A TECHNICAL SEMINAR REPORT ON

Research on Electrical Vehicle Charging

Safety Warning Based on A-LSTM Algorithm
A dissertation submitted in partial fulfilment of the
Requirements for the award of the degree of

BACHELOR OF TECHNOLOGY
in

INFORMATION TECHNOLOGY
Submitted by

G RUSHI BHARGAV (20B81A1235)

Under the esteemed guidance of

Mrs. G. Shailaja
Assistant Professor, IT Department
CVR College of Engineering

DEPARTMENT OF INFORMATION TECHNOLOGY

CVR COLLEGE OF ENGINEERING

ACCREDITED BY NBA, AICTE & Affiliated to JNTU-H
Vastunagar, Mangalpalli (V), Ibrahimpatnam (M), R.R. District, PIN-501510
2023-2024
 DEPARTMENT OF INFORMATION TECHNOLOGY
CERTIFICATE

This is to certify that the Technical Seminar entitled “Research on Electric Vehicle
Charging Safety Warning Based on A-LSTM Algorithm” is a bonafide work done and
submitted by G RUSHI BHARGAV (20B81A1235), during the academic year 2023-2024, in
partial fulfilment of requirement for the award of Bachelor of Technology degree in Information
Technology from Jawaharlal Nehru Technological University Hyderabad, is a bonafide record
of work carried out by them under my guidance and supervision.

Certified further that to my best of the knowledge, the work in this dissertation has not
been submitted to any other institution for the award of any degree or diploma.

INTERNAL GUIDE HEAD OF THE DEPARTMENT

G. Shailaja Dr. Bipin Bihari Jayasingh
Assistant Professor, IT Department Head and Professor, IT Department

TECHNICAL SEMINAR COORDINATOR

Dr. K. Sarangam
Sr. Assistant Professor, IT Department
 ACKNOWLEDGEMENT
The satisfaction of completing this project would be incomplete without mentioning our
gratitude towards all the people who have supported us. Constant guidance and encouragement
have been instrumental in the completion of this technical seminar.
First and foremost, we thank the Chairman, Principal, Vice Principal for availing
infrastructural facilities to complete the technical seminar in time.

We offer our sincere gratitude to our internal guide G. Shailaja, Assistant Professor, IT
Department, CVR College of Engineering for her immense support, timely co-operation and
valuable advice throughout the course of our technical seminar.
We would like to thank the Head of Department, Professor Dr. Bipin Bihari Jayasingh,
for his meticulous care and cooperation throughout the technical seminar.

We are thankful Dr.K.Sarangam, technical seminar Coordinator, Assistant Professor, IT

Department, CVR College of Engineering for his supportive guidelines and for having provided
the necessary help for carrying forward this technical seminar without any obstacles and
hindrances.

We also thank the Seminar Review Committee Members for their valuable suggestions.
 DECLARATION

We hereby declare that the project report entitled “Research on Electric Vehicle Charging Safety
Warning Based on A-LSTM Algorithm” is an original work done and submitted to IT
Department, CVR College of Engineering, affiliated to Jawaharlal Nehru Technological
University Hyderabad, Hyderabad in partial fulfilment of the requirement for the award of
Bachelor of Technology in Information Technology and it is a record of bonafide project work
carried out by us under the guidance of G. Shailaja, Assistant Professor, Department of
Information Technology.

We further declare that the work reported in this technical seminar has not been submitted,
either in part or in full, for the award of any other degree or diploma in this institute or any other
Institute or University.

_________________________

Signature of the Student

(G Rushi Bhargav)

(20B81A1235)
 ABSTRACT

Accidents involving electric vehicle fires have increased as the number of electric
vehicles has grown recently. The issue of charging safety is a key barrier to the growth
of the electric vehicle sector because these accidents have resulted in large financial
losses for car owners and charging facility operators. The approach for resolving the issue
of electric car charging safety through an electric vehicle charging safety warning system
is suggested in this research. The suggested solution uses an adaptive optimization of
long short-term memory neural network (A-LSTM) to forecast voltage changes
throughout the whole charging process by using the vehicle’s daily historical charging
data. The warning threshold adjustment method is established by the difference between
the predicted voltage data and the actual voltage data, which is dynamically adjusted as
the charging process progresses. Finally, a real-time warning model for vehicle charging
alert is developed. The daily charging data of electric vehicles is used in the paper to
verify the precision of data prediction and the accuracy and timeliness of the model. The
study’s findings demonstrate that the early warning model suggested in this paper can
quickly send out early warning signals to safeguard the safety of car charging and can
identify aberrant charging data.
 List of Figures
Figure 1: Accident Ratio
Figure 2: Flow chart of EV charging safety warning model
Figure 3: Pearson coefficient between parameters
Figure 4: Charging temperature curve of vehicle motor and battery in January and August
Figure 5: Pearson coefficient after improved temperature expression
Figure 6: EV charging data prediction error
Figure 7: Comparison of prediction results
Figure 8: EV charging data forecast results
Figure 9: Comparison of R2 values tested by different algorithms
Figure 10: Histogram of prediction residuals for different algorithms
Figure 11: Dynamic warning threshold
Figure 12: Dynamic warning threshold after adjustment
Figure 13: Dynamic threshold warning accuracy rate
Figure 14: Fixed threshold warning accuracy
Figure 15: Time-limit of warning for different models
 List of Tables

TABLE 1 : EV charging safety warning factor

TABLE 2. Summary of EV Charging Data
TABLE 3. Results of MAE with different parameters of A-LSTM
TABLE 4. A-LSTM RMSE results with different parameters
TABLE 5. Hyperparameter setting of A-LSTM model
 Table of Contents

Content Pg. No

1. Introduction………………………………………………….…. 1
2. Motivation………………..………………………………….… 3
3. Literature survey …………………………………...…………. 4
4. Methodology……………………………………...……………... 15
4.1 LSTM Algorithm…………….……………………………… 18
4.2 A- LSTM Algorithm ……………………………….………... 20
5. Result ……………………………………..……………………. 21
6. Conclusion………………………………………………..…… 33
References………………………...………………………….... 34
Appendix ………………………..…………………………….... 37
 1. INTRODUCTION:
In recent years, the world has witnessed a growing global awareness of the
ecological and energy crises. The imperative to reduce our carbon footprint and conserve
finite oil resources has led to a surge in the adoption of electric vehicles (EVs). EVs offer
substantial advantages over traditional fuel vehicles, with their potential to mitigate carbon
emissions and diminish oil consumption. Governments and automotive companies
worldwide have recognized the significance of EVs, resulting in a continuous increase in
the number of EVs on the road.
However, amidst the EV industry's promising growth, concerns about safety have
arisen, particularly regarding the frequent occurrences of spontaneous combustion and fires
in EVs. These safety incidents have inflicted substantial economic losses on vehicle owners
and charging facility operators, in addition to impeding the overall development of the EV
industry. To address these challenges, the integration of big data research into the realm of
EV safety has emerged as a pivotal pathway.
With the full implementation of national big data strategies, the automotive industry
is experiencing a discernible shift towards digitalization and intelligence. The fusion of big
data with EV technology holds the potential to revolutionize automotive safety regulations,
further catalysing the high-quality development of China's EV industry.
The existing research landscape primarily focuses on safety warning and fault
diagnosis during the EV charging process. Traditionally, researchers have approached EV
fault diagnosis and warning by constructing intricate electrochemical models of batteries.
These models often require extensive calibration and are primarily tailored to specific fault
types, making them less versatile for practical applications.
For example, Seo et al. employed a recursive least squares method to detect internal
short-circuit faults in batteries based on an equivalent circuit model, utilizing battery open-
circuit voltage and state of charge (SOC) as inputs. Similarly, Zhang et al. developed an
electrochemical model of power batteries to monitor charging faults by comparing
simulated charge response information with actual charge status data. Tran et al. adopted
an equivalent circuit model for lithium-ion batteries, considering SOC, temperature, and
state of health (SOH) to estimate battery parameters with high precision. Some scholars
have resorted to expert system approaches for fault detection and early warning, such as
Song's comprehensive evaluation index system for charging safety based on expert scoring
and the grey correlation degree method.
Nonetheless, these methods often involve complex modelling work and are not
always suitable for real-world applications due to the strong nonlinearity of EV battery
systems. Moreover, most of these methods are based on laboratory data, with limited
research on daily EV charging data.

