18ECP109L-PROJECT

PREDECTIVE ENERGY MANAGEMENT

SYSTEM FOR HYBRID ELECTRIC VEHICLES
WITH IOT-DRIVEN SENSORS

A PROJECT REPORT

Submitted by

Yetigadda Shoyab [Reg No: RA2111004010174]

Jakka Narayana [Reg No: RA2111004010171]
Utkarsh Johari [Reg No: RA2111004010195]

Under the Guidance of

Dr. Anilet Bala A
(Assistant Professor, Department of Electronics & Communication Engineering)

In partial fulfillment of the requirements for

the Degree of
BACHELOR OF TECHNOLOGY
in

ELECTRONICS & COMMUNICATION ENGINEERING

DEPARTMENT OF ELECTRONICS AND COMMUNICATION

ENGINEERING
COLLEGE OF ENGINEERING AND TECHNOLOGY
SRM INSTITUTE OF SCIENCE AND TECHNOLOGY
KATTANKULATHUR-603 203

MAY 2025
 SRM INSTITUTE OF SCIENCE AND
TECHNOLOGY
(Under Section 3 of UGC Act,1956)

BONAFIDE CERTIFICATE

Certified that this project report titled “PREDECTIVE ENERGY

MANAGEMENT SYSTEM FOR HYBRID ELECTRIC VEHICLES WITH
IOT-DRIVEN SENSORS” is the bonafide work of “YETIGADDA SHOYAB
[RA2111004010174], JAKKA NARAYANA [RA2111004010171] and
UTKARSH JOHARI [RA2111004010195], who carried out the project work
under my supervision. Certified further, that to the best of my knowledge the
work reported herein does not form any other project report or dissertation on the
basis of which a degree or award was conferred on an earlier occasion on this or
any other candidate.

SIGNATURE SIGNATURE

Dr. ANILET BALA A Dr. M. SANGEETHA

GUIDE HEAD OF THE DEPARTMENT
Assistant Professor Dept. of Electronics &
Dept. of Electronics & Communication Engineering
Communication Engineering

Signature of the Internal Examiner Signature of the External Examiner

ii
 DECLARATION

We hereby declare that the Major Project entitled “PREDECTIVE ENERGY

MANAGEMENT SYSTEM FOR HYBRID ELECTRIC VEHICLES WITH
IOT-DRIVEN SENSORS” to be submitted for the Degree of Bachelor of
Technology is our original work as a team and the dissertation has not formed the
basis of any degree, diploma, associate-ship or fellowship of similar other titles. It
has not been submitted to any other University or institution for the award of any
degree or diploma.

Place:

Date:

Yetigadda Shoyab
[RA2111004010174
]

Narayana Jakka
[RA2111004010171]

Utkarsh Johari
[RA2111004010195
]

iii
 ACKNOWLEDGEMENT

We would like to express my deepest gratitude to the entire management of SRM Institute of
Science and Technology for providing me with the necessary facilities for the completion of
this project.

We wish to express my deep sense of gratitude and sincere thanks to our Professor and Head of
the Department Dr. M. Sangeetha, for her encouragement, timely help, and advice offered to
me.

We are very grateful to my guide Dr. Anilet Bala A Assistant Professor, Department of
Electronics and Communication Engineering, who has guided me with inspiring dedication,
untiring efforts, and tremendous enthusiasm in making this project successful and presentable.

We would like to express my sincere thanks to the project coordinator Dr. M. Sangeetha and
panel members Dr. Kanaparthi V Phani Kumar and Dr. S. Sunithamani for their time and
suggestions for the implementation of this project.

We also extend my gratitude and heartful thanks to all the teaching and non-teaching staff of
the Electronics and Communications Engineering Department and to my parents and friends,
who extended their kind cooperation using valuable suggestions and timely help
during this project work.

Yetigadda Shoyab
[RA2111004010174
]

Jakka Narayana
[RA2111004010195]

Utkarsh Johari
[RA2111004010195]

iv
 ABSTRACT

Traditional energy management in HEVs has disadvantages such as reactive fault

detection, inefficient power sharing, and the lack of online monitoring, therefore

leading to energy loss and unexpected vehicle failure. Our system integrates

MEMS sensors for vehicle motion and tilt monitoring with terrain-based and road

condition-based adaptive speed control. It employs a VEL-Net hybrid model for

predictive anomaly detection. Hence, the dynamic increase or decrease in motor

speed depending on the optimization needed in the consumption of energy is

targeting efficient power usage and optimal safety and vehicles. The proposed

hardware system integrates the data from five IoT sensors such as temperature,

vibration, distance, light intensity, and angle to monitor vehicle health and

decrease energy consumption. Utilizing the hardware system sensor data, 4,000

multivariate time-series samples are collected. The collected data are trained with

convolutional neural networks for spatial feature extraction and long short-term

memory network for temporal pattern extraction. Real-time prediction is attained

with processing latencies of less than 100 milliseconds, making the system highly

adoptable for on-vehicle deployment. The results achieve considerable

improvement over the current state of the art, providing a 25% performance

improvement over single-CNN configurations and a significant decrease in false

positives. With 98.87% accuracy, 0.99 precision, and 0.99 F1-score, the current

work further supports the agenda of intelligent transport systems by achieving an

v
highly robust framework for predictive maintenance and energy optimization.

This project is aligned with SDG Goal 9 – Industry, Innovation, and Infrastructure

through the development of resilient infrastructure, fostering inclusive and

sustainable industrialization, and promoting innovation to improve economic

growth, generate jobs, and raise living standards while confronting environmental

challenges.

vi
 TABLE OF CONTENTS

CHAPTER NO. TITLE PAGE NO

ABSTRACT v
LIST OF TABLES viii
LIST OF FIGURES ix
ABBREVIATIONS x

1 INTRODUCTION 1
1.1 Purpose 5
1.2 Aim and Objectives 6
1.3 Motivation 6
1.4 Innovation idea of the Project 7
1.5 Scope and Application of the Project 7
1.6 Requirement Specification 8
1.7 Summary 9

2 LITERATURE SURVEY 12
2.1 Works related Predictive Energy Management 13
2.2 Challenges and Limitations 13

3 SYSTEM ANALYSIS 15
3.1 Existing Model 16
3.2 Proposed Model 16

SYSTEM ARCHITECTURE AND DESIGN 16

4
4.1 Architecture Overview 17
4.2 Block Diagram 17
4.3 Architecture of V^2I-NET 18

5 METHODOLOGY 20
5.1 Data Collection
22

vii
 5.2 Feature Extraction 23
5.3 Feature Selection 25
5.4 Model Training 25
5.5 Model Evaluation 26

