Ain Shams Engineering Journal xxx (xxxx) xxx

Contents lists available at ScienceDirect

Ain Shams Engineering Journal

journal homepage: www.sciencedirect.com

Battery management solutions for li-ion batteries based

on artificial intelligence
Yara Khawaja a, Nathan Shankar b, Issa Qiqieh c,⇑, Jafar Alzubi c, Omar Alzubi d, M.K. Nallakaruppan e,
Sanjeevikumar Padmanaban f
a
Faculty of Engineering and Technology, Applied Science Private University, Amman, Jordan
b
School of Electrical Engineering, Vellore Institute of Technology, India
c
Faculty of Engineering, Al-Balqa Applied University, Al-Salt, Jordan
d
Prince Abdullah bin Ghazi Faculty of Information and Communication Technology, Al-Balqa Applied University, Al-Salt, Jordan
e
School of Information Technology and Engineering, Vellore Institute of Technology, India
f
Department of Electrical Engineering, IT and Cybernetic, University of South-Eastern Norway, Porsgrunn, Norway

a r t i c l e i n f o a b s t r a c t

Article history: The automobile industry is currently undergoing a paradigm change from conventional, diesel, and
Received 7 January 2023 gasoline-powered vehicles to hybrid and electric vehicles of the second generation. Lithium-ion (Li-
Revised 14 February 2023 ion) batteries have sparked the automotive industry’s interest for quite some time. One of the most cru-
Accepted 23 February 2023
cial components of an electric car is the battery management system (BMS). Since the battery pack is an
Available online xxxx
electric vehicle’s most significant and expensive component, it must be carefully monitored and con-
trolled. The precise measurement and calculation of the many states of a Li-ion battery’s cells, such as
Keywords:
the State of Health (SOH) and State of Charge (SOC) is a difficult procedure as they cannot be monitored
Artificial Intelligence
Battery Management System
directly. This paper examines various methodologies and approaches for estimating the SOC and SOH of
Lithium-ion Batteries Li-ion batteries using Artificial Intelligent methods. Six machine learning algorithms are intensively uti-
Neural Network lized to investigate the Li-ion battery state estimation. The employed methods are linear, random forest,
State of Charge gradient boost, light gradient boosting (light-GBM), extreme gradient boosting (XGB), and support vector
State of Health machine (SVM) regressors. In comparison to all other models employed in this study, the discharge pre-
diction made using random forest exhibits significantly greater performance at a low loss of accuracy. For
instance, with the highest R2-score of 0.999, the random forest regressor achieves only 0.0035, 0.0013,
and 0.0097 for mean and median absolute error, and root means squared error (RMSE), respectively.
We showed that the state estimation of Li-ion batteries can be precisely predicted using AI methods,
which can be combined with a battery management system to improve electric vehicle performance.

2023 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Ain Shams Uni-
versity. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/
by-nc-nd/4.0/).

1. Introduction of the transportation sector. Due to the irregular nature of renew-

Because of overexploitation in several sectors, particularly
transportation and energy, worldwide stocks of fossil fuels are
rapidly depleting. Overexploitation of fossil fuels produces massive
volumes of CO2 and other Green House Gas Emissions (GHGE),
which has had a significant impact on the environment and con-
tributed to climate change. The GHGE can be decreased by up to
40% [1] with the use of renewable energy and the electrification
⇑ Corresponding author.
E-mail addresses: y_khawaja@asu.edu.jo (Y. Khawaja), nathanshankar465@g-
mail.com (N. Shankar), i.qiqieh@bau.edu.jo (I. Qiqieh), j.zubi@bau.edu.jo (J. Alzubi),
o.zubi@bau.edu.jo (O. Alzubi), nallakaruppan.mk@vit.ac.in (M.K. Nallakaruppan),
sanjeevi_12@yahoo.co.in (S. Padmanaban).
able energy sources such as wave, wind, tidal, and solar, an energy
storage system (ESS) is used to make the supply to the customer
more reliable [2–8]. (SEE Table 1.).
The Electric Vehicle (EV) as shown in Fig. 1 is thought to be the
answer to reducing the hazardous pollution emissions from auto-
mobiles. Additionally, because electric vehicles can be utilized as
energy storage systems to store energy from renewable energy
sources, they can engage actively with the electrical grid [9]. This
is known as vehicle-to-grid (V2G) interaction. In recent years,
many chemistries of energy storage systems (ESSs) have been
approved for use in transportation [10]. Li-ion batteries, nickel–
cadmium batteries, and lead acid batteries are the most commonly
used batteries in EVs. However, Li-ion batteries have grown in

https://doi.org/10.1016/j.asej.2023.102213
2090-4479/
2023 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Ain Shams University.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: Y. Khawaja, N. Shankar, I. Qiqieh et al., Battery management solutions for li-ion batteries based on artificial intelligence, Ain
Shams Engineering Journal, https://doi.org/10.1016/j.asej.2023.102213
Y. Khawaja, N. Shankar, I. Qiqieh et al. Ain Shams Engineering Journal xxx (xxxx) xxx

