Korean Government (MSIT) under grant 2022R1A2C2006688, and in part by the Basic
Science Research Program through the National Research Foundation of Korea (NRF) funded by the
Ministry of Education (2022R1C1C1005975).
Data Availability Statement: The datasets presented in this article are not readily available because
the data are part of an ongoing study.

122
Energies 2024, 17, 940

Conflicts of Interest: All of the authors declare that they have no known competing financial interests
or personal relationships that could have appeared to influence the work reported in this paper.

References
1. Enerdata. Global Energy Trends 2023; Enerdata: Grenoble, France, 2023.
2. Suryadevara, R.; Parsa, L. Full-Bridge ZCS-Converter-Based High-Gain Modular DC-DC Converter for PV Integration with
Medium-Voltage DC Grids. IEEE Trans. Energy Convers. 2019, 34, 302–312. [CrossRef]
3. Lu, Z.; Li, C.; Zhu, A.; Luo, H.; Li, C.; Li, W.; He, X. Medium Voltage Soft-Switching DC/DC Converter with Series-Connected
SiC MOSFETs. IEEE Trans. Power Electron. 2021, 36, 1451–1462. [CrossRef]
4. Jin, H.; Chen, W.; Hou, K.; Shao, S.; Shu, L.; Li, R. A Sharing-Branch Modular Multilevel DC Transformer with Wide Voltage
Range Regulation for DC Distribution Grids. IEEE Trans. Power Electron. 2022, 37, 5714–5730. [CrossRef]
5. IEEE Standard 1709-2018; IEEE Recommended Practice for 1 kV to 35 kV Medium-Voltage DC Power Systems on Ships. IEEE:
New York, NY, USA, 2018.
6. Yao, J.; Chen, W.; Xue, C.; Yuan, Y.; Wang, T. An ISOP Hybrid DC Transformer Combining Multiple SRCs and DAB Converters to
Interconnect MVDC and LVDC Distribution Networks. IEEE Trans. Power Electron. 2020, 35, 11442–11452. [CrossRef]
7. Reddy, P.L.S.K.; Obulesu, Y.P. Design and Development of a New Transformerless Multi-port DC–DC Boost Converter. J. Electr.
Eng. Technol. 2023, 18, 1013–1028. [CrossRef]
8. Zhao, B.; Song, Q.; Li, J.; Sun, Q.; Liu, W. Full-Process Operation, Control, and Experiments of Modular High-Frequency-Link DC
Transformer Based on Dual Active Bridge for Flexible MVDC Distribution: A Practical Tutorial. IEEE Trans. Power Electron. 2017,
32, 6751–6766. [CrossRef]
9. Sheng, J.; Chen, C.; Lu, R.; Li, C.; Xiang, X.; Li, W.; He, X. Control Optimization of Modular Multilevel Resonant DC Converters
for Wide-Input-Range MVdc to LVdc Applications. IEEE Trans. Power Electron. 2022, 37, 5284–5298. [CrossRef]
10. Ferreira, J.A. The multilevel modular DC converter. IEEE Trans. Power Electron. 2013, 28, 4460–4465. [CrossRef]
11. Jovcic, D. Bidirectional, high-power DC transformer. IEEE Trans. Power Deliv. 2009, 24, 2276–2283. [CrossRef]
12. Li, W.; He, X. Review of Nonisolated High-Step-Up DC/DC Converters in Photovoltaic Grid-Connected Applications. IEEE
Trans. Ind. Electron. 2011, 58, 1239–1250. [CrossRef]
13. Stieneker, M.; De Doncker, R.W. Medium-voltage DC distribution grids in urban areas. In Proceedings of the 2016 IEEE 7th
International Symposium on Power Electronics for Distributed Generation Systems (PEDG), Vancouver, BC, Canada, 27–30 June
2016; IEEE: Vancouver, BC, Canada, 2016; pp. 1–7. [CrossRef]
14. Luth, T.; Merlin, M.M.C.; Green, T.C.; Hassan, F.; Barker, C.D. High-Frequency Operation of a DC/AC/DC System for HVDC
Applications. IEEE Trans. Power Electron. 2014, 29, 4107–4115. [CrossRef]
15. Wang, X.; Peng, Y.; Chai, J.; Xia, Y.; Wei, W.; Yu, M. An Ideal DC Transformer for Active DC Distribution Networks Based on
Constant-Transformation-Ratio DABC. IEEE Trans. Power Electron. 2020, 35, 2170–2183. [CrossRef]
16. Zhao, B.; Song, Q.; Liu, W.; Sun, Y. Overview of Dual-Active-Bridge Isolated Bidirectional DC–DC Converter for High-Frequency-
Link Power-Conversion System. IEEE Trans. Power Electron. 2014, 29, 4091–4106. [CrossRef]
17. Alhurayyis, I.; Elkhateb, A.; Morrow, J. Isolated and Nonisolated DC-to-DC Converters for Medium-Voltage DC Networks: A
Review. IEEE J. Emerg. Sel. Top. Power Electron. 2021, 9, 7486–7500. [CrossRef]
18. Xiang, X.; Gu, Y.; Qiao, Y.; Zhang, X.; Li, W.; He, X.; Green, T.C. Resonant Modular Multilevel DC–DC Converters for Both High
and Low Step-Ratio Connections in MVDC Distribution Systems. IEEE Trans. Power Electron. 2021, 36, 7625–7640. [CrossRef]
19. Zhao, B.; Li, X.; Wei, Y.; An, F.; Sun, J.; Liu, G.; Zhao, Y. Modular Hybrid-Full-Bridge DC Transformer with Full-Process Matching
Switching Strategy for MVdc Power Distribution Application. IEEE Trans. Ind. Electron. 2020, 67, 3317–3328. [CrossRef]
20. Du, S.; Wu, B.; Tian, K.; Xu, D.; Zargari, N.R. A Novel Medium-Voltage Modular Multilevel DC–DC Converter. IEEE Trans. Ind.
Electron. 2016, 63, 7939–7949. [CrossRef]
21. Gray, P.A.; Lehn, P.W.; Yakop, N. A Modular Multilevel DC–DC Converter with Flying Capacitor Converter Like Properties. IEEE
Trans. Ind. Electron. 2022, 69, 6774–6783. [CrossRef]
22. Xing, D.; Hu, B.; Li, X.; Cheng, Q.; Wang, J.; Atcitty, S. Comparison of Modular Multilevel Converter Based High Voltage-
Step-Down Ratio DC/DC Converters for Energy Storage System Applications. IEEE J. Emerg. Sel. Top. Power Electron. 2023.
[CrossRef]
23. Falin, J.; Aguilar, A. Maximize power density with three-level buck-switching charges. Analog. Des. J. Q 2021, 1–8.
24. Meynard, T.A.; Foch, H. Multi-level conversion: High voltage choppers and voltage-source inverters. In Proceedings of the PESC
‘92 Record, 23rd Annual IEEE Power Electronics Specialists Conference, Toledo, Spain, 29 June–31 July 1992; 1992; Volume 1,
pp. 397–403. [CrossRef]
25. Stupar, A.; McRae, T.; Vukadinović, N.; Prodić, A.; Taylor, J.A. Multi-Objective Optimization of Multi-Level DC–DC Converters
Using Geometric Programming. IEEE Trans. Power Electron. 2019, 34, 11912–11939. [CrossRef]
26. Ye, Z.; Lei, Y.; Liao, Z.; Pilawa-Podgurski, R.C.N. Investigation of Capacitor Voltage Balancing in Practical Implementations of
Flying Capacitor Multilevel Converters. IEEE Trans. Power Electron. 2022, 37, 2921–2935. [CrossRef]
27. Villar, G.; Alarcon, E. Monolithic Integration of a 3-Level DCM-Operated Low-Floating-Capacitor Buck Converter for DC-DC
Step-Down Conversion in Standard CMOS. In Proceedings of the 2008 IEEE Power Electronics Specialists Conference, Rhodes,
Greece, 15–19 June 2008. [CrossRef]

123
Energies 2024, 17, 940

28. Denniston, N.; Massoud, A.M.; Ahmed, S.; Enjeti, P.N. Multiple-Module High-Gain High-Voltage DC–DC Transformers for
Offshore Wind Energy Systems. IEEE Trans. Ind. Electron. 2011, 58, 1877–1886. [CrossRef]
29. Athab, H.; Yazdani, A.; Wu, B. A transformerless DC-DC converter with large voltage ratio for MV DC grids. IEEE Trans. Power
Deliv. 2014, 29, 1877–1885. [CrossRef]
30. Abbasi, M.; Emamalipour, R.; Cheema, M.A.M.; Lam, J. A Constant-Frequency High-Voltage Gain Resonant Converter Module
with Semiactive Phase-Shifted Voltage Multiplier for MVdc Distribution. IEEE J. Emerg. Sel. Top. Power Electron. 2022, 10,
3603–3616. [CrossRef]
31. Zheng, L.; Kandula, R.P.; Divan, D. Current-Source Solid-State DC Transformer Integrating LVDC Microgrid, Energy Storage,
and Renewable Energy into MVDC Grid. IEEE Trans. Power Electron. 2022, 37, 1044–1058. [CrossRef]
32. Li, L.; Jiang, Z.; Jiang, Y. A SEPIC Flyback DC-DC Converter with GaN Switching Device. J. Electr. Eng. Technol. 2023, 18,
2759–2769. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

124
 energies
Article
An Optimized Switching Strategy Based on Gate Drivers with
Variable Voltage to Improve the Switching Performance of SiC
MOSFET Modules
Jixiang Tan * and Zhongfu Zhou

Department of Electronic and Electrical Engineering, Faculty of Science and Engineering, Swansea University Bay
Campus, Swansea SA1 8EN, UK; z.zhou@swansea.ac.uk
* Correspondence: 979921@swansea.ac.uk

Abstract: This paper proposes an optimized switching strategy (OSS) based on a silicon carbide (SiC)
MOSFET gate driver with variable voltage, which allows simultaneous variations in several different
parameters to optimize the switching performance of semiconductor devices. As a relatively new
device, the SiC MOSFET shines in the field of high power density and high-frequency switching; it has
become a popular solution for electric vehicles and renewable energy conversion systems. However,
the increase in voltage and current slope caused by high switching speeds inevitably increases the
overshoot and oscillation in a circuit and can even generate additional losses. The principle of
this new control strategy is to change the voltage and current in the turn-on and turn-off stages by
changing the gate driver’s voltage. That is, we reduced the drive’s voltage after a certain time delay
and maintained it for a period of time, thus directly controlling the slopes of di/dt and dv/dt. This
study focused on the optimization of the SiC MOSFET by changing the time delay preceding the
decrease in the voltage of the gate driver, analyzing and calculating the optimal time delay before
the decrease in the voltage of the gate driver, and verifying the findings using LTspice simulation
software. The simulated results were compared and analyzed with hard-switching strategies. The
results showed that the proposed OSS can improve the switching performance of SiC MOSFETs.

Keywords: optimal-switching strategy; gate driver circuit; silicon carbide MOSFET; voltage and
current overshoot; LTspice

1. Introduction
Silicon carbide (SiC) MOSFETs offer a range of advantages over silicon-based switches,
including faster switching, higher efficiency, higher operating voltages, and higher tem-
peratures, enabling smaller and lighter designs [1,2]. Thus, significant progress has been
made in the power semiconductor industry. These advantages allow SiC MOSFETs to be
applied in a range of automotive and industrial applications. However, the excessively fast
switching speed of SiC MOSFETs causes high values of di/dt and dv/dt [3]. At the same
time, due to the existence of parasitic inductance and parasitic resistance in circuits, SiC
MOSFETs are prone to overshoots and oscillations in current and voltage during switching,
thus generating additional switching losses. These losses can even cause device damage.
Therefore, reducing or even eliminating the overshoot, oscillation, and electromagnetic
interference (EMI) of silicon carbide MOSFETs during switching is a priority in improving
their working efficiency [4].
Current and voltage overshoots are the biggest obstacles in SiC semiconductor appli-
cations. Stray inductance and parasitic capacitance are the main contributors to current and
voltage overshoots [5,6]. When a SiC MOSFET works at a high frequency, the switching
speed becomes too fast and the di/dt and dv/dt slopes increase. Thus, current and voltage
overshoots are amplified due to stray inductance and parasitic capacitance [7]. When the
voltage and current overshoots exceed the breakdown voltage of the SiC MOSFET and the

Energies 2023, 16, 5984. https://doi.org/10.3390/en16165984 125 https://www.mdpi.com/journal/energies

Energies 2023, 16, 5984

maximum recovery current of the freewheeling diode, the SiC is destroyed. However, the
parasitic elements in the actual circuit cannot be eliminated [8–11]; so, the overshoot can
only be eliminated by other means.
Solving the problems of overshoots and oscillations is key to improving the efficiency
of SiC MOSFETs.
In the traditional approach, slowing the switching speed in the case of high gate
resistance [12] can alleviate the SiC MOSFET’s overshoots and oscillations. However, this
solution means the power losses become greater with longer switching times [6]. Adding an
RC snubber circuit is also a common method [13,14] to eliminate overshoot and oscillation
problems. Although highly efficient snubber circuits have been proposed, snubber circuits
affect the overall efficiency of the system. In particular, extra devices increase the power
loss of the overall circuit. Active gate driver (AGD) solutions have been developed to
increase the efficiency of power devices. The main advantages of these gate drivers are
reduced oscillations and overshoots. However, the extra components not only increase the
volume of the SiC system but also increase the circuit’s power loss [15–21]. In [21], a new
active gate driver was proposed that could effectively suppress overshoots and oscillation
and reduce losses without compromising EMI. The main strategy of the proposed AGD was
to reduce the current and voltage slope by reducing the gate driver’s voltage. However, its
main disadvantages were that the process of calculating the parameters of the transformer
was complicated, an additional printed circuit board was required, and the implementation
of the circuit was complicated and expensive. As such, configurable digital gate drivers
(DGDs) are the latest technological development in this sphere [22]. The strengths of
DGDs are their programmability, support for a wide variety of power devices, and ease
of use [23]. A key element of this technology is the ability to configure the turn-on and
turn-off processes, which provides a series of steps to control the voltage level at specific
times [24–28]. This allows designers to digitally configure the turn-on and turn-off curve
through software without requiring changes to the hardware.
In this paper, an optimized switching strategy based on gate drivers with variable
voltage (OSS) was proposed to improve the switching performance of SiC MOSFETs. This
switching strategy was based on and optimized for an AGD in [21] and a DGD in [22].
There is no doubt that the AGD performed well in [21]. However, the gate driver circuits
were too complex and too large. The ease of operation of the DGD in [22] is very attractive.
Therefore, this article aimed to combine the advantages of the voltage gate drivers of
AGDs and DGDs in waveforms to further optimize the algorithm and control and reduce
overshoots and oscillations in SiC MOSFETs through an OSS. Finally, we used the LTspice
simulation software to verify the feasibility of our switching strategy. Compared with
AGDs and DGDs, an OSS can control more variables to achieve more precise and optimal
circuit control. Due to space limitations, this article focused on the impact of delay time
on switching performance and lists the formula to calculate the delay time for a given
calculated decrease in voltage. At the same time, the decrease in the voltage and the time
of this decrease were also considered. The development potential of the OSS is significant.
Compared with other switching strategies, which can only change one or two variables, the
OSS can control three variables simultaneously, namely, the delay time before a decrease
in voltage, the decrease in voltage itself, and the time duration of the decrease in voltage.
More variables mean better control performance. However, due to various limitations, this
paper only considered the influence of the delay time before a decrease in voltage on switch
performance and analyzed the relationship between voltage drop and switch performance.
In our next work, we will produce more simulations and experimental results to prove the
advanced nature of the OSS.
This paper is organized as follows. Sections 2 and 3 present the control principle,
circuit model, and related formulations of the proposed switching strategy. In Section 4, the
simulation setup and simulation results are introduced and analyzed. Finally, in Section 5,
the conclusions and a discussion of this simulation are given.

