e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Contents lists available at ScienceDirect

e-Prime - Advances in Electrical

Engineering, Electronics and Energy
journal homepage: www.elsevier.com/locate/prime

State of charge estimation techniques of Li-ion battery of electric vehicles

Abhishek Singh a, *, Kirti Pal a, C.B. Vishwakarma b
a
School of Engineering, Electrical Engineering Department, Gautam Buddha University, Greater Noida 201312, India
b
Associate professor, Harcourt Buttler Technical University, Kanpur, India

A R T I C L E I N F O A B S T R A C T

Keywords: The Lithium-ion batteries are widely utilized in the electric car, bus, and two-wheeler industries because of their
Electric vehicle high energy density, low cost, extended lifespan, high power density, and stable voltage. One of the essential
Lithium-ion battery systems that must be present in any electric vehicle (EV) is the battery management system (BMS). One major
State of charge
input to BMS is state of charge to ensure the battery’s durability, safety, and reliable operation. The state-of-
Machine learning
Equivalent circuit model
charge (SoC) estimation of EV batteries plays a crucial role in optimizing their performance and extending
Battery management system their lifespan. As batteries are nonlinear and time-variant devices, estimating the state of charge or instantaneous
remaining charge within a battery is a particularly challenging task. This paper covers a deep understanding of
SoC estimation techniques for BMS. The two main approaches to explaining estimation of instantaneous
remaining charge are model-based which relies on various battery models and their mathematical equations to
explain the battery characteristics. The second approach is data-driven which studies large measured battery
data sets to understand the behavior of running algorithms. Model-based approaches are based on series-parallel
combinations of resistance and capacitance electrical circuits, while data-driven approaches are based on neural
networks and machine learning algorithms. The review highlights the strengths and limitations of each tech­
nique, suggesting that hybrid approaches could yield more robust results. It emphasizes the importance of future
research in integrating multiple information sources and developing standard evaluation procedures to enhance
SoC estimation accuracy and its practical application in EVs.

accuracy, computational complexity, and real-world implementation

1. Introduction feasibility. A myriad of methodologies has been proposed, each har­
nessing different principles and mathematical models to estimate SoC
The transition towards sustainable transportation solutions has wit­ based on diverse battery parameters.
nessed a remarkable surge in the adoption of EVs worldwide. As the
centerpiece of EVs’ energy storage systems, lithium-ion batteries play a 1.1. Literature review
pivotal role in determining their performance, range, and overall effi­
ciency. A fundamental challenge in optimizing EV operation lies in the The combustion of diesel and gasoline engines increases the level of
precise and reliable estimation of the SoC of these lithium-ion batteries. pollution in the environment and these combustion vehicles can be
The state of charge refers to the available capacity within a battery, minimised by electric vehicles. EVs are becoming increasingly popular
representing the proportion of its total energy storage capability that has because they emit zero carbon emissions [1–4]. Battery is the main
been utilized. Accurate SoC estimation is critical for effective battery component of an EV, and LiB meet the EV’s requirements to deliver
management, enabling safe and efficient charging and discharging maximum power, safety, high-energy density, long life cycle and reli­
strategies, maximizing battery life, and predicting the remaining range ability, as well as better temperature tolerance, a high charging rate, and
of the EV. However, SoC estimation for lithium-ion batteries is a com­ a low self-discharging rate [5]. In [6] author has suggested LiB as the
plex task due to the non-linear and dynamic characteristics of these primary energy storage component for EVs. For EVs to run safely,
energy storage devices, influenced by varying operating conditions, dependably, and effectively, a smart BMS must keep track of the bat­
ageing effects, and environmental factors. tery’s condition. An instantaneous remaining charge, or SoC, is the main
Over the years, extensive research efforts have been dedicated to parameter in the BMS, and this BMS gives information regarding the
devising SoC estimation techniques that strike a balance between presently available energy in the battery and predicts the next working

* Corresponding author.
E-mail address: [email protected] (A. Singh).

https://doi.org/10.1016/j.prime.2023.100328
Received 7 June 2023; Received in revised form 7 September 2023; Accepted 16 October 2023
Available online 28 October 2023
2772-6711/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Nomenclature α0 intercept on Y axis

α1 regression coefficient
LiB lithium-ion battery ε random error
SoC state of charge ibat battery current
BESS battery energy storage system fi activation function
BMS battery management system Wj,k weight
EV electric vehicle θj,k bias
IRC instantaneous remaining charge Oj output of the middle layer
Q battery rated capacity ƞ battery efficiency
R resistance K Kernel
C capacitance xi support vector
AI artificial intelligence ML machine learning
Yi predicted output OCV open circuit voltage
PDE partial differential equation ECIM electrochemical impedance model
ECM equivalent circuit model

duration of the battery for sustained operation [7]. The SoC must be characteristics of LiB are high voltage, long life, low self-discharge rate,
accurate for EV range determination to be smooth, and it is a very useful highly reliable, fast charging, and high energy density [19,20]. There is
input to BMS to ensure connected cells are working in a safe operating some differentiation between the different EV batteries like Li-ion
range and to increase battery life cycle. The SoC, can be assessed using (Lithium Ion), ZnBr (zinc-bromine), VR (vanadium redox), NaS (so­
cell-measurable variables like current and voltage. An accurate SoC dium sulphate), and lead-acid in terms of their voltages, depth of
calculation remains difficult due to battery characteristics and many discharge, life cycle, and round-trip efficiency. After comparison in
operating variables like self-discharge rate, regeneration, hysteresis, cell Table 1, LiB is a good option for electrical vehicles [20].
ageing, and temperature, [8]. Internal battery action can be defined by The main issue in EVs using LiB is accurate SoC estimation, since
an electrochemical model, and its mathematical equations explain the according to SoC the person can plan charging and discharging of EVs.
battery’s inner characteristics. In [9,10], like the Peukert equation, The BMS of EV completely depends on SoC estimation. This paper covers
which can explain some dynamic behavior but cannot handle its thirteen methods of SoC estimation techniques broadly classified in
nonlinear behavior is explained. The mathematical models’ drawbacks three categories conventional, data driven and hybrid approach.
can be minimised by using AI model in which neurons are given weight
instead of state variables [11]. By using resistors, capacitors, and a
voltage source, design an electrical equivalent circuit whose character­ 1.2. Objectives and outcomes of the paper
istics are like the batteries. A high-voltage capacitor reacts like an open
terminal voltage across the battery and reacts as an ideal voltage source, The primary objective of this review paper is to present a compre­
while the rest circuit elements show the internal resistance and terminal hensive overview of the state of charge estimation techniques used in
voltage. On behalf of open circuit voltage, impedance, and other pa­ electric vehicles. By critically analyzing the existing literature, the
rameters, SoC can be estimated by a lookup table. This method explains article seeks to address the following objectives:
the behavior of batteries with the help of tabulation relationships. The
model-based equivalent circuit is used for simulation in EVs and battery a. The paper will categorize and describe different SoC estimation
management systems [12]. The instantaneous remaining charge shows, techniques, including AI and machine learning methods, open circuit
energy remaining in the battery energy storage system [13,14]. voltage approaches, equivalent circuit models, electrochemical
The IRC does not represent any physical value, hence, it cannot be impedance models, and more. It will elucidate the underlying prin­
measured directly. It is measured in terms of electrical parameters ciples, assumptions, mathematical models and implementation as­
voltage, current with temperature, which is represented as a ratio of the pects of each technique.
battery’s instantaneous remaining charge to its rated charge capacity b. Compare the performance of different techniques in terms of accu­
[15]. In fact, rated charge capacity does not remain constant throughout racy, computational complexity, robustness, and real-time imple­
the battery life, it depends upon various factors like the battery envi­ mentation feasibility.
ronment, age of the battery, temperature, pressure, and critical chemical c. Identify the challenges and limitations faced by each technique and
reactions that occur inside the battery. A variety of sensors, including discuss possible strategies for improvement.
amperometric sensors (which measure current by the oxidation and d. Highlight the implications of SoC estimation accuracy on battery
reduction processes of electrochemical reactions) and conductometric management, vehicle performance, and user experience.
sensors (which measure the electrolytic conductivity), can be used to e. Analyse the impact of state of charge estimation on EV range pre­
monitor the electrochemical behavior of a battery. Having a smart BMS diction, energy efficiency, and battery health.
is essential for monitoring the cell clusters and making decisions about
the battery’s, pressure, condition of health, power level, and SoC, as well Outcomes of the paper:
as when the battery need charging and when it should be turned off
[16]. In BMS, SoC is the essential parameter that affects the other Table 1
functions and works as an input for other measurements like cell Different battery characteristics [20].
balancing, health state and power state [17]. The BMS enhances the S. Type of Voltage Depth of Life Round trip
battery age, increases the battery efficiency, avoids damages from No Battery (V) discharge in% cycle efficiency (%)
overvoltage by triggering cut off at favourable conditions and provides
1 Li-Ion 4.3 95 10,000 96
high power density with help of accurate SoC estimation [18]. 2 ZnBr 1.8 100 10,000 70
Electric vehicle-based automobile industries require a battery with 3 VR 1.4 100 13,000 70
characteristics that a LiB can successfully fulfil. Some of the 4 NaS 2.08 100 5000 80