1
 In recent years, the proliferation of big data and machine learning techniques has
ushered in innovative approaches to fault monitoring and early warning in EVs. Zhao et al.
introduced a fault diagnosis method for EV battery systems based on big data statistical
methods and machine learning algorithms, providing a comprehensive battery fault
diagnosis model. Zhang et al. devised a charging warning model using improved grey wolf
optimization-backpropagation neural networks to identify abnormal charging voltage
conditions. Xia et al. proposed a short-circuit fault diagnosis method based on voltage
profile correlation coefficients, utilizing neural networks and recursive sliding windows for
enhanced fault detection. Gao et al. employed adaptive deep belief networks for electric
vehicle charging process fault warning, ensuring accurate fault prediction during charging.
These methods have demonstrated their effectiveness in handling time-series EV charging
data but are primarily based on laboratory data.
However, daily EV charging data poses unique challenges due to its voluminous and
highly dimensional nature, making it less amenable to certain deep learning algorithms. To
address these challenges, we propose a novel approach based on the Long Short-Term
Memory (LSTM) algorithm for vehicle charging safety warnings. LSTM excels in handling
time-series data and can predict future trends based on current and historical data, making
it well-suited for the intricacies of vehicle charging data.
In this paper, we aim to construct a comprehensive EV charging safety warning
model using the LSTM algorithm. Our approach begins with an analysis of factors
influencing vehicle charging safety, combining the characteristics of vehicle charging data
with the specific charging attributes of lithium-ion batteries. Using Pearson correlation
analysis, we filter charging safety warning factors based on their correlation with vehicle
charging data. We then introduce the A-LSTM prediction algorithm, designing a prediction
model for charging data. By analysing historical data, we generate predicted values for
electric vehicle charging data.
To create a real-time warning model for EV charging safety, we integrate the A-
LSTM prediction algorithm and design a dynamic warning threshold method, combining
predicted values, actual data, and charging standards. This method ensures accurate and
timely warnings, contributing to enhanced EV charging safety. We validate the accuracy
and timeliness of our charging safety warning model through an analysis of daily charging
data.
In summary, this paper addresses the pressing need for improved safety in the
electric vehicle charging process. We leverage big data research and machine learning
algorithms, with a particular focus on the LSTM algorithm, to develop a robust and
adaptable EV charging safety warning model. By doing so, we aim to contribute to the
continued growth of the EV industry while safeguarding the interests of EV owners and
charging facility operators.

2
 2. MOTIVATION
As I embark on my academic and career journey, I find myself drawn to the
intersection of technology, sustainability, and innovation. The research paper titled
"Enhancing Electric Vehicle Charging Safety through Big Data and LSTM-Based
Predictive Models" resonates deeply with my academic and career goals for several
compelling reasons.
First and foremost, this paper addresses a pressing global issue—the ecological and
energy crises. In an era where the world is grappling with the consequences of climate
change and finite oil resources, electric vehicles (EVs) have emerged as a beacon of hope.
They offer a tangible solution to reduce carbon emissions and conserve our planet's precious
resources. The paper underscores the significance of EVs in this context, emphasizing their
potential to bring about substantial positive change.
The safety of EVs during the charging process is a critical concern, and this paper
tackles it head-on. The frequency of spontaneous combustion and fires in EVs poses not
only economic challenges but also a significant hindrance to the widespread adoption of
this promising technology. By delving into safety issues and proposing innovative
solutions, this paper aligns perfectly with my desire to contribute to making EVs safer and
more accessible to all.
One aspect that particularly excites me is the integration of big data and machine
learning, specifically the Long Short-Term Memory (LSTM) algorithm, to enhance safety
in EV charging. This fusion of cutting-edge technology, data science, and automotive
engineering represents the future of innovation. It not only aligns with my academic
interests but also equips me with sought-after skills and knowledge that are highly relevant
in today's data-driven world.
Moreover, the potential real-world impact of this research is immense. By
contributing to the development of a robust and adaptable EV charging safety warning
model, I can actively participate in shaping the automotive industry and advancing
sustainability efforts. This aligns perfectly with my aspiration to make meaningful
contributions to society through my academic and career pursuits.
Furthermore, this paper's multidisciplinary nature, drawing from fields such as
environmental science, engineering, data science, and machine learning, mirrors my own
diverse interests and educational background. It presents an exciting opportunity to bridge
these fields and engage in interdisciplinary research, fostering a well-rounded perspective
that I believe is essential in today's complex world.

3
 3. LITERATURE SURVEY
In recent years, the world has witnessed a growing global awareness of the
ecological and energy crises. The imperative to reduce our carbon footprint and conserve
finite oil resources has led to a surge in the adoption of electric vehicles (EVs). EVs offer
substantial advantages over traditional fuel vehicles, with their potential to mitigate carbon
emissions and diminish oil consumption. Governments and automotive companies
worldwide have recognized the significance of EVs, resulting in a continuous increase in
the number of EVs on the road.
An electric vehicle (EV) is a type of vehicle that is powered by electricity instead of
traditional internal combustion engines (ICEs) that run on gasoline or diesel. Electric
vehicles use one or more electric motors to propel the vehicle, drawing electricity from an
onboard battery or an external source (such as charging stations) to generate the power
needed for operation. At their core, electric vehicles are a departure from the traditional
internal combustion engine (ICE) vehicles that have dominated the automotive landscape
for over a century. Instead of relying on gasoline or diesel to power an internal combustion
engine, EVs harness the power of electricity to propel themselves.
The fundamental principle behind the operation of an electric vehicle is the use of
one or more electric motors to drive the wheels. These electric motors are powered by
electricity stored in an onboard battery pack. Unlike conventional vehicles, there is no
tailpipe emitting harmful exhaust gases. Instead, EVs produce zero tailpipe emissions,
contributing significantly to the reduction of greenhouse gas emissions and improving
urban air quality.
There are several types of electric vehicles, including:
1. Battery Electric Vehicles (BEVs): These vehicles run solely on electricity and are
equipped with a large battery pack to store and provide electrical energy to the motor(s).
BEVs are charged by plugging into electrical outlets or charging stations. They produce
zero tailpipe emissions.
2. Plug-in Hybrid Electric Vehicles (PHEVs): PHEVs combine an electric motor and a
gasoline or diesel engine. They can operate in electric-only mode for a certain range, after
which the internal combustion engine kicks in to extend the vehicle's range. PHEVs can be
charged through electrical outlets.
3. Hybrid Electric Vehicles (HEVs): HEVs have both an electric motor and an internal
combustion engine, but they cannot be charged externally. The electric motor assists the
gasoline or diesel engine in improving fuel efficiency.
Electric vehicles offer several advantages, including reduced greenhouse gas
emissions, lower operating costs (as electricity is generally cheaper than gasoline or diesel),
quieter operation, and regenerative braking, which can increase energy efficiency.
However, EV adoption is influenced by factors such as battery technology, charging
infrastructure availability, and cost considerations.