Table 1 which is dependent on a precise SOC estimate. When cells are

List of Abbreviations and Symbols. grouped in large numbers in either series or parallel modules in
Abbreviations a battery stack to provide increased power output based on
SOC State of Charge requirements, the pack’s safety is critical [17]. If Li-ion batteries
SOH State of Health are utilized beyond their safety tolerances, they can deteriorate
Li-ion Lithium-ion and ultimately become a hazard.
BMS Battery Management System A Battery Management System (BMS) is a set of software and
ESS Energy Storage System
EV Electric Vehicle
hardware designed to improve a battery’s charge and discharge
FLS Neuro-Fuzzy Systems cycles while also extending its life [15]. For our needs, the BMS cal-
ES Expert Systems culates and monitors two crucial metrics. The first is the State of
SVM Support Vector Machine Health (SOH), which depicts the battery’s performance in compar-
EA Evolutionary Algorithm
ison to its past and anticipated future. The State of Charge (SOC)
HIS Hybrid Intelligent Systems
ANN Artificial Neural Network refers to the amount of charge in a battery throughout a charge
AI Artificial Intelligence or discharge cycle. Neither of the parameters can be measured
XGB Extreme Gradient Boosting directly [16]. There are several methods for determining the SOH
Light- Light Gradient Boosting and SOC of a battery pack. Data-driven methods, model-based
GBM
methods, or even advanced sensing-based methods can be used
SEI Solid Electrolyte Interphase
V2G Vehicle-to-Grid to estimate SOC and SOH.
RMSE Root Means Squared Error When it comes to the design of BMS in EVs, accurate battery
MSE Means Squared Error pack parameter estimations are crucial and important; they also
ML Machine learning
provide some important supplementary data, such as the remain-
GHGE Green House Gas Emissions
Symbols ing life or useable time [17]. BMSs also prevent Li-ion batteries
Eoc Open circuit voltage from being overcharged or discharged. Because the Li-ion battery
Vterm Battery terminal voltage is a complex, nonlinear, and time-varying electrochemical system,
R0 Series resistance in battery equivalent circuit its performance varies as operating conditions change, such as
Rm mth parallel resistance in RC branch in battery equivalent circuit
charge–discharge current, aging, and fluctuations [18]. Since direct
Cm mth parallel capacitance in RC branch in battery equivalent
circuit interaction with a sensor is not a possibility, precise calculation of
iðtÞ applied current to the battery equivalent circuit SOC and SOH for a Li-ion battery is a difficult task.
im ðtÞ Current flowing through the mth RC branches The traditional battery management approaches have two fun-
K Gain of the transfer function damental flaws. The first disadvantage is that you must use num-
bi Numerator parameters of the transfer function
ai Denominator parameters of the transfer function
bers from ‘‘baseline” models, which represent the average of the
BðsÞ Battery transfer function laboratory results [19]. In real-world applications, batteries exhibit
v ðsÞ Battery voltage in the Laplace domain a wide range of attributes over a wide range of operating situa-
iðsÞ Battery current in the Laplace domain tions. As a result, the accuracy of traditional models is limited
[20]. Furthermore, as batteries age, their properties alter, increas-
ing the model’s ineptitude even more. The second disadvantage
is that the parameters of the battery must be computed while
the battery is ‘‘offline,” or out of operation, for the duration of
the measurements [21]. For state estimate and prediction in
Radiator
BMS, current research is aiming to use numerous additional
options such as Computational Intelligence, Artificial Intelligence
(AI), and others [22].
The importance of artificial intelligence and its following com-
ponents has long been recognized. As a result, during the function-
Fuel
ing of parts incorporating this technology, manual involvement is
Storage the last thing that may be requested or observed. These machines
accelerate activities and processes while ensuring precision and
accuracy making them a valued and helpful tool. As established
Electric earlier due to the increased popularity of Li-ion batteries, it is
Motor
Power
imperative to develop different algorithms to estimate SOC and
Electronics SOH due to the non-linear nature of a Li-ion battery. In our work
Box Battery
we leverage this opportunity to make the following key
contributions:
Fig. 1. A Standard Electric Vehicle.
1. We perform an extensive literature review of the diverse AI
popularity as a result of their increased dependability, power den- technologies employed in battery management systems.
sity, energy density, efficiency, longer lifespan, reduced discharge 2. We investigate six AI strategies: ANN, FLS, ES, SVM, EA, and HIS
rates, and high efficiency [11]. Another significant element con- to develop an advanced AI strategy for accurate prediction of
tributing to the widespread commercialization of Li-ion batteries battery’s SOC and SOH in EVs.
is their decreasing manufacturing costs, which supports their use 3. We perform a comparative analysis of the different AI tech-
in a variety of industries [12]. niques employed based on several metrics to determine the
Li-ion batteries have been employed in the ESSs ranging in size best AI strategy for accurate estimation of battery’s state in EVs.
from a few kilowatt-hours in household systems to multi-
megawatt batteries in power grids [13]. Despite its potential for The rest of the paper is organized as follows. Section 2 delves
usage in energy storage solutions, Li-ion batteries have a few lim- into the internal model of the battery. Section 3 examines the fun-
itations, including the need for a battery pack’s safe operating zone, damentals of BMS and analyses the various AI approaches used in
2
Y. Khawaja, N. Shankar, I. Qiqieh et al.
BMS and their merits and demerits. Section 4 gives a detailed anal-
ysis of the different AI approaches used with their merits and
demerits followed by Section 5 which details the results and anal-
ysis for a sample battery dataset. Section 6 gives a comparative
analysis of the different AI techniques used in Section 5.
2. Lithium-Ion battery model
Currently, battery modeling for SOC determination is routinely
created using a variety of equivalent circuit (RC network) models,
each with its own set of material properties and accuracy criteria.
The generic model, on the other hand, is based on the assumption
that the internal resistance remains constant during charge and
discharge cycles. Therefore, the correctness of this model is debat-
able. The battery deterioration model based on capacity fading was
simulated and created while taking the SOH estimate into account.
These model parameters were mostly determined by the physical
properties of the individual anode and cathode. However, in a
dynamic setting, external factors such as ambient temperature
and discharge current load will cause these stationary models to
be erroneous.
In this investigation, a fourth-order electrochemical model was
used due to its ability to correctly record the battery’s complicated
terminal voltage behavior. As indicated in Fig. 2, ten characteristics
must be determined, including the open circuit voltage source
(Eoc), series resistance (R0), and the resistance and capacitance of
each RC network (Rm, Cm), where m = 1:4.
At time t, the terminal voltage (Vterm) may be computed as
follows:
V term ðnÞ ¼ Eoc ðSOC n Þ  iðt ÞR0  i1 ðtÞR1  i2 ðtÞR2  i3 ðtÞR3  i4 ðtÞR4
ð1Þ
where iðt Þ denotes the applied current and im ðtÞ (where m = 1:4)
denotes the currents flowing through the mth RC branches, respec-
tively. The voltage varies as a function of current change at the
SOC-defined operating point, and the voltage Eoc for the electro-
chemical model in Fig. 2 is determined by a fourth-order transfer
function.
b0 þ b1 s1 þ b2 s2 þ b3 s3 þ b4 s4
BðsÞ ¼ K ; ð2Þ
1 þ þa1 s1 þ a2 s2 þ a3 s3 þ a4 s4
where:
 K is the gain of the transfer function
 bi " i = 0,. . .4 – numerator parameters of the transfer function
 ai " i = 0,. . .4 – denominator parameters of the transfer function
 BðsÞ is the battery transfer function.
In the Laplace domain, the battery voltage is given by the
expression:
v ðsÞ ¼ Eoc þ BðsÞiðsÞ; ð3Þ
where:
C1 C2 C3 C4
R0
Eoc
R1 R2 R3 R4
Fig. 2. Equivalent Circuit Model of the battery.
Ain Shams Engineering Journal xxx (xxxx) xxx
 v ðsÞ is the battery voltage in the Laplace domain [V].
 iðsÞis the battery current in the Laplace domain [A].
3. AI strategies in battery management systems
A Battery Management System (BMS) is a software and hard-
ware system that regulates the battery for effective functioning
[23]. A BMS is made up of various functional units, such as a cell
voltage balance, fuel gauge monitor, cut-off field effect transistor,
a cell voltage monitor, a state machine, temperature monitors,
and a real-time clock [24]. There are several varieties of BMS-
integrated chips on the market. The functional pieces are organized
differently for different systems; they might range from a simple
analog front end with a microcontroller capable of balancing and
monitoring to a stand-alone fully integrated solution capable of
running autonomously.
The BMS in EVs may incorporate a variety of actuators, controls,
and sensors. BMSs are responsible for safeguarding batteries, oper-
ating batteries within acceptable parameters of voltage, current,
and temperature, and accurately monitoring battery parameters
[25]. In terms of hardware structure, three basic types of topologies
have been used: modular architectures, centralized, and dis-
tributed. Richter and Meissner presented a layer structure for mon-
itoring and managing the status of a battery [26]. According to
Gold, BMSs can be classified based on their different functionalities
[27]. These concepts could be used to create a broad framework
with basic functionality. Various sensors located within the battery
pack collect data at the monitoring layer [28]. The BMS is linked to
all of the battery-pack components as well as the vehicle control
processor BMSs have a long history of using safety [28]. However,
the suggested BMS uses more sensors than the existing safety cir-
cuitry designs, allowing for advancements like the addition of pre-
cise warnings and controls to avoid overcharge, over-discharge,
and overheating. To monitor and measure battery characteristics
such as cell voltage, current, and temperature of a battery, a system
of sensors is required. However, the viability of these measures
outside of laboratory contexts is hampered by both space limita-
tions and device expense. Therefore, it is important to measure
current, temperature, and voltage to enhance the capabilities of
state tracking in practical applications. Modeling, data collection
(via data acquisition system DAQ), and data storage are essential
components for obtaining the desired data.
The charge–discharge procedure is controlled by a subsystem
called the charge controller. For balancing cells or monitoring
internal resistance, a variable resistor could be required. Cell bal-
ancing management is still an important design aspect that can
be improved to equalize the cells of the battery pack and accu-
rately determine the battery’s state. Since most BMS subsystems
are standalone modules, internal data transmission is necessary.
The main method of data transfer inside the BMS is communica-
tion over a Controller Area Network (CAN) bus. More information
can be gathered by building intelligent batteries such that micro-
chips embedded in the battery can communicate with the users
and the chargers. Furthermore, in charging systems, radio, and
communication technologies are gradually being implemented
that facilitate communication between the battery and charger.
As temperature variations affect cell imbalance, reliability, and
performance, a thermal management module is essential. It is
therefore crucial to reduce the temperature disparity between
cells, which must be regulated and run under the right tempera-
Vt ture circumstances.
The BMS software is the brains of the entire system since it
manages all equipment functions and evaluates sensor data for
decision-making and status predictions. The software of a BMS
should be able to handle control switching, sample rate tracking
3
Y. Khawaja, N. Shankar, I. Qiqieh et al.
in the sensor module, cell balance management, and even the con-
struction of dynamic safety circuits. In addition, for continuous
updates and control of battery functions, web-based data analysis
and processing are required. Robust and reliable automated data
analysis is vital for success since it determines state estimation
and problem detection. The user will see this information via a
user-friendly interface with the necessary suggestions. The calcula-
tion of SOH and SOC will be a part of a capability evaluation that
also displays the battery’s life status and establishes operational
restrictions using cutting-edge techniques like fuzzy systems, arti-
ficial neural networks, state-space-based models, etc.
Cell balancing is performed to enhance the performance of the
battery without overcharging and over-discharging, its purpose is
to bring cells with similar SOC levels together. The controller will
manage the charging procedure using a detailed program based on
the level of charge retained by each cell. Accordingly, a key compo-
nent of enhancing the balancing is an accurate SOC calculation of
every cell. To deliver battery fault warnings and identify out-of-
tolerance conditions, smart data evaluation is required. Historical
data will be kept and will show the pre-alarm situation before
potential issues. The interface must provide users with access to
the key BMS data. Depending on the SOC of the battery, the remain-
ing range should be displayed on the dashboard. The consumers
must also be informed about the prediction and estimation of the
said battery through alarms and replacement recommendations.
According to Fig. 3, a BMS’s functionality may be categorized as
follows [29]:
1. Protection: against extreme temperatures, over-discharge,
overcurrent, overcharge, and short circuits.
2. High-voltage Control and Sensing: measure temperature, volt-
age, current, temperature, thermal management, control, con-
tactor, pre-charge, and ground-fault detection.
3. Diagnostics: state-of-life (SOL) estimation, SOH estimation, and
abuse detection.
4. Performance management: power-limit computation, balance/
equalize cells, and SOC estimation.
5. Interface: data recording, reporting, communications, and range
estimation.
There are several traditional ways, but AI-based strategies are
gaining popularity for improving dynamic responsiveness and
dependability. BMSs are currently installed with a fixed control
algorithm, without adjustments or modifications during the sys-
tem’s lifetime. [30]. However, the batteries have a relatively long
life span, with many warranties ranging from 5 to 10 years. The
operational environment may vary during the system’s lifespan
due to such a long period. These changes may need adjustments
to the control technique to improve performance characteristics.
The significance of SOH and SOC for the BMS extends beyond the
accuracy of its calculations [31]. A BMS is responsible for not only
communicating data about the batteries to the user, but also for
Fig. 3. Advanced Functioning Blocks of a BMS.
4
Ain Shams Engineering Journal xxx (xxxx) xxx
protecting the battery. The estimates of SOH and SOC, particularly
in Li-ion batteries, influence the instructions provided to the bat-
tery [32]. Fig. 4. Below shows the different AI techniques that can
be used for state estimation.
As a result, these parameters are in control of preventing over-
charge or over-discharge scenarios that might cause the batteries
to explode. The usage of neural networks and fuzzy logic is the
solution to these challenges [33]. This section investigates the
machine learning algorithms in BMS applications and compares
them. Moreover, a robust literature review of the most recent
research in the literature regarding state estimating is illustrated
in Table 2. Table 3.
3.1. Artificial neural network (ANN)
The Artificial Neural Network (ANN) is made up of nodes or
neurons, which are analogous to human brains. Fundamentally,
neural network (NN) employs a multiple-layer structure with
input, output, and hidden layers comprising neurons with system
requirements [34]. The NN is capable of self-adaptation and self-
learning. A backtracking search algorithm (BSA) and back propaga-
tion neural network (BPNN) can be utilized to increase NN perfor-
mance [35]. To train the network, a back-propagation technique
based on experimental data is utilized. Several inputs are taken
from previously recorded data and applied to the input layer, each
of which is multiplied by a weight before the final product is
added. The summation is then passed via a sigmoid function, com-
monly known as the activation function. The algorithm changes
the network weights to minimize the MSE in the network’s out-
come. The NN can be used to effectively estimate battery states.
The hybrid pulse power characterization (HPPC) test is used to
identify the battery parameters of the first-order equivalent circuit
model. Then, it can be used to train the multiple-layer back prop-
agation neural network, allowing the structure and parameters of
the network including weights and thresholds to be determined.
SOH is also estimated using this well-trained neural network as
well [36]. Choi et al. [37] implement a multilayer perceptron
(MLP) model that employs a number of nonlinear parameters.
For model design, 300Li-ion battery data are utilized, and the
model is validated by categorizing the data into training and test
data. The information on the capacity predicted by the MLP model
was utilized for SOC estimation, and the accuracy of the SOC esti-
mate was evaluated by comparing it to the SOC measurement
value.
Artificial Neural
Network
Hybrid Intelligent Neuro-Fuzzy
Systems Systems
Artificial
Intelligence
Techniques
Evolutionary
Expert Systems
Computing
Support Vector
Machine
Fig. 4. Different AI Techniques.
Y. Khawaja, N. Shankar, I. Qiqieh et al. Ain Shams Engineering Journal xxx (xxxx) xxx