2
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

This review paper will significantly impact the research community 2. SoC estimation classification
in the field of electric vehicles and battery management systems:
The SoC indicates the amount of instantaneous remaining charge in a
a. Enhanced Understanding: Researchers will gain a comprehensive battery storage system. It is a ratio of the present-time available charge
understanding of the diverse SoC estimation techniques, enabling (Qavailable) to the total battery capacity charge (Qrated) [21,22]. The SOC
them to make informed decisions while choosing appropriate is measured in terms of voltage and current. An accurate SoC mea­
methods for their specific applications. surement is not possible because Qrated is not the same throughout the
b. Identification of Research Gaps: The paper will shed light on existing battery’s life as it varies depending on different parameters such as
research gaps and challenges in the field of SoC estimation. This will temperature, pressure, battery age, and the various chemical changes in
inspire further research and development efforts to address these the battery [23].
limitations.
Qavailable
c. Improved Battery Management: By understanding the strengths and SoC = (1)
Qrated
weaknesses of different SoC estimation methods, battery manage­
ment system designers can optimize energy usage, prolong battery SoC classification shown in Fig. 1, which is mainly based on two
life, and enhance safety. methods: one is conventional model-based and the other is data-driven
d Advancements in EV Range Prediction: Accurate SoC estimation is techniques. In model-based technique to design a battery equivalent
critical for predicting EV range reliably. This review will encourage circuit which characteristics are nearly similar as battery static and
the development of more accurate and precise range prediction dynamic characteristics. Data-driven is the study of large measured
models. battery data to understand the behavior of running algorithms, while
e. Real-World Implementation: Researchers and practitioners will gain machine learning techniques are used for the learning process to un­
insights into the practical implementation challenges of different SoC derstand the battery’s attitude regarding data correlation and some set
estimation techniques, guiding the development of efficient real- of rules between the battery data [24–26].
time solutions.
f. Future Research Directions: By highlighting potential avenues for 2.1. SoC estimation by using coulomb counting (CC)
future research, the review paper will stimulate innovation and
collaboration in advancing SoC estimation techniques for electric In this method SOC obtained by the integration of current from t0 to t
vehicles. is given as:

The aim of this review paper is to comprehensively survey and 1

∫t
evaluate the existing state of charge estimation techniques specifically S(t) = S(to ) − (η.i(t) − isd )dt (2)
Q
tailored for lithium-ion batteries in electric vehicles. By critically to

analyzing the strengths, limitations, and real-world applicability of Where S(t0) denotes the beginning of SoC at time t0, whereas S(t)
these techniques, intend to provide researchers, engineers, and policy­ displays SoC at a certain time t [27]. Efficiency is represented by the
makers with invaluable insights to foster advancements in battery symbol ƞ, the battery’s instantaneous charging or discharging current, i
management systems and EV technology. (t), the self-discharging current, isd, and the total battery capacity, Q
[28].
1.3. Organization of the paper
2.2. SoC estimation by using open circuit voltage
The paper is divided in five sections. An EV’s SoC estimation needs
are focussed in introduction section, SoC estimation classification and After a sufficient break to allow the battery to approach equilibrium,
conventional methods are described in section two, data driven tech­ the SoC can be measured using the OCV method. The OCV curves were
niques and hybrid methods are described in section three. Summary, predicted using the linear regression method at various charging cur­
advantages, disadvantages of all methods with future directions for re­ rents. From battery to battery, the SoC and OCV differ. Although this
searchers are explained in discussion section. The last section gives method is relatively straight forward and highly accurate, it requires a
conclusion of the literature survey done for SoC estimation techniques lot of rest time to attain equilibrium [29,31]. And this rest time is also
required for BMS of EVs. affected by the nearby surroundings. Furthermore, due to the battery’s

Fig. 1. Classification of SOC estimation.