4
 The electric vehicle market has been growing steadily, with advancements in battery
technology leading to longer driving ranges, faster charging times, and increased
affordability. EVs are seen as a crucial component in reducing carbon emissions and
combating climate change, as they offer a more environmentally friendly alternative to
traditional fossil-fuel-powered vehicles.
Advantages of Electric Vehicles
The rise of electric vehicles is not merely a response to environmental concerns but
also driven by their numerous advantages:
1.Reduced Emissions : EVs produce zero tailpipe emissions, contributing to a cleaner and
healthier environment.
2.Lower Operating Costs: Electricity is typically cheaper than gasoline or diesel, resulting
in lower fuelling costs.
3.Quiet Operation: Electric motors are notably quieter than internal combustion engines,
reducing noise pollution.
4.Regenerative Braking: EVs can capture and store energy during braking, enhancing
energy efficiency.
5.Technological Advancements: Ongoing advancements in battery technology are
increasing driving ranges, reducing charging times, and enhancing overall performance.
As we delve deeper into the world of electric vehicles, it becomes evident that they
represent not only a mode of transportation but a catalyst for change in the automotive
industry and a vital component in the global transition toward a more sustainable and
environmentally conscious future. The growth of EVs is propelled by innovation,
environmental stewardship, and the collective commitment to reduce our carbon footprint
while embracing the electrifying possibilities of the road ahead.
The global surge in electric vehicle (EV) adoption is ushering in a new era of
sustainable transportation. With the promise of reduced emissions, lower operating costs,
and advancements in battery technology, EVs are increasingly becoming a mainstream
choice for environmentally conscious consumers. However, as the EV market expands,
ensuring the safety of EV charging processes is paramount. This article delves into the
critical aspects of charging safety for electric vehicles, highlighting the importance of
comprehensive safety measures and the challenges that need to be addressed.
Charging Safety Concerns:
1. Electrical Safety: EVs draw a substantial amount of electrical power during charging,
necessitating the use of high-voltage systems. Ensuring electrical safety is paramount to
prevent electric shock hazards during charging. Proper insulation, grounding, and
compliance with safety standards are crucial in this regard.

5
2. Battery Safety: The battery is the heart of an electric vehicle, and its safety during
charging is of utmost importance. Overcharging, undercharging, or exposure to extreme
temperatures can lead to battery degradation or even thermal runaway, resulting in fires or
explosions. Battery management systems (BMS) play a vital role in monitoring and
regulating charging to prevent such incidents.
3. Charging Infrastructure: The design and maintenance of charging infrastructure,
including charging stations and cables, must adhere to safety standards. Faulty or poorly
maintained equipment can pose safety risks. Regular inspections and maintenance are
essential.
Safety Measures:
1. Standardization: The development and adherence to safety standards specific to EV
charging, such as ISO 15118 and IEC 61851, are fundamental. These standards ensure
consistency and safety across different charging technologies and platforms.
2. Battery Management Systems (BMS): EVs are equipped with BMS that monitor the state
of charge, temperature, and voltage of individual cells within the battery. BMS plays a
critical role in preventing overcharging, overheating, and over-discharging, enhancing
overall safety.
3. Overcurrent and Overvoltage Protection: EV charging equipment includes protection
mechanisms to safeguard against overcurrent and overvoltage conditions. These systems
shut down charging in case of abnormalities, preventing damage to the vehicle and charging
infrastructure.
4. Emergency Shutdown: Charging stations are equipped with emergency shutdown
mechanisms that allow users to quickly stop charging in case of an emergency or safety
concern.
5. User Education: Properly educating EV owners and users on safe charging practices is
essential. This includes guidelines on compatible charging equipment, proper cable
management, and avoiding unauthorized modifications.
Challenges:
1. Diversity of Charging Infrastructure: The rapid expansion of EV charging networks
introduces variability in equipment and standards. Achieving uniform safety across diverse
charging systems can be challenging.
2. Battery Technology: As battery technology evolves, ensuring the compatibility of new
batteries with existing charging infrastructure and safety standards remains a concern.
3. Regulatory Framework: Developing and enforcing comprehensive safety regulations
specific to EV charging can be a complex and evolving process.

6
 Charging safety is a foundational pillar of the electric vehicle revolution. As the
EV market continues to grow, stakeholders across the industry, including manufacturers,
regulators, and charging infrastructure providers, must collaborate to uphold the highest
safety standards. By prioritizing electrical and battery safety, adhering to standardized
protocols, and investing in research and development, the industry can ensure that EV
charging remains a secure and reliable process. Through these collective efforts, electric
vehicles will continue to gain the trust and confidence of consumers, accelerating the
transition to a cleaner and more sustainable transportation future.
Early warning systems (EWS) are vital tools for risk reduction and disaster
management, playing a pivotal role in enhancing preparedness, response, and mitigation
efforts. These systems provide timely and actionable information about imminent threats,
such as natural disasters, pandemics, financial crises, and more, enabling governments,
organizations, and individuals to take proactive measures to minimize potential harm. This
article explores the significance of early warning systems, their various applications,
challenges, and the role they play in safeguarding lives and assets.
The Importance of Early Warning:
1. Risk Reduction: Early warning systems are designed to identify hazards and assess
potential risks, allowing stakeholders to reduce vulnerabilities and strengthen resilience. By
offering advance notice of impending threats, EWS enable communities to prepare and
adapt, ultimately saving lives and reducing economic losses.
2. Timely Response: Rapid response is often the key to minimizing the impact of disasters.
Early warnings empower authorities to initiate evacuation plans, mobilize resources, and
implement disaster preparedness measures well in advance of a crisis.
3. Preventative Measures: EWS provide the foundation for preventive actions, including
infrastructure improvements, land-use planning, and public awareness campaigns. By
acting on early warnings, communities can take measures to prevent disasters or reduce
their severity.
4. Decision Support: EWS supply decision-makers with critical information to make
informed choices during crises. These decisions may involve resource allocation, policy
adjustments, and coordination of response efforts.
5. Community Engagement: Engaging communities in early warning processes fosters a
sense of ownership and responsibility. Communities that are well-informed and prepared
are more resilient and better equipped to respond effectively.
Challenges and Considerations:
1. Data Quality and Availability: The effectiveness of EWS relies on accurate and up-to-
date data. Ensuring data quality and accessibility can be challenging, particularly in
resource-constrained regions.

7
2. Technological Infrastructure: Developing and maintaining the technological
infrastructure necessary for EWS can be costly. Smaller and less economically developed
regions may face limitations in this regard.
3. Community Engagement: Building trust and ensuring active community participation in
early warning systems can be a complex process. Effective communication and education
are essential.
4. False Alarms: Over-reliance on early warning systems can lead to "alert fatigue" if false
alarms are frequent. Striking the right balance between caution and alarm is critical.
Early warning systems are invaluable tools in contemporary risk management and
disaster preparedness. By providing timely, accurate, and actionable information, EWS
empower individuals, communities, and organizations to make informed decisions and take
preventive measures. As the global community faces increasingly complex and interrelated
challenges, from natural disasters to health crises and financial instability, the importance
of robust early warning systems cannot be overstated. The continued development,
improvement, and accessibility of these systems are essential for enhancing resilience and
reducing the human and economic costs of crises and disasters worldwide.
Artificial Intelligence (AI) represents the pinnacle of human innovation, a
technological frontier that has reshaped industries, expanded the boundaries of what
machines can accomplish, and ignited the imagination of researchers, developers, and
society at large. At its core, AI is the pursuit of creating machines and software systems that
can mimic human intelligence, enabling them to perform tasks, solve problems, and make
decisions with a level of proficiency and autonomy that was once solely the domain of
human beings.
Key Components of AI
1. Machine Learning
2. Deep Learning
3. Natural Language Processing (NLP)
4. Computer Vision
Applications of AI:
AI's impact is felt across diverse sectors:
1. Healthcare: AI aids in disease diagnosis, drug discovery, personalized treatment plans,
and the analysis of medical imaging data.
2. Finance: In finance, AI algorithms improve risk assessment, fraud detection, algorithmic
trading, and customer service through chatbots.
3. Autonomous Systems: Self-driving cars, drones, and industrial robots benefit from AI's
ability to perceive and navigate their environments.