Table 2
Extensive Review of the Diverse AI Technologies Employed in Battery Management Systems.

The functionality of the AI technique AI Outcomes from the Technique Employed
Technique
employed
Advanced battery management system HIS  Cheap non-recurring engineering costs.
ANN  Low power consumption and improved
FLS performance.
 High operation speed and huge data storage
capacity.
 Large reconfigurable logic.
Estimation Model for Lithium-Ion Batteries EA  Increased precision and efficiency.
ANN  Dynamic
 Cheap
 Global estimation performance that is
optimal
Electric vehicle Li-ion battery SOC Estimations ANN  Non-linearity is exploited.
 Close results are obtained.
System-based simulation environment for designing and HIS  Optimization of the BMS algorithms in
optimizing the BMS, which includes a lithium-ion various test conditions.
battery model and CAN communication interface.  System simulations run quickly.
Using ANFIS MATLAB for SOC Estimation AI  Easier to implement.
ANN  Accurate SOC value.
FLS
Enhancing the Performance of Electric Vehicle AI  Lowering emissions.
 The cost is negligible.
 Infrastructure recharging and safety.
State of charge estimation ANN  ANN can estimate the state of charge
accurately.
 ANN can self-learn battery dynamics.
SOC and SOH estimation HIS  Improved health.
ANN  Higher accuracy.
FLS  Non-linearity exploitation.
 Data globalization.
Analysis of the cost of utilizing battery electric vehicle HIS  Advantageous to battery life.
(BEV) batteries as energy storage in power grids  Most of the cost of V2G battery wear would
be attributable to DoD and cycle.
Real-Time Fault Diagnosis Capability for Electric Vehicles HIS  The users’ safety and enhanced lifespan of
the battery cells.
State of charge estimation ANN  Online real-time implementation.
 Accurate SOC values.
 Elimination of the relative SOC prediction
error and voltage measurement noise.
Surface Charge Estimation Using an Electrochemical HIS  The bulk and interface solid concentrations
Model-Based Extended Kalman Filter are estimated, allowing the critical surface
charge and SOC to be calculated.
 High efficacy.
Modeling and estimation of SOC of Li-Ion batteries using ANN  Rapid convergence and good accuracy.
neural networks and the extended Kalman filter (EKF)  Could be applied to a wide range of
rechargeable batteries.
Design of a Nonlinear Autoregressive with External Input ANN  Mean squared error of less than 1e-6
(NARX) artificial neural network to estimate the SOC of  The proposed model’s precision is very
an EV using real data extracted from the car significant.
 Employs sophisticated machine-learning
tools.
A probabilistic neural network (PNN) is used to estimate ANN  Simple to use.
the SOH of Li-ion batteries.  Extremely fast for medium-sized databases.
 This method can improve the estimation of
the battery’s SOC.
SOC estimation using Back propagation Neural Network HIS
with search-based optimization algorithm ANN  Less Complexity.
 Model-free strategy.
 More precise.
 Temperature sensitivity is included.
Challenges Ref.
 Parameters selection and data [29]
collection.
 It is difficult to track the life of a [31]
cycle.
 Can only be used on new cells
 It is dependent on the accuracy of
the current sensor and can
accumulate errors.
 Temperature variations affect [39]
training data.
 The cells in a battery pack do not [64]
have the same capacitance or
nominal voltage.
 For a similar battery, training data [65]
is required.
 Environment of packaging. [66]
 Choosing a Battery Technology.
 The initial cost of using battery [67]
systems.
 Expensive elements.
 A large number of cycles must be [68]
trained to estimate battery
availability and energy.
 V2G is reliant on driving [69]
efficiency.
 Cell aging and recycling. [70]
 One must compromise between [71]
accuracy and computing time to
suit each specific application.
 A relatively narrow range of [72]
charge states.
 The current and voltage of each [73]
cell must be measured.
 When the battery ages, the
trained NN may no longer produce
acceptable results.
 Restricted range The vehicle (and [74]
its traction battery)
is not in a controlled environment.
 When it comes to classifying new [75]
cases, multilayer perceptron
networks outperform them.
 PNNs necessitate more memory
space to store the model.
 Real-time implementation [76]
necessitates the use of a high-end
processor.
 The implementation of the coding
requires the use of skilled
engineers.

3.2. Neuro-Fuzzy logic systems (FLS)
Fuzzy logic controllers (FLSs) in general are insufficient for
estimating the SOC and SOH of a battery pack. Only one research
team used a fuzzy approach to determine the SOC and SOH of
the battery pack [38]. Previously, the majority of research on
adaptive battery control has relied solely on neural network
models or fuzzy logic. However, in recent years, research has
begun to tilt toward combining these two domains. The ability
to counteract the limits of one with the constraints of the other
has caused a change in study and application toward employing
neuro-fuzzy logic [39].