126
Energies 2023, 16, 5984

2. Operation Principle of the OSS

The working principle of the OSS is to improve the switching performance of the SiC
MOSFET by controlling the voltage drop delay time, voltage drop, and voltage drop time.
Compared with traditional AGDs and DGDs, the advantage of an OSS is that it has more
control parameters, which means that the semiconductors can be optimally controlled. In
this section, the general schematic circuit and working principle of the proposed OSS will
be introduced in detail. Figure 1 shows a circuit schematic of the SiC MOSFET and Figure 2
shows a timing chart of the SiC MOSFET with an optimized switching strategy.

Figure 1. Circuit schematic of the SiC MOSFET under test.

Figure 2. Timing chart of the SiC MOSFET with the OSS.

2.1. Operation Modes

As shown in Figure 1, the OSS circuit was mainly composed of three parts: the gate
drive circuit, silicon carbide MOSFET, and DC bus power supply. The model of the SiC
MOSFET, including junction capacitances (Cgd , Cgs , Cds ) and module intrinsic parasitic
inductances (Ls, Lg, Ld), is also depicted in Figure 1.

127
Energies 2023, 16, 5984

In [21], the gate drive voltage used by the AGD was 20 V/−5 V; also, 15 V and 0 V
were used as dropped voltages to optimize the switching performance of the SiC MOSFET.
Because it used a switching bridge, it could generate four gate drive voltages (i.e., 20 V, 15 V,
0, and −5 V) from two drive power supplies (20 V and 5 V) through different combinations.
In this design, the gate drive voltage waveform design of the AGD was still used. The
purpose was to control variables and only explore the influence of the voltage drop delay
time (td ) on the performance of SiC MOSFETs.

2.2. Operation Principle

Figure 2 shows the timing chart of the SiC MOSFET with an optimized switching
strategy. The whole process could be split into two steps: turn on and turn off. The blue
interval represents the gate driver voltage in the stage of the voltage drop delay time.

2.2.1. Turn-On Stage

When the pulse width modulation (PWM) signal changed from low to high at t0 , a
high value gate drive voltage VGG of 20 V was applied to the SiC MOSFET. Then, a high
gate current was generated to charge the input capacitance Ciss = Cgs + Cgd with Rg , which
meant the VGS started to increase. When the VGS reached the threshold voltage VTH at t1 ,
the ID began to conduct. After a time delay td1 , the lower VGG was activated and a lower
gate current was generated because the VGG was reduced to 15 V.
In this interval, ID continued to rise and a current peak value IOS appeared due to the
freewheeling diode effect. The VGS reached the Miller plateau voltage and stopped rising.
At the same time, the VDS started to drop. After the voltage fall time tvd1 , the VGS came out
of the Miller plateau and continued to rise. Finally, the SiC MOSFET was fully turned on at
t4 . From [21], the current slope of ID and the voltage slope of the VDS during the turn-on
stage could be expressed as:

dID VGG − VTH − ID /gm

= (1)
dt (Ciss ·Ron )/gm + Ls

dVDS V − VMiller
= − GG (2)
dt CGD ·Ron
where gm is the transconductance of the SiC MOSFET and Cgd is the gate-drain capacitance
of the SiC MOSFET.

2.2.2. Turn-Off Stage

When the signal from the voltage source changed from high to low at t5 , a negative
voltage (i.e., −5 V) was generated to discharge the Ciss with Rg . The VGS started to drop
from 20 V until it reached the Miller voltage at t6 . The VDS rose rapidly and caused a
voltage overshoot VOS due to stray inductance. After a delay time td2 , the VGG increased
to 0 V and remained there until ID was fully turned off at t8 . After t8 , the gate voltage
stabilized at −5 V to ensure that the SiC MOSFET was always in the off state. The voltage
slope could be calculated as:

dID VTH + (ID /gm ) − VGG

= gm · (3)
dt (Ciss ·Roff )/gm + Ls

dVDS V − VMiller
= GG (4)
dt CGD ·Roff
According to (1)–(4), changing the instantaneous slope of current and voltage could
be achieved by changing the VGG . In the turn-on stage, the change was reducing the gate
diver voltage while in the turn-off stage, the change was increasing the gate diver voltage.

128
Energies 2023, 16, 5984

3. Calculation for Voltage Drop Delay Time

An OSS can change the switching performance through unlimited control of the
voltage drop delay time, voltage drop, and voltage drop time, aiming to obtain the best
data for the switching performance through countless permutations and combinations.
However, this is difficult to carry out. In fact, the best data can be derived by equation. The
main research direction of this paper was to explore the influence of the voltage drop delay
time on the switching performance and to explore the influence of different voltage drops
and different voltage drop times on the switching performance. Thus, the content of this
chapter focuses on deriving the best voltage drop delay time by equation.

3.1. Equivalent Slope for Current and Voltage

The slope of the current and voltage is an important factor affecting the overshoot and
oscillation during the switching process of the SiC MOSFET. It can be seen from Figure 2
that the IOS in the turn-on stage and the VOS in the turn-off stage were obvious and large.
However, the slope of the voltage and current in the SiC MOSFET with the OSS method
was not constant because of the variable gate drive voltage; so, the equivalent slope was
proposed and applied in the calculation, with the aim of writing the voltage drop delay
times, td1 and td2 , into the equation.
In Figure 3, when the VGG was 20 V, the slope of the current was dID /dt1 ; when
the VGG was 15 V, the slope of the current was dID /dt2 . Similarly, dVDC /dt1 and dVDC
/dt2 represented the value of the voltage slope under the VGG of −5 and 0 V. In addition,
ton represented the increase in the current ID and toff represented the increase in the
voltage VDS ; ton1 and toff1 were the rising times under dID /dt1 and dVDC /dt1 , respectively.
Moreover, dID /dt represented the actual current slope with OSS and dVDC /dt was the
current slope with OSS, in actuality. Additionally, dID /dteq was the equivalent slope at the
turn-on stage and dVDC /dteq represented the equivalent slope at the turn-off stage; ton2 and
toff2 were the actual running times for the turn-on stage and turn-off stage. In [21], the slope
of the current was used in both the turn-on and turn-off stages because it was convenient
for the subsequent calculations. In this paper, in order to fit reality, the voltage slope was
used in the turn-off stage. According to the equivalent principle shown in Figure 3, the
following equation could be obtained:
 
IL = |dID /dt1 |·ton + |dID /dt1 |·(td1 − ton ) + |dID /dt2 |·(ton2 − td1 ) = dID /dteq ·ton2 (5)

and the equivalent rising time ton2 of the current can be expressed as:

ton2 = (IL − (|dID /dt1 | − |dID /dt2 |)·td1 )/|dID /dt2 | (6)

Figure 3. Schematic diagram of the equivalent slope. (a) Current slope on the turn-on stage. (b) Volt-
age slope on the turn-off stage.

129
Energies 2023, 16, 5984

Thus, according to (5) and (6), the value of the equivalent current slope at the turn-on
stage can be expressed as:
 
dID /dteq  = IL ·|dID /dt2 |/(IL − (|dID /dt1 | − |dID /dt2 |)·td1 ) (7)

Similarly, at the turn-off stage, the equations of toff2 and voltage slope are:
 
VDC = |dVDS /dt1 |·toff + |dVDS /dt1 |·(td2 − toff ) + |dVDS /dt2 |·(toff2 − td2 ) = dID /dteq ·toff2 (8)

toff2 = (VDC − (|dVDS /dt1 | − |dVDS /dt2 |)·td2 )/|dVDS /dt2 | (9)

 
dVDS /dteq  = VDC ·|dVDS /dt2 |/VDC − (|dVDS /dt1 | − |dVDS /dt2 |)·td2 (10)

3.2. Power Losses in the Turn-On and Turn-Off Stages

The turn-on and turn-off losses of the power device could be calculated by (11) and
(12), which were derived in []:

EON = Eon,dID /dt + Eon,dVDS /dt + EIrr + ELs (11)

EOFF = Eoff,dID /dt + Eoff,dVDS /dt + ELs (12)

where ELs is the energy generated by the current passing through LS and EIrr is the energy
loss of the device generated by the reverse recovery effect, which can be calculated by
(14) [29]:
1
ELs = ·LS ·Iloop 2 (13)
2
 $ 
EIrr = IL · Qrr /|dID /dt| + Qrr ·VDC ·(1 − σ) (14)

where Lloop is the stray inductance in the power loop, Qrr is the reverse recovery charge,
and σ is the overshoot ratio, which can be defined as:

VOS Lloop ·|dID /dt|

σ= = (15)
VDC VDC

During the turn-on transient processes, the energy losses during the current and
voltage changes can be expressed as:

1 IL
Eon,dID /dt = ·IL · VDC − Lloop ·|dID /dt | · (16)
2 |dID /dt|

1 VDC
Eon,dVDS /dt = ·IL ·VDC · ·(1 − σ)2 (17)
2 |dVDS /dt|
Therefore, according to Formulas (11) and (13)–(17), the energy loss during the turn-on
stage can be expressed as:
! " ! "
VDC ·IL 2 1 + (1 − σ)2 Qrr 1
EON = · + IL · + Qrr ·VDC ·(1 − σ) − ·LS ·Iloop 2 (18)
2 |dID /dt| |dID /dt| 2

Similarly, according to (12), the energy loss can be expressed as:

IL ·VDC 2 1 + (1 + σ)2 1
EOFF = · + · LS ·Iloop 2 (19)
2 |dVDS /dt| 2

130
Energies 2023, 16, 5984

3.3. Calculation of the Voltage Drop Delay Time

In order to simplify the calculation model, it was assumed that the change of power
consumption was only related to overshoot and was positively correlated. Therefore,
during the turn-on stage of the SiC MOSFET, the change in power consumption is positively
correlated with the change in current overshoot [21]. Similarly, in the turn-off phase, the
change of power consumption is positively correlated with the change of voltage overshoot.
The formula can be expressed as:

|EON (td1 ) − EON (0)| |I (t ) − IOS (0)|

= OS d1 (20)
|EON (ton2 ) − EON (0)| |IOS (ton2 ) − IOS (0)|

|EOFF (td2 ) − EOFF (0)| |VOS (td2 ) − VOS (0)|

= (21)
|EOFF (toff2 ) − EOFF (0)| |VOS (toff2 ) − VOS (0)|
In the turn-on phase, the most important factor affecting power consumption is the
overshoot of the current; so, Equation (21) can be abbreviated as:

IL 2 ·VDC 1 − σ
EON = · (22)
2 |dID /dt|

Because the main cause of current overshoot is the reverse recovery effect [21], the
current overshoot can be expressed as:
$
IOS = Irr = Qrr ·|dID /dt| (23)

According to (25) and (26), (11) can be expressed as:

  &
dID /dteq  Q ·|dI /dt1 |
= $ rr  D  (24)
|dID /dt1 | Q ·dI /dt 
rr D eq

Similarly, in the turn-off stage, the formula can be obtained:

VDC 2 ·IL 1+σ

EOFF = · (25)
2 |dVDS /dt|
 
dVDS /dteq  Lloop ·|dVDS /dt1 |
=   (26)
|dVDS /dt1 | Lloop ·dVDS /dteq 
According to (1)–(4), (7), and (24), an equation expressed by td1 could be obtained.
Similarly, according to (1)–(4), (10), and (26), an equation expressed by td2 could be obtained.
Therefore, the optimal voltage drop delay times, td1 and td2 , could be calculated.

4. Simulation Verification
In order to evaluate the performance of the proposed OSS, a double-pulse simulation
test of the SiC MOSFET was performed using the software LTspice. The power device
used in the test was CREE’s 1.2 KV/115 A SiC MOSFET module (C3M0016120D). In the
simulation, the tested module was first modeled in LTspice to ensure that the simulation
results would be closer to the experimental data and then the circuit was built and simulated.
The detailed parameters of the tested modules in the experiment are shown in Table 1.
The comparison of the OSS’s performance under different time delays was simulated to
verify whether the calculated time delay was the best. To further compare the performance
of the OSS, it was compared with hard switching (HS) and a digital gate driver (DGD) from
Agileswitch.

131
Energies 2023, 16, 5984

Table 1. Parameters of the SiC MOSFET C3M0016120D (TCJ = 25 ◦ C).

Symbol Parameter Value Unit

VDS Drain—Source Voltage 1200
VGS Gate—Source Voltage −8/+19 V
VTH Gate Threshold Voltage 2.5
Ciss Input Capacitance 6085
Coss Output Capacitance 230 pF
Crss Reverse Transfer Capacitance 13
tr Rise Time 28 ns
tf Fall Time 27 ns
Drain-Source On-State
RDS(on) 16 mΩ
Resistance
RG(int) Internal Gate Resistance 2.6 Ω
ID Continuous Drain Current 115 A
Qrr Reverse Recovery Charge 604 nC
gm Transconductance 53 S

4.1. SiC MOSFET Device

In order to facilitate future experiments based on this simulation, this simulation used
the existing silicon carbide semiconductor module in the laboratory as the test equipment.
The detailed parameters of the tested modules in the experiment are shown in Table 1.
Figure 4 shows the device.

Figure 4. C3M0016120D device.

4.2. Simulation Modeling

To improve the simulation, we chose LTspice as the simulation software. As shown
in Figure 5, and according to Figure 1, a simulation circuit was built. The main purpose
of this simulation was to verify the effect on the switching performance of the OSS and
compare it with other switching strategies. It was firstly particularly important to verify
whether it was necessary to study the OSS in depth. A simulation can obtain more accurate
results and save time. Once the excellent switching performance of the OSS was confirmed,
the next step was to build and experiment with actual circuits. According to (24), (26), and
data from Table 1, we calculated an optimized delay time td1 of 21 ns and td2 of 41 ns; tvd1
and tvd2 used the rise time (tr ) and fall time (tf ) in the device data sheet for the simulation,
respectively.
Table 2 shows the parameters of some components in the analog circuit. Among them,
C1 and C4 represent Cds , C2 and C5 represent Cdg , and C3 and C6 represent Cgs . In order
for the oscillation in the simulation results to be closer to the actual situation, LS , LD , and
LG were adjusted.