3
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

hysteresis properties, meticulous voltage monitoring is necessary. The Dual Polarization model [42]. Table 2 presents six types of model
OCV method’s operational cycle is depicted in Fig. 2. For the Li-ion equivalent circuit diagram and their state space equations.
battery to depolarize, it must first be fully charged before being
allowed to rest [30,32]. Depolarize the LiB, then pulse current through it 2.5. SoC estimation by using ECIM
until it is completely discharged. Then, after each relaxation period,
depolarize the battery by relaxing it for a set period. At this point, the The constant phase element, which depicts charge transfers at the
OCV of the LiB is measured. To measure OCV when in charging mode, electrode and electrolyte interface, along with Li diffusion in electrodes
the same procedure must be followed. [33]. and the development of polarisation resistance and capacitance, are
non-destructive tools used to describe the internal dynamics of batteries
2.3. SoC estimation by using electrochemical model [46,47]. Electrochemical impedance spectroscopy is employed to take a
closer look at the impedance behavior of batteries based on an appro­
The SoC estimation by an electrochemical model shown in Fig. 3 is priately constructed ECIM [48]. In electrochemical impedance spec­
based on electrode theory. This model is based on the mathematical troscopy studies, the battery is activated with a single sinusoidal
modelling of the electrical and chemical properties of the battery in alternating current across a broad frequency range, the recorded voltage
terms of PDE [34]. And these PDE explain the battery dynamics like, Li response can then be used to create a complex impedance frequency
diffusion and potential gradient. No doubt, these equations are very spectrum [49].
complex for real-time SoC estimation. Researchers always try to convert
these higher-order PDEs into lower-order PDEs [35–38]. Fig. 3(b) and 3. SoC estimation by data-driven technique
(c) shows the Li-ion battery charging and discharging concept, in which
battery charge by an external dc supply, anode is kept positive and In data-driven techniques previously measured data of battery are
cathode is kept negative. Li+ and electron release from LiMnO2, but required to understand the behavior of running algorithm and deter­
electron passes through external wire while Li+ passes through elec­ mine the network parameters with the help of self-learning methodology
trolyte and separator as shown in Fig. 3(b), separator allow only Li+ and [50–52]. This technique is fully different from the model-based tech­
these Li+ ion store in graphite structure. Fig. 3(c) shows, discharge of Li nique in that there is no need for parameter estimation such as voltage,
battery via load connected across the terminals of the battery. About 5 % current, and temperature [53–55]. This approach needs ML techniques
of Li+ react with solvent of electrolyte and form solid electrolyte to determine the relationship between the measured data. Recently, AI
interface which also protect the battery and rest of 95 % Li+ are utilized & ML are being apply in various places like trading in stock market,
in battery. medical diagnosis and robotics, finance sector, image processing, clas­
sification and regression, and types of ML are shown in Fig. 5, [56–62].
2.4. SoC estimation by using model-based techniques
3.1. Supervised learning
The model-based techniques also known as ECM. In this method an
equivalent circuit of a battery is developed in the form of electrical A supervised learning is used to train a model by past collected data.
circuit by series-parallel combination of resistor and capacitor elements The data is divided into training samples data about 70 % and remaining
to achieve a similar behavior of battery dynamics [39]. And this ECM sample data is used for validating the output. All algorithms use the
model is shown in Fig. 4. Under any current excitation, the ECM should training dataset to discover various patterns and corelation between the
be ideal to achieve actual battery voltage. Some characteristics of Li-Ion data and finally apply to the test dataset to make predictions or cate­
batteries, like the hysteresis effect and the Warburg effect, are not well gorise data. Fig. 6 below shows the workflow of supervised machine
defined by the circuit element. To increase the precision of voltage learning algorithms.
simulation, an exact model with hysteresis must be created [40,41]. One
method for estimating SoC involves identifying the ECM parameters, 3.2. Support vector machine (SVM)
while another involves calculating a predefined SoC to achieve OCV
before calculating the LiB voltage using the ECM. Some equivalent cir­ The SVM is becoming into a potent tool for dealing with regression
cuit models of battery’s are Rint model, Thevenin model, RC model, issues in nonlinear systems [60,63]. The SVM converts a nonlinear

Fig. 2. Flowchart of OCV based SoC estimation method.

4
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Fig. 3. (a) Li-Ion Battery [35], (b) Electrochemical model of Li-ion battery during charge, (c) Electrochemical model of Li-ion battery during discharge.

Fig. 4. Model-based SoC estimation.

model into a linear model using various kernel functions and regression
∑
N
procedures [64]. The support vector regressor, which has been designed Yi = (W.K(xi , x) + θ) (3)
for regression issues, is one type of SVM. Fig. 7 illustrates the SVM al­ n=1
gorithm model that is used in [50].

5
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Table 2
SoC estimation by different equivalent circuit models.
Types of models Equivalent circuit model Model state equations Refs.

Rint model UL = Uoc − IL R0 [9,12,17]

)
RC model Ub Rc Uc Re ( Re Rc [12,21,22]
UL = + − IL Rt −
Re + Rc Re + Rc Re + Rc
dUb (Ub − Uc + IL Rc )
=
dt Cb (Rc + Re )
dUc (Uc − Ub + IL Re )
=
dt Cc (Rc + Re )

Thevenin Model UL = − UTh + Uoc − IL R0 [12,34,35,42]

dUTh (− UTh ) (IL )
= +
dt CTh RTh CTh

Dual Polarization UL = Uoc − UPa − UPc − IL R0 [12,43]

dUPa (− UPa ) (IL )
= +
dt CPa RPa CPa
dUPc (− UPc ) (IL )
= +
dt CPc RPc CPc

PNGV Model UL = UOC − U1 − U2 − R0 IL [12,44]

dU2 − U2 IL
= +
dt C2 R2 C2
dU1 IL
=
dt C1

GNL Model UL = UOC − Uo − U1 − U2 − R0 IL [12,45]

dU2 − U2 UL IL (Ro + RS )
= + +
dt R2 C2 RS C2 RS C2
dU1 − U1 UL IL (Ro + RS )
= + +
dt R1 C1 RS C1 RS C1
dUo UL IL (Ro + RS )
= +
dt RS Co RS Co

by the trained model and less chance of the error occurring [65].
3.2.1. Data classification in SVM using hyperplane
SVM lie in a supervised learning category and supervised learning is
3.2.2. Data classification in SVM by using kernel selection
used for classification and regression analyzis which is shown in Fig. 8.
The Support vector regression (SVR) was investigated in [67–71], to
Here, a decision boundary, or hyperplane, is employed to classify the
forecast SoC of LiB as a function of cell, voltage, current and tempera­
data into two groups using SVM [64]. The maximal margin hyperplane
ture. The authors used LiB to derive the static dataset and dynamic stress
(between two classes of the support vectors) indicates good prediction

6
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Fig. 5. Types of ML.

3.3. Linear regression (LR)

This technique belongs to the supervised learning class which is

shown in Fig. 10, it is an automated learning technique that can predict
SoC from a high volume of LiB as a function of battery current, battery
voltage, and battery temperature [67]. For continuous variable output
like pass/fail or good/bad, linear regression algorithms become a good
choice. LR creates a relationship between an independent variable (X)
and a dependant variable (Y) in terms of simple linear regression,
multiple LR, and logistic regression [76]. A simple linear regression
equation is given as
Y = α0 + α1 X + ε (4)
In this model input data labelled to create training data [77,78]. In
Fig. 10 shows how the model line is created by labelling the training
data. Once the model is trained it can be used to forecast the unknown
data in labelled form.

3.4. Logistic linear regression

Fig. 6. Supervised learning model for data classification. This technique is shown in Fig. 11 and used for categorized data such
as fraud detection, disease diagnosis, spam, no spam, and yes/no in
form. If probability is 1 data occurs, 0 probability no data occurs and 0.5
probability shows unclassified data.

3.5. Neural network

The neural network as shown in Fig. 12, which takes inspiration from
the human brain, typically undergoes self-adaptation and learning. In
order to train the model, neural networks (NN) are employed for
nonlinear and complex input-output data sets [79–81]. Through hidden
neurons, NN provides non-linear input output mapping. The battery’s
instantaneous input temperature, current and voltage are included in
the input layer of this three-layer device, while the output layer shows
the battery’s SoC in real time [82–86]. Neurons and weight in hidden
layers enabled strong connections between the input and output layers.
NN builds a nonlinear battery model without being aware of the bat­
tery’s core structure by using this input-output relationship [79]. In
[73–75] neural network is used for SoC estimation of LiB of EV.
Fig. 7. SVM algorithm model [50].
∑( )
SoCi = fi Wj,k Oj + θj,k (5)
k
test cycle. They used 10-fold cross validation to train the SVM using
kernel to predict SoC values. Kernal selection for nonlinear data in SVM 3.6. Ensemble bagging (EBa)
is very important which shows in Fig. 9, convert low dimensional feature
space to high dimensional feature space. The EBa technique shown in Fig. 13, in which data S1, S2, and S3 are

7
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Fig. 8. Support Vector Machine Techniques [66].