5
Y. Khawaja, N. Shankar, I. Qiqieh et al.
Table 3
Comparative Analysis of Regression of various AI techniques.
Model R2_score Mean
absolute error
Linear 0.9897 0.0715
Random 0.9999 0.0035
Forest
Gradient Boost 0.9994 0.0193
XGB 0.9994 0.0195
Light-GBM 0.9999 0.0048
SVM 0.9885 0.0599
The usage of neural networks and fuzzy logic reduces the
demand for experimental data. As an alternative to static solutions,
neural networks, and fuzzy logic provide dynamic solutions to the
majority of battery management system challenges [40]. The lack
of a data set to train the system regarding the operating circum-
stances at the time of system deployment limits the use of deep
neural networks in BMSs [41]. This data, on the other hand, maybe
acquired continually and in real-time. The neuro-fuzzy approach’s
goal is to use a learning mechanism with input–output training
data to alter the fuzzy system’s parameters. Sugeno fuzzy rules
are used in the framework of ANFIS. It has five layers. Layer 1 is
made up of input variables. Layer 2 is a membership layer that reg-
ulates the membership function’s weights. This layer accepts input
values from the first layer and specifies membership values to rep-
resent fuzzy collections of related input variables. Layer 3 is known
as the rule layer, and it contains the items from the previous layer.
This layer determines the amount of activation for each rule. The
fourth layer is the defuzing layer. The output values obtained from
rule extraction are provided by the definition layer. Layer 5 is the
output layer, which gathers all of the preceding layer’s inputs
and turns the fuzzy classification results into a clear value [42].
The efficiency of a neuro-fuzzy system was projected to outper-
form that of merely neural networks [43]. The performances of
these two systems, however, cannot be compared since they do
not provide quantifiable estimates of their accuracy. The research
on neuro-fuzzy systems is also insufficiently thorough. Because
of the scarcity of previous implementations, future research will
be time-consuming [44].
3.3. Expert systems (ES)
The expert systems (ESs) are made up of a feature data repository,
a knowledge base, an explanation machine, a reasoning machine, a
knowledge acquisition mechanism, and a human–machine interac-
tion interface [45]. The fault symptom information after feature
extraction is necessary to construct a feature information database
to acquire the unified form of fault symptom information [46]. Most
energy management systems, for example, require the existence of
logic that executes a state estimator and security analysis on a regu-
lar clock interval, with an override if switching actions occur within a
few seconds of the start. Furthermore, the operator can request the
execution of the state estimator and security analysis at any time.
Actions can be made based on the sort of switching action that has
occurred with an expert system directing the state estimator and
security analysis. What the difference in load level is from the last
run, and so on. For the operators to be able to use vast volumes of
analog data, significant numeric-intensive algorithms must be used
to analyze it. Analog measurement data collected from a transmis-
sion system is far more helpful to an operator when it has been pro-
cessed by a state estimator than when it has been supplied to the
operator as raw data. The state estimator can identify and remove
6
Ain Shams Engineering Journal xxx (xxxx) xxx
Median MSE RMSE
absolute error
0.0371 0.0203 0.142
0.0013 0.0001 0.0097
0.0072 0.0025 0.04955
0.0076 0.0026 0.05074
0.0023 0.0001 0.0111
0.0323 0.0127 0.1126
false data while also offering more accurate estimates for all trans-
mission network quantities.
3.4. Support vector Machine (SVM)
Support vector machine (SVM)-based SOC estimate techniques
were introduced in [47–52]. For instance, in [47], the SOC of the
battery was calculated as a function of voltage and current. Tem-
perature, voltage, current, and measurements, on the other hand,
were utilized as inputs in [48] and [49] to train the SVM model
and estimate the SOC. This theory has been demonstrated to be
superior to the common theory of empirical risk reduction used
by conventional neural networks. A quadratic programming prob-
lem, which has the global optimal solution, is solved by the learn-
ing method of SVMs, although neural networks may have local
minimization issues, SVM may generalize better as a result of this
characteristic. It is not immediately clear, yet, that SVMs are sup-
posed to completely replace ANN. For instance, multilayer feedfor-
ward networks (MLPs), compete directly with SVMs.
A two-step approach of SVM training and virtual testing is
applied to battery data obtained with a typical EV current profile
to calculate the SOH figures of merit, capacity, and resistance
[53]. An SVM is trained using current, temperature, and SOC as
inputs and voltage as output in the first step. The resulting mode-
l acts as a voltage look-up table for hypothetical current/tempera-
ture SOC input based on the actual current/temperature SOC
profile from the relevant real standard test [54]. As in genuine
standard performance testing, the virtual test result may then be
utilized to calculate resistance and capacity [55].
3.5. Evolutionary algorithm (EA)
To address the growing concerns in BMS, Evolutionary Compu-
tation (EA) is recommended. To magnify on a nanoscale and
develop a model for lithium-ion battery capacity, multigene
genetic programming (GP) is presented [56]. The second case study
uses a Non-dominated Sorting Genetic Algorithm II (NSGAII) to
optimize battery enclosure design at the pack level [57]. The vali-
dation results demonstrate NSGAII’s remarkable optimization
function as well as the enclosure’s increased performance. Further-
more, when paired with other artificial intelligence systems, the
results demonstrate that NSGAII performs better [58]. The findings
imply that EC may be implemented into an EV system to monitor
its performance and ensure its safety. The motivating factors for
using EA are as follows: i) EA is a traditional derivative or
gradient-based optimization method, ii) EA is very good at solving
non-convex optimization problems with multiple local optima and
finally, iii) EA is very flexible in addressing multi-objective opti-
mization problems. A typical EA is an iterative and population-
based algorithm [59] that begins its search for the best answer
with a random population. The original population is then sub-
Y. Khawaja, N. Shankar, I. Qiqieh et al.
jected to an iterative process including numerous operations such
as reproduction, fitness evaluation, and selection until particular
stopping criteria are fulfilled.
3.6. Hybrid intelligent systems (HIS)
For plug-in hybrid electric and electric cars, an intelligent pre-
dictive battery management system is presented as an ideal energy
and time management solution [60]. This research proposes a sys-
tem to aid drivers in choosing an optimal route and driving profile
to save travel time and energy consumption. It investigated and
proved the benefits of the predictive intelligent battery manage-
ment system for improving battery energy usage and journey
duration using both analysis and simulation [61]. Because of the
possibility of nonlinear response, this hybrid model is based on
clustering; where the modeling dataset is separated into groups
with comparable features. Each cluster received a unique regres-
sion model, to improve overall model performance. The number
of clusters is determined by evaluating various topologies. This pic-
ture depicts a generic hybrid model with n clusters and n models,
each with its own set of parameters. To develop a hybrid model,
the modeling process might be separated into the following steps:
clustering phase, regression modeling phase for each cluster, selec-
tion of the best local model (per cluster), and selection of the best
topology for the hybrid model.
To establish groups with comparable traits, the k-means
method was utilized during the clustering step. The ANN method
was chosen for the regression modeling phase because of its ability
to forecast the output of nonlinear systems with a simple internal
setup. Even though this is a hybrid system, the model performs
better when the regression models are intelligent systems rather
than standard regression methods.
4. Future prospects and inference
Lithium-ion batteries are one of the major achievements in
electrochemistry, and their uses as ESS have grown dramatically
[62]. With no major competition, Li-ion batteries have established
themselves in the consumer electronics and electric automobile
sectors. They will soon dominate the automotive transportation
and renewable energy storage industry by utilizing clever adjust-
ments in the various functional blocks of the BMS [63].
One can tap into the latent potential and opportunity of battery