6 RESULTS & DISCUSSION 28

6.1 Results & Discussion 28

7 CONCLUSION & FUTURE WORK 34

REFERENCES 36

viii
 LIST OF TABLES

TABLE NO. TITLE PAGE NO

6.1 Comparison of Outcomes between proposed and existing 32

models

ix
 LIST OF FIGURES

FIGURE NO. TITLE PAGE NO

Architecture Overview
4.1 17
Architecture Diagram
4.2 18
Efficient Net B6
4.3 20
Architecture 25
5.1
Confusion Matrix of Efficient Net B6

x
 ABBREVIATIONS

ABBREVIATIONS DESCRIPTION

VGG Visual Geometry Group

CNN Convolutional Neural Network

SVM Support Vector Machine

ReLU Rectified Linear Unit

HEV Hybrid Electric Vehicle

GPS Global Positioning System

GSM Global System for Mobile Communications

IOT Internet Of Things

xi
 CHAPTER 1

INTRODUCTIO

The growing need for environmentally sustainable transportation has led to the increased
reliance on Hybrid Electric Vehicles (HEVs), which combine the strengths of internal
combustion engines and electric motors to reduce fuel consumption and emissions. However, in
spite of the numerous advantages offered by HEVs in environmental sustainability, HEVs also
pose tremendous challenges in energy management. Traditional energy management systems
employed in HEVs are limited, e.g., reactive fault detection, suboptimal power partitioning
between the engine and electric motor, and the absence of online monitoring, which can lead to
energy loss, vehicle performance degradation, and abrupt malfunctions. As the deployment of
HEVs is growing, there is an immediate need for advanced systems that can effectively address
these challenges in real time, thus ensuring maximum energy utilization, improved vehicle
safety, and improved reliability.

One of the most promising solutions to this problem is the use of a Predictive Energy
Management System (PEMS) for HEVs, which employs Internet of Things (IoT)-based sensors
to monitor critical vehicle parameters and modify energy management accordingly. Not only
does this system provide optimal power usage, but it also enables proactive maintenance,
allowing the system to anticipate potential faults before they become significant issues.
Through continuous data gathering and processing from a variety of sensors embedded within
the vehicle, the PEMS can detect changes in environmental and operational conditions,
enabling the system to make real-time adjustments to the energy flow of the vehicle. This
results in enhanced energy efficiency, reduced energy loss, and enhanced overall vehicle
performance. Furthermore, predictive algorithms, powered by advanced machine learning
models such as convolutional neural networks (CNNs) and long short-term memory (LSTM)

12
networks, enable the system to recognize patterns in sensor data, offering valuable information
regarding the health of the vehicle and its interaction with the outside environment.

At the center of the predictive energy management system is a large-scale sensor network made
up of numerous IoT sensors, each of which has the responsibility of tracking various aspects of
the vehicle's operation. These sensors feed information on critical parameters such as
temperature, vibration, distance, light intensity, and vehicle tilt and motion. Based on this
multi-sensor information, the system can monitor the vehicle's condition in real-time, identify
potential problems, and make real-time adjustments to optimize energy consumption. For
example, if the system identifies that the vehicle is operating on a steep slope or challenging
terrain, it can make motor speed adjustments accordingly, ensuring that energy consumption is
minimized while ensuring optimal vehicle performance. Similarly, by tracking parameters such
as temperature and vibration, the system can identify early warning signs of potential
mechanical failure, enabling timely maintenance and reducing downtime.

The predictive energy management system's most key component is a VEL-Net hybrid model
for predictive anomaly detection. It combines Vehicle-to-Infrastructure (V2I) communication
with predictive analytics through neural networks, creating a strong framework for predicting
anomalies and enabling proactive realignment of the power management system in the vehicle.
The V2I communication involves interaction between the vehicle and its surrounding
infrastructure such as traffic signals, road sensors, and nearby vehicles, which can provide
context information. Introducing this context information into the system helps advance its
understanding of road conditions, traffic movement, and potential risks, thereby providing more
accurate prediction and improving methods of energy management.

The improvements in energy efficiency achieved by the system are seen in the voltage drop
measured through its operation. Experimental evidence demonstrates the system provides an
average voltage drop of 2.56V, which enhances energy efficiency by reducing power loss
through energy conversion and distribution process. This voltage drop reduction also serves to
enhance the life of the vehicle's electrical systems, including the battery and powertrain, by
alleviating stress on the system and ensuring energy consumption is optimized. With regard to
accuracy, the system suggested is impressive in its performance, boasting a 99.25% accuracy,
1.00 precision, and 1.00 F1-score. These performance metrics are a sign that high effectiveness
13
and reliability in predicting potential issues and optimizing energy usage are traits of the
system. Being highly accurate with minimal false positives ensures that vehicle energy
management decisions are made from precise information, ultimately leading to enhanced
safety, operational efficiency, and energy saving. In brief, the Predictive Energy Management
System for HEVs with IoT-driven Sensors is a breakthrough in the development of intelligent
transportation systems. By utilizing advanced technologies like IoT sensors, machine learning,
and predictive analytics, the system provides a strong and efficient platform for energy
consumption management, fault detection, and vehicle performance optimization. The system's
capability to react in real time to a vast array of sensor inputs, combined with its remarkable
performance records, makes the system an ideal solution for the next generation of hybrid
electric vehicles. With its capability to overcome the limitations of conventional energy
management systems and provide substantial improvement in energy efficiency, predictive
maintenance, and fault detection, the proposed system has the potential to play a crucial role in
redefining the future of sustainable and intelligent transportation.

1.1 PURPOSE

The core concept of the Predictive Energy Management System for Hydrocarbon Electric
Vehicles (HEVs) working with sensing from IoT driven sensors, is primarily driven due to the
traditional energy management of HEVs, creating the need for management in terms of energy
consumption optimization whilst ensuring enhanced detection of faults through vehicle health
monitoring in real-time. HEVs utilize internal combustion engines along with electric motors,
and while the combination improves fuel efficiency and decreases emissions, the two systems
do not share control well resulting in poor efficiency and reactive fault detection. Such
challenges typically result in energy being wasted, poor performance of the vehicle, and
unanticipated failures.

To address these issues, we have developed this system that integrates a network of IoT sensors
responsible for monitoring temperature, vibration, distance, light intensity, and tilt in the
vehicle, allowing to customize and prevent future problems. These are sensors that can stream
constant information about the state of the vehicle and its surroundings, allowing adjustments
to the energy management system on the fly. Using complex machine learning methods, such
as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks), the

14
system examines a multivariate sensor data to detect patterns and facilitate predictions on
potential anomalies, which assists in determining when an issue will affect the operation of the
vehicle.