8
4. Customer Service: Chatbots and virtual assistants provide efficient and personalized
customer support.
5. Entertainment: AI enhances gaming experiences through realistic simulations, character
behaviours, and content recommendations on streaming platforms.
Challenges and Considerations:
1. Ethical Concerns: As AI systems gain autonomy, ethical considerations regarding
decision-making, bias, and privacy come to the forefront.
2. Transparency: Developing transparent and interpretable AI models remains a challenge,
particularly in critical applications like healthcare and law.
3. Job Displacement: Automation driven by AI has the potential to disrupt employment
sectors, necessitating workforce adaptation.
4. Data Privacy: The collection and use of vast amounts of data raise concerns about
individual privacy and data security.
AI's evolution is relentless, with ongoing research and innovation pushing boundaries. The
future holds exciting prospects:
1. Human-AI Collaboration: AI will augment human capabilities, leading to innovative
collaborations across industries.
2. AI for Good: AI can address global challenges like climate change, healthcare
accessibility, and poverty alleviation.
3. Responsible AI: Ethical and responsible AI development will be paramount, emphasizing
fairness, transparency, and accountability.
In conclusion, Artificial Intelligence represents a transformative force that is shaping
our world. Its potential for innovation, efficiency, and problem-solving is limitless. While
challenges exist, the responsible and ethical development of AI promises a future where
humans and machines work harmoniously to create a better, more connected, and intelligent
world.
Machine Learning (ML) is a transformative field within the broader realm of
artificial intelligence (AI) that empowers computers to learn from data and make intelligent
decisions without explicit programming. At its core, ML leverages algorithms and statistical
models to enable systems to recognize patterns, draw insights, and improve their
performance through experience, much like the way humans learn.
Key Concepts of Machine Learning:
1. Data: Data is the lifeblood of ML. It encompasses vast and diverse information that is
used for training, validating, and testing ML models. This data can be structured (e.g.,
tables) or unstructured (e.g., text, images, audio).

9
2. Algorithms: ML algorithms are mathematical constructs that process data to uncover
patterns and relationships. These algorithms are tailored to specific tasks, from
classification and regression to clustering and reinforcement learning.
3. Training: ML models learn by processing large volumes of labelled or unlabelled data
during training. Through this iterative process, models adjust their parameters to improve
accuracy and predictive capabilities.
Types of Machine Learning:
1. Supervised Learning: In supervised learning, models are trained using labelled data,
where input-output pairs are provided. The model learns to map inputs to correct outputs,
making it suitable for tasks like classification and regression.
2. Unsupervised Learning: Unsupervised learning involves training on unlabelled data to
discover patterns and structure within the data. Clustering and dimensionality reduction are
common applications.
3. Reinforcement Learning: Reinforcement learning employs a reward-based system, where
an agent learns to make decisions by interacting with an environment. It is widely used in
gaming, robotics, and autonomous systems.
In conclusion, Machine Learning represents a revolutionary field that leverages
data-driven intelligence to unlock unprecedented possibilities. As it continues to advance,
ML promises to drive innovation, solve complex problems, and transform industries,
ultimately reshaping the way we interact with technology and the world around us.
A neural network, also known as an artificial neural network (ANN) or simply a
neural net, is a fundamental component of artificial intelligence (AI) that is inspired by the
structure and function of biological neural networks, particularly those found in the human
brain. Neural networks are a class of machine learning models used for various tasks,
including pattern recognition, classification, regression, and decision making.
Types of Neural Networks:
1. Feedforward Neural Networks (FNNs): In FNNs, data flows in one direction, from the
input layer to the output layer. They are used for tasks like classification and regression.
2. Convolutional Neural Networks (CNNs): CNNs are designed for tasks involving
structured grid data, such as image and video analysis. They use convolutional layers to
automatically detect patterns.
3. Recurrent Neural Networks (RNNs): RNNs are well-suited for sequential data, such as
time series and natural language. They have feedback loops that allow them to maintain a
memory of past inputs.
4. Long Short-Term Memory Networks (LSTMs): A type of RNN, LSTMs are specialized
in handling long sequences and are often used in tasks like speech recognition and language
modelling.

10
5. Gated Recurrent Unit Networks (GRUs): Similar to LSTMs, GRUs are used for
sequential data but have a simpler architecture, making them computationally efficient.
In summary, neural networks are a foundational technology in artificial intelligence,
enabling machines to learn and make decisions from data. Their ability to model complex
patterns and relationships has led to significant advancements in various fields, making
them a critical component of modern AI systems.
The Long Short-Term Memory (LSTM) algorithm stands out as a groundbreaking
innovation in sequential data modelling. LSTMs are a type of recurrent neural network
(RNN) designed to overcome the limitations of traditional RNNs when dealing with long
sequences and learning long-term dependencies. In this comprehensive exploration, we
delve into the inner workings of the LSTM algorithm, its architecture, applications, training
methods, and why it has become a cornerstone in various fields, from natural language
processing to time series forecasting.
LSTM Architecture
1. Cell State: LSTMs maintain a cell state that serves as a memory repository. The cell state
can selectively store, read, and write information over long sequences.
2. Gates: LSTMs employ three types of gates—input, forget, and output gates—each
responsible for controlling the flow of information into and out of the cell state.
3. Activation Functions: Activation functions, such as the sigmoid and hyperbolic tangent
(tanh), play a crucial role in regulating gate outputs and cell state updates.
4. LSTM Units: LSTM units consist of interconnected components, including gates and
activation functions, working in concert to process and manage sequential data.
LSTM Training and Learning
1. Backpropagation Through Time (BPTT): LSTMs are trained using BPTT, a variant of
backpropagation that accounts for sequential dependencies by unrolling the network over
time.
2. Gradient Clipping: To mitigate gradient explosion during training, gradient clipping is
often applied to ensure stable convergence.
3. Initialization Strategies: Initialization of LSTM parameters is essential for effective
training. Common strategies include orthogonal initialization and the use of small random
values.
The Long Short-Term Memory (LSTM) algorithm has revolutionized the world of
sequential data modelling by providing a robust solution for capturing long-term
dependencies. Its architecture, consisting of memory cells, gates, and activation functions,
has made it indispensable in fields ranging from natural language processing to finance.
However, challenges remain, such as overfitting and interpretability, necessitating ongoing
research and innovation. As we continue to unlock the potential of LSTMs and their

11
applications, the future promises even more sophisticated solutions for mastering the
complexities of sequential data.
The A-LSTM (Attention-Based Long Short-Term Memory) algorithm is a
cutting-edge development in the field of machine learning and deep learning, particularly
in the domain of sequence modelling and time-series forecasting. A-LSTM builds upon the
foundational LSTM (Long Short-Term Memory) architecture by incorporating attention
mechanisms, which allow the model to focus on relevant parts of the input sequence and
improve its ability to capture dependencies and patterns in the data. In this explanation, we
will delve into the A-LSTM algorithm, its architecture, advantages, and applications.
A-LSTM extends LSTM by incorporating attention mechanisms, which are inspired
by human cognitive processes. Attention mechanisms enable the model to focus on specific
parts of the input sequence at different time steps, dynamically adjusting its focus as it
processes the sequence. This attention mechanism addresses some of the limitations of
traditional LSTMs and enhances the model's performance in various tasks.
A-LSTM introduces two critical components:
1. Attention Mechanism: The attention mechanism allows the A-LSTM model to weigh the
importance of different elements in the input sequence when making predictions. It assigns
higher attention scores to relevant parts of the sequence, making the model more adaptive
and context-aware.
2. Long Short-Term Memory (LSTM) Cells: A-LSTM retains the LSTM cells, which are
responsible for maintaining the memory of past information in the sequence. LSTM cells
are equipped with gates that control the flow of information and gradients during training.
As the field of deep learning continues to evolve, A-LSTM represents a significant
step forward in enhancing sequence modelling and time-series forecasting. Future research
is likely to focus on refining attention mechanisms and integrating A-LSTM into more
complex architectures for even more accurate predictions and insights across a wide range
of applications. The adaptability and context-awareness of A-LSTM hold promise for
solving increasingly complex real-world problems.
Daily charging data in the context of electric vehicles (EVs) refers to the
information and metrics associated with the process of charging an electric vehicle on a
day-to-day basis. This data plays a crucial role in monitoring and managing the charging
behavior of EVs, optimizing charging infrastructure, ensuring safety, and promoting the
efficient use of electric energy. In this explanation, we will explore the significance of daily
charging data, its components, and its relevance in the rapidly evolving landscape of electric
mobility.
Components of Daily Charging Data:
1.Charging Time: The duration for which an EV is connected to a charging station,
measured in hours and minutes.