life cycle management by using Machine Learning (ML) and Neural
Networks. The functioning of the battery packs used in EVs can be
optimized by combining modern electronics with testing and
training data. Predictive consciousness derived from experimental
laboratory data can be utilized to forecast battery life, diagnose
problems and their causes, and correct errors before they occur.
5. Results and analysis
This section describes the results obtained by using six machine
learning algorithms for the regression analysis such as, linear
regressor, random forest regressor, gradient boost regressor, light
gradient boosting machine (light-GBM) regressor, extreme gradi-
ent boosting (XGB) regressor, and SVM regressor.
The assumptions made for the Li-ion battery model are as
follows:
1. Operating temperature of the battery is 23 C since the battery
is assumed to be maintained at room temperature.
2. Battery starts degrading during charging cycles.
3. Initial SOC is 20%, and zero SOC is neglected in the dataset.
4. Initial SOH is 1x10-84
7
Ain Shams Engineering Journal xxx (xxxx) xxx
The dataset for this particular work is acquired from GitHub
which is an open-source online library where users may work on
or adopt open-source code projects, fork code, discuss ideas, and
even improvise existing code. Several datasets were obtained and
based on the parameter required for the research, the required
dataset was filtered out. The accuracy of the dataset was estimated
by running it through linear regression and random forest.
Among those models, the experiments in this section are con-
ducted for both the linear and the random forest regression mod-
els. However, a comparative analysis to evaluate the capability of
the six AI models on SOC and SOH predictions is introduced in
the following section.
The formula for the linear regressor is given as,
Ci = f (Bi, Y) + E (4).
The Ci is the dependent variable which is the output and f (Bi) is
the independent variable. Y is the unknown parameters and E is
the error rate. Thus, the Ci is estimated from the independent input
classes with an error rate of E.
The random forest regression analysis takes the average of the
sub-sample of the decision tree and uses this averaging to improve
the accuracy by reducing the over-fitting of the results. The random
forest regression model controls the maximum sample parameter
when the bootstrap is true and then each model will produce a data-
set. Since we use both models, we can interrelate the regression
model with the linear aggression model. The random forest model
provides better accuracy when compared with the linear regression
model. It is used in the proposed system to validate the regression of
the linear model when we map the exponential response of both
these models given in Fig. 5 and Fig. 6. The output of the models
when captured as a CSV file provides similar responses of both linear
and random forest models in the order of the exponents.
From the dataset further analysis was done on the charging and
discharging cycles as observed in Fig. 7 where the positive half of
the cycle depicts charging while the negative half of the cycle rep-
resents discharging of the Li-ion battery in the EV. Fig. 7 is used as
the instruction signal to estimate the SOC and SOH.
Fig. 8 illustrates the SOC of the battery, while Fig. 9 shows the
SEI thickness. The capacity retention of a battery is a measurement
of its capacity to hold onto stored energy during a prolonged open-
circuit rest time. The length of the rest interval, the temperature of
the cell during that time, and the cell’s past performance all affect
retained capacity. The layout of the cell also has an impact on
capacity retention. Several electrode formulation techniques are
used to create nickel-.
Fig. 5. Linear Regression Model.
Y. Khawaja, N. Shankar, I. Qiqieh et al.
Fig. 6. Random Forest Model.
Fig. 7. Charging and Discharging Cycles.
Fig. 8. SOC of the battery.
cadmium cells. Due to the differences between each method,
the capacity retention properties of the electrodes produced
by each process may vary. Capacity retention is also influenced
by other cell design variables such as electrode spacing, elec-
trode surface area, and separator material. Here since the bat-
tery is maintained at 23 C we notice the retentivity is not
affected drastically while the vehicle is not being charged.
From Fig. 9 it can be observed that the solid electrolyte inter-
phase (SEI) layer thickness goes on increasing with time. On elec-
trode surfaces, the breakdown products of electrolytes create a
8
Ain Shams Engineering Journal xxx (xxxx) xxx
Fig. 9. SEI of the battery.
passivation layer known as the solid electrolyte interphase (SEI).
To ensure that electrochemical reactions continue and that the
electrolyte doesn’t degrade further, the SEI permits Li + flow while
blocking electrons [77]. The inherent consumption of the anode
and electrolyte during the development of the SEI results in poor
efficiency and the SEI effectively stops Li from coming into contact
with the solvent again, rendering Li dynamically stable in some
organic solvents. The SEI can change how Li-ions are distributed
between the anode and the bulk electrolyte. We propose a novel
method for SOH assessment based on the physical interpretation
of the model parameters that combines parameters of the impe-
dance spectrum by constructing the ratio of the SEI resistance to
the sum of the SEI resistance and the charge transfer resistance.
This ratio, which describes the charge-transfer efficiency at the
electrode surfaces, should consistently decline with SOH. Apart
from the SEI thickness other factors such as overcharging and cycle
count affect the SOH of the battery.
6. Comparative analysis
The regression analysis is performed using various machine
learning algorithms as described below. This work used linear
regressor, random forest regressor, gradient boost regressor,
light-gradient boost regressor, and SVM regressor. Since the data
is numeric, it is more suitable to perform a regression analysis to
determine the correlation and variance among the features. Vari-
ous features that are taken into consideration are SOC, SOH, Time,
SEI_after, and SEI_delta.
The SEI-delta is the dependent variable while the other vari-
ables are independent. The target variable SEI-delta is used to
determine the following parameters, such as R2_score, mean_abso-
lute_error, median_absolute_error, mean squared error (MSE), and
the root mean squared error (RMSE). The R2_score metric is used
to measure how much variance is explained in predictions by a
dataset [78]. The measured parameters and the specific values
measured are tabulated in Table III.
Based on the performance comparisons, the random forest
model performs much better with the highest R2_score and lowest
of MSE, RMSE, mean, and median absolute errors. In comparison
with the random forest, the light-gradient boost regressor per-
forms better than the rest of the methods with the next decent
score and error performance. The response of all these models is
plotted in Fig. 10.
This comparison shows all of the models provided a decent per-
formance of R2_score with negligible error rates. Among all of
them, the random forest regressor provides the highest R2_score
with the lowest possible error rates and emerged as the best of
all predictions.
Y. Khawaja, N. Shankar, I. Qiqieh et al.
Fig. 10. Comparative Analysis of Various Machine Learning Regression models for SoH and SOC Analysis.
7. Conclusion
The battery management system along with its two main
parameters, the state of charge and state of health, has been com-
prehensively discussed in this proposed work. The primary goal of
this work is to identify the best AI technology for battery manage-
ment systems. Therefore an extensive literature review of the tar-
geted AI technologies is conducted, highlighting the challenges
encountered. This proposed work focused on Li-ion batteries as a
preferred energy storage solution for electric vehicles due to their
longer lifespan, reduced discharge rates, excellent efficiency, and
power and energy densities. The key parameters and their roles
in a battery management system’s safe operation and operating
cycle to ensure safe and reliable battery operation in electric vehi-
cles have also been described.
Six well-known AI technologies are applied for the battery man-
agement system state of charge and state of health estimation.
Detailed results are presented for the linear regression model
and random forest, showing that the random forest model outper-
forms the linear regression by obtaining more accurate dataset.
Additionally, a comparative analysis is conducted between all the
algorithms including linear, random forest, gradient boost, light-
gradient boosting machine, and support vector machine regressors.
Similarly, random forest has demonstrated much superior perfor-