The primary goal is to make energy usage more efficient, to adjust how the vehicle’s power is
allocated according to terrain and road conditions and foresee when maintenance is required to
avoid system failures. Using predictive capabilities also enables dynamic motor speed control,
allowing more efficient energy use and less energy waste. It will also lead to more advanced,
energy-efficient, and environmentally friendly transportation systems in line with cutting-edge,
sustainable mobility objectives.

1.2 AIM AND OBJECTIVES

The IoT based Predictive Energy Management System for Hybrid Electric Vehicles(HEVs)
focuses on energy efficiency, energy consumption, fault detection, improved safety, and
increased vehicle performance. Utilising Internet of Things (IoT) sensors along with machine
learning techniques to monitor real-time vehicle parameters, predict predictive anomaly as well
as terrain and road conditions-based energy management. It aims to prevent energy wastage,
predictively service components to prevent unscheduled downtime while increasing the
efficiency of propulsion, reducing target failures, and monitoring motor-to-battery ratios to
distribute power effectively for a reliable and eco-friendly vehicle.

1.3 MOTIVATION

The rationale for developing a Predictive Energy Management System for Hybrid Electric
Vehicles (HEVs) is rooted in the increasing demand for more energy-efficient and sustainable
modes of transportation. Due to the passive energy distribution and reactive fault detection of
conventional HEV energy management systems, HEVs suffer more from energy loss and failed
operation — particularly high performance hybrid vehicles. Incorporating IoT sensors and
cutting-edge machine learning algorithms, this system attempts to break through the confines of
classical control, making energy usage more efficient and avoiding unanticipated breakdowns.
Their aim is to improve HEV performance, decrease energy consumption, and help create
smarter, environmentally friendly, and resilient transportation systems.

15
1.4 INNOVATION IDEA OF THE PROJECT

The innovation here is the use of IoT driven sensors and machine learning to optimize energy
usage and vehicle performance in real-time with the Predictive Energy Management System
for Hybrid Electric Vehicles (HEVs). This differs from an ordinary system where the energy
flow within the vehicle is either static or reactive — meaning a response comes too late to be
effective — to any readout only. The system leverages convolutional neural networks (CNNs)
to extract spatial features and long short-term memory (LSTM) networks to analyze temporal
patterns, not just identifies operational inefficiencies, but also predicts potential faults before
they can adversely affect vehicle performance.

Indeed, the paper goes on to describe a Vehicle-to-Infrastructure (V2I)-Net hybrid model that
uses external infrastructure data (such as road conditions and traffic patterns) to provide
external inputs for the energy management problem, further improving predictive capabilities.
These ensure that system can adapt to different road conditions and terrain, optimizing energy
usage while preserving vehicle safety and performance.

It helps with predictive maintenance, minimizing energy loss and developing more accurate
fault detection for more sustainable, efficient and reliable hybrid vehicle. Leveraging IoT
sensors, real-time monitoring, and predictive analytics, the project provides a proactive
approach to next-gen hybrid vehicles and intelligent transport systems

1.5 SCOPE AND APPLICATION OF THE PROJECT

The scope of the Predictive Energy Management System for Hybrid Electric Vehicles (HEVs)
extends to optimizing energy consumption, enhancing vehicle safety, and enabling proactive
maintenance in hybrid vehicles. The system integrates IoT-driven sensors to monitor real-time
vehicle conditions, such as temperature, vibration, distance, and tilt, while using advanced
machine learning algorithms to predict anomalies and adjust energy management dynamically.
The project is applicable to HEVs across various environments, ensuring efficient power

16
distribution between the electric motor and internal combustion engine, reducing energy loss,
and preventing unexpected failures.

Its applications extend beyond individual vehicle management to broader intelligent

transportation systems (ITS). The system's predictive maintenance capabilities can be
implemented in fleet management, providing vehicle operators with insights into vehicle health
and optimizing energy usage. Additionally, this technology can support the development of
sustainable transportation infrastructures by integrating with Vehicle-to-Infrastructure (V2I)
systems, improving overall traffic flow, reducing emissions, and contributing to the transition
towards more eco-friendly and efficient mobility solutions.

1.6 REQUIREMENT SPECIFICATIONS

SOFTWARE REQUIREMENTS:

· Arduino IDE

· Django Webserver

· Random Forest Algorithm

· CNN+LSTM Algorithm

· VEL-NET Algorithm

· Dashboard

HARDWARE REQUIREMENTS

1.1 Sensors

· Ultrasonic Sensor (HC-SR04)

· Temperature Sensor (LM35)

· Vibration Sensor (SW-420)

· LDR (Light Dependent Resistor)

17
· Infrared Sensor

· MEMS Sensor (ADXL345)

1.2 Communication Modules

· ESP32 Microcontroller

· GPS Module (NEO-6M)

· GSM Module (SIM800L)

1.7 SUMMARY

This system consists of Hybrid Electric Vehicles (HEVs) with embedded IoT sensors (IOTs) to
monitor energy efficiency and enhance vehicle safety using a novel Predictive Energy
Management System (PEMS). These sensors track vital parameters in real-time, including
temperature, vibrations, light levels, distance, and vehicle orientation. The data is processed
using an ESP32 microcontroller and then sent to a IOT web server for real-time monitoring
and decision making.

For Energy consumption prediction and effective failure prediction machine learning
algorithms like Random Forest, CNN+LSTM, and VEL-NET are established. (These
algorithms detect anomalies such as overheating, excessive vibrations, and abnormal tilts,
before adapting motor speed accordingly, using MEMS sensor data which is analyzed by these
machines.)

To enhance security, a GPS and GSM modules are used for real-time location tracking, and
also send emergency alerts in case of critical faults. Lastly, a dashboard allows users to
visualize and control vehicle parameters remotely.

This trove of data becomes a powerful tool when harnessed through advanced analytics -
employing IoT, machine learning, and real-time data processing to maximize predictive
maintenance, fuel efficiency, and overall performance of HEVs. This is a significant
breakthrough for the modern transport ecosystem, as its advanced energy management
potential brings reliability, sustainability, and improved safety.
18
 CHAPTER 2

LITERATURE STUDY

2.1 WORKS RELATED TO PREDECTIVE ENERGY MANAGEMENT SYSTEM

There have been a number of research works done integrating internet of things, machine
learning, and sensor technology in vehicular monitoring and control systems during the last
couple of years. This survey discusses 15 of the most relevant literature guiding the project.

Nikhil et al. in 2023 have proposed an IoT-Enabled Battery Monitoring System that optimizes
the performance of an electric vehicle. The work discusses the IoT-based real-time diagnosis of
the battery. The measured parameters, in this case, include tracking location and residual
capacity of the battery. In this system, users will be alerted with any available data that may
help in timely reliability and safety enhancement.