12
2. Charging Energy: The amount of electric energy transferred to the vehicle's battery
during the charging session, typically measured in kilowatt-hours (kWh).
3. Charging Cost: The monetary cost associated with the energy consumed during charging,
which depends on electricity rates and tariffs.
4. Charging Location: The physical location where the EV is charged, such as home
charging, workplace charging, public charging stations, or fast-charging stations.
5. Charging Efficiency: The ratio of energy delivered to the vehicle's battery to the energy
consumed from the grid, which can vary depending on charging equipment and conditions.
6. Charging Start and End Times: The specific times at which the charging session
commenced and concluded.
Significance of Daily Charging Data:
1. Range and Battery Management: Daily charging data provides insights into an EV's
charging patterns, allowing drivers to manage their vehicle's range effectively and plan their
daily commutes with confidence.
2. Cost Analysis: By tracking daily charging costs, EV owners can calculate the economic
benefits of electric mobility compared to traditional gasoline-powered vehicles.
3. Charging Infrastructure Planning: Utilities and infrastructure providers can utilize this
data to identify high-demand areas for charging stations and plan for infrastructure
expansion.
4. Grid Load Management: Grid operators can use daily charging data to manage the load
on the electrical grid more efficiently, especially during peak demand periods.
5. Energy Efficiency: Analysing daily charging data helps in optimizing charging profiles
and improving the overall energy efficiency of EVs.
As the adoption of electric vehicles continues to grow, daily charging data will
become increasingly valuable for optimizing charging networks, enhancing user
experiences, and supporting the transition to sustainable transportation. Innovations in data
analytics, machine learning, and real-time data processing will further unlock the potential
of daily charging data in shaping the future of electric mobility. Additionally, data sharing
initiatives and collaborations among stakeholders will play a pivotal role in leveraging this
data to its fullest potential.
According to statistics, the accident ratios for these scenarios are shown in Figure, with the
highest proportion of combustion accidents occurring when EVs are being charged, parked,
or driven.

13
 Figure 1: Accident Ratio
FACTORS AFFECTING THE SAFETY OF EV CHARGING
INTERNAL SHORT CIRCUIT IN BATTERY: current will be generated inside the battery
in a short period of time, causing the internal temperature of the battery to rise sharply,
which will lead to charging accidents caused by overcharge or over discharge, mechanical
damage and self-triggered internal short circuits
BATTERY OVERCHARGE: Overcharging of a battery can be a dangerous condition
where the charging equipment continues to supply energy to the battery for an extended
period of time. This can lead to safety accidents.
BATTERY SEPARATOR AND ELECTROLYTE MATERIAL: The battery separator
serves the critical function of isolating the positive and negative electrodes to prevent short
circuit accidents that may arise from the penetration of the separator during the
electrochemical reaction
ATTERY PACK CONSISTENCY: Due to differences in production technology and daily
usage, individual battery parameters of EV battery packs may become inconsistent, and
there may be differences between individual batteries due to internal decay effects of the
battery pack. The inconsistency of battery packs can lead to differences in SOC, voltage,
SOH, etc. between individual cells, which can seriously affect normal use

14
 4.METHODOLOGY

The safety warning model is developed based on the characteristics of EV charging

data, comprising five distinct steps. The first four steps involve offline processing to
establish EV charging warning thresholds, while the fifth step involves online comparison
of EV charging data with the warning thresholds to achieve status detection and warning.
The model operates as follows:

Step 1: EV charging data processing:

Firstly, the collected EV historical charging data are subjected to de-hybridization

operation to remove missing values and great outliers from the data, and then the data are
normalized according to equation to make the EV historical charging data vector EVhis a
standard data mapped to [−1, 1].

This can prevent the subsequent calculation errors caused by data changes, and at
the same time can improve the operation speed and prediction accuracy of the early
warning model.

EVhis-input is the normalized historical data value, which is used as the standard input data
in the subsequent steps.
EVhis is the original charging data
EVhis-max and EVhis-min are the maximum and minimum values of the corresponding data in
the original charging data.

Step 2: EV charging warning factor calculation

First, the EV charging safety warning factor rwf is determined by considering the
EV charging history data types and the EV charging safety influencing factors.

Then the corresponding data in EVhis, the EV charging history data set, are extracted
according to the warning factor, and the correlation coefficient ck among the influencing
factors is calculated by the Pearson correlation coefficient formula shown in equation (2)
and equation (3).

15
 i is the data number of the warning factor.
n is the number of input data
x and y are two different warning data

Finally, the correlation coefficients among the factors are compared, and the one
with the strongest correlation with other factors is selected as the charging safety warning
factor rwf

Table 1 : EV charging safety warning factor

cxy−i shown in Table 1 indicates the correlation coefficient between warning data x and y, x
and y are the average value of the corresponding warning data set, k is the warning factor
serial number, m is the number of warning factors, and m is taken as 6 according to Table
1

Step 3: EV charging data prediction

Extract EV standard charging data EVhis-input to select suitable parameters as input

EVinput, take the predicted data of charging factor as output EVpre, and build A-LSTM deep
network for regression calculation of the data.

Observe the prediction results, gradually adjust the hyperparameters of the

algorithm, explore the balance between the amount of input data and the prediction
accuracy of the algorithm, while ensuring the optimal prediction accuracy and prediction
time, determine the optimal hyperparameters of the algorithm, form the A-LSTM deep
learning network model, and then use the A-LSTM algorithm to fit the data, as shown in
equation (4) and equation (5) to obtain the EV warning factor prediction vector EVpre and
EV warning factor pressure residual vector EVre, which lay a good foundation for the
subsequent charging warning threshold setting.

16
Step 4: EV dynamic warning threshold setting
According to the EV prediction data residual EVre obtained in step 3 and the
‘‘Electric Vehicle Safety Requirements (GB 18384-2020)’’, the vehicle charging voltage
threshold array EVthr is established with the data characteristics.

Then the vehicle charging process is divided into 4 different warning regions
according to the SOC value, I, II, III and IV, considering the changing characteristics of the
vehicle charging data, and each region is set up with the warning The threshold adjustment
factor w, w is dynamically adjusted according to different regions, and equation (6)
constructs the dynamic warning threshold EVthr-dy for EV charging warning factor.

where j is the charging warning region serial number, taking the value of [1] and [4]
wj is the warning threshold adjustment factor of region j
λ is the adjustment factor, taking the value between [−1.15, 1.15] according to the specific
data variation

Step 5: EV charging safety warning

Input EV real-time charging data EVrt, record the initial SOC of EV charging as
SOC0 and the highest voltage of vehicle battery as V0, determine the vehicle warning area
according to the initial state of vehicle charging, then select the dynamic warning threshold
of the corresponding area and monitor the vehicle charging status, when the vehicle
charging status is abnormal, is abnormal, i.e., the real-time charging voltage, charging
current and temperature changes are different from the safety model built based on the
historical normal charging data, it is considered that there is a safety problem in EV
charging at this time, and the early warning is carried out in time and measures are taken to
deal with it.

The specific EV charging safety warning rules are as follows:

The first level is EV normal charging state, the vehicle charging data in this state is
below the warning threshold, which is the most ideal charging state and no alarm will be
made;

The second level is EV warning state, the EV SOC in this state is below 60% and
the maximum battery voltage is higher than the warning threshold, the state of this level
indicates that the EV charging is abnormal and protective measures should be taken in
time to make it converge to normal charging;

17
 The third level for the electric vehicle alert state, the state of the vehicle SOC in
60%-80%, the vehicle single battery maximum voltage is higher than the warning
threshold, at this time the vehicle state may have risk, should be alarmed and timely
measures to prevent the vehicle from danger;

The fourth level is EV dangerous state, the state of the vehicle SOC is above 80%,
the highest voltage of the vehicle single battery is higher than the warning threshold and
lasts for a long time, the EV continues to charge in this state is very likely to burn the car
accident, should promptly cut off the power and stop charging.