mance for the discharge prediction compared to all other models
used in this work in terms of higher R2_score (0.9999), lower mean
absolute error (0.0035), median absolute error (0.0013), means
squared error ((0.0001), and root means squared error (0.0097).
Accordingly, i it is concluded from the results achieved that the
Li-ion battery state can be accurately estimated using random for-
est regression. With the increasing size of the dataset, a similar
application with more enhanced features and samples may require
higher computing power and complexity, thus requiring the pres-
ence of a deep neural network or federated learning, which could
be addressed in future implementations.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared
to influence the work reported in this paper.
Acknowledgement
The authors would like to thank Applied Science Private Univer-
sity (Jordan) and Al-Balqa Applied University (Jordan) for their
funding and support.
9
Ain Shams Engineering Journal xxx (xxxx) xxx
References
[1] Abdallah L, El-Shennawy T. Reducing Carbon Dioxide Emissions from
Electricity Sector Using Smart Electric Grid Applications. J Eng
2013;2013:1–8. doi: https://doi.org/10.1155/2013/845051.
[2] Sharma P. Rani Chinnappa Naidu, Optimization techniques for grid-connected
PV with retired EV batteries in centralized charging station with challenges
and future possibilities: A review. Ain Shams Eng J 2023;14(7):. doi: https://
doi.org/10.1016/j.asej.2022.101985101985.
[3] Khawaja Y, Qiqieh I, Alzubi J, Alzubi O, Allahham A, Giaouris D. Design of cost-
based sizing and energy management framework for standalone microgrid
using reinforcement learning. Sol Energy 2023;251:249–60. doi: https://doi.
org/10.1016/j.solener.2023.01.027.
[4] Aldbaiat B, Nour M, Radwan E, Awada E. Grid-Connected PV System with
Reactive Po wer Management and an Optimized SRF-PLL Using Genetic
Algorithm. Energies 2022;15:2177. doi: https://doi.org/10.3390/en15062177.
[5] Khawaja Y, Allahham A, Giaouris D, Patsios C, Walker S, Qiqieh I. An Integrated
Framework for Sizing and Energy Management of Hybrid Energy System Using
Automata. Appl Energy 2019;250:257–72. doi: https://doi.org/10.1016/j.
apenergy.2019.04.185.
[6] Y. Khawaja, D. Giaouris, H. Patsios and M. Dahidah, ‘‘Optimal cost-based model
for sizing grid-connected PV and battery energy system,” 2017 IEEE Jordan
Conference on Applied Electrical Engineering and Computing Technologies
(AEECT), Aqaba, Jordan, 2017, pp. 1-6, DOI: 10.1109/AEECT.2017.8257779.
[7] Osama Ayadi, Reem Shadid, Abdullah Bani-Abdullah, Mohammad Alrbai,
Mohammad Abu-Mualla, NoorAldeen Balah, Experimental comparison
between Monocrystalline, Polycrystalline, and Thin-film solar systems under
sunny climatic conditions, Energy Reports, Volume 8, Supplement 9, 2022,
Pages 218-230, ISSN 2352-4847, https://doi.org/10.1016/j.egyr.2022.06.121.
[8] M. Bani Khalid, A. Qandil, N. Beithou, H.Sß . Aybar, Renewable hydrogen driven
CHCP device, International Journal of Hydrogen Energy, Volume 47, Issue 4,
2022, Pages 2208-2219, https://doi.org/10.1016/j.ijhydene.2021.10.148.
[9] Ravi SS, Aziz M. Utilization of Electric Vehicles for Vehicle-to-Grid Services:
Progress and Perspectives. Energies (Basel) 2022;15:589. doi: https://doi.org/
10.3390/en15020589.
[10] Abdi H, Mohammadi-ivatloo B, Javadi S, Khodaei AR, Dehnavi E. Energy Storage
Systems. Distributed Generation Systems, Elsevier; 2017, p. 333–68.
https://doi.org/10.1016/B978-0-12-804208-3.00007-8.
[11] Hannan MA, Hoque MdM, Hussain A, Yusof Y, Ker PJ. State-of-the-Art and
Energy Management System of Lithium-Ion Batteries in Electric Vehicle
Applications: Issues and Recommendations. IEEE Access 2018;6:19362–78.
https://doi.org/10.1109/ACCESS.2018.2817655.
[12] Changki Choi, Seongyun Park, Jonghoon Kim, Uniqueness of multilayer
perceptron-based capacity prediction for contributing state-of-charge
estimation in a lithium primary battery, Ain Shams Engineering Journal,
Volume 14, Issue 4, 2023, 101936, https://doi.org/10.1016/j.asej.2022.101936.
[13] Hesse H, Schimpe M, Kucevic D, Jossen A. Lithium-Ion Battery Storage for the
Grid—A Review of Stationary Battery Storage System Design Tailored for
Applications in Modern Power Grids. Energies (Basel) 2017;10:2107. doi:
https://doi.org/10.3390/en10122107.
[15] Kumar B, Khare N, Chaturvedi PK. Advanced battery management system
using MATLAB, Simulink IEEE International Telecommunications Energy
Conference (INTELEC). IEEE 2015;2015:1–6. doi: https://doi.org/10.1109/
INTLEC.2015.7572447.
[16] Chang W-Y. The State of Charge Estimating Methods for Battery: A Review.
ISRN Applied Mathematics 2013;2013:1–7. doi: https://doi.org/10.1155/2013/
953792.
[17] Arora S, Abkenar AT, Jayasinghe SG, Tammi K. Battery Management System:
Charge Balancing and Temperature Control. Heavy-Duty Electric Vehicles,
Elsevier; 2021, p. 173–203. https://doi.org/10.1016/B978-0-12-818126-
3.00005-1.
Y. Khawaja, N. Shankar, I. Qiqieh et al.
[18] Shankar N, Chitra A, Banerjee D, Sharma V, Zhutshi K, Razia SW, et al.
Innovations in Power and Advanced Computing Technologies (i-PACT). IEEE
2021;2021:1–7. doi: https://doi.org/10.1109/i-PACT52855.2021.9697046.
[19] Gabbar H, Othman A, Abdussami M. Review of Battery Management Systems
(BMS) Development and Industrial Standards. Technologies (Basel) 2021;9:28.
doi: https://doi.org/10.3390/technologies9020028.
[20] Hossain Lipu MS, Miah MdS, Ansari S, Wali SB, Jamal T, Elavarasan RM, et al.
Smart Battery Management Technology in Electric Vehicle Applications:
Analytical and Technical Assessment toward Emerging Future Directions.
Batteries 2022;8:219. https://doi.org/10.3390/batteries8110219.
[21] Cabrera J, Vega A, Tobajas F, Deniz V, Fabelo HA. Design of a reconfigurable Li-
Ion Battery Management System (BMS). XI Tecnologias Aplicadas a la
Ensenanza de la Electronica (Technologies Applied to Electronics Teaching)
(TAEE). IEEE 2014;2014:1–6. doi: https://doi.org/10.1109/TAEE.2014.6900162.
[22] Feng Y, Cao Z, Shen W, Yu X, Han F, Chen R, et al. Intelligent battery
management for electric and hybrid electric vehicles: A survey. 2016 IEEE
International Conference on Industrial Technology (ICIT), IEEE; 2016, p. 1436–
41. https://doi.org/10.1109/ICIT.2016.7474969.
[23] Lelie M, Braun T, Knips M, Nordmann H, Ringbeck F, Zappen H, et al. Battery
Management System Hardware Concepts: An Overview. Appl Sci 2018;8:534.
doi: https://doi.org/10.3390/app8040534.
[24] Uzair M, Abbas G, Hosain S. Characteristics of Battery Management Systems of
Electric Vehicles with Consideration of the Active and Passive Cell Balancing
Process. World Electric Vehicle Journal 2021;12:120. doi: https://doi.org/
10.3390/wevj12030120.
[25] Balasingam B, Ahmed M, Pattipati K. Battery Management Systems—
Challenges and Some Solutions. Energies (Basel) 2020;13:2825. doi: https://
doi.org/10.3390/en13112825.
[26] Meissner E, Richter G. Battery Monitoring and Electrical Energy Management. J
Power Sources 2003;116:79–98. doi: https://doi.org/10.1016/S0378-7753(02)
00713-9.
[27] Gold S. A PSPICE macromodel for lithium-ion batteries. The Twelfth Annual
Battery Conference on Applications and Advances, IEEE; n.d., p. 215–22.
https://doi.org/10.1109/BCAA.1997.574106.
[28] Xing Y, Ma EWM, Tsui KL, Pecht M. Battery Management Systems in Electric
and Hybrid Vehicles. Energies (Basel) 2011;4:1840–57. doi: https://doi.org/
10.3390/en4111840.
[29] Yin H, Zhou W, Li M, Ma C, Zhao C. An Adaptive Fuzzy Logic-Based Energy
Management Strategy on Battery/Ultracapacitor Hybrid Electric Vehicles. IEEE
Trans Transp Electrif 2016;2:300–11. doi: https://doi.org/10.1109/
TTE.2016.2552721.
_
[30] Ozadowicz A, Grela J. Energy saving in the street lighting control system—a
new approach based on the EN-15232 standard. Energy Effic 2017;10:563–76.
doi: https://doi.org/10.1007/s12053-016-9476-1.
[31] Huang S-C, Tseng K-H, Liang J-W, Chang C-L, Pecht M. An Online SOC and SOH
Estimation Model for Lithium-Ion Batteries. Energies (Basel) 2017;10:512. doi:
https://doi.org/10.3390/en10040512.
[32] Venugopal P, T. V. State-of-Health Estimation of Li-ion Batteries in Electric
Vehicle Using IndRNN under Variable Load Condition. Energies (Basel)
2019;12:4338. https://doi.org/10.3390/en12224338.
[33] Obert J, Trevizan RD, Torres-Castro L, Preger Y. Ensemble Learning, Prediction
and Li-Ion Cell Charging Cycle Divergence. IEEE Open Access Journal of Power
and Energy 2021;8:303–15. doi: https://doi.org/10.1109/OAJPE.2021.3100004.
[34] Alzubaidi L, Zhang J, Humaidi AJ, Al-Dujaili A, Duan Y, Al-Shamma O, et al.
Review of deep learning: concepts, CNN architectures, challenges,
applications, future directions. J Big Data 2021;8:53. doi: https://doi.org/
10.1186/s40537-021-00444-8.
[35] Scrosati B, Garche J. Lithium batteries: Status, prospects and future. J Power
Sources 2010;195:2419–30. doi: https://doi.org/10.1016/j.
jpowsour.2009.11.048.