Fig. 9. Selection of kernel for nonlinear data in SVM.

Fig. 10. Linear regression [77].

randomly selected from bagging data, trained using their respective

classification models M1, M2, and M3, and then predicted to have a
strong output on their behalf. This process is known as "bagging." This
approach reduces variation and guards against overfitting [50].
Fig. 11. Logistic linear regression.
3.7. Ensemble boosting (EBo)

The EBo technique shown in Fig. 14, in which bias error (mismatched

8
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

data) is sequentially removed while selecting data S1 from the boosting

data. It is more likely that the data will be filled during the selection of
S2 and the subsequent sequential operation when the weight of mis­
matched data is increased before going on to the next step. The M1, M2,
and M3 models are weak even if they together make up a powerful
predictive model because they were trained on the individual sets of S1,
S2, and S3 data [50].

3.8. Hybrid method

In a hybrid approach two or more than two algorithms are used

together to improve the battery model’s effectiveness and accuracy. In
addition to producing trust worthy and efficient outcomes, the method
also lowers the price of the battery management system [87,88]. The
technique, however, involves extremely intricate mathematical calcu­
lations that call for a sizeable memory device introduced by [89].
A nonlinear system, like a LiB, can have unknown model parameters
that must be determined using the genetic algorithm (GA), an optimi­
Fig. 12. Three layers architecture of NN [79]. sation technique with biological roots. The GA generates a random set of
chromosomes and uses biological operations like crossover, selection,
and mutation to discover the best values. The second order ECM pa­
rameters of a battery were also determined using the GA [90–95]. The
bacterial foraging algorithm (BFA) is an optimisation method inspired
by nature and based on the cooperative foraging behavior of Escherichia
coli bacteria. Due to its simplicity and high efficiency, the BFA has been
utilised to successfully address a variety of technical and mathematics
issues [96,97]. Lib unknown properties were also estimated using the
BFA [98]. In PSO nature inspired algorithm is used to find the optimal
sulution of complex engineering problem. The same approach is used by
[99–108] to find the unknown parameter of LiB with model based SoC
estimation.

4. Discussion and future directions

In this paper total 108 research papers are reviewed for SoC esti­
mation of LiB of EVs. The broad classification of SoC estimation tech­
niques studied here is divided in two sections conventional and data
driven techniques. In conventional approach five types of SoC estima­
tion techniques are studied,out of five methods, mostly research has
been done on euivalent circuit model based method. In second data
Fig. 13. Ensemble Bagging [50]. driven techniques, machine learning and hybrid approach are studied
and discussed in detail, out of two methods mostly researchers are
working on machine learning techniques. As shown in Fig. 15 in this
literature survey 20 % papers are reviewed on machine learning tech­
niques, 18 % papers are reviewed on OCV techniques and 17 % papers
are reviewed on hybrid techniques as these three techniques are the
most popular techniques.
Both conventional and modern(AI & ML) approach have their ad­
vantages, disadvantages they are listed in Table 3. SoC estimation by
modern machine learning techniques and their advantages and disad­
vantages has been discussed in Table 4.
In Table 5 strengths, weaknesses and application of different SoC esti­
mation techniques are discussed in detail. As shown in Table 3, OCV, CC
and electrochemical model techniques are simple in implementation with
high accuracy. The major disadvantage is intial SoC estimation calcula­
tions is difficult and take time for calculation. Equivalent Circuit model and
Electrochemical Impedance model have major disadvantage that battery
electrochemical reactions are not counted in SoC estimation. Data driven
methods are fast, high accuracy, easy to handle non linear data and can
work with multi input system.Once the network is trained they are able to
predict output for any complex non linear battery characteristics.
Comparative analyzis of different EV state of charge (SoC) estimation
techniques, discussed in this paper are Data driven, Open Circuit
Voltage, Equivalent Circuit Model, Electrochemical Impedance Model
their strengths, weaknesses, and applications are discussed here in
Fig. 14. Ensemble Boosting [50]. detail.

9
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Fig. 15. Percentage of SoC estimation techniques used in this literature survey.

Table 3
SoC estimation by conventional methods.
Conventional SoC Refs. Advantages Disadvantages
Estimation Techniques

Open Circuit Voltage [5,7,8,12,14,18–20, Accuracy high Long time require to reach equilibrium state
29–33] Simple implementation
[39–44]
Coulomb Counting [5,7,8,12,14,18–20,24, Simple implementation Difficult to find initial SOC
25,27,28]
[39–44]
Electrochemical model [21,22,34–38] Model depends on Partial differential equations. PDE computation are complex for SoC estimation.
[39–44]
Equivalent Circuit model [9,12,17,21,22,34,35] Model characteristics replicates as battery. Expensive computation.
[39–45] Model parameter cannot be same throughout the
battery life.
Electrochemical Impedance [46–49] Polarization resistance and capacitance describe the internal Lack of detailed explanation of battery
model dynamics of battery. electrochemical reaction.

Table 4
SoC estimation by modern machine learning techniques.
Techniques Refs. Advantages Disadvantages

Support Vector [63,64,50, • It converts low dimension non-linear data to high dimension linear data by using • High computation time
Machine 64–66] different kernel function • Parameters adjustment needed more
• Accuracy level high trial
Linear Regression [67–74] • Apply where output is continuously varying • More time required to train the model
• Whole training data is needed to train the
model
Logistic Linear [67,69,70,72] • It is used for categorized data such as fraud detection, disease diagnosis, • 0.5 probability shows unclassified data.
Regression
Neural Network [79–86] • Trained data without knowing initial SoC and battery internal Structure. • Complex non-linear model
• Can work with non-linear condition during charging or discharging of battery. • Large memory required
Ensemble Bagging [50] • Good prediction rate as compares to SVM • Overfitting problem in training data
Ensemble Boosting [50] • It converts weak learner model to strong learner model • Training data effected by input feature
• Good working with multiple input

In summary, each SoC estimation technique has its advantages and techniques offer adaptability and accuracy for advanced EVs, while
limitations, and the choice of method depends on factors such as the simpler methods like OCV are suitable for basic applications. ECM and
level of accuracy required, available resources, computational capabil­ EIM provide improved accuracy but may require additional effort in
ities, and the specific application scenario. AI and machine learning terms of parameter estimation and equipment setup.

10
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Table 5
Comparative analysis of different EV state of charge (SoC) estimation techniques.
SoC estimation techniques Strengths Weaknesses Applications

Data driven Reduce complexity, good understanding to relate nonlinear parameters, Huge data needed, Used in advanced EVs,
Characteristics varying according to the environment High computational time and Handle multiple sensors and their
cost, connectivity easily,
Lack of interpretability in deep Real-time and adaptive SoC
learning model estimation

Open circuit voltage Directly related to SoC Less accurate at dynamic load, Low-cost basic battery
Simple and widely used ageing effect, management system
Temperature effect

Equivalent circuit model More accurate, Previous parameters knowledge Suitable for mid-range EV’s
Handle different battery chemistries and ageing effects needed,
High computational cost and
time

Electrochemical Good study of battery health and ageing Sophisticated measuring devices In research of battery performance,
Impedance Model is needed, health and age
Challenges in real-time
implementation

4.1. Future research directions for SoC estimation Overall, the future of electric vehicle state of charge estimation
techniques holds tremendous potential for advancements in accuracy,
Given the rapid advancements in technology, there are several po­ efficiency, and usability. Continued research, collaboration between
tential future scopes for improving and enhancing EV state of charge academia and industry, and technological innovation will play essential
estimation techniques: roles in realizing these advancements.