132
Energies 2023, 16, 5984

Figure 5. Simulation circuit in Ltspice.

Table 2. Parameters of the SiC MOSFET and associated circuit in LTspice.

Parameters Value
C1 , C4 227 pF
C2 , C5 13 pF
C3 , C6 6072 pF
C7 100 μF
LD 100 nH
LG 20 nH
LS 1 nH
Lload 60 μH
RDS(on) 1Ω
RG 2.6 Ω
VBUS 500 V
VTC 25 V

Because the double-pulse simulation of the SiC MOSFET was carried out using the
software LTspice, it was necessary to simulate the double-pulse signal, as shown in Figure 6.
Because of the convenience of the simulation software, the voltage waveform from the
voltage source could be edited directly. When the SiC MOSFET was simulated with the
OSS, the waveform could also be adjusted directly at the voltage source, which required
corresponding devices in the actual experiments.

Figure 6. Switching waveforms for the SiC MOSFET with double-pulse switching.

133
Energies 2023, 16, 5984

4.3. Simulation Results

In order to verify the feasibility of the proposed OSS, the simulation was run with the
gate resistance 2.6 Ω, both at the turn-on stage and turn-off stages; the dc bus voltage was
500 V and the drain current was 30 A. The voltage drop delay times td1 of 21 ns and td2 of
41 ns were calculated by Equations (11) and (12). Meanwhile, as visible from Table 1, the
tvd1 was 28 ns and the tvd2 was 27 ns.
It can be seen in Figure 7a that when the turn-on signal arrived, first, a high VGG (i.e.,
20 V) was applied to the gate terminal. After the delay time td1 , which was calculated to be
21 ns, the IDS started to rise and the VGG dropped to 15 V. After the delay time tvd1 (i.e.,
28 ns), the high VGG restarted to charge the SiC MOSFET. At the turn-off stage, the SiC
MOSFET was first discharged by −5 V. After the delay time td2 , which was calculated to be
41 ns, the VDS started to rise from 0 V and the VGG became 0 V. After the delay time tvd2
(i.e., 27 ns), the VDS reached the peak value and the negative VGG was discharged again.

(a) (b)

Figure 7. Switching waveforms for the SiC MOSFET with the OSS. (a) Turn-on stage, (b) turn-off
stage.

As mentioned in Section 2, the voltage drop delay times, td1 and td2 , are very important
because they can affect the switching performance of the SiC MOSFET. In order to verify
the relationship between the delay time and the performance of the switch, and to further
the accuracy of the calculation results in the previous chapter, more experiments were
carried out. As shown in Figure 8, around the calculation result (i.e., td1 = 21 ns td2 = 41 ns)
at an interval of 10 ns, a total of four other data points were taken before and after each
transition and were simulated. Then, the simulation results were compared.
For the turn-on transition, because the calculation result of td1 was 21 ns, the five
testing data points were 11 ns, 21 ns, 31 ns, 41 ns, and 51 ns, respectively. The reason td1
was not set to 0 was that this waveform was equal to the switching strategy of the digital
gate driver. It can be seen from Figure 8a that there was no obvious difference and the VDS
and IDS waveforms generated by the five different td1 were not notably different. When td1
was 11 ns, the overshoot of the IDS was the smallest but the overshoot produced by the VDS
was the largest.
For the turn-off transition, since the calculation result of td2 was 41 ns, the five test
data points were 31 ns, 41 ns, 51 ns, 61 ns, and 71 ns, respectively. It can be seen from
Figure 8b that there was an obvious difference and the VDS and IDS waveforms produced
by the five different td1 showed little difference. When td1 was 41 ns, the overshoot and
oscillation of the IDS were the smallest; the overshoot and oscillation of the VDS were also
the smallest.
As the delay time increased or decreased, the calculation results in both cases were
the best. However, a trade-off between the VDS and IDS could be achieved when a td1 of
about 21 ns at turn on and td2 of about 41 ns at turn off were, respectively, applied.

134
Energies 2023, 16, 5984

(a) (b)

Figure 8. The switching performance of the SiC MOSFET with different gate voltage drop time delays.
(a) Turn-on stage, (b) turn-off stage.

4.4. Comparison with Other Switching Strategies

After determining the optimal time delay of the OSS, it was necessary to compare the
OSS with other switching strategies. First, we compared the OSS with the hard-switching
(HS) strategy to confirm its advantages and improvements. Then, we compared it with a
digital gate driver (DGD), which is widely used in teaching and research, to further verify
the advantages of the OSS. A detailed experimental comparison of these different switching
strategies follows.

4.4.1. Compared to the Hard-Switching Strategy

In order to demonstrate the excellent switching performance of the OSS in terms of
overshoot and oscillation, more simulations were performed and the results were compared
with the switching performance of hard switching. It can be seen from Figure 9a that during
the turn-on stage, the peak value of the IDS dropped from 70.04 A to 60.23 A, a decrease of
14%, and the oscillation was also significantly reduced. However, the overshoot of the VDS
increased, which will increase the power loss; although, this will not affect the peak value
of the VDS .

(a) (b)

Figure 9. Switching performance comparison of the SiC MOSFET between HS and the OSS. (a) Turn-
on stage, (b) turn-off stage.

Unlike in the turn-on process, the OSS successfully reduced the oscillation and peak
value during the turn-off process, not only regarding the IDS but also the VDS . It can be

135
Energies 2023, 16, 5984

seen from Figure 9b that the overshoot of the VDS reduced from 704.56 V to 577.52 V, a drop
of 18.03%.
Combining the two processes of turn on and turn off, it can be clearly seen that the OSS
effectively reduced the overshoot and oscillation of the IDS and VDS in the SiC MOSFET
switch, thereby reducing the overshoot and power consumption of the entire circuit.

4.4.2. Compared to Digital Gate Driver

The digital gate drive system from Agileswitch is a product with a wide range of
applications. It can control the voltage drop and voltage drop time of the gate voltage
without limitation to affect the switching performance of semiconductors. The control
strategy of the DGD is also very advanced. However, the existing DGD control strategy is
still unable to design more complex gate drive waveforms, especially the voltage drop after
a delay time, like an AGD. Therefore, it was necessary to compare the OSS with existing
digital gate driver strategies.
Figure 10 shows the VDS and IDS comparison of the OSS and DGD at the same turn-on
time. Compared with the hard-switching strategy, the overshoot of the DGD was reduced
by 19%, which was better than the OSS’s 14%. Although the DGD was slightly better at
reducing the IDS overshoot and oscillation, the overshoot of the VDS was larger, which
seriously increased the power losses of the circuit. Although the two switching strategies
were both effective in reducing overshoot, they increased power consumption and were
low in cost performance. This is why the first generation of DGDs cannot adjust the pulse
shape during turn on.
Figure 10 shows the comparison of the VDS and IDS for the OSS and DGD at the same
turn-off time. Compared with the hard-switching strategy, the DGD had no advantage and
even increased power consumption due to the longer switch-off process.
In order to better compare two different switching strategies, the closing time of the
DGD was extended from 68 ns to 88 ns and was then simulated and compared. The new
waveform comparison is in Figure 11. Compared with Figure 10b, the DGD effectively
suppressed the overshoot and oscillation of the VDS and the overshoot was reduced by 11%.
Compared with the OSS, the DGD was inferior. Continuing to extend the turn-off time of
the DGD will make the optimization more effective; but, this will inevitably increase power
consumption and reduce the frequency of the semiconductor device.

(a) (b)

Figure 10. Switching performance comparison of the SiC MOSFET between the DGD, HS, and OSS.
(a) Turn-on stage, (b) turn-off stage.

136
Energies 2023, 16, 5984

Figure 11. Turn-off switching performance comparison of the SiC MOSFET between the DGD, HS,
and OSS when the the DGD time delay was 88 ns.

4.5. Comparison with Different Drop Voltages

The OSS can become more complex and diverse. The previous chapter mainly consid-
ered the influence of the voltage time delay on switching performance; this section attempts
to study different voltage drops.
It can be seen from Figure 12 that, whether in the on state or the off state, the larger
the voltage drop, the less overshoot. Because the increase in the voltage drop reduces the
slope of the current and voltage, this proves that the formula in Section 3 is valid. However,
during the turn-off process, with the increase in the voltage drop, the oscillation of the
voltage and current was obviously increased.

(a) (b)

Figure 12. Switching performance of the SiC MOSFET with different drop voltages. (a) Turn-on stage,
(b) turn-off stage.

4.6. Comparison with Different Voltage Drop Times

In the previous simulation, in order to simplify the complexity of the simulation, the
voltage drop time (tvd ) used the rise time (tr ) and fall time (tf ) in the device datasheet.
However, in the OSS, the voltage drop time is also an important parameter, similar to the
delay time and voltage drop. Therefore, in this subsection, the effect of the voltage drop
time on the switching performance was studied by simulation.
For the turn-on transition, because the tr was 28 ns, five testing data points were
chosen: 18 ns, 28 ns, 38 ns, 48 ns, and 58 ns. It can be seen from Figure 13a that different
voltage drop times did not have much influence on the switching performance at the
turn-on stage.

137
Energies 2023, 16, 5984

(a) (b)

Figure 13. Switching performance of the SiC MOSFET with different voltage drop times. (a) Turn-on
stage, (b) turn-off stage.

For the turn-off transition, because the tf was 27 ns, five testing data points were
chosen: 17 ns, 27 ns, 37 ns, 47 ns, and 57 ns. It can be seen from Figure 13b that in the
turn-off stage, different voltage drop times had an impact on the switching performance.
When the voltage drop time was 27 ns, the switching achieved the best performance. As the
voltage drop time increased, the suppression effect of the OSS on voltage oscillations began
to weaken. On the contrary, as the voltage drop time decreased, the voltage oscillation was
suppressed; but, a larger reverse current was generated.

4.7. Power Loss Comparison

To determine whether a switching strategy is good or bad, in addition to observing
whether it effectively reduces overshoot and oscillation, another important criterion is
whether power consumption increases. Due to the advanced nature of LTspice, the product
of current and voltage can be quickly integrated directly from the simulated circuit. Table 3
records the power losses from simulation models in one turn-on and turn-off round.

Table 3. Power losses for the simulations in LTspice.

Subsection Switching Strategy Power Losses (μJ)

HS 494.43
OSS 520.21
4.4 DGD 704.52
DGD (extended) 681.03
Voltage drop 7 V 675.11
4.5 Voltage drop 6 V 604.05
Different voltage drop Voltage drop 5 V 536.85
Voltage drop 4 V 473.72
Voltage drop time −10 ns 491.56
Voltage drop time 0 ns 539.26
4.6
Voltage drop time +10 ns 582.92
Different voltage drop time
Voltage drop time +20 ns 606.19
Voltage drop time +30 ns 686.85
Voltage drop delay time −10 ns 559.25
Voltage drop delay time 0 ns 518.93
4.3
Voltage drop delay time +10 ns 582.87
Different voltage drop delay time
Voltage drop delay time +20 ns 658.26
Voltage drop delay time +30 ns 743.88

It can be seen from Table 3 that, compared with HS and the DGD, the power consump-
tion of the OSS was slightly higher than that of HS. However, combined with Figure 11, the
optimization effect of the OSS remained obvious.

138
Energies 2023, 16, 5984

In the comparison of different voltage drops, it can be concluded that the smaller the
voltage drop, the smaller the power consumption. However, combined with Figure 12, it
can be seen that the smaller the voltage drop, the smaller the suppression effect of the OSS
on the circuit overshoot. Similarly, in the comparison of different voltage drop times, it can
be concluded that the shorter the voltage drop time, the smaller the power consumption.
However, combined with Figure 13, it can be seen that the smaller the voltage drop, the
smaller the effect of the OSS on circuit oscillation suppression.
Finally, in the comparison of different voltage drop delay times, it can be concluded
that when the voltage drop time is the calculated value, the power consumption is close to
the minimum.

5. Conclusions and Discussion

In this paper, an optimal-switching strategy was proposed for improving the switching
performance of high-power SiC MOSFETs under hard-switching conditions. In addition,
considering the trade-off between switching loss and switching overshoot, the voltage drop
delay time was analyzed and calculated. By optimizing the voltage drop delay time of
the turn-on and turn-off stages, the OSS can effectively minimize overshoot and suppress
oscillation. The simulation results showed that the OSS can reduce the current overshot
at the turn-on stage and minimize the voltage overshoot at the turn-off stage. In addition,
compared with the hard-switching strategy, the current overshoot in the turn-on phase
decreased by 14% and the voltage overshoot in the turn-off phase decreased by 18.03%.
This showed that the OSS achieved a more comprehensive control strategy and better
switching performance. Compared with digital gate drivers, although each has its own
advantages, the OSS is more advanced and suffers less power losses.
The next step in this research is to make the actual circuit and test it. The contribution
of this paper is limited due to simulation constraints. The optimization of the OSS in
this paper was limited to the control voltage drop delay time; the control strategy could
not be fully tested. Only the influence of a single variable of the OSS on the switching
performance was tested. In future studies, multivariable control, such as delay time, voltage
drop, voltage drop time, etc., could be tested in an actual circuit and jointly controlled
to optimize semiconductor switching performance. However, we can confirm that the
proposed OSS is promising compared to other methods.