In summary, the EV charging safety warning model is constructed, and the

specific model operation block diagram is shown in Figure 2.

FIGURE 2. Flow chart of EV charging safety warning model.

4.1 LSTM algorithm

The LSTM neural network is a variant of Recurrent Neural Network (RNN) that
addresses the problems of gradient vanishing and explosion. It replaces the hidden layer of
the original RNN with LSTM units, which contain input, output, and forget gates. The
forget gate regulates the amount of historical input by controlling which information is
retained and which is discarded. The activation function of all three gates is the sigmoid
function, which produces values between 0 and 1. The gates learn to weight the historical
input, current input, and historical output, thereby achieving the memory function of
historical input and output.

18
 The LSTM unit is constructed as shown in (8) to (13), with the candidate LSTM
memory cell state value represented by:

x(t) is the input data of the historical charging of the EV at the current moment,
h(t −1) is the output of the LSTM unit at the previous moment,
ωx and ωhc are the connection weights corresponding to the two inputs x(t) and the output
h(t − 1),
C˜ (t) is the memory unit reference,
bc is the bias of the network

Values of the input gates of the LSTM network:

ωxi, ωhi and ωci are the input data for the current moment of EV historical charging, the
previous moment LSTM cell output and the previous moment cell output connection
weights to the input gate, respectively,
bi is the bias of the input gate.

The values of the forgotten gates of the LSTM network:

ωxf , ωhf and ωcf are the EV historical charging input data at that moment, the previous
moment LSTM cell output and the previous moment cell output connection weights to the
forgetting gate, respectively;
bf is the bias of the forgetting gate.

In this way, the current LSTM memory cell state value:

where ⊗ denotes the residence product operation.

And the value of the output gate of the LSTM network:

19
ωxo, ωho and ωco are the connection weights of the current moment’s input, the previous
moment’s LSTM cell output and the previous moment’s cell output to the output gate,
respectively
bo is the bias of the output gate

It is feasible to determine the output of the LSTM memory cell at a given time t by
combining equations (8) to (13)

4.2 A-LSTM ALGORITHM

To enhance the precision of the LSTM model’s predictions and mitigate its error, this study
utilizes the error correlation linear analysis approach and proposes the A-LSTM
(Adaptation LSTM) model. This model involves developing a relationship equation (13)
that correlates the input variables with the historical prediction error.

eh represents the LSTM historical prediction error;

f (x1, . . . , xn) is a primary function on the input, (x1, . . . , xn denotes the input).

The prediction model after error correction is shown in equation (14).

h(x1, . . . , xn) is the established LSTM prediction model

f (x1, . . . , xn) is the error linear correction function,
hA-LSTM is the final output of the A-LSTM algorithm.

20
 5. RESULT AND ANALYSIS

A. EV CHARGING DATA
The EV battery system used in this study consists of 18650 type ternary lithium-ion
batteries. The EV is composed of 92 individual cells connected in series to form one group,
and a total of 32 groups are connected in parallel to form the battery pack of the EV. Each
individual cell has a rated voltage of 3.7V and a rated capacity of 2.2Ah. Therefore, the
rated voltage and capacity of the EV battery pack are 328.5V and 69Ah, respectively.
Throughout the entire charging process, the EV adopts a typical three-stage charging
method, with slow charging before the battery pack’s SOC reaches 80% using constant
current charging. The charging termination voltage of the vehicle’s battery system is 377V,
and the discharge termination voltage is 276V. The operating temperature range for
charging is between 0 to 50◦C, and for discharging, it is between −20 to 60◦C. The
maximum continuous charging current allowed is 35A, and the maximum continuous
discharge current is 103A. The charging protection voltage for each individual cell is 4.1V,
and the discharge protection voltage is 3V.

The experimental EV operating data is presented in Table 2, which includes fundamental

information on vehicle operation and data related to the vehicle battery system.

TABLE 2. Summary of EV Charging Data.

21
B. EV CHARGING DATA PRE-PROCESSING

In this study, the researchers collected a substantial dataset consisting of 153,247 vehicle
charging records spanning a year. This dataset contained crucial information related to
vehicle-pile communication and other essential data types needed for a comprehensive
depiction of the vehicle charging scenarios. However, issues such as null values, abnormal
readings, and data duplications were observed due to factors like sensor errors and data
transmission issues. Additionally, frequent data collection led to multiple charging records
corresponding to the same State of Charge (SOC) value, necessitating data de-aggregation.

Following data pre-processing, the researchers focused on extracting pertinent charging

data and calculating correlation coefficients between various factors, including real-time
SOC, initial SOC during charging, State of Health (SOH) of the battery pack, charging
current, maximum single-cell temperature, and maximum single-cell voltage. The aim was
to assess the interrelationships between these factors and determine the charging safety
influence factor (rwf).

The results, summarized in Figure 3, revealed key insights. Notably, the real-time SOC of
the battery pack exhibited a strong correlation with both the electric vehicle (EV) charging
current and the highest single-cell voltage within the battery pack. Battery pack health, on
the other hand, displayed a robust correlation with the initial SOC during charging.
Surprisingly, the maximum temperature of the single cell showed weaker correlations with
other factors, challenging previous literature findings. It was deduced that this particular
variable was more influenced by ambient temperature variations, highlighting its sensitivity
to external conditions.

FIGURE 3. Pearson coefficient between parameters

22
In this study, the impact of ambient temperature on battery performance was examined by
considering temperature variations between winter and summer. The analysis incorporated
both EV motor and battery temperature changes, revealing substantial differences between
these temperature profiles in January and August. This stark contrast in temperatures
underscores the importance of ambient temperature in battery behavior, explaining the
weaker correlation between temperature and other factors. To address this, the study used the
stable motor temperature as a reference for ambient temperature and reconstructed the battery
temperature by calculating the temperature difference between the two. The revised
temperature data significantly improved the correlation coefficients with other factors, except
for SOH, as the original SOH data covered only a limited range (94% to 96.5%).

Upon closer examination of the correlation coefficients in Figure 5, it became evident that
single cell voltage exhibited stronger correlations with other factors when considered in
combination. Consequently, single cell voltage was identified as a key warning factor,
offering valuable insights into battery performance and safety considerations during
charging. This improved understanding of temperature's influence on battery behavior
contributes to more accurate predictive models for enhanced charging safety and efficiency.

FIGURE 4. Charging temperature curve of vehicle motor and battery in January and August.

23
.
FIGURE 5. Pearson coefficient after improved temperature expression.

C. CHARGING VOLTAGE PREDICTION

Constructing a robust algorithm model is contingent upon defining crucial

hyperparameters for the A-LSTM algorithm. These hyperparameters encompass the
determination of neural unit counts in the input and output layers, the specification of the
number of layers in the hidden layer, and the selection of the neural unit count within the
hidden layer. To evaluate the predictive performance of the A-LSTM algorithm across
various parameter configurations, the study adopts the root mean square error (SRMSE or
RMSE) and the mean absolute percentage error (sMAE or MAE) as assessment criteria.
Formulas for these evaluation criteria are presented in equations (16) and (17). These
chosen evaluation metrics provide a comprehensive measure of the algorithm's prediction
accuracy, facilitating a robust assessment of its performance under varying hyperparameter
settings.

24
 TABLE 3. Results of MAE with different parameters of A-LSTM.

TABLE 4. A-LSTM RMSE results with different parameters.

To identify the optimal parameters for the A-LSTM model, a systematic exploration
is conducted, varying the number of hidden layers from 1 to 4 and the number of neurons
within each layer from 50, 100, 150, to 200. The performance of the model is evaluated
using prediction results and assessment metrics for each parameter combination. The
results, presented in Tables 3 and 4, reveal that the A-LSTM model performs optimally
when configured with three hidden layers, each containing 100 hidden units. Accordingly,
the A-LSTM model's architecture is finalized, featuring three hidden layers, each
comprising 100 units. Additional hyperparameters are detailed in Table 5, contributing to
the establishment of an effective predictive model for charging behavior.