(i) Big Data and Cloud Integration: With the increasing adoption of 5. Conclusion
connected and autonomous vehicles, EVs can generate vast
amounts of data. Leveraging big data analytics and cloud In this review paper comprehensively explored and evaluated
computing can enable real-time analysis of various parameters various SoC estimation techniques for EVs. The focus was primarily on
affecting SoC, leading to more precise estimations. data-driven techniques, model-based approaches, electrochemical
(ii) Fusion of Multiple Sensors: Combining data from various sensors, methods, and open circuit voltage methods. Through an in-depth anal­
such as voltage, current, temperature, and GPS, can result in a ysis of the literature, it is evident that SoC estimation remains a critical
more comprehensive and reliable SoC estimation. Sensor fusion aspect in the development and operation of EVs, as accurate and reliable
techniques, including Kalman filters and particle filters, can be SoC information is crucial for optimizing battery performance, extend­
employed to integrate data from diverse sources efficiently. ing battery life, and enhancing the overall efficiency of EVs.
(iii) Adaptive and Self-Learning Algorithms: Developing adaptive al­ The study discovered that from the conventional approaches model-
gorithms that can adapt to changing battery characteristics and based estimation is mostly used by researchers since quantitative anal­
environmental conditions can lead to more accurate SoC esti­ ysis and record keeping are simple and easy to use, prone to ageing,
mations throughout the lifetime of the battery. Incorporating self- temperature, sensor drift, and outside influences. The model-based
learning mechanisms can allow the system to continually estimation methodologies produce reliable results that are correct. The
improve its estimation capabilities. adaptive filter-based solution is not very resilient and has a very high
(iv) Inclusion of Environmental Factors: State of charge estimation degree of complexity. The data-driven techniques, such as machine
techniques can be enhanced by considering external factors, such learning algorithms and neural networks, have shown promising results
as weather conditions, road conditions, and traffic patterns. in estimating SoC accurately with real-time data.
Including these factors in the estimation process can provide Overall, each SoC estimation technique has its advantages and lim­
more realistic SoC predictions. itations, and the choice of the appropriate method depends on the spe­
(v) Calibration-Free Techniques: Current SoC estimation methods cific application and requirements. A combination of multiple
often require periodic calibration, which can be cumbersome and techniques, forming hybrid SOC estimation strategies, could potentially
time-consuming. Future research can focus on developing yield more robust and accurate results, thereby enhancing the perfor­
calibration-free techniques that maintain accuracy without mance and reliability of EVs.
frequent calibration routines. In conclusion, this review paper sheds light on the current state-of-
(vi) Wireless Charging Integration: As wireless charging technology the-art in SoC estimation for electric vehicles and highlights the chal­
advances and becomes more prevalent, SoC estimation tech­ lenges and opportunities in this field. As the global transition towards
niques will need to adapt to account for potential differences in sustainable transportation accelerates, advancements in SoC estimation
charging efficiency and optimize charging strategies accordingly. techniques are crucial for ensuring the widespread adoption and success
(vii) Standardization and Interoperability: The EV industry can benefit of electric vehicles in the future.
from standardized SoC estimation methodologies, ensuring
interoperability between different EV models and charging Declaration of Competing Interest
infrastructure. Common standards would improve the accuracy
and reliability of SoC estimations across the board. The authors declare that they have no known competing financial
(viii) Safety and Reliability: Ensuring the safety and reliability of SoC interests or personal relationships that could have appeared to influence
estimation systems will remain a crucial aspect of future research. the work reported in this paper.
EV manufacturers and researchers will need to address issues
related to cybersecurity, fault tolerance, and fail-safe mechanisms
to build trust in SoC estimation technologies.

11
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

Data availability Energy 39 (1) (2012) 310–318, https://doi.org/10.1016/j.energy.2012.01.009.