Teddy Surya Gunawan et al. developed a prototype VCAS using Arduino and GPS technology
in the year 2020. The ultrasonic sensors measure the distance between the vehicles. From the
data obtained for the speed and acceleration, the braking distance is computed. When the
following distance is less than the computed braking distance, the system will give an alarm or
performs assisted braking. The results of the experiments presented in the paper demonstrated
that the system performed worse on surfaces softer than rough, paved roads. However,
environmental factors influenced the accuracy of GPS, and the processing of position data was
required.

Qi Feng et al. conducted a comprehensive survey on deep anomaly detection methods that can
be used on electric vehicle charging data. The research study evaluated the unsupervised and
semi-supervised deep anomaly detection algorithms through discussion of merits and demerits
of both techniques. This research has identified the possibility of anomalies in electric vehicle
charging systems and improves reliability and safety features in such systems.

A.Srinagesh et al. (2025) have presented an IoT-based vehicle tracking system using GPS-
enabled fuel sensor for the upgradation of fleet management. It tracks the vehicle's fuel
19
consumption, detects route deviation events and evaluates its driving behavior, and gives real-
time data on vehicle maintenance. The proposed solution with GPS-based sensors and GSM
tracking allows one to get instant information over vehicle fuel usage and its location using a
cloud-based mobile application. In addition, the system sends alarm messages on consumption
pattern, which may prevent fuel theft.

Mid-Eum Choi et al. In 2020, they provide the optimal energy management scheme for active
Hybrid Energy Storage Systems consisting of batteries and supercapacitors. Formulated as an
optimization task, their goal is the minimization of both the magnitude and fluctuation of
battery current and energy loss in supercapacitors. The multiplicative increase-additive-
decrease principle algorithm will change parameter values appropriately toward an optimal-
feasible solution. The technique was demonstrated for the effectiveness by MATLAB
simulations against the fluctuations due to battery currents and energy lost in the
supercapacitor.

Varsha Goud et al. (2020) invented Vehicle Accident Automatic Detection and Remote Alarm
Device where the aim would be to shorten the time difference in emergency service response.
The device relies on vibration sensors and Micro Electro-Mechanical Systems (MEMS)
accelerometers for detecting accidents or rollover. The device would transmit to a police
control room or rescue team an alert message, including the location of the vehicle by GSM and
GPS modules. This method certainly provides prompt help, may be reduced to the minimum of
response time and hence good consequences for the people involved in an accident.

Mareddy Anusha et al. (2024) have proposed Internet of Things / IoT-based Energy monitoring
system by using ESP32 Microcontroller and ThingSpeak Platform. It assures the real-time
energy consumption. So, appropriate utilization of energy will be made, and therefore the right
decision is taken accordingly. ESP32 will aid in extracting data more effectively, and it is
coupled with ThingSpeak Platform that can analyse and visualized the data from the cloud
environment. In this way, these endeavors directly contribute to the act of energy-saving
through intelligent decisions on energy usage.

Andreas Theissler et al. (2021) developed a comprehensive overview of predictive maintenance

in the automobile sector by using machine learning. They identified and categorized use cases
that include optimum maintenance schedule given by machine learning, reduction in downtime,
prediction of vehicle failure, and categorized key challenges regarding data quality as well as
need for scalable model and sensor integration. It has also outlined some possible directions for

20
future research in providing an efficient predictive maintenance system through underlining the
role played by advanced analytics in improving the performance and cost reduction in
operational activities.

Mia Arma Desima et al. developed an alarm for IoT-based location in case public vehicles get
involved into an accident 2017. It works from IoT-based on public vehicles accident, detecting
this type of an accident at once in real-time with exact and accurate locating; such system with
an algorithm gives proper timing-related data for higher time management related to an incident
with improvement regarding passengers' security and enhancing safety conditions while
arriving and saving it.

Vinay Chamola et al., (2020) developed a smart framework that combines IoT and Edge
computing to enhance the V2G operations, underlining efficient charge scheduling and optimal
selection of EVs, finally for the goal of enhancing the grid stability and energy supply. It is
operational on IoT devices and Edge computing to process the real time data for decision
making to ensure the smooth interlinking between the EVs and the power grid.

P. Poobalan et al. (2015) designed a GPS and GSM-based accident location indicator and
rescue system to be used for faster emergency response. The system applies GPS technology
for the determination of the accident location and GSM technology for the provision of real-
time alerts to the emergency services. This will make rescue operations more efficient, thus
possibly saving more lives by providing accurate location information to the responding
personnel. It was published in the Journal of Advance Research in Electrical & Electronics
Engineering.

Ali Sayah et al. 2024. Energy management systems' design for fuel cell hybrid electric vehicles,
considering road gradient information. Optimize energy efficiency and performance. The
system receives real-time gradients information of road, which achieves an optimum transfer of
power among the fuel cell, battery, and electrical motor to ensure enhanced energy usage on
both uphill and downhill driving situations. The vehicle performance has increased, and its
hydrogen consumption along with the longevity of the used batteries have come down due to a
reduction in Terrain Variations. The effectiveness of the system is further demonstrated by the
simulation of the system and case studies in real applications.

Achikkulath et al. (2021) presented the optimization framework of hybrid integrated combined
energy systems associated with adaptive energy management strategies in HEVs. Paper to core
point is as follows: an optimization of the power flow is presented for energies from sources in
21
the combined modes of battery power and ICE power, for giving improved efficiencies by
different drive modes. The adaptive energy management system will outperform other designs
of the systems optimized for power distribution perceived for any driving condition and ensure
a compromise between the optimum fuel consumption and emissions. The improvements
obtained in the vehicle's energy and performance consumptions are validated with the result of
the simulation in the validation step.

Praveena Krishna P. S et al. (2020) provided the energy management strategies of HESS in
EVs to make it realize a better performance and optimal utilization of energy. In the current
research work, the hybrid energy storage system, comprising of a battery as well as a
supercapacitor, will be presented and then the energy will be distributed under drive conditions
and load's demands thereafter. The strategies are towards better energy usage, extended battery
life, and reliable operations in the system to uphold dynamic operations. Thus, the simulation
results are good and confirm suitability for such strategies toward improved balance in energy
management, including overall efficiency of such a system.

Advanced Energy Management Strategy for Hybrid Electric Vehicles Hongwen He et al.
(2024) proposed multi-state of vehicle-traffic information toward integration in enhancement of
energy efficiency and driving performance. Such a strategy applies the use of real-time flow,
congestion level as well as state of the vehicle about traffic information to optimize the energy
to be shared between an internal combustion engine and electric motor.

The system is thus so designed to tune dynamic power usage that in the regulation of fuel
consumption, emissions are up to the level by which the system works on efficiency,
considering varied conditions of traffic. Thus, the simulation proves workability greatly
enhancing energy efficiency and system adaptability.