Author Contributions: Methodology, J.T.; validation, J.T.; writing—original draft preparation, J.T.;
writing—review and editing, Z.Z. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Data Availability Statement: The data are not publicly available.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Castellazzi, A.; Fayyaz, A.; Romano, G.; Yang, L.; Riccio, M.; Irace, A. SiC power MOSFETs performance, robustness and
technology maturity. Microelectron. Reliab. 2016, 58, 164–176. [CrossRef]
2. Biela, J.; Schweizer, M.; Waffler, S.; Kolar, J.W. SiC versus Si Evaluation of Potentials for Performance Improvement of Inverter
and DC-DC Converter Systems by SiC Power Semiconductors. IEEE Trans. Ind. Electron. 2011, 58, 2872–2882. [CrossRef]
3. Oswald, N.; Anthony, P.; McNeill, N.; Stark, B.H. An Experimental Investigation of the Tradeoff between Switching Losses and EMI
Generation with Hard-Switched All-Si, Si-SiC, and All-SiC Device Combinations. IEEE Trans. Power Electron. 2014, 29, 2393–2407.
[CrossRef]
4. Hazra, S.; Ankan, D.; Cheng, L. High Switching Performance of 1700-V, 50-A SiC Power MOSFET Over Si IGBT/BiMOSFET for
Advanced Power Conversion Applications. IEEE Trans. Power Electron. 2016, 31, 4742–4754.
5. Chen, H.; Divan, D. High speed switching issues of high power rated silicon-carbide devices and the mitigation methods. In
Proceedings of the IEEE Energy Conversion Congress and Exposition (ECCE), Montreal, QC, Canada, 20–24 September 2015.
6. Wang, J.; Chung, H.S.; Li, R.T. Characterization and Experimental Assessment of the Effects of Parasitic Elements on the MOSFET
Switching Performance. IEEE Trans. Power Electron. 2013, 28, 573–590. [CrossRef]

139
Energies 2023, 16, 5984

7. Sayed, H.; Zurfi, A.; Zhang, J. Investigation of the effects of load parasitic inductance on SiC MOSFETs switching performance. In
Proceedings of the IEEE International Conference on Industrial Technology (ICIT), Toronto, ON, Canada, 22–25 March 2017.
8. Nayak, P.; Hatua, K. Modeling of switching behavior of 1200 V SiC MOSFET in presence of layout parasitic inductance. In
Proceedings of the IEEE International Conference on Power Electronics, Drives and Energy Systems (PEDES), Trivandrum, India,
14–17 December 2016.
9. Liu, T.; Ning, R.; Wong, T.Y.; Shen, Z.L. Modeling and analysis of SiC MOSFET switching oscillations. IEEE J. Emerg. Sel. Top.
Power Electron. 2016, 4, 747–756. [CrossRef]
10. Yang, B.; Ge, Q.; Zhao, L.; Zhou, Z.; Cui, D. Influence of parasitic elements of busbar on the turn-off voltage oscillation of SiC
MOSFET half-bridge module. In Proceedings of the IECON 2017—43rd Annual Conference of the IEEE Industrial Electronics
Society, Beijing, China, 29 October–1 November 2017.
11. Fabre, J.; Ladoux, P.; Piton, M. Characterization and implementation of Dual-SiC MOSFET modules for future use in traction
converters. IEEE Trans. Power Electron. 2015, 30, 4079–4090. [CrossRef]
12. Rujas, G.; Lopez, V.M.; Mir, L.; Nieva, T. Gate driver for high power’ SiC modules: Design considerations, development and
experimental validation. IET Power Electron. 2018, 11, 977–983. [CrossRef]
13. Liang, M.; Li, Y.; Chen, Q.; Lu, Y. Research on an improved DC-side snubber for suppressing the turn-off overvoltage and
oscillation in high speed SiC MOSFET application. In Proceedings of the IEEE Energy Conversion Congress and Exposition
(ECCE), Cincinnati, OH, USA, 1–5 October 2017.
14. Jahdi, S.; Alatise, O.; Ortiz Gonzalez, J.A.; Bonyadi, R.; Ran, L.; Mawby, P. Temperature and switching rate dependence of crosstalk
in Si-IGBT and SiC power modules. IEEE Trans. Ind. Electron. 2016, 63, 849–863. [CrossRef]
15. Zhang, Z.; Wang, F.; Tolbert, L.M.; Blalock, B.J. Active Gate Driver for Crosstalk Suppression of SiC Devices in a Phase-Leg
Configuration. IEEE Trans. Power Electron. 2014, 29, 1986–1997. [CrossRef]
16. Riazmontazer, H.; Mazumder, S.K. Optically Switched-DriveBased Unified Independent dv/dt and di/dt Control for Turn-Off
Transition of Power MOSFETs. IEEE Trans. Power Electron. 2015, 30, 2338–2349. [CrossRef]
17. Yamaguchi, K.; Sasaki, Y.; Imakubo, T. Low loss and low noise gate driver for SiC-MOSFET with gate boost circuit. In Proceedings of
the IECON 2014—40th Annual Conference of the IEEE Industrial Electronics Society, Dallas, TX, USA, 29 October–1 November 2014.
18. Riazmontazer, H.; Rahnamaee, A.; Mojab, A.; Mehrnami, S.; Mazumder, S.K.; Zefran, M. Closed-loop control of switching
transition of SiC MOSFETs. In Proceedings of the IEEE Applied Power Electronics Conference and Exposition (APEC), Charlotte,
NC, USA, 15–19 March 2015.
19. Nayak, P.; Hatua, K. Active gate driving technique for a 1200 V SiC MOSFET to minimize detrimental effects of parasitic
inductance in the converter layout. IEEE Trans. Ind. Appl. 2017, 54, 1622–1633. [CrossRef]
20. Paredes, A.; Ghorbani, H.; Sala, V.; Fernandez, E.; Romeral, L. A new active gate driver for improving the switching performance
of SiC MOSFET. In Proceedings of the IEEE Applied Power Electronics Conference and Exposition (APEC), Tampa, FL, USA,
26–30 March 2017.
21. Yang, Y.; Wen, Y.; Gao, Y. A Novel Active Gate Driver for Improving Switching Performance of High-Power SiC MOSFET
Modules. IEEE Trans. Power Electron. 2019, 34, 7775–7787. [CrossRef]
22. Katada, R.; Hata, K.; Ymauchi, Y.; Wang, T.W. Digital Gate Driving (DGD) is Double-Edged Sword: How to Avoid Huge Voltage
Overshoots Caused by DGD for GaN FETs. In Proceedings of the IEEE Energy Conversion Congress and Exposition (ECCE),
Vancouver, BC, Canada, 10–14 October 2021.
23. Miyazaki, K.; Abe, S.; Tsukuda, M.; Omura, I.; Wada, K.; Takamiya, M.; Sakurai, T. General-purpose clocked gate driver IC with
programmable 63-level drivability to optimize overshoot and energy loss in switching by a simulated annealing algorithm. IEEE
Trans. Ind. Appl. 2017, 53, 2350–2357. [CrossRef]
24. Anthony, P.; McNeill, N.; Holliday, D. High-Speed Resonant Gate Driver with Controlled Peak Gate Voltage for Silicon Carbide
MOSFETs. IEEE Trans. Ind. Appl. 2014, 50, 573–583. [CrossRef]
25. Chen, L.; Peng, F.Z. Switching loss analysis of closed-loop gate drive. In Proceedings of the Applied Power Electronics Conference
and Exposition (APEC), Palm Springs, CA, USA, 21–25 February 2010.
26. Regnat, G.; Jeannin, P.O.; Frey, D.; Ewanchuk, J.; Mollov, S.V.; Ferrieux, J.P. Optimized power modules for silicon carbide MOSFET.
IEEE Trans. Ind. Appl. 2018, 54, 1634–1644. [CrossRef]
27. Baliga, B.J. Fundamentals of Power Semiconductor Devices; Springer: Berlin/Heidelberg, Germany, 2008.
28. Liao, X.; Li, H.; Yao, R.; Huang, Z.; Wang, K. Voltage Overshoot Suppression for SiC MOSFET-Based DC Solid-State Circuit
Breaker. IEEE Trans. Compon. Packag. Manuf. Technol. 2019, 9, 649–660. [CrossRef]
29. Lobsiger, Y. Closed-Loop IGBT gate driver and current balancing. Ph.D. Thesis, Department of Information Technology and
Electrical Engineering, ETH-Zurich, Seedorf, Switzerland, 2014.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

140
 energies
Review
Review of Surface Charge Accumulation on Insulators in DC
Gas-Insulated Power Transmission Lines: Measurement and
Suppression Measures
Fangwei Liang 1 , Hanhua Luo 1 , Xianhao Fan 1, *, Xuetong Li 1 and Xu Wang 2

1 Department of Electrical Engineering, Tsinghua University, Beijing 100084, China;

liangfangwei@tsinghua.edu.cn (F.L.); lhh22@mails.tsinghua.edu.cn (H.L.);
li-xt19@mails.tsinghua.edu.cn (X.L.)
2 School of Mechanical Electronic & Information Engineering, China University of Mining & Technology,
Beijing 100084, China; zqt2200404117@student.cumtb.edu.cn
* Correspondence: xianhao_fan@163.com

Abstract: Gas-insulated power transmission lines (GILs) can replace cables and overhead transmis-
sion lines, playing an important role in DC transmission systems. However, the influence of surface
charge accumulation on insulation reliability cannot be ignored as the operational voltage of the
DC GIL increases. In this paper, the measurement methods for the insulator surface potential are
summarized, including, dust maps, the Pockels effect method, and the electrostatic probe method.
Then, a typical surface charge inversion algorithm is introduced. The main influencing factors of
surface charge accumulation are analyzed, such as the applied voltage, insulation gas, insulator
shape, and temperature. The charge accumulation pathway is revealed. Furthermore, methods for
inhibiting the accumulation of surface charges and promoting the dissipation of accumulated charges
are introduced to reduce the surface charges on insulators. Finally, the development direction of DC
GIL insulators is predicted. We anticipate that the online monitoring of surface charge distribution,
clarifying the percentage of charge accumulation pathways, and optimizing the insulator casting
process will be the research directions for the insulator surface charge topic in the future. This article
provides a comprehensive understanding of the surface charges of GIL insulators and a reference for
the insulation design of DC GILs.

Keywords: gas-insulated power transmission line (GIL); surface potential measurement; surface charge
inversion algorithm; surface charge accumulation mechanism; surface charge suppression measure

1. Introduction
Primary energy is mainly distributed in the northwest of China, while most electric en-
ergy is consumed in the central and eastern regions. Therefore, it is necessary to implement
high-capacity and long-distance power transmission to meet the increasing demand for
electricity [1–3]. Compared with AC, high-voltage DC transmission has advantages such
as low losses, a large transmission capacity, and convenience for grid interconnection. DC
transmission has been widely used in the long-distance transmission of electric energy [4–6].
Ultra-high voltage (UHV) and high-voltage direct current (HVDC) power transmission sys-
tems have rapidly developed since the 1950s [7,8]. The DC system is of great significance for
China to optimize its energy allocation [9,10]. However, the HVDC transmission lines are
facing new challenges, such as high altitudes, large drops, harsh meteorological conditions,
low electromagnetic radiation, and strict environmental protection rules [11,12], and the
overhead transmission lines and power cables cannot meet these requirements. Under such
power supply requirements, gas-insulated transmission lines (GILs) with a SF6 or SF6 /N2
gas mixture as the insulation medium have many advantages, such as a large transmission
capacity, small footprint, high operational stability, and environmental friendliness [13–15].

Energies 2023, 16, 6027. https://doi.org/10.3390/en16166027 141 https://www.mdpi.com/journal/energies

Energies 2023, 16, 6027

GILs can be directly buried underground or installed in tunnels. Therefore, GILs are an
excellent choice for large-scale power transmission in addition to overhead transmission
lines and power cables [16].
DC GILs will play an important role with the development of DC transmission [17–19].
However, the GILs in operation are basically a AC systems. There are only a few DC GILs
due to the difference between AC and DC insulation systems [20,21]. This is because of the
different electric field distributions caused by the surface charge accumulation on the insu-
lator [22]. For AC systems, the distribution of the electric field depends on the permittivity.
However, the DC system is mainly determined by the resistivity of the material [23,24].
The electric field distribution gradually transitions from the initial capacitive electric field
to a steady resistive electric field due to surface charge accumulation [25,26]. Furthermore,
the internal insulation system of a DC GIL consists of insulators and compressed gas (e.g.,
SF6 or a SF6 /N2 gas mixture) [16,27]. However, the SF6 in GILs is one of the greenhouse
gases limited by the Kyoto Protocol. Its greenhouse effect is 23,500 times greater than that
of CO2 [28]. The arc extinguishing capability of insulation gas is not required. This is
because there are no switchgears in GILs [16]. Furthermore, the SF6 /N2 gas mixture has
the advantages of good insulation strength, stable chemical properties, and lasting mixing
characteristics. The SF6 /N2 gas mixture has been widely used in second-generation GILs to
reduce the use of SF6 [29]. Moreover, the cost can be further reduced due to the low price of
N2 . For example, a 20% SF6 /80% N2 gas mixture has 69% of the insulation strength of pure
SF6 under the same gas pressure. The pressure of the gas mixture needs to be increased by
45% in order to achieve the same insulation strength. However, the use of SF6 is reduced
by 71% [16]. Therefore, the impact of greenhouse gases has been reduced. As a result, the
insulator has become the weakest component in DC GILs [30,31].
The insulator is the core component of a GIL [32]. GIL insulators are mainly divided
into basin insulators and tri-post insulators. High-voltage conductors are supported by
insulators. The conductor is mounted on the axis of the GIL. Additionally, high-voltage
conductors are kept insulated from the ground by the insulator [33–36]. The gas seal of the
pipeline is achieved by a basin insulator with a sealing ring. The installation position of
the tri-post insulator can be easily adjusted. This is because the grounded electrode of the
tri-post insulator has a wheel. GIL insulators are mainly composed of epoxy resins with
high resistivity [37–39]. As a result, the accumulated surface charge cannot dissipate for a
long time. Therefore, the operation of the GIL is threatened due to the flashover triggered
by the accumulated surface charges [40–43].
The critical boundary parallel to the insulator surface was selected as the flashover
pathway by Kumara et al. The line originated from the maximum electric field on the
cathode. A flashover voltage model considering the influence of surface charges was
established by determining the initiation, development, and maintenance of the streamer.
Under Gaussian distribution, the pulse flashover voltage was decreased by nearly 11%
when the maximum positive charge density was 125 μC/m2 [44]. Furthermore, the
DC ±500 kV gas-insulated switchgear (GIS) developed by the Kansai Electric Power
Company, Shikoku Electric Power Company, and Power Development Company of Japan
has been successfully applied in the Anan converter station. However, the actual opera-
tion voltage is only half of the rated voltage due to the influence of the insulator surface
charges [45–47]. This means that the surface charge accumulation of insulators leading to a
decrease in the flashover voltage is the issue restricting the development of DC GILs [48–51].
In this paper, the surface charge measurement technology, inversion algorithm, accu-
mulation mechanism, and inhibitory strategies for GIL insulators are summarized. Addi-
tionally, the research trends in insulator surface charges are identified. This article provides
a reference for the surface charge accumulation of DC GIL insulators.

2. Surface Potential Measurement

The study of surface charges should be based on the accurate measurement of the sur-
face potential distribution on the GIL insulator [52]. The measurement of surface potential

142
Energies 2023, 16, 6027

is different from that of the voltage for a high-voltage conductor. This is because a certain
insulation gap must be maintained between the probe and the measured surface to prevent
the leakage of the induced charge onto the probe. At present, the main measurement meth-
ods for insulator surface potential include dust maps, the Pockels effect, and electrostatic
probes. [53,54].