ISSN 0360-5442.
[23] W. Liu, T. Placke, K.T. Chau, Overview of batteries and battery management for
No data was used for the research described in the article. electric vehicles, Energy Rep. 8 (2022) 4058–4084.
[24] H.W. You, J.I. Bae, S.J. Cho, J.M. Lee, S.H. Kim, Analysis of equivalent circuit
models in lithium-ion batteries, AIP Adv. 8 (12) (2018), https://doi.org/10.1063/
1.5054384. DecArt. no. 125101.
References [25] O. Ali, M.K. Ishak, A.B. Ahmed, M.F.M. Salleh, CAi Ooi, M.F.A.J. Khan, I. Khan,
On-line WSN SoC estimation using Gaussian process regression: an adaptive
[1] D.R. Song, S.Y. Zheng, S. Yang, J. Yang, M. Dong, M. Su, Y.H. Joo, Annual energy machine learning approach, Alex. Eng. J. 61 (12) (2022) 9831–9848, https://doi.
production estimation for variable-speed wind turbines at high-altitude sites, org/10.1016/j.aej.2022.02.067. ISSN 1110-0168.
J. Mod. Power Syst. Clean Energy (2020), https://doi.org/10.35833/ [26] X. Liu, J. Yang, L. Wang, J. Wu, Bayesian information criterion-based data-driven
MPCE.2019.000240. state of charge estimation for lithium-ion battery, J. Energy Storage 55 (2022).
[2] X.S. Zhang, B. Yang, T. Yu, et al., Dynamic surrogate model-based optimization [27] C. Lin, A. Tang, J. Xing, Evaluation of electrochemical models based battery state-
for MPPT of centralized thermoelectric generation system under heterogeneous of-charge estimation approaches for electric vehicles, Appl. Energy 207 (2017)
temperature differences, IEEE Trans. Energy Convers. 35 (2) (2020) 966–976. 394–404, https://doi.org/10.1016/j.apenergy.2017.05.109. ISSN 0306-2619.
[3] B. Yang, X.S. Zhang, T. Yu, H.C. Shu, Z.H. Fang, grouped grey wolf optimizer for [28] L.H. Saw, Y. Ye, A.A. Tay, et al., Integration issues of LiB into electric vehicles
maximum power point tracking of doubly-fed induction generator-based wind battery pack, J. Clean. Prod. 113 (113) (2016) 1032–1045.
turbine, Energy Convers. Manag. 133 (2017) 427–443. [29] Y.J. Wang, J.Q. Tian, Z.D. Sun, L. Wang, R.L. Xu, M. Li, Z.H. Chen,
[4] B. Yang, T. Yu, H.C. Shu, X.S. Zhang, K.P. Qu, L. Jiang, Democratic joint A comprehensive review of battery modeling and state estimation approaches for
operations algorithm for optimal power extraction of PMSG based wind energy advanced battery management systems, Renew. Sustain. Energy Rev. 131 (2020),
conversion system, Energy Convers. Manag. 159 (2018) 312–326. https://doi.org/10.1016/j.rser.2020.110015.
[5] M.S. HossianLipu, M.A. Hannan, A. Hussain, A. Ayob, H.M. Mohamad, T.F. Saad, [30] H. Cai, Q.Y. Chen, Z.J. Guan, J.H. Huang, Day-ahead optimal charging/
N.T.D. Karim, Data-driven state of charge estimation of lithium-ion batteries: discharging scheduling for electric vehicles in microgrids, Protect. Control Mod.
algorithms, implementation factors, limitations and future trends, J. Clean. Prod. Power Syst. 3 (1) (2018) 93–107.
277 (2020), 124110. [31] P. Pillai, S. Sundaresan, P. Kumar, K.R. Pattipati, B. Balasingam, Open-circuit
[6] Li S., He H., Su C., Zhao P., Data driven battery modeling and management voltage models for battery management systems: a review, Energies 15 (18)
method with aging phenomenon considered, Appl Energy, 275, 2020, 115340, (2022).
ISSN 0306-2619, 10.1016/j.apenergy.2020.115340. [32] W.H. Li, R. Monika, B. Julia, J. Dominik, S. Dominik, U.S. Dirk, Digital twin for
[7] M.A. Hannan, M.S.H. Lipu, A. Hussain, A. Mohamed, A review of LiB state of battery systems: cloud battery management system with online state-of-charge
charge estimation and management system in electric vehicle applications: and state-of-health estimation, J. Energy Storage 30 (2020), 101557.
challenges and recommendations, Renew. Sustain. Energy Rev. 78 (2017), [33] R. Xiong, J. Cao, Q. Yu, et al., Critical review on the battery state of charge
https://doi.org/10.1016/j.rser.2017.05.001. Pages 834-854, ISSN1364-0321. estimation methods for electric vehicles, IEEE Access 6 (2018) 1832–1843.
[8] L. Lu, X. Han, J. Li, J. Hua, M. Ouyang, A review on the key issues for LiB [34] H.W. You, J.I. Bae, S.J. Cho, J.M. Lee, S.H. Kim, Analysis of equivalent circuit
management in electric vehicles, J. Power Sources 226 (2013) 272–288, https:// models in lithium-ion batteries, AIP Adv. 8 (12) (2018). DecArt. no. 125101.
doi.org/10.1016/j.jpowsour.2012.10.060. ISSN 0378-7753. [35] A. Hentunen, T. Lehmuspelto, J. Suomela, ‘Time-domain parameter extraction
[9] Y. Hu, S. Yurkovich, Y. Guezennec, B.J. Yurkovich, Electro-thermal battery model method for Thévenin-equivalent circuit battery models, IEEE Trans. Energy
identification for automotive applications, J. Power Sources 196 (1) (2011) Convers. 29 (3) (2014) 558–566. Sep.
449–457. Jan. [36] X.B. Han, M.G. Ouyang, L.G. Lu, J.Q. Li, Simplification of physics-based
[10] F. Luo, H.H. Huang, L.P. Ni, T. Li, Rapid prediction of the state of health of retired electrochemical model for LiB on electric vehicle. Part II: pseudo two-dimensional
power batteries based on electrochemical impedance spectroscopy, J. Energy model simplification and state of charge estimation, J. Power Sources 278 (2015)
Storage 41 (2021), 102866, https://doi.org/10.1016/j.est.2021.102866. ISSN 814–825.
2352-152X. [37] C. Lin, A.H. Tang, Simplification, and efficient simulation of electrochemical
[11] L.W. Kang, X. Zhao, J. Ma, A new neural network model for the state-of-charge model for Li-ion battery in EVs, Energy Procedia 104 (2016) 68–73.
estimation in the battery degradation process, Appl. Energy 121 (2014) 20–27, [38] D. Grazioli, O. Verners, V. Zadin, D. Brandell, A. Simone, Electrochemical