[36] Yang D, Wang Y, Pan R, Chen R, Chen Z. A Neural Network Based State-of-
Health Estimation of Lithium-ion Battery in Electric Vehicles. Energy Procedia
2017;105:2059–64. doi: https://doi.org/10.1016/j.egypro.2017.03.583.
[37] Choi C, Park S, Kim J. Uniqueness of multilayer perceptron-based capacity
prediction for contributing state-of-charge estimation in a lithium primary
battery. Ain Shams Eng J 2023;14:. doi: https://doi.org/10.1016/j.
asej.2022.101936101936.
[38] Chitnis MS, Pandit SP, Shaikh MN. Electric Vehicle Li-Ion Battery State of
Charge Estimation Using Artificial Neural Network. In: 2018 International
Conference on Inventive Research in Computing Applications (ICIRCA), IEEE. p.
992–5. doi: https://doi.org/10.1109/ICIRCA.2018.8597234.
[39] Dai H, Guo P, Wei X, Sun Z, Wang J. ANFIS (adaptive neuro-fuzzy inference
system) based online SOC (State of Charge) correction considering cell
divergence for the EV (electric vehicle) traction batteries. Energy
2015;80:350–60. doi: https://doi.org/10.1016/j.energy.2014.11.077.
[40] Martinez DA, Poveda JD, Montenegro D. Li-Ion battery management system
based in fuzzy logic for improving electric vehicle autonomy. IEEE Workshop
on Power Electronics and Power Quality Applications (PEPQA). IEEE
2017;2017:1–6. doi: https://doi.org/10.1109/PEPQA.2017.7981677.
[41] Chen C, Zhang P, Zhang H, Dai J, Yi Y, Zhang H, et al. Deep Learning on
Computational-Resource-Limited Platforms: A Survey. Mob Inf Syst
2020;2020:1–19. doi: https://doi.org/10.1155/2020/8454327.
[42] Abdalla SA, Abdullah SS, Kassem AM. Performance enhancement and power
management strategy of an autonomous hybrid fuel cell/wind power system
based on adaptive neuro fuzzy inference system. Ain Shams Eng J 2022;13:.
doi: https://doi.org/10.1016/j.asej.2021.101655101655.
10
Ain Shams Engineering Journal xxx (xxxx) xxx
[43] Talpur N, Abdulkadir SJ, Alhussian H, Hasan MH, Aziz N, Bamhdi A. Deep
Neuro-Fuzzy System application trends, challenges, and future perspectives: a
systematic survey. Artif Intell Rev 2023;56:865–913. doi: https://doi.org/
10.1007/s10462-022-10188-3.
[44] Shihabudheen KV, Pillai GN. Recent advances in neuro-fuzzy system: A survey.
Knowl Based Syst 2018;152:136–62. doi: https://doi.org/10.1016/
j.knosys.2018.04.014.
[45] Choi SY, Kim SH. Knowledge Acquisition and Representation for High-
Performance Building Design: A Review for Defining Requirements for
Developing a Design Expert System. Sustainability 2021;13:4640. doi:
https://doi.org/10.3390/su13094640.
[46] Qiu W, Qiu Z. Design for symmetrical management of storage battery expert
system based on single battery. In: 2006 International Conference on
Mechatronics and Automation. p. 1141–6. doi: https://doi.org/10.1109/
ICMA.2006.257786.
[47] Feng Y, Xue C, Han Q-L, Han F, Du J. Robust Estimation for State-of-Charge and
State-of-Health of Lithium-Ion Batteries Using Integral-Type Terminal Sliding-
Mode Observers. IEEE Trans Ind Electron 2020;67:4013–23. doi: https://doi.
org/10.1109/TIE.2019.2916389.
[48] Aslan E, Yasa YA. In: Review on The Battery State of Charge Estimation
Methods For Electric Vehicle Battery Management Systems. p. 281–5.
[49] Liu K, Li Y, Hu X, Lucu M, Widanage WD. Gaussian Process Regression With
Automatic Relevance Determination Kernel for Calendar Aging Prediction of
Lithium-Ion Batteries. IEEE Trans Industr Inform 2020;16:3767–77. doi:
https://doi.org/10.1109/TII.2019.2941747.
[50] Chen Y, Gao C, Yang W, Wang Z, Yu Z, Xia H. A Recurrent Gaussian Process
Regression Model with Composite Kernel for Industrial Process Quality Prediction.
In: 2019 International Conference on Networking and Network Applications
(NaNA), IEEE. p. 229–34. doi: https://doi.org/10.1109/NaNA.2019.00048.
[51] Zhao J, Guo H, Wang L, Han M. Computer Modeling of the Eddy Current Losses of
Metal Fasteners in Rotor Slots of a Large Nuclear Steam Turbine Generator Based
on Finite-Element Method and Deep Gaussian Process Regression. IEEE Trans
Ind Electron 2020;67:5349–59. doi: https://doi.org/10.1109/TIE.2019.2931487.
[52] Thenaisie G, Park C-H, Lee S-G. A Real-Time Entropy Estimation Algorithm for
Lithium Batteries Based on a Combination of Kalman Filter and Nonlinear
Observer. IEEE Trans Ind Electron 2020;67:8034–43. doi: https://doi.org/
10.1109/TIE.2019.2945283.
[53] Junping W, Quanshi C, Binggang C. Support vector machine based battery
model for electric vehicles. Energy Convers Manag 2006;47:858–64. doi:
https://doi.org/10.1016/j.enconman.2005.06.013.
[54] Klass V, Behm M, Lindbergh G. Capturing lithium-ion battery dynamics with
support vector machine-based battery model. J Power Sources
2015;298:92–101. doi: https://doi.org/10.1016/j.jpowsour.2015.08.036.
[55] Klass V, Behm M, Lindbergh G. A support vector machine-based state-of-
health estimation method for lithium-ion batteries under electric vehicle
operation. J Power Sources 2014;270:262–72. doi: https://doi.org/10.1016/j.
jpowsour.2014.07.116.
[56] Garg A, Yun L, Shaosen S, Goyal A, Niu X, Gao L, et al. A combined
experimental-numerical framework for residual energy determination in
spent lithium-ion battery packs. Int J Energy Res 2019;43:4390–402. doi:
https://doi.org/10.1002/er.4564.
[57] Cui X, Panda B, Chin CMM, Sakundarini N, Wang C, Pareek K. An application of
evolutionary computation algorithm in multidisciplinary design optimization
of battery packs for electric vehicle. Energy Storage 2020;2. doi: https://doi.
org/10.1002/est2.158.
[58] Morais H, Sousa T, Castro R, Vale Z. Multi-Objective Electric Vehicles
Scheduling Using Elitist Non-Dominated Sorting Genetic Algorithm. Appl Sci
2020;10:7978. doi: https://doi.org/10.3390/app10227978.
[59] Ali AF, Tawhid MA. A hybrid particle swarm optimization and genetic algorithm
with population partitioning for large scale optimization problems. Ain Shams
Eng J 2017;8:191–206. doi: https://doi.org/10.1016/j.asej.2016.07.008.
[60] Rajalakshmi M, Razia Sultana W. Intelligent Hybrid Battery Management
System for Electric Vehicle. Artificial Intelligent Techniques for Electric and
Hybrid Electric Vehicles, Wiley; 2020, p. 179–206. https://doi.org/10.1002/
9781119682035.ch10.
[61] Abdul-Hak M, Al-Holou NITS. based Predictive Intelligent Battery Management
System for plug-in Hybrid and Electric vehicles. IEEE Vehicle Power and
Propulsion Conference. IEEE 2009;2009:138–44. doi: https://doi.org/10.1109/
VPPC.2009.5289858.
[62] el Kharbachi A, Zavorotynska O, Latroche M, Cuevas F, Yartys V, Fichtner M.
Exploits, advances and challenges benefiting beyond Li-ion battery
technologies. J Alloys Compd 2020;817:. doi: https://doi.org/10.1016/
j.jallcom.2019.153261153261.
[63] Das CK, Bass O, Kothapalli G, Mahmoud TS, Habibi D. Overview of energy
storage systems in distribution networks: Placement, sizing, operation, and
power quality. Renew Sustain Energy Rev 2018;91:1205–30. doi: https://doi.
org/10.1016/j.rser.2018.03.068.
[64] Buccolini L, Ricci A, Scavongelli C, DeMaso-Gentile G, Orcioni S, Conti M.
Battery Management System (BMS) simulation environment for electric
vehicles. 2016 IEEE 16th International Conference on Environment and
Electrical Engineering (EEEIC), IEEE; 2016, p. 1–6. https://doi.org/10.1109/
EEEIC.2016.7555475.
[65] Manish Mawatwal, Aayush Mohanty, Anitha G. S. State of Charge Estimation
for Rechargeable Lithium-ion Battery using ANFIS MATLAB. International
Journal of Engineering Research And 2020;V9. https://doi.org/10.17577/
IJERTV9IS030565.
Y. Khawaja, N. Shankar, I. Qiqieh et al.
[66] Abdul-Hak M, Al-Holou N, Mohamm U. Predictive Intelligent Battery
Management System to Enhance the Performance of Electric Vehicle. Electric
Vehicles - Modelling and Simulations, InTech; 2011. https://doi.org/10.5772/
17099.
[67] Ismail M, Dlyma R, Elrakaybi A, Ahmed R, Habibi S. Battery state of charge
estimation using an Artificial Neural Network. 2017 IEEE Transportation
Electrification Conference and Expo (ITEC), IEEE; 2017, p. 342–9. https://doi.
org/10.1109/ITEC.2017.7993295.
[68] Watrin N, Blunier B, Miraoui A. Review of adaptive systems for lithium
batteries State-of-Charge and State-of-Health estimation. 2012 IEEE
Transportation Electrification Conference and Expo (ITEC), IEEE; 2012, p. 1–
6. https://doi.org/10.1109/ITEC.2012.6243437.
[69] Zhou C, Qian K, Allan M, Zhou W. Modeling of the Cost of EV Battery Wear Due
to V2G Application in Power Systems. IEEE Trans Energy Convers
2011;26:1041–50. doi: https://doi.org/10.1109/TEC.2011.2159977.
[70] Gao Z, Chin C, Chiew J, Jia J, Zhang C. Design and Implementation of a Smart
Lithium-Ion Battery System with Real-Time Fault Diagnosis Capability for
Electric Vehicles. Energies (Basel) 2017;10:1503. doi: https://doi.org/10.3390/
en10101503.
[71] Eddahech A, Briat O, Vinassa JM. Adaptive voltage estimation for EV Li-ion cell
based on artificial neural networks state-of-charge meter. 2012 IEEE
International Symposium on Industrial Electronics, IEEE; 2012, p. 1318–24.