2.1. Dust Map

Dust maps, also known as Lichtenberg figures, were proposed by Lichtenberg in
1777 [52]. Firstly, the red lead oxide and yellow sulfur are selected. The powder is uniformly
suspended in gas. The charged powder with different polarities is absorbed on the insulator
surface due to the electrostatic attraction of the surface charges. As a result, the visual
analysis of the charge distribution is realized. However, the surface charge distribution
cannot be quantitatively measured by a dust map [55]. Since the dust particles are adsorbed
on the measured surface, it is not possible to achieve repeated measurements through a
dust map [56]. The movement of charged dust particles is also affected by gravity and
buoyancy [57]. Furthermore, the original surface charge distribution may be changed. This
means that the accuracy of dust maps needs to be further improved. Based on a dust map
established by quartz and feldspar powder, the surface charge distribution characteristics
with a hollow ring structure are shown in Figure 1 [58].

Figure 1. Dust map of surface charge distribution with a hollow ring [58].

2.2. Pockels Effect

The Pockels effect method was proposed by Takada in 1991. The principle was based
on the electro-optical effect of Pockels crystals, such as Bi12 SiO20 and Bi4 GeO12 [12,59]. The
refractivity of Pockels crystals varies linearly with the electric field [60]. Once the electric
field inside the crystals is changed due to the accumulated surface charge, the phase of
the incident laser is delayed. The phase delay of light is transformed into a difference
in light intensity. Then, the electric field distribution inside the crystal is obtained. As a
result, the surface charge density is calculated, as shown in Figure 2 [61]. The real-time and
quantitative measurement of surface charges on certain thin and transparent materials can
be realized by the Pockels effect [18]. Furthermore, the method is only applicable to the
measurement of insulator surface charges under an AC voltage or impulse voltage due to
the electrostatic relaxation of Pockels crystals [12].

143
Energies 2023, 16, 6027

Figure 2. Schematic of the Pockels effect [61].

2.3. Electrostatic Probe

The electrostatic probe method is a non-contact measurement technique [62,63]. There
are two commonly used electrostatic probes: capacitive electrostatic probes based on
electrostatic induction and Kelvin electrostatic probes designed according to the principle
of electric field compensation [64,65]. The capacitive electrostatic probe was proposed
by Davies in 1967 [66]. The quantitative measurement of surface charges can be realized
by a voltage-following circuit. The data acquisition unit of the probe is composed of a
high-input-impedance operational amplifier [67]. In contrast to capacitive probes, Kelvin
probes are active probes. Furthermore, Kelvin probes are not affected by the surrounding
gas. Kelvin probes are widely used due to their higher accuracy. Their measurement
principle is shown in Figure 3 [64].

Figure 3. Measurement principle of a Kelvin electrostatic probe [64].

The inductive electrode in the probe is kept vibrating by an oscillator. As a result,

the capacitance between the electrostatic probe and the measured surface is changed. If
the voltage on the probe is not equal to the measured surface potential, a displacement
current is generated on the induction electrode. The voltage on the probe is adjusted until
the induced current becomes zero. At this moment, the voltage on the probe is equal to the
measured surface potential [68]. Since there is no potential difference between the probe
and the measured surface, the partial discharge triggered by the electrostatic probe during
the measurement is avoided [57]. This method is an ideal measurement method for surface
charges. Typical products include the Trek-341B [69], Trek-347 [70,71], and Monroe 244A
electrostatic voltmeter [72,73].

144
Energies 2023, 16, 6027

3. Surface Charge Inverse Algorithm

For an electrostatic probe, only the surface potential can be directly measured. There-
fore, it is essential to calculate the surface charge density by the inversion algorithm [52].
There have been a number of inversion algorithms proposed in the past, such as the linear
algorithm, λ function method, improved λ function method, analytical method, and ϕ
function method [1].

3.1. Linear Algorithm

The linear algorithm was proposed by Davies in 1967. The linear relationship between
the surface charge and the induced voltage on the probe is established based on the
equivalent measurement circuit, as shown in Figure 4 [1]. The measurement circuit of a
capacitance probe consists of two parallel branches, where Cd is the capacitance of the
dielectric, Cg is the capacitance formed by the gap between the grounded electrode and
the dielectric, Cp is the capacitance between the probe and the measured surface, Cm is the
capacitance of the electrostatic probe to the ground, and A is the effective area measured by
the probe.

ǻaǼ ǻbǼ
Figure 4. Measurement through capacitive electrostatic probes: (a) measurement diagram, (b) equiv-
alent circuit [1].

When the gap between the electrostatic probe and the measured surface is small
enough, the electric field between them is approximately uniform. Based on the principle
of capacitor voltage division, the linear relationship between the surface charge density σ
and induced voltage on the probe Ug can be obtained [74].

σ = M·Ug (1)

where M is the scaling coefficient.

The calculation of the linear algorithm is very simple. However, the method is only
suitable for thin films smaller than 5 mm. This is mainly because the equivalent circuit is
not always consistent during the measurement. Furthermore, the influence of the surface
charge outside the measurement area on the results cannot be ignored [67].

3.2. λ Function Method

The relationship between the induced charge on the probe and the surface charges
was established by Perdersen et al. based on the λ function [75].

q= λσdA (2)
A0

where q is the induced charge on the electrostatic probe, A0 is the measured area, and λ is
the solution of the Laplace equation for the insulation system.

145
Energies 2023, 16, 6027

Compared with the linear algorithm, the λ function method has a higher accuracy.
However, the method is only suitable for geometrically symmetric insulation systems. Fur-
thermore, the influence of the measurement position of the electrostatic probe is ignored [1].

3.3. Improved λ Function Method

The improved λ function method was proposed by Kumada et al. in 2004 with
reference to digital image processing technology [76]. The convolution operation in the
spatial domain is transformed by two-dimensional Fourier transform into the product
calculation in the frequency domain.

Φ(μ, ν) = H(μ, ν)·σ (μ, ν) (3)

The calculation of the inversion matrix is avoided. The surface charge density in
the frequency domain is solved by division. Finally, the surface charge density σ(x, y) in
the space domain is obtained by inverse Fourier transform. Thus, the calculation of the
surface charge is greatly simplified. Furthermore, the Wiener filter is used in the frequency
domain for noise reduction to further improve the accuracy. However, this method is only
applicable to shift-invariant systems, such as large-area flat plates and long cylindrical
tubes [77].

3.4. Analytical Method

The analytical method was proposed by Ootera et al. [78]. The method considering
all surface charges was based on the theory of electrostatic fields. When the high-voltage
electrode is grounded, the insulator surface is divided into n grid units based on the finite
element method. Once the mesh is fine enough, the surface charge can be considered
uniformly distributed in any cell. According to the superposition theorem of potential, the
potential Vi of any element i satisfies the following formula:
⎡ ⎤⎡  ⎤ ⎡ ⎤
P11 P12 ··· ··· P1n σ1 V1
⎢ P21 P22 ··· ··· P2n ⎥ ⎢ σ2 ⎥ ⎢ V2 ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ .. .. .. ⎥⎢ .. ⎥ ⎢ .. ⎥
⎢ . . ⎥ ⎢ ⎥ ⎢ ⎥
⎢ . ⎥⎢ . ⎥ = ⎢ . ⎥ (4)
⎢ . .. .. ⎦⎢
⎥ .⎥ ⎢ . ⎥
⎣ .. . . ⎣ .. ⎦ ⎣ .. ⎦
Pn1 Pn2 ··· ··· Pnn σn Vn

1 1
Pij = dS (5)
4πε 0 rij
Si

where ε0 is the vacuum permittivity; σi and Si are the apparent surface charge density and
area, respectively; and rij is the distance between elements i and j.
Based on the dielectric interface conditions, the relationship between the apparent
surface charge density and free surface charge density σ can be written as follows:
⎡ ⎤⎡  ⎤ ⎡ ⎤
F11 F12 ··· ··· F1n σ1 σ1
⎢ F21 F22 ··· ··· ⎢ 
F2n ⎥⎢ σ2 ⎥ ⎢ σ2 ⎥
⎥ ⎥ ⎢
⎢ ⎥
⎢ .. .. .. ⎥⎢ .. ⎥ ⎢ .. ⎥
⎢ . . . ⎥ ⎢.⎥=⎢.⎥ (6)
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ . .. .. ⎥ ⎢.⎥ ⎢.⎥
⎣ .. . . ⎦⎣ .. ⎦ ⎣ .. ⎦
Fn1 Fn2 ··· ··· Fnn σn σn
⎧ ε 2 +ε 1
⎨ 2ε 0  i=j
Fij = rij ·n j (7)
(
⎩ 2ε − ε ) 1
1 4πε 0 rij 3
dS i = j
Si

146
Energies 2023, 16, 6027

where subscripts 1 and 2 represent the insulator side and the gas side, respectively; ε is
the permittivity of the material; and n is the unit normal vector in the positive direction
pointing from the insulator to the gas.
Therefore, the surface charge density is:

σ = TV (8)

The inversion matrix T is only determined by the material and geometry of the model.
Therefore, a surface charge density of 1 μC/m2 is defined for the element to solve the surface
potential distribution on the insulator. As a result, the inversion matrix T is obtained by
moving the unit charge [79].

3.5. ϕ Function Method

Based on the λ function method, the relationship between the induced voltage on the
probe and the surface charge was constructed by Faircloth et al. [80]. The insulator surface
is divided into many small regions based on the finite element method. The surface charge
of all regions contributes to the surface potential.

V = σΦ (9)

For a particular insulation system, the inversion matrix Φ only depends on the geo-
metric size and material. The Φ matrix is solved by finite element software, such as ANSYS
and COMSOL Multiphysics. Firstly, an inversion calculation model is built. The unit
charge is defined on the surface of the measured insulator to calculate the surface potential
distribution. As a result, the complete Φ matrix is solved. The ϕ function method takes the
motion of the electrostatic probe into account, thus improving the accuracy. However, the
amount of calculation increases with the dimensions of the inversion matrix.

4. Surface Charge Accumulation Mechanism

The surface charge accumulation of a DC GIL insulator is affected by the applied
voltage, insulation gas, insulator shape, temperature, etc.

4.1. Applied Voltage

Under a DC voltage, the surface charge density increases with the applied time. The
surface charge distribution is similar under different voltage levels [39]. Furthermore, the
polarity of surface charges is reversed due to the increase in the DC voltage when only
natural radiation is considered for the generation of charged particles in gas [25]. Since the
conductivity of the insulation gas and insulator is a function of the electric field [81], the
surface charge accumulation is also affected by the electric field.

4.2. Insulation Gas

In order to reduce the use of SF6 , the insulation gas of second-generation GILs is a
SF6 /N2 gas mixture [16]. Therefore, the surface charge distribution of insulators in 0.5 MPa
pure SF6 and a 20% SF6 /80% N2 gas mixture were compared by Wang et al. [82]. The results
showed that the amount of surface charge was larger in the 20% SF6 /80% N2 gas mixture
under −30 kV. This was mainly because the partial discharge in the gas increased with the
decrease in the SF6 content. Furthermore, the accumulation of surface charges on insulators
in the pure SF6 had a significant polarity effect. That is to say, the surface charges under
a negative voltage were obviously larger than those under a positive voltage. However,
no significant polarity effect was observed in the 20% SF6 /80% N2 gas mixture. This was
because the pure SF6 was more sensitive to metal particles. The ionization coefficient of
the gas sharply increased due to the enhancement in the local electric fields. As a result,
the sources of the surface charges increased [83,84]. For the 20% SF6 /80% N2 gas mixture,
a polarity effect was not observed. This was mainly because the sensitivity of the partial
discharge to the electric field distortion was reduced by the N2 .

147
Energies 2023, 16, 6027

4.3. Insulator Shape

The accumulation of surface charges is inevitably affected by the geometry of the
insulator. Early on, cylindrical insulators were used as simplified models for GIL insulators.
Then, cylindrical insulators with shielded electrodes were adopted by Wang et al. [62]. In
order to truly reflect the operational conditions of GIL insulators, the circular insulator
was proposed by Li et al. [70]. Furthermore, proportionally reduced conical insulators
were adopted by Deng et al. [29]. Bowl-shaped insulators were designed to achieve the
self-adaptation of surface charges based on nonlinear materials [85]. In addition, an actual
GIL basin-type insulator was researched by Qi et al., as shown in Figure 5 [86].

ǻaǼ ǻbǼ ǻcǼ ǻdǼ

Figure 5. Different insulator shapes: (a) cone-like frustum, (b) cone, (c) bowl type, (d) basin-
type [29,70,85,86].

Based on a surface charge transport model considering the microscopic parameters

of the gas, the influence of the insulator shape was analyzed by Ma et al. The initial
capacitive electric field distribution was compared with the steady-state resistive electric
field distribution [87]. The results showed that the surface charge distribution was similar
to the normal component of the initial capacitive electric field. The amount of surface
charge accumulated on the conical insulator was the largest.

4.4. Temperature
Due to the Joule heating of the operational current, the temperature of the internal
conductor in a GIL may rise to 90 ◦ C [88]. Furthermore, the volume conductivity of epoxy
insulators is a function of the temperature. The partial discharge in a gas is also affected
by the temperature [89]. Therefore, the influence of the temperature on the surface charge
accumulation cannot be ignored. The temperature distribution inside a GIL was calculated
by Zhou et al. based on thermal convection, radiation, and conduction [90]. The results
showed that the temperature distribution inside the GIL gradually decreased from the
center conductor to the grounded shell. The temperature gradient was 30 K under a current
of 3450 A. The peak of surface charge density was 7.68 nC/cm2 , which increased by 60%.
Additionally, high-temperature oil circulation was used to heat the internal conductor of
the GIL. As a result, a surface potential measurement platform for DC GIL insulators under
a temperature gradient was established, as shown in Figure 6 [91].
Once the temperature gradient was 70 ◦ C, the average value of the surface potential
increased from 278 V to 1670 V under a positive voltage. Under a negative voltage, the
average surface potential increased nearly twofold. Tang et al. pointed out that the mean
free path of electrons increased with the temperature. Therefore, the initial voltage of the
corona decreased. As a result, the positive charge density increased under the temperature
gradient [41].
The above factors affecting insulator surface charges are mainly due to the differences
in the charge accumulation pathway. There are three pathways to accumulating charge
on an insulator surface: through the insulator volume, insulation gas, and along the
insulator surface [92]. For the ideal case of ignoring the insulation defects inside the GIL,
the surface resistivity of the insulator is large enough. The current along the insulator
surface is small enough to be ignored. Therefore, the charge accumulated through the
insulator volume is uniformly distributed [18]. The conduction current is determined

148
Energies 2023, 16, 6027

by the volume conductivity of the insulator. For conical insulators, charges of the same
polarity accumulate on the insulator surface near the inner conductor [36]. Since space
charges are easily generated due to the uneven distribution of volume conductivity, the
space charge migrates to the insulator surface, resulting in surface charge accumulation.
That is to say, it follows the volume conductivity model proposed by Cooke et al. [12].