TABLE 5. Hyperparameter setting of A-LSTM model.

25
 FIGURE 6. EV charging data prediction error.

FIGURE 7. Comparison of prediction results.

From Figures 6 and 7, it is evident that the proposed algorithm yields high accuracy in
prediction the maximum voltage of the individual battery during the charging of EVs.

This facilitates the prediction of the entire voltage change process during vehicle
charging. The variation in initial SOC from 0 to 99.9% during charging is illustrated in
Fig. 8.

26
 FIGURE 8. EV charging data forecast results.

To verify the prediction accuracy of the A-LSTM model proposed in this paper, we selected
BP neural network, RNN neural network, LSTM neural network, and A-LSTM neural
network to fit the charging voltage of the vehicle. The BP neural network was set up with
3 fully connected layers, while the RNN, LSTM, and A-LSTM neural networks were set
up with 3 layers, and the number of hidden units in each layer was set to 100, respectively

As illustrated in Figure 9, the results demonstrate that the A-LSTM algorithm achieved
exceptional prediction accuracy, with its prediction results evenly distributed around the
centre of the graph and a maximum R2 value of 0.99781. This outstanding performance
signifies the algorithm's capability to accurately predict voltage variations during EV
charging, thereby providing reliable prediction data essential for setting charging warning
thresholds. Moreover, when comparing the prediction residuals of the A-LSTM algorithm
with those of the LSTM algorithm, as shown in Figure 10, it becomes evident that the A-
LSTM model consistently outperforms the LSTM model. The residuals of the A-LSTM
algorithm are concentrated within a narrower range, highlighting its superior prediction
accuracy. These findings underscore the effectiveness of the A-LSTM algorithm in
predicting EV charging behavior, offering valuable insights for charging safety and
management.

27
 FIGURE 9. Comparison of R2 values tested by different algorithms.

D. EV CHARGING SAFETY WARNING EFFECT

In this study, the establishment of vehicle charging warning thresholds was facilitated
through the utilization of a dynamic threshold model, which takes into account the
predictive charging values. Specifically, a charging dataset was selected, and the dynamic
threshold construction approach was employed, dividing the charging curve into four
distinct regions based on voltage change characteristics.

In the first region (warring area I) with a State of Charge (SOC) between 10% and 30%,
the voltage undergoes a rapid rise while maintaining a low level. To address this scenario,
the warning threshold was set to ±(1.12%∼1.16%) of the normal value. Moving to the
second region (warring area II) encompassing SOC levels of 30% to 60%, the voltage
exhibits a slower increase. Consequently, the warning threshold in this region was
configured as ±(1.08%∼1.12%) of the normal value. Region III (SOC 60%∼80%) features
higher voltage levels with a further slowed increase, posing an elevated risk of
overcharging. Here, the threshold was defined as ±(1.05%∼1.08%) of the normal value.
Lastly, in Region IV (SOC 80%∼100%), where the battery nears completion, and voltage
is close to its peak, making it susceptible to overcharging, the threshold was established as
±(1.03%∼1.05%) of the normal value. Any deviation from these thresholds would trigger
a warning, with the threshold range further contracting in the presence of abnormal battery
behavior. This dynamic threshold approach offers enhanced flexibility in warning vehicles
during charging, ensuring safety and maximizing the security of the vehicle charging point.

28
FIGURE 10. Histogram of prediction residuals for different algorithms.

FIGURE 11. Dynamic warning threshold.

29
 FIGURE 12. Dynamic warning threshold after adjustment.

To evaluate the accuracy of the early warning model, three types of simulated
vehicle fault state data were generated using the fault data setting method outlined in Table
6. These fault states introduced variations in the voltage data (V err) compared to normal
charging data (Vnor) according to the specified ranges. Specifically, equations (20), (21),
and (22) were used to simulate fault voltage data with different levels of deviation from
normal charging values.
To assess the model's accuracy, the warning accuracy (wa) was calculated using
equation (19), where Ndet represents the number of fault data detected by the model, and
Nerr represents the total number of fault data points in the study. The warning accuracy is
expressed as a percentage and reflects the model's ability to correctly identify fault states
among the simulated data.

To validate the accuracy of the proposed charging warning model, predictions from
the BP neural network, RNN neural network, LSTM neural network, and A-LSTM neural
network were utilized to establish both fixed and dynamic warning thresholds. The warning
models were then executed sequentially, and the resulting warning accuracy was recorded.
The outcomes of this analysis are visualized in Figures 13 and 14, providing insights into
the model's performance under various thresholding strategies and highlighting its
effectiveness in detecting charging anomalies.

30
 FIGURE 13. Dynamic threshold warning accuracy rate.

FIGURE 14. Fixed threshold warning accuracy.

The proposed charging warning model in this study demonstrates remarkable

accuracy in identifying vehicle charging faults for fault types 2, 3, and 4, achieving an
impressive average accuracy rate of 99.0%. A comparison between Figure 13 and Figure
14 reveals the efficacy of dynamic warning thresholds in enhancing charging safety
warnings across all four fault types. Notably, for fault types 2 and 4, the dynamic threshold
improves the warning accuracy by an average of 2.51%. In the case of the third charging
fault type, the dynamic threshold exhibits a substantial average improvement of 5.84%
compared to using a fixed threshold. This improvement stems from the model's ability to
diligently monitor and issue warnings throughout the subsequent charging process by
dynamically adjusting the warning threshold following fault detection.

31
 Furthermore, the assessment of the warning model extends to the measurement of
warning time, represented by 1SOC, where a lower 1SOC value signifies more timely fault
detection. The dynamic threshold warning model, as depicted in Figure 15, excels in
providing early warnings with a short lead time. For the first fault type, where the SOC is
low, the dynamic threshold is set more leniently to prevent false alarms, resulting in a
required warning 1SOC of 3.8%. Conversely, for fault type 3, characterized by higher SOC
and battery voltage, the dynamic threshold is set more rigorously to maximize vehicle
safety, necessitating a smaller required warning 1SOC of 0.8%. In comparison with fixed
thresholds, the dynamic threshold approach introduced in this paper offers exceptional
warning timeliness across various fault scenarios, underscoring its effectiveness in
enhancing charging safety and preventing potential hazards.

FIGURE 15. Time-limit of warning for different models

32
 6. CONCLUSION

In summary, this paper presents a robust state monitoring and fault warning approach
tailored for electric vehicle charging processes, leveraging deep learning techniques
from the charging-side perspective. The primary objective is to enhance the safety of
electric vehicle charging operations while fostering the convergence of automotive
safety technology with an array of available data resources, aligning with the broader
national big data strategy. The proposed method exhibits the capacity to efficiently
monitor diverse physical data associated with EV charging, facilitate timely fault
warnings, and mitigate the risk of false alarms stemming from erroneous charging data.
This innovative approach holds significant promise in advancing the safety and
reliability of electric vehicle charging systems, ultimately contributing to the broader
sustainable mobility landscape.

Nonetheless, it's important to acknowledge certain limitations within this study, notably
in terms of data acquisition related to State of Health (SOH) and the absence of
comprehensive life cycle EV charging data. Future research endeavours should delve
deeper into these areas, seeking to address these limitations by exploring strategies for
more comprehensive data gathering and analysis. By doing so, we can further refine and
optimize the proposed methodology, ensuring its applicability and effectiveness in real-
world electric vehicle charging scenarios while continually advancing the field of
electric vehicle charging safety and monitoring.

33
 REFERENCES

[1] X. Feng, Y. Pan, X. He, L. Wang, and M. Ouyang, ‘‘Detecting the internal short circuit
in large-format lithium-ion battery using model-based faultdiagnosis algorithm,’’ J. Energy
Storage, vol. 18, pp. 26–39, Aug. 2018, doi: 10.1016/j.est.2018.04.020.

[2] B. Mao, C. Zhao, H. Chen, Q. Wang, and J. Sun, ‘‘Experimental and modeling analysis
of jet flow and fire dynamics of 18650-type lithiumion battery,’’ Appl. Energy, vol. 281,
Jan. 2021, Art. no. 116054, doi: 10.1016/j.apenergy.2020.116054.