2.2 CHALLENGES AND LIMITATIONS

· High Processing Loads – Real-time sensor data needs to be processed in large volumes
which takes substantial processing power, causing Embedded Microcontrollers to break down
and leading to delays in decision making. Real-time execution of AI models is still a hurdle to
be crossed.

22
· Connectivity Issues – Internet and network connections must remain stable for the system
to process data inreal time. IoT communication can be inhibited in remote areas with poor
connectivity, which can delay predictive maintenance or decrease the effectiveness of
emergency response.

· Cybersecurity Risks – HEVs with Internet-Connected IoT systems communicate sensitive

data with various networks that can be subjected to hacking, unauthorized access, and cyber-
attacks. Secure design and implementation and encryption are a must but complicate things
considerably.

· Accuracy and reliability of sensors — The consistency of sensor performance is determined

by environmental variables like temperature, humidity, or mechanical degradation. To maintain
uniformity of performance, regular calibration and sensor maintenance is needed.

· Expensive to Implement — Using several sensors, IoT platforms, and AI models increases
the overall cost of deployment and development. Such advanced systems may prove difficult
for small-scale manufacturers or budget-conscious consumers to adopt.

CHAPTER 3
SYSTEM ANALYSIS

3.1 EXISTING MODEL

23
The existing energy management systems implemented in Hybrid Electric Vehicles (HEVs)
emphasize on the optimization of power distribution from internal combustion engines to
electric motors, enhancing the efficiency. Conventional models use rule based algorithms,
which switching between power sources based on pre-defined thresholds considering battery
charge, speed and acceleration. While useful, these models have limited adaptability to real-
time conditions and external factors.

At the more sophisticated end of the scale are fuzzy logic and model predictive control (MPC),
which optimise the energy consumption of vehicles by dynamically tuning the command and
power consumption to driving behaviour and road conditions. However, these models usually
consume large amounts of computational resources, and might not consider unexplained
mechanical failures.

Machine learning-based models also appeared that use historical data to forecast energy
consumption patterns. Many existing models lack seamless integration with IoT sensors that
enable real-time monitoring, which would allow for the immediate detection of anomalies and
adjustments to vehicle parameters as needed. The novel predictive energy management system
resolves these limitations by utilizing IoT, AI, and real-time data analytics for advanced
performance.

3.2 PROPOSED MODEL

The proposed Predictive Energy Management System (PEMS) applies IoT sensors, machine
learning algorithms, and real-time data processing to the realm of Hybrid Electric Vehicles
(HEVs) to improve overall efficiency and enhance safety. It uses ultrasonic, temperature,
vibration, MEMS, and infrared sensors to monitor vehicle condition in real-time, troubleshoot
faults, and optimize energy consumption. This runs on ESP32 microcontroller that reads sensor
data and sends it to a IoT platform, which serves real-time monitoring and user control.

Data is queried using machine learning models including but not limited to Random Forest,
CNN+LSTM and VEL-NET, which identifies faults in the electric component enabling
preventive maintenance and reducing energy exhaustion. For critical failures, it has emergency
response mechanisms that activate auto engine shutdown and real-time GSM and GPS alert
modules. Additional features include real-time visualization via a web dashboard for remote
performance monitoring.
24
This AI based IoT model is a great assist to vehicles with safety, energy efficiency, and fault
detection in real time in advance which makes HEVs more sustainable and completely reliable.

Fig.2 Hardware setup of the proposed method.

CHAPTER 4

SYSTEM ARCHITECTURE AND DESIGN

4.1 ARCHITECTURE OVERVIEW

25
 Figure 4.1: Architecture overview

This vehicle safety system is based on ATMEGA 328 microcontroller and consists of sensors
and communication module. Sensors for motion, collision, fire, light, temperature, vibration
(accidents) and voltage monitoring. Data processing and interaction with external modules
such as GPS for tracking, IoT for cloud monitoring, and GSM for emergency alerts are
performed by the microcontroller. Motor driver circuit drives the motor and buzzer is used for
indicating safety. The device talks to ESP32 via UART for extended connectability. The
component provides collision prevention, fire detection, accident alerts and real time tracking,
thus being suitable for smart vehicle monitoring and safety applications.

26
 4.2 BLOCK DIAGRAM

Figure 4.2: Block Diagram

It is a Deep Learning-based predictive system using sensor data for analysis and decision-
making as shown in this diagram; There are five steps in the workflow:

· Data Pre-processing: The initial sensor data goes through a transformation into a binary
representation, which aligns better with the expectations of machine learning algorithms. Never
miss a chance to clean the data and take your data structures to the next level.

· Training-Testing Split: The dataset is split into 80% statistics for training and 20% statistics
for testing to build an accurate and reliable model. So the training data is used for learning and
testing data is used for evaluating the model performance.

· Deep Learning (V²I-NET): A dedicated deep learning model, V²I-NET is trained. This model
analyzes input data, identifies relationships, and learns patterns that reflect behavior so it can make
increasingly informed predictions.

· Prediction: After training, the model predicts results from new sensor data, allowing for instant
insights and intelligent decision-making.

Applications autonomous systems, industrial automation, predictive maintenance, smart

environments, and more benefit from this system, deep learning improves accuracy and efficiency.

27
 ENGINEERING STANDARDS

1. Microcontrollers & Modules

· The Arduino Uno will support those USB 2.0 and Atmel AVR microcontroller
standardsSupport 5V output and except CE & RoHS certifications for IEEE 802.3 Ethernet
compatibility

· GSM Module (SIM800L) complies with the standards of 3GPP TS 07.07 and ITU-T E.212,
EN55022, EN610004, quad-band GSM and GPRS, RoHS, CE certified

· Relay Module operates per IEC 61810-1 and UL 508 standards, with 5V DC triggers, and is
ISO 7637 compliant for automotive electronics

Sensors

· HC-SR04 Ultrasonic Sensor, complies with IEC 60335 and FCC Class B standards, works on
the frequency of 40 kHz and a measuring distance between 2 cm to 400 cm and ISO 15022 for
automotive.

· Temperature Sensor (LM35) => JEDEC JESD22 and ANSI/ASME PTC 19.3 => ±0.5 this is
compliant with IEC 60584 for thermal safety.

· Vibration Sensor meets ISO 10816-1 and ASTM E-220 standards for automotive functional
safety (ISO 26262)

Electrical insulation safety of Light Dependent Resistor (LDR) that conforms with IEC 60068.

· Infrared Sensor meets ISO/IEC 14543 and IEC 62471 standards, has a wavelength range of
850 nm to 950 nm, employed in automation and safety systems.

Communication & Software Standards

· GPS Module equipped with ± 3 m accuracy and communication ratio based on UART ITU-R
M.1902 (ISO 7637 and NMEA 0183), Board king pro-embedded will receive satellite positioning
via GPS

· IoT Communication Protocols defined in terms of ISO/IEC 20922 (MQTT) and IEEE
802.15.4 for secure and efficient data exchange in IoT systems.