Figure 6. Surface potential measurement system under a temperature gradient [91].

The conduction current through the gas is affected by the generation rate, recombi-
nation coefficient, mobility, diffusion coefficient, and other factors [7]. Insulation defects
are inevitable within a DC GIL, such as protrusions on the electrode, metal particles on the
insulator surface, and the gas-electrode-insulator tri-junction, since the corona discharge
of insulation defects is the main cause of surface charge accumulation. The randomly
distributed charge spots and charge bands are shown in Figure 7 [18]. This means that the
insulator surface charge distribution is uneven [18]. Furthermore, a normal electric field
model was proposed by Knecht et al. [93]. It showed that the normal electric field on the
insulator surface is the cause of charge accumulation.

Figure 7. Surface charge distribution pattern: 1. uniform distribution, 2. charge spots, 3. charge
clouds [18].

149
Energies 2023, 16, 6027

The conduction current along the surface is closely related to the surface conductivity
of the material [94]. For insulators with surface modification, the surface charge distribution
depends on the conduction current along the surface. The tangential electric field model
was proposed by Nakanishi et al. [95]. This model points out that there is an exponential
relationship between surface conductivity and tangential electric field, resulting in uneven
material surface. Therefore, the discontinuity of conducting current along the surface is the
main cause of surface charge accumulation.
The three accumulation pathways of surface charges exist simultaneously. However,
the proportion of each part varies significantly under different tests [28]. For the vacuum,
corona discharges from the tri-junction of the cathode are the main source of insulator
surface charges [96]. Yu et al. pointed out that the surface charge density is closely related
to the secondary electron energy emission curve [97]. When the partial discharge in the gas
is ignored, the conduction current through the insulator volume is the main accumulation
pathway for surface charges [87]. For high-resistivity insulation systems, the accumulation
of surface charges is determined by the conduction current in the gas side. The main
accumulation pathway gradually changes from the gas to the insulator when the volume
resistivity is less than 5 × 1017 Ω ·cm [22]. Furthermore, the generation rate of ion pairs
in the gas is one of the key factors. The accumulation pathway changes to the conduction
current in the gas with the increase in the ion pair generation rate [48,98].
For basin insulators, a unified surface charge accumulation mechanism was proposed
by Li et al. The electric field is divided into three levels based on the leakage current,
as shown in Figure 8 [99–101]. For the low electric field, there are many factors, such
as residual static charges on the surface, bulk charges inside the insulator, and charged
particles in the gas. Each factor may become the core. Thus, charge accumulation under
low electric fields can occur in a variety of ways. Under a medium electric field, charges of
the same polarity injected through the insulator are the main source. Since the ionization in
the gas side is enhanced due to insulation defects under a high electric field, the influence
of charged particles in the gas on the surface charge accumulation increases. Therefore, the
source of the surface charge is electric-field-dependent.

Figure 8. Field-dependent surface charge accumulation model [101].

5. Surface Charge Suppression Strategy

The methods to reduce the surface charge include inhibiting the accumulation of
surface charges and promoting the dissipation of accumulated charge, for example, by
doping modification, structural optimization, and surface treatment [102]. Since the sur-
face charge accumulation is determined by the conduction current flowing through the
insulator when the volume resistivity of the insulator is lower than the critical value of

150
Energies 2023, 16, 6027

the system [103], increasing the volume resistivity of insulators is an effective method to
inhibit surface charge accumulation. The volume conductivity of epoxy resin was reduced
through doping with fullerene C60 by Zhang et al. The surface charge on the C60-epoxy
composite insulators was reduced due to the hollow cage structure of C60 [104]. That is to
say, C60 has extremely strong electrophilic properties. Free electrons in the polymer matrix
could be captured.
Nanoparticle doping is the most commonly used modification method for insulation
materials, such as Al2 O3 , TiO2 , SiO2 , BN, and AlN [105–112]. This is because a charge
accumulation center is introduced by nanoparticles, and the carrier concentration is reduced.
As a result, the material conductivity is decreased [113]. Since the surface charge is usually
derived from partial discharge in gas for Al2 O3 epoxy resin insulators [114–116], the normal
electric field should be reduced as much as possible during the optimization of insulator
shape [87]. For example, a conical insulator was adopted, and the flat bottom was basically
parallel to the electric field line. As a result, there was only a small charge on the bottom [36].
Furthermore, a bowl-shaped insulator was designed by He et al. The normal electric field
on the insulation area was designed to be negligible in order to reduce the surface charge,
as shown in Figure 9 [7]. Therefore, the surface charge accumulation could be reduced by
doping modification and structure optimization. However, the influence of the insulation
surface is usually ignored. This is because the surface resistivity of GIL insulators is greater
than the volume resistivity. The conduction current along the insulator surface is negligible.

ǻaǼ ǻbǼ

Figure 9. Electric field distribution around insulators: (a) cone, (b) bowl-type [7,36].

For accumulated surface charges, surface treatment is used to promote dissipation

and reduce the surface charge. The main dissipation pathways include: recombination
with charged particles in gas, dissipation through the insulator volume, and dissipation
along the insulator surface [117]. As partial discharges are random, there is usually little
recombination between heteropolar charged particles in gas. For GIL epoxy insulators, the
volume conductivity is in the order of 10−17 S/m [118]. A surface conductivity of 10−18 S
has been measured on epoxy resin insulators. As a result, it is impossible for the surface
charge to dissipate autonomously for a long period of time. For example, the dissipation
curve of the surface potential within 1200 s is approximately horizontal. The dissipation of
charge is negligible [119].
The dissipation of surface charges through an insulator material is promoted by the
increase in the volume conductivity. However, the insulation size needs to be increased
due to the decrease in insulation strength. Therefore, an ideal charge dissipation strategy is
to increase the surface conductivity through surface treatment. The surface treatment of
insulator materials is used to increase the surface conductivity by 1–2 orders of magnitude,
including through surface fluorination, plasma surface treatment, and coating. Thus, the
amount of surface charge is reduced by accelerating the dissipation rate.

151
Energies 2023, 16, 6027

5.1. Fluoridation
An epoxy resin sheet was fluorinated in a gas mixture of 12.5%F2 /87.5%N2 by An
et al., as shown in Figure 10. A fluorinated layer of micron size was formed. The mobility
of the carrier was improved when the fluorination reaction time was less than 60 min. That
is to say, the rate of surface charge dissipation along the surface was accelerated [120].

Figure 10. SEM images of the cross-section of fluorinated samples [73].

5.2. Plasma Surface Treatment

Plasma was used by Shao et al. to improve surface conductivity. According to
Figure 11, an atmospheric plasma injection device deposited SiOx on the surface of the
epoxy resin. Inorganic groups were introduced, such as Si-O-Si and Si-OH [121,122].
Therefore, plasma treatments could also result in shallow trap energy levels on the surface.
As a result, the surface charge could rapidly dissipate along the insulator surface [123].

Figure 11. Atmospheric pressure plasma jet device [121].

5.3. Coating
Non-uniform coatings are also functional gradient materials. A functional gradient
coating was prepared by Li et al. [124]. The surface conductivity with a gradient was
formed by magnetron sputtering ZnO or fluoridation on the epoxy resin surface, as shown
in Figure 12.

152
Energies 2023, 16, 6027

ǻaǼ ǻbǼ

Figure 12. Functional gradient coating on the insulator surface: (a) magnetron sputtering,
(b) fluorination [124,125].

The effect of the concentration and particle diameter of the SiC in the coating on the
surface charges was analyzed by Tang et al. [50]. The results showed that the surface
trap density was increased due to the SiC/EP coating. Therefore, more surface charges
accumulated. Since the surface trap energy level could also be decreased by the coating,
the dissipation rate of charges was accelerated. This indicates that the regulation of SiC/EP
coatings is attributed to the interaction between the trap density and trap energy level.

6. Conclusions and Suggestions for Future Work

The surface charge measurement methods, charge accumulation mechanisms, and
suppression measures for DC GIL insulators were summarized. In order to measure the
insulator surface potential, dust patterns, the Pockels effect method, and the electrostatic
probe method can be implemented. Then, a surface charge inversion algorithm should be
adopted to calculate the surface charge density. Based on the surface charge distribution
characteristics, many factors (such as the applied voltage, insulation gas, insulator shape,
and temperature) have an impact on the surface charge accumulation. For surface charge
suppression, the main solution is inhibiting the accumulation of surface charges and
promoting the dissipation of accumulated charges. Finally, suggestions for future research
in this area are as follows:
(1) With regard to surface potential measurement, Kelvin electrostatic probes have a
high accuracy. However, the application of Kelvin electrostatic probes in online measure-
ments is limited by their lower measurement range. Furthermore, capacitive electrostatic
probes have a simple structure and lower cost. Capacitive probes are expected to be applied
for the online monitoring of insulator surface charges after the measurement accuracy is
further improved.
(2) Focusing on the surface charge inversion algorithm, the environmental noise
during the data acquisition cannot be ignored. For online measurement, the changes in the
original electric field caused by the probe should be considered to ensure the accuracy of the
measurement. Furthermore, a 3D display of the insulator surface charge distribution could
be established by the surface charge inversion algorithm combined with data transmission,
storage, and image processing.
(3) For the surface charge accumulation mechanism, the proportions of the three
pathways of charge accumulation should be clarified. The dominant pathway of charge
accumulation should be revealed to provide a theoretical basis for the suppression of
surface charges. For the surface charge transport model, the influence of partial discharges
in the gas and the surface trap distribution on the insulator cannot be ignored. Additionally,
the type of carrier in the gas must be further distinguished rather than simply dividing it
into positive ions, negative ions, and free electrons.
(4) In terms of surface charge suppression, the application of coatings on the GIL
insulator surface needs to be further tested in terms of adhesion strength and material
aging. The effect of coating detachment on insulation should be paid more attention.

153
Energies 2023, 16, 6027

Since the material doping modification of an insulator is limited by the casting techniques
available, the interface due to the curing properties of different materials should not be
ignored. Therefore, understanding the transition region between different doped materials
will be one of the challenges for subsequent research.

Author Contributions: Writing—original draft preparation, F.L. and H.L.; writing—review and
editing, X.F.; visualization, X.L. and X.W. All authors have read and agreed to the published version
of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China under
grant No. 51677061.
Data Availability Statement: The data supporting the findings of this study are available from the
corresponding author upon reasonable request.
Acknowledgments: The authors would like to thank Jiayi Liu from Hunan University for the help
with the writing.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Lin, C.J.; Li, C.Y.; Zhang, G.X.; Hu, J.; He, J.L. Review and prospect of surface charge inversion algorithm of Cone-Type spacer
based on surface potential measurements. Proc. CSEE 2016, 36, 6654–6663. (In Chinese)
2. Wang, Q.; Xie, G.; Tian, H.; Peng, Z.; Yang, X. Impact of temperature on the transient DC field distribution of ±1100 Kv UHVDC
wall bushing. IEEE Trans. Power Deliv. 2021, 36, 1449–1457. [CrossRef]
3. Ma, G.; Zhou, H.; Liu, S.; Wang, Y.; Zhao, S.; Lu, S.J.; Li, C.; Tu, Y. Measurement and simulation of charge accumulation on a disc
spacer with electro-thermal stress in SF6 gas. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1221–1229.
4. Wang, J. Research on the Moving Mechanisms of Charged Metal Particles in DC GIL and Suppressing Methods. Ph.D. Dissertation,
North China Electric Power University (Beijing), Beijing, China, 2017. (In Chinese).
5. Ma, G.; Zhou, H.; Lu, S.; Wang, Y.; Liu, S.; Li, C.; Tu, Y. Effect of material volume conductivity on surface charges accumulation on
spacers under DC electro-thermal coupling stress. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1211–1220. [CrossRef]
6. Liu, Q.; Yan, J.D.; Hao, L.; Zhang, B.; Liu, S. Charge transport and accumulation around a spacer insulator for application in
HVDC wall bushing. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 281–293. [CrossRef]
7. Li, C.; Lin, C.; Jun, H.; Liu, W.; Li, Q.; Zhang, B.; He, S.; Yang, Y.; Liu, F.; He, J.L. Novel HVDC spacers by adaptively controlling
surface charges-Part I: Charge transport and control strategy. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1238–1247.
8. Du, B.X.; Li, J. Effects of ambient temperature on surface charge and flashover of heat-shrinkable polymer under polarity reversal
voltage. IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 1190–1197. [CrossRef]
9. Zhang, Y.L. Observation of Surface Charge for Composite Insulator and Reasearch of Its Influence Factor. Master’s Thesis, Hunan
University, Changsha, China, 2014. (In Chinese)
10. Zhang, Y.; Lu, L.C.; Yang, X.; Qi, B.; Yang, Z.D.; Dang, Z.P.; Yan, F.Q.; Li, C.R. Exploration on the characteristics of new ceramic
materials for insulators used in SF6 gas insulated HVDC equipment. Proc. CSEE 2021, 41, 174–182. (In Chinese)
11. Wang, J.; Li, Q.M.; Li, B.T.; Chen, C. Motion analysis of spherical conducting particle in DC GIL considering the influence of
inelastic random collisions and SF6 /N2 gaseous mixture. Proc. CSEE 2015, 35, 3971–3978. (In Chinese)
12. Zhang, B.Y.; Zhang, G.X. Review of charge accumulation characteristics at gas-solid interface in DC GIL, Part I: Measurement and
mechanisms. Trans. China Electrotech. Soc. 2018, 33, 4649–4662. (In Chinese)
13. Du, B.X.; Du, Q.; Li, J.; Li, A.; Fu, M.L.; Xiao, W. Surface charge dynamic behaviors of Epoxy/BN composite with high thermal
conductivity for gas insulated transmission pipeline. High Volt. Eng. 2018, 044, 2646–2653. (In Chinese)
14. Xue, J.; Li, Y.; Dong, J.; Chen, J.; Li, W.; Deng, J.; Zhang, G. Surface charge transport behavior and flashover mechanism on
alumina/epoxy spacers coated by SiC/epoxy composites with varied SiC particle size. J. Phys. D Appl. Phys. 2020, 53, 155503.
[CrossRef]
15. Pan, C.; Tang, J.; Wang, D.; Luo, Y.; Zhuo, R.; Fu, M. Decay characters of charges on an insulator surface after different types of
discharge. Plasma Sci. Technol. 2017, 19, 075503. [CrossRef]
16. Koch, H.; Goll, F.; Magier, T.; Juhre, K. Technical aspects of gas insulated transmission lines and application of new insulating
gases. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1448–1453. [CrossRef]
17. Wang, J.; Li, B.T.; Li, Q.M.; Liu, S.H.; Ma, G.M. Impact of linear metal particle on surface charge accumulation of post insulator
within DC GIL. Trans. China Electrotech. Soc. 2016, 31, 213–222. (In Chinese)
18. Xue, J.; Wang, H.; Liu, Y.; Li, K.; Liu, X.; Fan, X.; Deng, J.; Zhang, G.; Guo, B. Surface charge distribution patterns of a truncated
cone-type spacer for high-voltage direct current gas-insulated metal-enclosed transmission line/gas-insulated metal-enclosed
switchgear. IET Sci. Meas. Technol. 2018, 12, 436–442. [CrossRef]