[3] A. Barré, F. Suard, M. Gérard, M. Montaru, and D. Riu, ‘‘Statistical analysis for
understanding and predicting battery degradations in real-life electric vehicle use,’’ J.
Power Sources, vol. 245, pp. 846–856, Jan. 2014, doi: 10.1016/j.jpowsour.2013.07.052.

[4] X. Feng, S. Zheng, D. Ren, X. He, L.Wang, H. Cui, X. Liu, C. Jin, F. Zhang, C. Xu, H.
Hsu, S. Gao, T. Chen, Y. Li, T. Wang, H. Wang, M. Li, and M. Ouyang, ‘‘Investigating the
thermal runaway mechanisms of lithiumion batteries based on thermal analysis database,’’
Appl. Energy, vol. 246, pp. 53–64, Jul. 2019, doi: 10.1016/j.apenergy.2019.04.009.

[5] G. Zhang, X. Wei, X. Tang, J. Zhu, S. Chen, and H. Dai, ‘‘Internal short circuit
mechanisms, experimental approaches and detection methods of lithium-ion batteries for
electric vehicles: A review,’’ Renew. Sustain. Energy Rev., vol. 141, May 2021, Art. no.
110790, doi: 10.1016/j.rser.2021.110790.

[6] P. Bangalore and L. B. Tjernberg, ‘‘An artificial neural network approach for early fault
detection of gearbox bearings,’’ IEEE Trans. Smart Grid, vol. 6, no. 2, pp. 980–987, Mar.
2015, doi: 10.1109/TSG.2014.2386305.

[7] M.-K. Tran and M. Fowler, ‘‘A review of lithium-ion battery fault diagnostic
algorithms: Current progress and future challenges,’’ Algorithms, vol. 13, no. 3, p. 62, Mar.
2020, doi: 10.3390/a13030062.

[8] Y. Zhang, T. Li, X. Yan, L. Wang, J. Zhang, X. Diao, and B. Li, ‘‘Electric vehicle
charging fault monitoring and warning method based on battery model,’’ World Electr.
Vehicle J., vol. 12, no. 1, p. 14, Jan. 2021, doi: 10.3390/wevj12010014.

[9] M.-K. Tran, M. Mathew, S. Janhunen, S. Panchal, K. Raahemifar, R. Fraser, and M.

Fowler, ‘‘A comprehensive equivalent circuit model for lithium-ion batteries, incorporating
the effects of state of health, state of charge, and temperature on model parameters,’’ J.
Energy Storage, vol. 43, Nov. 2021, Art. no. 103252, doi: 10.1016/j.est.2021.103252.

[10] W. Song, ‘‘Research on integrated safety warning and protection system of electric
vehicle charging,’’ M.S. thesis, Nanjing Univ. Posts Telecommun., 2019, vol. 2.

[11] L. Qian, M. Zhao, and W. Zhang, ‘‘A method to design the security early warning
model of EV charging,’’ Adv. Power Syst. Hydroelectr. Eng., vol. 32, no. 12, pp. 114–119,
2016.

34
[12] Y. Zhao, P. Liu, Z. Wang, L. Zhang, and J. Hong, ‘‘Fault and defect diagnosis of
battery for electric vehicles based on big data analysis methods,’’ Appl. Energy, vol. 207,
pp. 354–362, Dec. 2017, doi: 10.1016/j.apenergy.2017.05.139.

[13] L. Zhang, T. Gao, G. Cai, and K. L. Hai, ‘‘Research on electric vehicle charging safety
warning model based on back propagation neural network optimized by improved gray
wolf algorithm,’’ J. Energy Storage, vol. 49, May 2022, Art. no. 104092, doi:
10.1016/j.est.2022.104092.

[14] B. Xia, Y. Shang, T. Nguyen, and C. Mi, ‘‘A correlation based fault detection method
for short circuits in battery packs,’’ J. Power Sources, vol. 337, pp. 1–10, Jan. 2017, doi:
10.1016/j.jpowsour.2016.11.007.

[15] D. Gao, Y. Wang, X. Zheng, and Q. Yang, ‘‘A fault warning method for electric
vehicle charging process based on adaptive deep belief network,’’ World Electr. Vehicle J.,
vol. 12, no. 4, p. 265, Dec. 2021, doi: 10.3390/wevj12040265.

[16] K. Zhang, Z.Yin, X.Yang, Z.Yan, andY. Huang, ‘‘Quantitative assessment of electric
safety protection for electric vehicle charging equipment,’’ in Proc. Int. Conf. Circuits,
Devices Syst. (ICCDS), Sep. 2017, pp. 89–94.

[17] Y. Jing, ‘‘Research on safety monitoring and fault warning method of electric vehicle
charging based on data mining technology,’’ M.S. thesis, North China Electr. Power Univ.,
vol. 3, 2022, doi: 10.27139/d.cnki.ghbdu.2021.000707.

[18] R. Yang, R. Xiong, S. Ma, and X. Lin, ‘‘Characterization of external short circuit faults
in electric vehicle Li-ion battery packs and prediction using artificial neural networks,’’
Appl. Energy, vol. 260, Feb. 2020, Art. no. 114253, doi: 10.1016/j.apenergy.2019.114253.

[19] L. Jiang, Z. Deng, X. Tang, L. Hu, X. Lin, and X. Hu, ‘‘Data-driven fault diagnosis
and thermal runaway warning for battery packs using realworld vehicle data,’’ Energy, vol.
234, Nov. 2021, Art. no. 121266, doi: 10.1016/j.energy.2021.121266.

[20] J. Lamb, C. J. Orendorff, L. A. M. Steele, and S. W. Spangler, ‘‘Failure propagation

in multi-cell lithium ion batteries,’’ J. Power Sources, vol. 283, pp. 517–523, Jun. 2015,
doi: 10.1016/j.jpowsour.2014.10.081.

[21] J. Hou, X. Feng, L. Wang, X. Liu, A. Ohma, L. Lu, D. Ren, W. Huang, Y. Li, M. Yi,
Y. Wang, J. Ren, Z. Meng, Z. Chu, G.-L. Xu, K. Amine, X. He, H. Wang, Y. Nitta, and M.
Ouyang, ‘‘Unlocking the self-supported thermal runaway of high-energy lithium-ion
batteries,’’ Energy Storage Mater., vol. 39, pp. 395–402, Aug. 2021, doi:
10.1016/j.ensm.2021.04.035.

[22] B. Liu, ‘‘Design of charging control flow based on electric vehicles,’’ Automobile
Parts, vol. 147, no. 9, pp. 6–9, 2020, doi: 10.19466/j.cnki.1674-1986.2020.09.002.

[23] Y. Wu and Y. Wang, ‘‘Review of internal short circuit of lithium-ion battery,’’ Mach.
Building Autom., vol. 49, no. 4, pp. 169–172, 2020.

35
[24] D. Ouyang, Y. He, M. Chen, J. Liu, and J. Wang, ‘‘Experimental study on the thermal
behaviors of lithium-ion batteries under discharge and overcharge conditions,’’ J. Thermal
Anal. Calorimetry, vol. 132, no. 1, pp. 65–75, Apr. 2018, doi: 10.1007/s10973-017-6888-
x.

[25] X. Cheng, R. Zhang, C. Zhao, F. Wei, J. Zhang, and Q. Zhang, ‘‘A review of solid
electrolyte interphases on lithium metal anode,’’ Adv. Sci., vol. 3, no. 3, Mar. 2016, Art. no.
1500213, doi: 10.1002/advs.201500213.

[26] Y. Liu, C. Zhang, J. Jiang, W. Zhang, and L. Zhang, ‘‘Research on capacity difference
identification method of lithium-ion battery pack,’’ Proc. CSEE, vol. 41, no. 4, pp. 1422–
1430, 2021, doi: 10.13334/J.0258-8013.PCSEE.200483.

36
 APPENDIX

Abbreviation Full Form

EV ELECTRIC VECHILE

37