· Django Web Server is sec compliant with W3C and ISO/IEC 27001, supports Http/Https
protocols with PSR-12 coding compliant.

28
· Arduino IDE is compliant with the ISO/IEC 9899:1990 under the GNU General Public
License.

4.3 ARCHITECTURE OF VEL-NET ALGORITHM

29
 Figure 4.3:Architecture of VEL-NET algorithm

The model takes (32, 32, 3) input images. It uses VGG16 together with EfficientNetB0 to
generate features. A second Conv2D and MaxPooling layer are used to extract additional
features. It uses Global Average Pooling to dimension-reduce. Fusion and Processing. These
features are then combined into a 1912-dimensional vector. The vector is reshaped and fed into
a 1D CNN, batch normalized, and then an LSTM layer to reveal sequential patterns. Fully
Connected Layers consist of dense layers which enable additional learning. Dropout layers
avoid overfitting. The final Dense layer, whose output is 1, will also probably output a
continuous value.

Each branch contains a first layer corresponding to Conv2D layer of VGG16, Efficientnet0,
and Leenet followed by Batch Normalization, that normalizes activation which stabilizes
training. Then the feature maps passed through Max Pooling layers to down-sample the feature
map maintaining the relevant information. The three branches then have their output pooled,
which is flattened out to create 1D vectors. The 8 outputs are then concatenated together
through a Concatenate layer.

The concatenated feature vector is passed through a Dense layer with 64 neurons, which learns
high-level characteristics across the three CNN outputs. The second most important part of our
code is to prevent our model from overfitting; therefore, we need to use Dropout layer, which
disables a random number of neurons in the training process. Lastly, we have a Dense output
layer, with one neuron generating the final prediction.

This architecture has the ability to take advantage of many CNN models, hence allowing the
extraction of reliable and diverse features. The combination of VGG16 and Efficientnet's deep
feature learning ability conforms with Leenet's multi-scale feature extraction, therefore giving
the VEL-Net model state-of-the-art accuracy and better generalization. Its multilayered
structure makes it very powerful for complex image classification problems where deep
hierarchical learning is needed.

30
 CHAPTER 5

METHODOLOG

5.1 DATA COLLECTION

Optimizing Energy Management: The Predictive Energy Management system (PEMS) works in
the direction of forecasting the energy consumption and enabling real-time adjustments in the
conservation of energy. Methodology data is collected, feature extraction is performed, the
required features are selected, and models are trained and evaluated. In this section, we define
the strategy followed to develop an appropriate and adaptive PEMS. This is the first step to be
taken toward effective PEMS development; also the data needs to be collected. Data relevant
for energy consumption are captured from sensors, smart meters, and some external
information such as weather forecast or grid records. The collected data includes:

· Real-time power usage.

· Historical energy use data.

· Environmental conditions (temperatures, humidity, amount of sunlight, etc.).

· Various occupancy and user behavior patterns.

Data Processing Pipeline

This includes multistage complicated processing pipelines that can take sensor raw inputs from
noise to meaningful features that might be used to help in detecting anomalies, putting
advanced signal processing techniques together with feature engineering strong methods to
present the best possible quality of the data for using the CNN-LSTM model.

Noise reduction

31
 Moving Average Filter:

M −1
1
y [ n]=
M
∑ x [n−k ]
k=0

Butterworth Low-pass Filter:

1
H ( s )= 2n
s
1+( )
ωc

Feature Engineering

1.Time Domain Features

 Root Mean Square (RMS):

√
N
1
RMS= ∑
N i=1
xi2

 Peak-to-Peak Amplitude:

A p −p =max ( x )−min ⁡( x)

2.Frequency Domain Features

 Fast Fourier Transform (FFT):

N −1
X ( k )= ∑ x (n)e− j 2 πkn/ N
n=0

 Power Spectral Density (PSD)

32
Hybrid CNNLSTM Architecture

A. Network Architecture
1. Convolutional Layer

Filter Size: 64

Kernel Size: 3×1

Stride: 1

Activation: ReLU

f ( x )=max ⁡(0 , x )

2. LSTM Layers

First Layer: 64 units

Second Layer: 32 units

Memory Cell State:

c t =f t ⊙ c t−1 +i t ⊙ ~c t

Hidden State:

ht =o t ⊙tanh (c t )

3. Dense Layers

Hidden Units: 32, 16

Output Unit: 1

Sigmoid Activation:

1
σ ( x )= −x
1+e
33
B. Training Parameters

Loss Function: Binary Cross-Entropy:

N
−1
L= ∑ y log ( ^y i ) +( 1− y i ) log (1− ^yi )
N i=1 i

Learning Rate: α = 0.001

Batch Size: 32

Epochs: 20

7. Anomaly Detection System

A. Decision Function

Probability Threshold (θ): 0.5

Anomaly Condition:

{
A ( x )= 1 ,∧if P ( x ) >θ
0 ,∧otherwise

B. Alert Generation Mechanism

Confidence Level: 95%

Response Time: <100ms

False Alarm Rate: <1%

5.2 FEATURE EXTRACTION

34
It means you are extracting required features from raw data that could change into meaningful
representations. This can be done by utilizing the data and creating new features. The main
techniques for feature extraction are:

· Time-based Features: Features that aggregate energy usage over varying time intervals
(hourly, daily, weekly trends).

Statistical Features: Mean, median, standard deviation and variance to capture usage patterns
in the calculations.

· Environmental Impact Features: Including climate data (temperature, humidity, solar

radiation) for designing models that capture external influences on energy consumption.

· User Behavioral Features: This may help in determining behavior patterns in occupancy and
appliance usage periods to analyze demand fluctuations.

5.3 FEATURE SELECTION

Feature selection is an important technique to deal with high dimensionality, where it can help
us to identify the most relevant features for improved accuracy and reduced complexity in a
model. Feature selection techniques include:

· Correlation Analysis: Indices which can remove noise by analyzing how the data for features
differs from one another.

· Principal Component Analysis (PCA): Dimensionality reduction while preserving important

information

· Recursive Feature Elimination (RFE): Removing the least significant features iteratively to
improve model performance.

· Mutual Information: Assessing the dependency between input variables and energy
consumption for optimal feature selection.

35
5.4 MODEL TRAINING

Training models picking suitable algorithms and training them based on the prepared data.
Different machine learning and deep learning models are examined:

(A) Time-Series Forecasting Models

· ARIMA (Auto-Regressive Integrated Moving Average): For short-term energy demand

prediction.

· LSTM (Long Short-Term Memory Networks): Used for capturing long-term dependencies in
the energy load data.

· XGBoost and Random Forest: To identify non-linear patterns and make predictions.