https://doi.org/10.1109/ISIE.2012.6237281.
[72] di Domenico D, Stefanopoulou A, Fiengo G. Lithium-Ion Battery State of Charge
and Critical Surface Charge Estimation Using an Electrochemical Model-Based
Extended Kalman Filter. J Dyn Syst Meas Control 2010;132. https://doi.org/
10.1115/1.4002475.
[73] Charkhgard M, Farrokhi M. State-of-Charge Estimation for Lithium-Ion
Batteries Using Neural Networks and EKF. IEEE Trans Ind Electron
2010;57:4178–87. doi: https://doi.org/10.1109/TIE.2010.2043035.
[74] Jiménez-Bermejo D, Fraile-Ardanuy J, Castaño-Solis S, Merino J, Álvaro-
Hermana R. Using Dynamic Neural Networks for Battery State of Charge
Estimation in Electric Vehicles. Procedia Comput Sci 2018;130:533–40. doi:
https://doi.org/10.1016/j.procs.2018.04.077.
[75] Lin H-T, Liang T-J, Chen S-M. Estimation of Battery State of Health Using
Probabilistic Neural Network. IEEE Trans Industr Inform 2013;9:679–85. doi:
https://doi.org/10.1109/TII.2012.2222650.
[76] Hannan MA, Lipu MSH, Hussain A, Saad MH, Ayob A. Neural Network
Approach for Estimating State of Charge of Lithium-Ion Battery Using
Backtracking Search Algorithm. IEEE Access 2018;6:10069–79. doi: https://
doi.org/10.1109/ACCESS.2018.2797976.
[77] Wang, A., Kadam, S., Li, H. et al. Review on modeling of the anode solid
electrolyte interphase (SEI) for lithium-ion batteries. npj Comput Mater 4, 15
(2018). https://doi.org/10.1038/s41524-018-0064-0.
[78] Abdelbasset WK, Elkholi SM, Ismail KA, Alshehri S, Alobaida A, Huwaimel B,
et al. Development a novel robust method to enhance the solubility of
Oxaprozin as nonsteroidal anti-inflammatory drug based on machine-
learning. Sci Rep 2022;12:13138. doi: https://doi.org/10.1038/s41598-022-
17440-4.
Yara Khawaja is currently an assistant professor at
Applied Science Private University, Jordan. She received
her Ph.D. degree in Renewable Energy/Electrical Engi-
neering from the School of Engineering, Newcastle
University, Newcastle upon Tyne – UK in 2019, and her
M.Sc. (with distinction) in Computer Science from
Prince Abdullah bin Ghazi Faculty of Information and
Communication Technology, Al-Balqa Applied Univer-
sity–Jordan in 2013. Yara received her B.Sc. degrees in
Power Engineering from Hijjawi Faculty of Engineering
Technology, Yarmouk University – Jordan, in 2007. Her
current research interests include sizing of Microgrids
containing distributed generation, energy management and control of Microgrid
using new methods such as Automata, and sizing and managing energy storage
systems.
Nathan Shankar is currently pursuing his B.Tech in EEE
from Vellore Institute of Technology, Vellore, and will
graduate in 2023. He has published around 2 papers in
international journals and conferences and has received
the best paper in one of the international conferences.
Besides this, he also has 1 preprint and 2 other accepted
papers currently undergoing publication. His research
areas include Robotics, Microcontrollers, the Internet of
Things (IoT), Power Electronics, and Solar PV. He has
worked on several projects that involve Robotics, IoT, AI,
and their applications in the different fields of Electrical
Engineering.
11
Ain Shams Engineering Journal xxx (xxxx) xxx
Issa Qiqieh is currently an assistant professor at Al-
Balqa Applied University, Jordan. He received his Ph.D.
degree in computer and electrical engineering from the
School of Engineering, Newcastle University, Newcastle
upon Tyne – UK in (2018), and his M.Sc. (with distinc-
tion) and B.Sc. degrees from Hijjawi Faculty of Engi-
neering Technology, Yarmouk University – Jordan, in
2007&2005, respectively. He is working as a full-time
lecturer in the Faculty of Engineering at Al-Balqa
Applied University from 2007 till now. He is also the
author/co-author of 15+ IEEE/ACM/Elsevier journal and
conference articles. His research interests include multi
and interdisciplinary environments involving approximate, power-adaptive, and
energy-efficient computing for intelligent embedded and IoT systems, as well as
research solutions in machine learning and cyber security.
Jafar A. Alzubi is an associate professor at Al-Balqa
Applied University, Jordan. Received his Ph.D. degree in
Advanced Telecommunications from Swansea Univer-
sity, Swansea – UK (2012). Master of Science degree
(Hons.) in electrical and computer engineering from
New York Institute of Technology, New York - USA
(2005). And Bachelor of Science degree (Hons.) in Elec-
trical Engineering, majoring in Electronics and Com-
munications, from the University of Engineering and
Technology, Lahore – Pakistan (2001). Jafar works and
researches in multi and interdisciplinary environment
involving Machine Learning, classifications and detec-
tion of Web scams, Internet of Things, Wireless Sensor Networks and its security,
and using Algebraic– Geometric theory in channel coding for wireless networks.
Managed and directed few projects funded by the European Union. A cumulative
research experience for over ten years resulted in publishing more than sixty papers
in highly impacted journals. Currently, he is IEEE senior member and serving as an
editor for IEEE Access Journal, and Wireless Sensor Networks area editor for Turkish
Journal of Electrical Engineering and Computer Sciences. In addition, he is editorial
board member and reviewer in many other prestigious journals in computer
engineering and science field. He also managed several special issues in high
impacted journals.
Dr. Omar A. Alzubi received his BSc degree in 1986
from The University of Jordan, Jordan. He obtained his
MSc degree with distinction in Computer and Network
Security from New York Institute of Technology, New
York, USA in 2006. In 2013, he received his Ph.D. degree
in Computer and Network Security from Swansea
University, Swansea, UK. Currently he is an associate
professor in the Computer Science Department at Al-
Balqa Applied University, Jordan. Dr. Alzubi research
interests include machine learning and cyber security.
His cumulative research experience for over ten years
resulted in publishing more than forty articles in highly
impacted journals. Dr. Alzubi is the vice dean of Prince Abdullah Bin Ghazi Faculty
of Information and Communication Technology. In addition, he is an editorial board
member and reviewer of many prestigious journals in computer science field.
Dr. Nallakaruppan M.K is working as the Assistant
Professor in School of Information Technology and
Engineering, Vellore Institute of Technology, Vellore,
Tamilnadu, India. He completed his bachelor degree in
B.E.Electronics and Communication Engineering in J.J.
College of Engineering and Technology. He completed
his master’s degree in M.E.Software Engineering in Anna
University. He completed his PhD in ‘‘Decision Support
Systems for IoT based Infrastructures” in VIT University,
Vellore. His area of expertise are IoT and smart net-
works, Machine learning, soft computing, Decision Sci-
ence and Deep learning. He has 31 International
publications in various International Journals under his area of interest.He has 31
international publications including 16 research papers with various reputed
publishers like MDPI, Springer and Inderscience , 9 conferences with various
reputed publishers like IEEE and springer and 3 book chapters including Elseiver
and Springer.
Y. Khawaja, N. Shankar, I. Qiqieh et al.
Sanjeevikumar Padmanaban (Member’12–Senior
Member’15, IEEE) received a PhD degree in electrical
engineering from the University of Bologna, Bologna,
Italy 2012. He is a Full Professor in Electrical Power
Engineering with the Department of Electrical Engi-
neering, Information Technology, and Cybernetics,
University of South-Eastern Norway, Norway. S. Pad-
manaban has authored over 750+ scientific papers and
received the Best Paper cum Most Excellence Research
Paper Award from IET-SEISCON’13, IET-CEAT’16, IEEE-
EECSI’19, IEEE-CENCON’19, and five best paper awards
from ETAEERE’16 sponsored Lecture Notes in Electrical
Engineering, Springer book. He is a Fellow of the Institution of Engineers, India, the
Institution of Electronics and Telecommunication Engineers, India, and the Insti-
12
Ain Shams Engineering Journal xxx (xxxx) xxx
tution of Engineering and Technology, U.K. He received a lifetime achievement
award from Marquis Who’s Who - USA 2017 for contributing to power electronics
and renewable energy research. He is listed among the world’s top 2 scientists
(from 2019) by Stanford University USA. He is an Editor/Associate Editor/Editorial
Board for refereed journals, in particular the IEEE SYSTEMS JOURNAL, IEEE Trans-
action on Industry Applications, IEEE ACCESS, IET Power Electronics, IET Electronics
Letters, and Wiley-International Transactions on Electrical Energy Systems, Subject
Editorial Board Member—Energy Sources—Energies Journal, MDPI, and the Subject
Editor for the IET Renewable Power Generation, IET Generation, Transmission and
Distribution, and FACETS Journal (Canada).