154
Energies 2023, 16, 6027

19. Gao, Y.; Wang, H.; Yuan, X.; Zhao, H.; Li, Z. Surface charge accumulation on a real size epoxy insulator with bouncing metal
particle under DC voltage. IEEE Trans. Plasma Sci. 2021, 49, 2166–2175. [CrossRef]
20. Gao, Y.; Wang, H.; Li, Z.; Yuan, X.; Zhao, H. Gaseous ionization dependence of surface charge pattern on epoxy insulator with
complex surface profile under DC voltage. IEEE Trans. Plasma Sci. 2021, 49, 1627–1635. [CrossRef]
21. Du, B.X.; Liang, H.C.; Li, J.; Ran, Z.Y. Electrical field distribution along SF6 /N2 filled DC-GIS/GIL epoxy spacer. IEEE Trans.
Dielectr. Electr. Insul. 2018, 25, 1202–1210. [CrossRef]
22. Lutz, B.; Kindersberger, J. Surface charge accumulation on cylindrical polymeric model insulators in air: Simulation and
measurement. IEEE Trans. Dielectr. Electr. Insul. 2011, 18, 2040–2048. [CrossRef]
23. Zhou, H.; Ma, G.; Wang, Y.; Li, C.; Tu, Y.; Ye, S.; Zhang, B.; Guo, X.; Yang, X. Surface charge accumulation on 500 Kv cone-type
GIS spacer under residual DC voltage. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1230–1237. [CrossRef]
24. Winter, A.; Kindersberger, J. Transient field distribution in gas-solid insulation systems under DC voltages. IEEE Trans. Dielectr.
Electr. Insul. 2014, 21, 116–128. [CrossRef]
25. Winter, A.; Kindersberger, J. Stationary resistive field distribution along epoxy resin insulators in air under DC voltage.
IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1732–1739. [CrossRef]
26. Hoshina, Y.; Yasuoka, T.; Takei, M. Influence of tiny metal particles on charge accumulation phenomena of GIS model spacer in
high-pressure SF6 gas. IEEE Trans. Dielectr. Electr. Insul. 2013, 20, 1895–1901. [CrossRef]
27. He, S.; Lin, C.; Hu, J.; Li, C.; He, J. Tailoring charge transport in epoxy based composite under temperature gradient using
K2 Ti6 O13 and asbestine whiskers. J. Phys. D Appl. Phys. 2018, 51, 215306. [CrossRef]
28. Zhang, B.; Li, X.; Wang, T.; Zhang, G. Surface charging characteristics of GIL model spacers under DC stress in C4 F7 N/CO2 gas
mixture. IEEE Trans. Dielectr. Electr. Insul. 2020, 27, 597–605. [CrossRef]
29. Wang, H.; Xue, J.Y.; Chen, J.H.; Deng, J.D.; Wan, B.Q.; Zhang, G.J. Influence of metal particles on surface charge accumulation
characteristics of GIL spacer in SF6 /N2 mixture. Trans. China Electrotech. Soc. 2018, 033, 4663–4671. (In Chinese)
30. Imano, A.M. Accumulation of surface charges on the particle contaminated spacer surface in compressed gas under impulse
voltage stress. J. Electrost. 2004, 61, 1–19. [CrossRef]
31. Li, C.; Xu, Y.; Lin, C.; Chen, G.; Tu, Y.; Zhou, Y.; Lei, Z.; Han, T.; Suraci, S.V.; Wang, J.; et al. Surface charging phenomenon on
HVDC spacers in compressed SF6 insulation and charge tailoring strategies. CSEE J. Power Energy Syst. 2019, 6, 83–99.
32. Li, C.Y.; Zhang, C.H.; Lv, J.Z.; Liang, F.W.; Liang, Z.D.; Fan, X.H.; Riechert, U.; Li, Z.; Liu, P.; Xue, J.Y.; et al. China’s 10-year
progress in DC gas-insulated equipment: From basic research to industry perspective. IEnergy 2022, 1, 400–433. [CrossRef]
33. Du, B.; Dong, J.; Liang, H.; Kong, X. Polarity reversal and over voltage affecting discharge inception of tri-post insulator in ±800
Kv GIL. IEEE Trans. Dielectr. Electr. Insul. 2022, 29, 223–230. [CrossRef]
34. Li, D.Y.; Zhang, B.Y.; Gao, W.Q.; Hou, Y.C.; Zhang, G.X. “Charge Speckles” pattern in surface charge distribution of insulator
under DC voltage. Trans. China Electrotech. Soc. 2018, 33, 77–83. (In Chinese)
35. Du, Q.D.; Zhang, Q.G.; Zhao, J.P.; Wu, Z.C.; Wang, C.Q. Influence of electric conductivity of materials on field and surface charge
distribution of Basin-type insulator. High Volt. Eng. 2018, 44, 95–101. (In Chinese)
36. Chen, X.; Guan, H.; Jiang, T.; Du, H.; Paramane, A.; Zhou, H. Surface charge dissipation and DC flashover characteristic of DBD
plasma treated epoxy resin/AlN nanocomposites. IEEE Trans. Dielectr. Electr. Insul. 2020, 27, 504–511. [CrossRef]
37. Li, X.; Wan, M.; Zhang, G.; Lin, X. Surface charge characteristics of DC-GIL insulator under multiphysics coupled field: Effects of
ambient temperature, load current, and gas pressure. IEEE Trans. Dielectr. Electr. Insul. 2022, 29, 1530–1539. [CrossRef]
38. Li, J.; Wang, Y.; Chen, H.; Liang, H.; Yao, H.; Zhang, C.; Du, B. Polishing orientation affecting surface charging and flashover
characteristics of GIL/GIS epoxy spacer. IEEE Trans. Dielectr. Electr. Insul. 2022, 29, 1625–1632. [CrossRef]
39. Du, Q.; Zhang, Q.; Wu, Z.; Ma, J.; Zhao, J. Surface charge distribution on DC basin-type insulator. IEEE Trans. Dielectr. Electr.
Insul. 2019, 26, 17–25. [CrossRef]
40. Ma, Y.Y.; Zhang, C.; Kong, F.; Wang, T.T.; Chen, G.Y.; Shao, T. Effect of plasma jet array assisted film deposition on epoxy resin
surface electrical characteristics. High Volt. Eng. 2018, 44, 3089–3096. (In Chinese)
41. Pan, C.; Tiang, J.; Wang, D.; Zhou, R.; Yang, D.; Ye, G.; Fu, M. Influence of temperature on the characteristics of surface charge
accumulation on PTFE model insulators. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 1210–1219. [CrossRef]
42. Mahmoodi, J.; Mirzaie, M.; Akmal, A.S. Contribution of surface charges on high-voltage DC silicon rubber insulators to DC
flashover performance. IET Gener. Transm. Dis. 2018, 12, 5851–5857. [CrossRef]
43. Amer, M.; Laninga, J.; McDermid, W.; Swatek, D.R.; Kordi, B. Surface charging and its effects on DC flashover strength of
insulating materials. IEEE Trans. Dielectr. Electr. Insul 2018, 25, 2452–2460. [CrossRef]
44. Kumara, S.; Serdyuk, Y.V.; Gubanski, S.M. Simulation of surface charge effect on impulse flashover characteristics of outdoor
polymeric insulators. IEEE Trans. Dielectr. Electr. Insul. 2010, 17, 1754–1763. [CrossRef]
45. Wang, F.; Fang, Z.; Qiu, Y.C. Study of charge accumulation on insulator surface in HVDC gas-insulated switchgear. Proc. CSEE
2005, 25, 105–109.
46. Tang, J.; Pan, C.; Wang, D.B.; Fu, M.L.; Zhuo, R. Development of studies about surface charge accumulation on insulating material
under HVDC. Trans. China Electrotech. Soc. 2017, 08, 16–27. (In Chinese)
47. Lorenzi, A.D.; Grando, L.; Pesce, A.; Bettini, P.; Specogna, R. Modeling of epoxy resin spacers for the 1 MV DC gas insulated line
of ITER neutral beam injector system. IEEE Trans. Dielectr. Electr. Insul. 2009, 16, 77–87. [CrossRef]

155
Energies 2023, 16, 6027

48. Luo, Y.; Tang, J.; Pan, C.; Lin, S.J.; Wang, Z.Q. The transition mechanism of surface charge accumulation dominating way in DC
GIS/GIL. Trans. China Electrotech. Soc. 2019, 34, 185–194. (In Chinese)
49. Chen, K.; Liang, X.D.; Liu, S. Experimental study on the decay process of surface charges on epoxy insulators under different
temperature in air. High Volt. Eng. 2018, 044, 723–1728. (In Chinese)
50. Xue, J.; Chen, J.; Dong, J.; Wang, H.; Li, W.; Deng, J.; Zhang, G. The regulation mechanism of SiC/epoxy coatings on surface
charge behavior and flashover performance of epoxy/alumina spacers. J. Phys. D Appl. Phys. 2019, 52, 405502. [CrossRef]
51. Amer, M.; Laninga, J.; McDermid, W.; Swatek, D.R.; Kordi, B. New experimental study on the DC flashover voltage of polymer
insulators: Combined effect of surface charges and air humidity. High Volt. 2019, 4, 316–323. [CrossRef]
52. Lin, C.; Li, C.; He, J.; Hu, J.; Zhang, B. Surface charge inversion algorithm based on bilateral surface potential measurements of
cone-type spacer. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 1905–1912. [CrossRef]
53. Xue, J.Y.; Wang, H.; Li, K.F.; Fan, X.F.; Liu, X.Y.; Liu, Y.Q.; Deng, J.B.; Zhang, G.J. Research on charge distribution characteristics on
spacer surface in DC GIL. Proc. CSEE 2018, 38, 6164–6172. (In Chinese)
54. Qi, B.; Gao, C.J.; Xing, Z.L.; Chen, Z.Z.; Zhao, J.K.; Li, C.R. Surface charge distribution on GIS insulator under DC/AC voltage.
IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 3173–3181. (In Chinese) [CrossRef]
55. Li, C.; Zhu, Y.; Zhi, Q.; Sun, J.; Song, S.; Connelly, L.; Li, Z.; Chen, G.; Lei, Z.; Yang, Y.; et al. Dust figures as a way for mapping
surface charge distribution-A review. IEEE Trans. Dielectr. Electr. Insul. 2021, 28, 853–863. [CrossRef]
56. Kumara, S. Surface charges on cylindrical polymeric insulators. IEEE Trans. Dielectr. Electr. Insul. 2012, 19, 1076–1083. [CrossRef]
57. Zhang, L.; Lin, C.; Li, C.; Suraci, S.V.; Chen, G.; Riechert, U.; Shahsavarian, T.; Hikita, M.; Tu, Y.; Zhang, Z.; et al. Gas-solid
interface charge characterization techniques for HVDC GIS/GIL insulators. High Volt. 2020, 5, 95–109. [CrossRef]
58. Li, C.; Zhu, Y.; Hu, J.; Li, Q.; Zhang, B.; He, J. Charge cluster triggers unpredictable insulation surface flashover in pressurized
SF6 . J. Phys. D Appl. Phys. 2021, 54, 015308. [CrossRef]
59. Wu, K.; Pan, C.; Meng, Y.; Cheng, Y. Dynamic behavior of surface charge distribution during partial discharge sequences.
IEEE Trans. Dielectr. Electr. Insul. 2013, 20, 612–619.
60. Mu, H.B.; Zhang, G.J.; Suzuki, S.; Tanaka, S. Surface charge distribution of polymeric insulating materials under HVAC. Proc.
CSEE 2010, 30, 130–136. (In Chinese)
61. Zhou, W. Study on Influencing Factors and Dissipation Characteristics of Charge Accumulation on the Surface of Insulators
under DC Voltage. Master’s Thesis, Hunan University, Changsha, China, 2018. (In Chinese)
62. Zhou, H.; Ma, G.; Wang, C.; Wang, G.; Zhang, G.; Tu, Y.; Li, C. Review of charge accumulation on spacer of gas insulated
equipment at DC stress. CSEE J. Power Energy Syst. 2019, 6, 496–517.
63. Takuma, T.; Yashima, M.; Kawamoto, T. Principle of surface charge measurement for thick insulating specimens. IEEE Trans.
Dielectr. Electr. Insul. 1998, 5, 497–504. [CrossRef]
64. Wang, F.; Qiu, Y.; Pfeiffer, W.; Kuffel, E. Insulator surface charge accumulation under impulse voltage. IEEE Trans. Dielectr. Electr.
Insul. 2004, 11, 847–854. [CrossRef]
65. Wang, F.; Liang, F.W.; Chen, S.; Tan, Y.; Zhong, L.; Sun, Q. Surface charge inversion method on cylindrical insulators based on
surface potentials measured online. IEEE Trans. Dielectr. Electr. Insul. 2021, 28, 192–197. [CrossRef]
66. Gao, Y. Study on the Dynamic Characteristics of Surface Charge of Polymer Dielectrics. Ph.D. Dissertation, Tianjing University,
Tianjin, China, 2009. (In Chinese)
67. Qi, B.; Gao, C.J.; Xing, Z.L.; Chen, Z.Z.; Zhao, J.K.; Li, C.R. Distribution characteristic for surface charge on GIS insulator under
DC/ AC voltage. Proc. CSEE 2016, 36, 5990–6001.
68. Zhang, Z.; Deng, B.; Li, C.; Li, Q.; Zhang, Z.; Yan, W. Multiphysics coupled modelling in HVDC GILs: Critical re-examination of
ion mobility selection. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 835–842. [CrossRef]
69. Kindersberger, J.; Lederle, C. Surface charge decay on insulators in air and sulfurhexafluorid-part II: Measurements. IEEE Trans.
Dielectr. Electr. Insul. 2008, 15, 949–957. [CrossRef]
70. Wang, Z.Y.; Wang, J.; Li, Q.M.; Liu, S.H. 3D Simulation and experimental study of metal particles’ effect in DC GIL on surface
charge accumulation. Proc. CSEE 2016, 36, 6718–6726. (In Chinese)
71. Du, B.X.; Li, A.; Li, J. Effects of AC and pulse voltage combination on surface charge accumulation and decay of epoxy resin.
IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 2368–2376. [CrossRef]
72. Mallem, H.; Ziari, Z.; Sahli, S. Temperature effect on electrical characteristics of negative DC corona charged polyimide films.
IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 421–427. [CrossRef]
73. An, Z.; Xiao, H.; Liu, F.; Zheng, F.; Lei, Q.; Zhang, Y. Improved resistance of epoxy resin to corona discharge by direct fluorination.
IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 2278–2287. [CrossRef]
74. Wang, F.; Qiu, Y.C.; Zhang, Q.G.; Chen, Q.G. Analysis of factors influence charge accumulation on insulator surface under
impulse voltage. Trans. China Electrotech. Soc. 2001, 16, 51–54. (In Chinese)
75. Rerup, T.O.; Crichton, G.C.; Mcallister, I.W. Using the λ function to evaluate probe measurements of charged dielectric surfaces.
IEEE Trans. Dielectr. Electr. Insul. 1996, 3, 770–777. [CrossRef]
76. Kumada, A.; Okabe, S. Charge distribution measurement on a truncated cone spacer under DC voltage. IEEE Trans. Dielectr.
Electr. Insul. 2004, 11, 929–938. [CrossRef]
77. Kumada, A.; Okabe, S.; Hidaka, K. Resolution and signal processing technique of surface charge density measurement with
electrostatic probe. IEEE Trans. Dielectr. Electr. Insul. 2004, 11, 122–129. [CrossRef]