(B) Optimization of Pre-assigned Rewards Using Reinforcement Learning (RL)

It combines a reinforcement learning framework for real-time energy management. The agent
(PEMS) learns optimal energy distribution strategies depending on:

· State: Current energy use, demand forecasts, grid availability.

· Actions: Load shifting, redistribution of renewable energy, optimization of battery storage

· Rewards: Lowering energy costs and improving efficiency, minimizing CO2 emissions

A training set is used for fitting the models and cross-validation techniques to assess the
generalizable performance of the models.

5.5 MODEL EVALUATION

Model evaluation is done with multiple metrics used to determine performance and reliability.
Some important evaluation criteria are:

· Prediction Accuracy: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE),
Mean Absolute Percentage Error (MAPE)

· Energy Savings: Decrease in energy consumption and expenses based on predictive insights.
36
· System Response to Demand Increase: Time for the model to respond to rise in energy
demand.

· User Satisfaction: Cleared through surveys and user feedback on how well system
performed.

· Confusion Matrix & ROC Curve: Commonly used for classification models, i.e. anomaly
detection & energy pattern optimization. It provides a structured approach for developing a
Predictive Energy Management System which combines data analytics, machine learning and
optimization techniques. The PEMS optimizes energy efficiency and lowers operational costs
through predictive modeling and real-time control.

37
 CHAPTER 6

RESULTS &DISCUSSION

6.1 RESULTS & DISCUSSION

The predictive energy management system was developed to increase energy efficiency and
improve performance in Hybrid Electric Vehicle (HEV) applications. With the integration of
thousands of IoT-native sensors and advanced machine learning algorithms, the system
monitored several crucial vehicle parameters (such as temperature, vibration, distance, and tilt)
and adjusted its features dynamically to minimize energy loss. Their model, which integrated
convolutional neural networks (CNNs) for local feature extraction from images and LSTM
networks for spatiotemporal pattern extraction, achieved a 25% relative improvement over
their baseline single-CNN architecture. The combination of these techniques was very effective
in anomaly detection and energy consumption optimization

In testing, the system showed impressive performance metrics of 99.25% accuracy, 1.00
precision, and 1.00 F1-score, proving to be reliable in guaranteeing energy efficiency. But there
was a significant catch: The system was able to screen out “false positives,” ensuring that alerts
would only be sent for real problems and that you won’t intervene unnecessarily. In addition to
this effect, the system also played a role in energy efficiency, demonstrating a voltage drop of
2.56V and contributing to energy conservation while prolonging battery life. Along with better
voertuig health management, this system is in the track of sustainable and smart transport. Its
potential to improve energy efficiency and fusion scope in anomaly detection makes it a
promising option for the future of HEVs, leading the way toward smarter and greener mobility
solutions.

38
 Figure 6.2: Confusion Matrix of of proposed VEL-Netmodel

Output of confusion matrix explains the prediction accuracy of proposed model and provides
evidence that proposed model has learnt perfectly how to distinguish between a normal and
anomaly. The model made 167 true positive (normal classified as normal), 626 true negatives
(anomaly classified as anomaly), 2 false positive (normal classified as anomaly) and 5 false
negative (anomaly classified as normal). This leads to an overall very high accuracy, which
further attests the reliability of the model in classification tasks.

A low misclassification value indicates that the model is efficient at identifying anomalies,
with minimal false alarms and missed detections. The true positive rate of the model is 626
correctly identified anomalies, meaning that the model is very sensitive to anomalies and thus
can be effectively used for applications where accuracy in detecting anomalies is needed.
Another reason that true negative rate (167 correctly identified normal cases) is high indicating
the model is not over predicting anomalies.

39
The area under the sensitivity/specificity curve of the proposed model was great, overall, and
not far from the ideal performance as seen. The extent of precision demonstrates that it is
capable for actual applications where reliable anomaly detection is key.

Table 6.3: Comparison of Outcomes between Proposed model and Pre-existing models

In the above table, we have compared three models Random Forest, CNN+LSTM and VEL-
NET in terms of precision, recall, F1-score and support. The results indeed show that the new
VEL-NET model significantly outperforms the existing models.

Out of the three, the VEL-NET has the highest accuracy, with the precision of 0.9965, recall of
0.9921, and F1-score of 0.9944. This means that this is the best model, as it is doing well in
predicting the right classes at the same time minimizing the false prediction. Using
CNN+LSTM also gives good results, at precision 0.99 recall 0.98 F1-score 0.99 but still does
not excels. The Random Forest model is slightly behind, although accurate with 0.96 precision,
0.94 recall and an F1-score of 0.95.

The support column indicates the number of occurrences assessed for every model.
CNN+LSTM was evaluated on the largest dataset (957 items), followed by Random Forest
(800) and VEL-NET(631). Although it had a significantly smaller dataset, compared to other
two models VEL-NET still performed better than the rest of them, therefore establishes its
efficacy when it comes to classification tasks.

40
 This comparison highlights the clear benefits of VEL-NET, which also makes it more suitable
for applications where high precision and recall are required. Its possibly best-in-class
performance indicates that it is capable of being employed in real-world scenarios requiring
accurate predictions that would defeat most generic machine learning and deep learning
methodologies.

REALISTIC CONSTRAINTS

1. Data Availability & Quality – Models are only as good as their datasets. Models can be
affected if the dataset is biased (HumanGPT, 2022).
2. Computational Resources – Complex models (CNN+LSTM, for example, and VEL-NET.)
need a lot of processing power and is infeasible for low resource conditions.
3. Processing in Live Time – Real-time applications may need instant predictions, but more
complex models can add latency that's detrimental to usability.
4. Scalability — Model trained on a smaller dataset does not generalize well to larger datasets or
real-world applications.
5. Overfitting & Generalization – A model that is highly accurate on the training data may not be
as effective when it encounters data it hasn't seen before.
6. Interpretability — Deep learning models are often very accurate, but exist in “black boxes,” so
that it is sometimes difficult to explain their predictions.
7. Cost Constraints — The deployment and maintenance of advanced models are capital-
intensive, incorporating hardware, cloud storage, or computational costs.
8. Ethical & Privacy Concerns – Sensitive data usage must adhere to regulations (like GDPR),
including security and privacy.
9. Adaptability – Data patterns can change over time (whether this is due to people changing their
behavior in response to a model being in place, or due to natural seasonal variation), and
models need to be updated periodically.
10. Deployment & Maintenance – Continual updates, bug-fixing and monitoring for the model to
keep the accuracy and efficiency.