https://doi.org/10.1016/j.apenergy.2014.01.066. ISSN 0306-2619. mechanical modelling of solid polymer electrolytes: impact of mechanical stresses
[12] H. He, R. Xiong, J. Fan, Evaluation of LiB equivalent circuit models for state of on Li-ion battery performance, Electrochim. Acta 296 (2019) 1122–1141.
charge estimation by an experimental approach, Energies 4 (2011) 582–598, [39] D.N. How, M.A. Hannan, M.S. Lipu, et al., State of charge estimation for
https://doi.org/10.3390/en4040582. lithiumion batteries using model-based and data-driven methods: a review, IEEE
[13] R. Morello, R. Di Rienzo, R. Roncella, R. Saletti, F. Baronti, Hardware-in-the-loop Access 7 (2019) 136116–136136.
platform for assessing battery state estimators in electric vehicles, IEEE Access 6 [40] P. Shen, M. Ouyang, X. Han, X. Feng, L. Liu, J. Li, Error analysis of the model-
(2018) 68210–68220, https://doi.org/10.1109/ACCESS.2018.2879785. based state-of-charge observer for lithium-ion batteries, IEEE Trans. Veh.
[14] W. Wang, X. Wang, C. Xiang, C. Wei, Y. Zhao, Unscented Kalman filter-based Technol. 67 (2018) 8055–8064.
battery SOC estimation and peak power prediction method for power distribution [41] Q. Yu, R. Xiong, C. Lin, Online estimation of state-of-charge based on the H
of hybrid electric vehicles, IEEE Access 6 (2018) 35957–35965, https://doi.org/ infinity and unscented Kalman filters for lithium-ion batteries, Energy Procedia
10.1109/ACCESS.2018.2850743. 105 (2017) 2791–2796, https://doi.org/10.1016/j.egypro.2017.03.600.
[15] M.S.H. Lipu, A. Hussain, M.H.M. Saad, A. Ayob, M.A. Hannan, Improved PagesISSN 1876-6102.
recurrent NARX neural network model for state of charge estimation of LiB using [42] S. Zhang, L. Yang, X. Zhao, et al., A GA optimization for lithium–ion battery
pso algorithm, in: Proceedings of the IEEE Symposium on Computer Applications equalization based on SOC estimation by NN and FLC, Int. J. Electr. Power Energy
& Industrial Electronics (ISCAIE), 2018, pp. 354–359, https://doi.org/10.1109/ Syst. 73 (2015) 318–328.
ISCAIE.2018.8405498. [43] C.F. Yang, X.Y. Wang, Q.H. Fang, H.F. Dai, Y.Q. Cao, X.Z. Wei, An online SOC and
[16] Xu Guoning, X. Du, Z. Li, X. Zhang, M. Zheng, Y. Miao, Y. Gao, Q. Liu, Reliability capacity estimation method for aged LiB pack considering cell inconsistency,
design of battery management system for power battery, Microelectron. Reliab. J. Energy Storage 29 (2020), 101250.
88–90 (2018) 1286–1292, https://doi.org/10.1016/j.microrel.2018.06.115. ISSN [44] C. Liu, W. Liu, L. Wang, et al., A new method of modelling and state of charge
0026-2714. estimation of the battery, J. Power Sources 320 (2016) 1–12.
[17] F. Zhang, et al., State-of-charge estimation based on microcontrollerimplemented [45] D.S. Yu, H. Chen, Analysis of the influences from battery impedance on
sigma-point Kalman filter in a modular cell balancing system for LiB packs, in: behaviours of switched reluctance drive, in: Proceedings of the International
Proceedings of the IEEE 16th Workshop on Control and Modeling for Power Conference on Advanced Mechatronic Systems, Luoyang, China, 2013,
Electronics (COMPEL), Vancouver, Canada, 2015, pp. 1093–5142, 12-15 July. pp. 455–460, 25-27 Sept.
[18] J. Meng, G. Luo, M. Ricco, M. Swierczynski, D.I. Stroe, R. Teodorescu, Overview [46] S.E. Li, B. Wang, H. Peng, et al., An electrochemistry-based impedance model for
of LiB modeling methods for state-of-charge estimation in electrical vehicles, lithium-ion batteries, J. Power Sources 258 (2014) 9–18.
Appl. Sci. 8 (2018) 659, https://doi.org/10.3390/app8050659. [47] U. Westerhoff, T. Kroker, K. Kurbach, Electrochemical impedance spectroscopy-
[19] J. Xie, J. Ma, K. Bai, Enhanced coulomb counting method for state-of-charge based estimation of the state of charge of lithium-ion batteries, J. Energy Storage
estimation of lithium-ion batteries based on peukert’s law and coulombic 8 (2016) 244–256.
efficiency, J. Power Electron. 18 (2018) 910–922, https://doi.org/10.6113/ [48] M.S. Yahia, H. Allagui, A. Mami, et al., The frequency behavior of the
JPE.2018.18.3.910. electrochemical model fuel cell by impedance spectroscopy, in: Proceedings of
[20] M.U. Ali, A. Zafar, S.H. Nengroo, S. Hussain, M. Junaid Alvi, H.J. Kim, Towards a the 7th International Renewable Energy Congress (IREC), Hammamet, Tunisia,
smarter battery management system for electric vehicle applications: a critical 2016, pp. 1–6, 22-24 March.
review of LiB state of charge estimation, Energies 12 (2019) 446, https://doi.org/ [49] Y. Lee, S. Park, S. Han, et al., Online embedded impedance measurement using
10.3390/en12030446. high-power battery charger, IEEE Trans. Ind. Appl. 51 (1) (2015) 498–508.
[21] X. Tang, Y. Wang, Z. Chen, A method for state-of-charge estimation of LiFePO4 [50] V. Chandran, C.K. Patil, A. Karthick, D. Ganeshaperumal, R. Rahim, A. Ghosh,
batteries based on a dual-circuit state observer, J. Power Sources 296 (2015) State of charge estimation of LiB for electric vehicles using machine learning
23–29, https://doi.org/10.1016/j.jpowsour.2015.07.028. ISSN 0378-7753. algorithms, World Electr. Veh. J. 12 (2021) 38, https://doi.org/10.3390/
[22] H. He, X. Zhang, R. Xiong, Y. Xu, H. Guo, Online model-based estimation of state- wevj12010038.
of-charge and open-circuit voltage of lithium-ion batteries in electric vehicles, [51] M.A. Hannan, M.S.H. Lipu, A. Hussain, M.H. Saad, A. Ayob, Neural network
approach for estimating state of charge of LiB using backtracking search