156
Energies 2023, 16, 6027

78. Ootera, H.; Nakanishi, K. Analytical method for evaluating surface charge distribution on a dielectric from capacitive probe
measurement-application to a cone-type spacer in ±500 kV DC-GIS. IEEE Trans. Power Delivery. 2002, 3, 165–172. [CrossRef]
79. Wang, D.B. The Dispersion of Surface Charge of PTFE under DC Voltage and Its Influence on Flashover Characteristics. Ph.D.
Dissertation, Chongqing University, Chongqing, China, 2015. (In Chinese)
80. Faircloth, D.C.; Allen, N.L. High resolution measurements of surface charge densities on insulator surfaces. IEEE Trans. Dielectr.
Electr. Insul. 2003, 10, 285–290. [CrossRef]
81. Liang, H.C.; Du, B.X.; Li, J. Electric field regulation and parameter optimization of surface nonlinear conductivity spacer for 500
kV DC-GIL. IEEE Trans. Dielectr. Electr. Insul. 2020, 27, 1330–1338. [CrossRef]
82. Wang, H.; Xue, J.Y.; Guo, B.H.; Deng, J.; Zhang, G. Comparison of time-dependent variation of surface charge accumulation in
SF6 /N2 and pure SF6 . In Proceedings of the 12th IEEE International Conference on the Properties and Applications of Dielectric
Materials, Xi’an, China, 20–24 May 2018.
83. Li, B.T. Study on the Mechanism of Metal Particles on Charge Accumulation on the Surface of Insulators in DC GIL. Master’s
Thesis, North China Electric Power University (Beijing), Beijing, China, 2016.
84. Guo, C.; Zhang, Q.; You, H.; Ma, J.; Li, Y.; Wen, T.; Qin, Y. Influence of electric field non-uniformity on breakdown characteristics
in SF6 /N2 gas mixtures under lightning impulse. IEEE Trans. Dielectr. Electr. Insul. 2017, 24, 2248–2258. [CrossRef]
85. Li, C.; Lin, C.; Yang, Y.; Zhang, B.; Liu, W.D.; Li, Q.; Hu, J.; He, S.; Liu, X.; He, J. Novel HVDC spacers by adaptively controlling
surface charges-part ii: Experiment. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1248–1258.
86. Qi, B.; Gao, C.J.; Xing, Z.L.; Li, C.R. The accumulation characteristic for surface charge on GIS insulator under switching impulse
voltage. Proc. CSEE 2017, 37, 4538–4547. (In Chinese)
87. Ma, G.; Zhou, H.; Li, C.; Jiang, J.; Chen, X. Designing epoxy insulators in SF6 -filled DC-GIL with simulations of ionic conduction
and surface charging. IEEE Trans. Dielectr. Electr. Insul. 2015, 22, 3312–3320. [CrossRef]
88. Li, C.Y.; Hu, J.; Lin, C.J.; He, J.L. The potentially neglected culprit of DC surface flashover: Electron migration under temperature
gradients. Sci. Rep. 2017, 7, 3271. [CrossRef] [PubMed]
89. Yan, W.; Li, C.; Lei, Z.; Han, T.; Zhang, Z.; Davide, F. Surface charging on HVDC spacers considering time-varying effect of
temperature and electric fields. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 1316–1324. [CrossRef]
90. Zhou, H.Y.; Ma, G.M.; Liu, S.P.; Li, C.R.; Zhao, S.J.; Tu, Y.P.; Qin, S.C. Study on surface charges accumulation on insulator and its
effects on the surface electrical field in DC-GIL with electro-thermal coupling model. Proc. CSEE 2017, 37, 1251–1259.
91. Zhou, H.Y.; Ma, G.M.; Zhao, S.J.; Liu, S.P.; Li, C.R.; Tu, Y.P.; Song, H.T.; Wu, Z. Effect of temperature on charge accumulation on
insulator in DC-GIL. Proc. CSEE 2016, 36, 6675–6681.
92. Li, C.Y.; Hu, J.; Li, C.J.; Zhang, B.Y.; Zhang, G.X.; He, J.L. Surface charge migration and DC surface flashover of surface-modified
epoxy-based insulators. J. Phys. D Appl. Phys. 2017, 50, 065301. [CrossRef]
93. Li, Z. Simulation of Surface Charge Accumulation on High Voltage DC Composite Insulator. Master’s Thesis, Hunan University,
Changsha, China, 2014. (In Chinese)
94. Wang, F.; Liang, F.W.; Zhong, L.; Chen, S.; Li, C.; Xie, Y. Short-time X-ray irradiation as a non-contact charge dissipation solution
for insulators in HVDC GIS/GIL. IEEE Trans. Dielectr. Electr. Insul. 2021, 28, 704–709. [CrossRef]
95. Liu, W.J. Design of Surface Charge Test and Measurement System of Insulator at Gas-Solid Interface. Master’s Thesis, Hunan
University, Changsha, China, 2008. (In Chinese)
96. Suzuki, K.; Kato, K.; Hakamata, Y.; Okubo, H. Real-time and high-speed measurements of charging processes on dielectric surface
in vacuum. IEEE Trans. Dielectr. Electr. Insul. 2003, 10, 563–568. [CrossRef]
97. Yu, K.K.; Zhang, G.J.; Zheng, N.; Raitses, Y.; Fisch, N.J. Monte Carlo simulation of surface-charging phenomena on insulators
prior to flashover in vacuum. IEEE Trans. Plasma Sci. 2009, 37, 698–704.
98. Tian, H.D.; Li, N.Y.; Wu, Z.H.; Guo, Z.; Peng, Z. Surface charge accumulation characteristics of DC GIL column insulators.
Trans. China Electrotech. Soc. 2018, 33, 38–48.
99. Zhang, Z.S.; Wang, Z.M.; Teyssedre, G.; Tohid, S.; Mohamadreza, A.B.; Chen, G.; Lin, C.; Zhang, B.; Uwe, R.; Lei, Z.; et al.
Gas-solid interface charge tailoring techniques: What we grasped and where to go. Nanotechnology 2021, 32, 122001. [CrossRef]
100. Li, Q.; Naderiallaf, H.; Lei, Z.; Wang, Y.; Liu, P.; Zhang, L.; Wang, Z.; Zhang, Z. Surface charge pattern analysis based on the
field-dependent charging theory: A review. IEEE Trans. Dielectr. Electr. Insul. 2020, 27, 257–269. [CrossRef]
101. Li, C.Y.; Lin, C.J.; Chen, G.; Tu, Y.P.; Zhou, Y.; Li, Q.; Zhang, B.; He, J.L. Field-dependent charging phenomenon of HVDC spacers
based on dominant charge behaviors. Appl. Phys. Lett. 2019, 114, 202904. [CrossRef]
102. Zhang, B.Y.; Zhang, G.X. Review of charge accumulation characteristics at gas-solid interface in DC GIL, part II: Charge control
and suppression strategy. Trans. China Electrotech. Soc. 2018, 33, 5–18.
103. Li, C.Y.; Lin, C.J.; Zhang, B.; Li, Q.; Liu, W.D.; Hu, J.; He, J.L. Understanding surface charge accumulation and surface flashover
on spacers in compressed gas insulation. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1152–1166. [CrossRef]
104. Zhang, B.Y.; Gao, W.Y.; Hou, Y.; Zhang, G.X. Surface charge accumulation and suppression on fullerene-filled epoxy-resin
insulator under DC voltage. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 2011–2019. [CrossRef]
105. Wang, S.; Li, A.; Xing, Y.Q.; Li, K.; Li, Z.K. Surface potential and breakdown characteristics of epoxy/AlN nanocomposites under
DC and pulse voltages. IEEE Trans. Appl. Supercond. 2019, 29, 1–5. [CrossRef]
106. Tu, Y.P.; Zhou, F.W.; Cheng, Y.; Jiang, H.; Wang, C.; Bai, F.J.; Lin, J. The control mechanism of micron and nano SiO2 /epoxy
composite coating on surface charge in epoxy resin. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1275–1284. [CrossRef]

157
Energies 2023, 16, 6027

107. Gao, Y.; Li, Z.Y.; Zhao, N.; Wang, M.H.; Han, T.; Liu, Y. Surface charging on epoxy/Al2 O3 nanocomposites under DC voltage
superimposed by repetitive pulses. IEEE Trans. Plasma Sci. 2019, 47, 3256–3265. [CrossRef]
108. Gao, Y.; Li, Z.Y.; Wang, M.; Du, B.X. Magnetic field induced variation in surface charge accumulation behavior on epoxy/Al2 O3
nanocomposites under DC stress. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 859–867. [CrossRef]
109. Du, B.X.; Zhang, J.W.; Gao, Y. Effect of nanosecond rise time of pulse voltage on the surface charge of epoxy/TiO2 nanocomposites.
IEEE Trans. Dielectr. Electr. Insul. 2013, 20, 321–328. [CrossRef]
110. Du, B.X.; Xiao, M. Influence of surface charge on DC flashover characteristics of epoxy/BN nanocomposites. IEEE Trans. Dielectr.
Electr. Insul. 2014, 21, 529–536. [CrossRef]
111. Du, B.X.; Wang, M.Y.; Li, J.; Xing, Y.Q. Temperature dependent surface charge and discharge behavior of epoxy/AlN nanocom-
posites. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 1300–1307. [CrossRef]
112. Chen, Y.; Wu, J. Investigation on relationship between breakdown strength enhancement of composites and dielectric characteris-
tics of nanoparticle. IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 927–934. [CrossRef]
113. Du, B.X.; Zhang, J.; Gao, Y. Effects of TiO2 particles on surface charge of epoxy nanocomposites. IEEE Trans. Dielectr. Electr. Insul.
2012, 19, 755–762. [CrossRef]
114. Ma, J.; Tao, F.; Ma, Y.; Zhao, K.; Li, H.; Wen, T.; Zhang, Q. Quantitative analysis on the influence of surface charges on flashover of
insulators in SF6 . IEEE Trans. Dielectr. Electr. Insul. 2021, 28, 274–281. [CrossRef]
115. Zhang, B.Y.; Wang, Q.; Qi, J.; Zhang, G.X. Measurement method and accumulation characteristics of surface charge distribution
on polymeric material under DC voltage. Proc. CSEE 2016, 36, 6664–6674.
116. LI, C.Y.; Lin, C.J.; Chen, G.; Zhang, Z.S.; Zhang, L.; He, J.L. Review of gas-solid interface charging phenomena of HVDC spacers.
Proc. CSEE 2020, 40, 2016–2025.
117. Xie, Q.; Liang, S.D.; Fu, K.X.; Liu, L.Z.; Huang, H.; Lu, F.C. Distribution of polymer surface charge under DC voltage and its
influence on surface flashover characteristics. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 2157–2168. [CrossRef]
118. Nakane, R.; Hayakawa, N.; Okubo, H. Time and space transition of DC electric field distribution based on emerging field by
accumulated charge in gas-solid composite insulation structure. IEEE Trans. Dielectr. Electr. Insul. 2021, 28, 440–447. [CrossRef]
119. Zhang, C.; Ma, Y.Y.; Kong, F.; Wang, R.X.; Ren, C.Y.; Tao, S. Surface charge decay of epoxy resin treated by AP-DBD deposition
and direct fluorination. IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 768–775. [CrossRef]
120. An, Z.; Liu, F.; Tang, Y.; Zheng, F.; Zhang, Y. Resistance of surface fluorinated epoxy resin to corona discharge in SF6 gas.
IEEE Trans. Dielectr. Electr. Insul. 2016, 23, 3659–3667. [CrossRef]
121. Lin, H.F.; Wang, R.X.; Xie, Q.; Zhang, S.; Shao, T. Rapid surface modification by plasma jet to promote surface charge decaying.
Trans. China Electrotech. Soc. 2017, 16, 262–270. (In Chinese)
122. Hu, D.; Ren, C.Y.; Zhang, C.; Qiu, J.T.; Kong, F.; Shao, T.; Yan, P. Effect of deposited film on the surface insulation characteristics of
FEP material by atmospheric pressure plasma jet. Proc. CSEE 2019, 39, 4633–4640. (In Chinese)
123. Kong, F.; Zhang, S.; Lin, H.F.; Ren, C.Y.; Zhang, C.; Shao, T. Effects of nanosecond pulse voltage parameters on characteristics of
surface charge for epoxy resin. IEEE Trans. Dielectr. Electr. Insul. 2018, 25, 2058–2066. [CrossRef]
124. Li, J.; Liang, H.C.; Du, B.X.; Wang, Z.H. Surface functional graded spacer for compact HVDC gaseous insulated system.
IEEE Trans. Dielectr. Electr. Insul. 2019, 26, 664–667. [CrossRef]
125. Du, B.X.; Ran, Z.Y.; Liang, H.C. Novel insulator with interfacial σ-FGM for DC compact gaseous insulated pipeline. IEEE Trans.
Dielectr. Electr. Insul. 2019, 26, 818–825. [CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.

158
 MDPI AG
Grosspeteranlage 5
4052 Basel
Switzerland
Tel.: +41 61 683 77 34

Energies Editorial Office

E-mail: energies@mdpi.com
www.mdpi.com/journal/energies

Disclaimer/Publisher’s Note: The title and front matter of this reprint are at the discretion of the
Guest Editors. The publisher is not responsible for their content or any associated concerns. The
statements, opinions and data contained in all individual articles are solely those of the individual
Editors and contributors and not of MDPI. MDPI disclaims responsibility for any injury to people or
property resulting from any ideas, methods, instructions or products referred to in the content.
Academic Open
Access Publishing

mdpi.com ISBN 978-3-7258-4790-7