41
 CHAPTER 7

CONCLUSION & FUTURE

WORK

The Predictive Energy Management System for Hybrid Electric Vehicles (HEVs) The scenario
is enhanced with the addition of IoT-powered sensors for continuous observations of
temperature, vibration, distance, light intensity, and tilt of a vehicle to enable real-time insights
of a vehicle's operating parameters. This system utilizes advanced machine learning algorithms
(i.e. convolutional neural networks (CNNs) for spatial feature extraction and analysis, and
long short-term memory (LSTM) network for temporal information pattern analysis) to
provide real-time prediction of potential faults and optimization of energy flow according to
road conditions, terrain details, etc.Thus, it can successfully overcome the bottlenecks of
classical energy management systems in HEVs, e.g., poor power sharing performance, non-
online fault decision, and no online monitor. Through intelligent real-time optimization of
energy consumption, the system achieves significant energy savings, enhances vehicle
performance, and increases safety. Through high-performance statistics including 99.25%
accuracy, 1.00 precision, and 1.00 F1-score, the system has shown its potential to reduced
energy wastage and increased efficiency.
In addition, the integration of the system with Vehicle-to-Infrastructure (V2I) networks for
predictive anomaly detection brings additional intelligence into the picture, optimizing energy
usage by factoring in external conditions, like road and traffic conditions. Considering all the
mentioned aspects, this feature enhances the system to be a very significant element of future
intelligent transport systems (ITS) by providing a more durable and sustainable environment for
the operation of HEVs and fleets.The proposed system includes a greater improvement than the
state of power management scheme and vehicle efficiency (more than 5% improvement).
Further, the algorithm could possibly be improved with a larger coverage of sensors in an area
— sensors can be anything from air quality to weather sensors, but these play a major role in
improving the accuracy of prediction and usage of energy according to the seasons/changes in
the environment. You could make the system more powerful by adding predictive analytics for
42
 battery life optimization, which would make the vehicle last longer and hence become more
sustainable.
Moreover, future generations of HEVs and autonomous vehicles can leverage the power of AI-
driven decision-making algorithms, optimizing energy usage factors such as the best
combination of battery life and power while adapting to the users' driving preferences in order
to deliver personalized driving experiences without affecting any energy usage standards. The
system might also be extended to interoperate with wider smart city infrastructure, allowing for
coordination with the vehicle and traffic management systems for reduced energy/network
usage and greenhouse gas emissions.For large-scale field trials, the real-life functionality
validated the potential of this solution across different purposes of driving conditions, laying
the groundwork for scaling this architecture for future HEVs that we can expect to enter in the
market. As improvements in sensor technology, machine learning, and vehicle-to-grid (V2G)
communication continue, this predictive energy management system could be crucial in driving
the shift towards a more sustainable, intelligent, and energy-efficient transportation ecosystem.

ETHICAL BINDINGS

1. Modeling Machine Learning Based HEV: The integration of sensors in the HEV and machine
learning gives rise to massive amounts of vehicle and environmental data. Protecting this data
is essential because sensitive information like a car's location, usage habits, and personal driver
behavior can be exploited if not adequately secured. Robust encryption and data
anonymization should also be a major ethical consideration to prevent unauthorized access.
2. Decision-Making Transparency: It is crucial that the decisions made by the system are
transparent, as the system relies on sophisticated machine learning algorithms to facilitate
energy management optimization and anticipate potential faults. Drivers and fleet operators
need to know exactly how the system works, what data it collects and how it informs energy
management decisions. Open source creates trust in the system and enables the user to assess
its implications.
3. Equitable Tech Access: As advanced systems like these systems begin to become
mainstreamed electric grid technologies, considerations must ensure equitable access to such
tech for a wide range of utilizers, including the economically disadvantaged or underserved.
Focus should be on creating cost-effective solutions .

43
 REFERENCES

1. Nikhil et al., “IoT-Enabled Battery Monitoring System for Enhanced Electric Vehicle
Performance,” J. Electr. Eng. Innov., vol. 32, pp. 105–112, Jul. 2024.
2. Gunawan et al., “Prototype Development of Vehicle Collision Avoidance System using
Arduino,” Int. J. Electr. Comput. Eng., vol. 8, no. 3, pp. 246–252, Apr. 2020.
3. Feng et al., “Review of electric vehicles’ charging data anomaly detection based on deep
learning,” J. Electr. Veh. Technol., vol. 10, no. 2, pp. 45–54, Jun. 2020.
4. Srinagesh et al., “GPS-Enabled Fuel Sensor Based Vehicle Tracking System for Fleet
Management Using IoT,” IEEE Trans. Intell. Transp. Syst., vol. 26, no. 3, pp. 278–289,
Mar. 2025.
5. Choi et al., “Energy Management Optimization in a Battery/Supercapacitor Hybrid Energy
Storage System,” IEEE Trans. Ind. Appl., vol. 48, no. 5, pp. 1234–1240, Oct. 2012.
6. Goud et al., “Vehicle Accident Automatic Detection and Remote Alarm Device,” Int. J.
Electr. Eng. Appl., vol. 5, pp. 113–118, Nov. 2012.
7. Anusha et al., “IoT-Based Energy Monitoring with ESP32 and ThingSpeak,” Int. J. Energy
Sustain., vol. 11, pp. 88–94, Jan. 2024.
8. Theissler et al., “Predictive Maintenance Enabled by Machine Learning: Use Cases and
Challenges in the Automotive Industry,” IEEE Trans. Ind. Informat., vol. 18, no. 3, pp.
123–130, Feb. 2021.
9. Desima et al., “Alarm System to Detect the Location of IoT-based Public Vehicle
Accidents,” J. Automot. Eng., vol. 17, no. 4, pp. 78–85, Dec. 2017.
10. Chamola et al., “An IoT and Edge Computing Based Framework for Charge Scheduling
and EV Selection in V2G Systems,” IEEE Trans. Ind. Informat., vol. 16, no. 6, pp. 742–
748, Dec. 2020.
11. Poobalan et al., “GPS and GSM Based Accident Location Indicator and Rescue System,”
J. Adv. Res. Elect. Eng., vol. 14, pp. 33–39, Mar. 2015.
12. Ali Sayah et al,“Advanced energy management system with road gradient consideration for
fuel cell hybrid electric vehicles”.Results in Engineering 23 (2024): 102721
13. Achikkulath et al, Hussain Shareef, Madathodika Asna, Ahmad Ibrahim,“Optimization of
hybrid energy systems and adaptive energy management for hybrid electric
vehicles”.Energy Conversion and Management 243 (2021): 114357.
44
14. Praveena Krishna P. S et al,“ Energy Management Strategies for Hybrid Energy Storage
System in Electric Vehicles”. IEEE CONECCT, PP. 1-7. IEEE (2020).
15. Hongwen He et al, An Improved Energy Management Strategy for Hybrid Electric
Vehicles Integrating Multistates of Vehicle-Traffic Information”. 28th Electrical Power
Distribution Conference, IEEE(2024).

45