12
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

algorithm, IEEE Access 6 (2018) 10069–10079, https://doi.org/10.1109/ Khan, Thailand, 2022, pp. 1–4, https://doi.org/10.1109/ECTI-
ACCESS.2018.27979761588. CON54298.2022.9795401.
[52] M.A. Hannan, M.S.H. Lipu, A. Hussain, et al., Toward enhanced state of charge [76] Yi Li, C. Zou, M. Berecibar, E. Nanini-Maury, J.C.W. Chan, P. van den Bossche,
estimation of lithium-ion batteries using optimized machine learning techniques, J. Van Mierlo, N. Omar, Random forest regression for online capacity estimation
Sci Rep 10 (2020) 4687, https://doi.org/10.1038/s41598-020-61464-7. of lithium-ion batteries, Appl. Energy 232 (2018) 197–210, https://doi.org/
[53] Y.J. Zheng, M.G. Ouyang, X.B. Han, L.G. Lu, J.Q. Li, Investigating the error 10.1016/j.apenergy.2018.09.182. ISSN 0306-2619.
sources of the online state of charge estimation methods for lithium-ion batteries [77] http://gerardnico.com/wiki/data_mining/simple_regression 2023.
in electric vehicles, J. Power Sources 377 (2018) 161–188. [78] C.J. Lin, R.C. Weng, S.S. Keerthi, Trust region Newton method for large-scale
[54] E. Chemali, P.J. Kollmeyer, M. Preindl, A. Emadi, State-of-charge estimation of Li- logistic regression, J. Mach. Learn. Res. 9 (2008).
ion batteries using deep neural networks: a machine learning approach, J. Power [79] M.A. Hannan, M.S.H. Lipu, A. Hussain, M.H. Saad, A. Ayob, ‘Neural network
Sources 400 (2018) 242–255. Oct. approach for estimating state of charge of lithiumion battery using backtracking
[55] M.S.H. Lipu, M.A. Hannan, A. Hussain, M.H.M. Saad, A. Ayob, M.N. Uddin, search algorithm, IEEE Access 6 (2018) 10069–10079.
Extreme learning machine model for state-of-charge estimation of LiB using [80] T.D. Sanger, Optimal unsupervised learning in a single-layer linear feedforward
gravitational search algorithm, IEEE Trans. Ind. Appl. 55 (4) (2019) 4225–4234. neural network, Neural Netw. 2 (6) (1989) 459–473. Jan.
Jul./Aug. [81] G.B. Huang, Y.Q. Chen, H.A. Babri, Classification ability of single hidden layer
[56] X. Hu, S.E. Li, Y. Yang, Advanced machine learning approach for LiB state feedforward neural networks, IEEE Trans. Neural Netw. 11 (3) (2000) 799–801.
estimation in electric vehicles, IEEE Trans. Transport. Electrific. 2 (2) (2016) May.
140–149. Jun[53]M. L. Giger, ‘‘Machine learning in medical imaging,’’ J. Amer. [82] H.T. Huynh, Y. Won, Regularized online sequential learning algorithm for single-
College Radiol., vol. 15, no. 3, pp. 512–520, Mar. 2018. hidden layer feedforward neural networks, Pattern Recognit. Lett. 32 (14) (2011)
[57] S.P. Chatzis, V. Siakoulis, A. Petropoulos, E. Stavroulakis, N. Vlachogiannakis, 1930–1935. Oct.
Forecasting stock market crisis events using deep and statistical machine learning [83] S. Tong, J.H. Lacap, J.W. Park, Battery state of charge estimation using a load-
techniques, Expert Syst. Appl. 112 (2018) 353–371. Dec. classifying neural network, J. Energy Storage 7 (2016) 236–243. Aug.
[58] Y. Xu, C. Yang, J. Zhong, N. Wang, L. Zhao, Robot teaching by teleoperation [84] L. Kang, X. Zhao, J. Ma, ‘A new neural network model for the stateof-charge
based on visual interaction and extreme learning machine, Neurocomputing 275 estimation in the battery degradation process, Appl. Energy 121 (May 2014)
(2018) 2093–2103. Jan. 20–27.
[59] M. Walter, S. Alizadeh, H. Jamalabadi, U. Lueken, U. Dannlowski, H. Walter, [85] H. Chaoui, C.C. Ibe-Ekeocha, ‘State of charge and state of health estimation for
S. Olbrich, L. Colic, J. Kambeitz, N. Koutsouleris, T. Hahn, D.B. Dwyer, lithium batteries using recurrent neural networks, IEEE Trans. Veh. Technol. 66
Translational machine learning for psychiatric neuroimaging, Prog. Neuro (10) (2017) 8773–8783. Oct.
Psychopharmacol. Biol. Psychiatry 91 (2019) 113–121. Apr. [86] D. Cui, B. Xia, R. Zhang, Z. Sun, Z. Lao, W. Wang, W. Sun, Y. Lai, M. Wang,
[60] B. Liu, H. Wang, M.L. Tseng, Z. Li, State of charge estimation for lithium-ion A novel intelligent method for the state of charge estimation of lithium-ion
batteries based on improved barnacle mating optimizer and support vector batteries using a discrete wavelet transform-based wavelet neural network,
machine, J. Energy Storage 55 (Part D) (2022), 105830, https://doi.org/ Energies 11 (4) (2018) 995. Apr.
10.1016/j.est.2022.105830. ISSN 2352-152X. [87] C. Chen, R. Xiong, W. Shen, A LiB-in-the-loop approach to test and validate
[61] D. Silver, A. Huang, C.J. Maddison, A. Guez, L. Sifre, G. van den Driessche, multiscale dual H infinity filters for state-of-charge and capacity estimation, IEEE
J. Schrittwieser, I. Antonoglou, V. Panneershelvam, M. Lanctot, S. Dieleman, Trans. Power Electron. 1 (33) (2017) 332–342.
D. Grewe, J. Nham, N. Kalchbrenner, I. Sutskever, T. Lillicrap, M. Leach, [88] M. Lagraoui, S. Doubabi, A. Rachid, SOC estimation of LiB using Kalman filter and
K. Kavukcuoglu, T. Graepel, D. Hassabis, Mastering the game of Go with deep Luenberger observer: a comparative study, in: Proceedings of the International
neural networks and tree search, Nature 529 (7587) (2016) 484–489. Jan. Renewable and Sustainable Energy Conference (IRSEC), Ouarzazate, Morocco,
[62] E. Badal-Valero, J.A. Alvarez-Jareño, J.M. Pavía, Combining Benford’s Law and 2014, pp. 636–641, 17-19 October.
machine learning to detect money laundering. An actual Spanish court case, [89] B. Yang, J.B. Wang, X.S. Zhang, T. Yu, W. Yao, H.C. Shu, F. Zeng, L.M. Sun,
Forensic Sci. Int. 282 (2018) 24–34. Jan. Comprehensive overview of meta-heuristic algorithm applications on PV cell
[63] J.N. Hu, J.J. Hu, H.B. Lin, X.P. Li, C.L. Jiang, X.H. Qiu, W.S. Li, State-of-charge parameter identification, Energy Convers. Manag. 208 (2020), 112595.
estimation for battery management system using optimized support vector [90] J. Xu, B. Cao, Z. Chen, Z. Zou, An online state of charge estimation method with
machine for regression, J. Power Sources 269 (2014) 682–693, https://doi.org/ reduced prior battery testing information, Int. J. Electr. Power Energy Syst. 63
10.1016/j.jpowsour.2014.07.016. ISSN 0378-7753. (2014) 178–184.
[64] W. Xuan, M. Lin, T. Wei, Q. Jialei, Z. Mengna, State of charge (SOC) estimation of [91] C. Lin, X. Zhang, R. Xiong, F. Zhou, A novel approach to state of charge estimation
NiMH battery based on least square support vector machines, Adv. Mater. Res. using extended kalman filtering for lithium-ion batteries in electric vehicles, in:
211–212 (2011) 1204–1209. Proceedings of the IEEE Transportation Electrification Conference & Expo 2014
[65] Y. Chen, B. Long, X. Lei, The battery state of charge estimation based weighted (ITEC Asia-Pacific), Beijing, China, 2014, 31 August–3 September.
least squares support vector machine, Power Energy Eng. Conf. 1 (2011) 1–4 [92] M.R. Khan, G. Mulder, J. Van Mierlo, An online framework for state of charge
(APPEEC), 2011 AsiaPac. determination of battery systems using combined system identification approach,
[66] V. Kecman, Support Vector Machines—An Introduction, Springer, 2005, J. Power Sources 246 (2014) 629–641.
pp. 1–47. [93] L. Chen, Z. Wang, Z. Lü, J. Li, B. Ji, H. Wei, H. Pan, A novel state-of-charge
[67] T. Hansen, C.J. Wang, Support vector-based battery state of charge estimator, estimation method of lithium-ion batteries combining the grey model and genetic
J. Power Sources 141 (2) (2005) 351–358. Mar. algorithms, IEEE Trans. Power Electron. 33 (2018) 8797–8807.
[68] X. Wu, L. Mi, W. Tan, J.L. Qin, M.N. Zhao, State of charge (SoC) estimation of ni- [94] S. Blaifi, S. Moulahoum, I. Colak, W. Merrouche, An enhanced dynamic model of
mh battery based on least square support vector machines. Advanced Materials battery using genetic algorithm suitable for photovoltaic applications, Appl.
Research, Trans Tech Publications, StafaZurich, Sweitzerland, 2011, Energy 169 (2016) 888–898.
pp. 1204–1209. [95] H. Mu, R. Xiong, H. Zheng, Y. Chang, Z. Chen, A novel fractional order model-
[69] J.Á. Antón, P.G. Nieto, F. de Cos Juez, F.S. Lasheras, M.G. Vega, M.R. Gutiérrez, based state-of-charge estimation method for lithium-ion battery, Appl. Energy
Battery state-ofcharge estimator using the SVM technique, Appl. Math. Model. 37 207 (2017) 384–393.
(2013) 6244–6253. [96] N.K. Jhankal, D. Adhyaru, Bacterial foraging optimization algorithm: a derivative
[70] J.A. Anton, P.G. Nieto, C.B. Viejo, J.V. Vilan, Support vector machines used to free technique, in: Proceedings of the Nirma University International Conference
estimate the battery state of charge, IEEE Trans. Power Electron. 28 (2013) on Engineering (NUiCONE), Gujarat, India, 2011, 8–10 December.
5919–5926. [97] A.K. Kar, Bio inspired computing–a review of algorithms and scope of
[71] H. Sheng, J. Xiao, Electric vehicle state of charge estimation: nonlinear applications, Expert Syst. Appl. 59 (2016) 20–32.
correlation and fuzzy support vector machine, J. Power Sources 281 (2015) [98] Y. Ma, J. Ru, M. Yin, H. Chen, W. Zheng, Electrochemical modeling and
131–137. parameter identification based on bacterial foraging optimization algorithm for
[72] S. El Fallah, J. Kharbach, Z. Hammouch, A. Rezzouk, M.O. Jamil, State of charge lithium-ion batteries, J. Appl. Electrochem. 46 (2016) 1119–1131.
estimation of an electric vehicle’s battery using deep neural networks: simulation [99] H. Han, H. Xu, Z. Yuan, Y. Zhao, State of charge estimation of li-ion battery in evs
and experimental results, J. Energy Storage 62 (2023), 106904, https://doi.org/ based on second order sliding mode observer, in: Proceedings of the IEEE
10.1016/j.est.2023.106904. ISSN 2352-152X. Transportation Electrification Conference & Expo 2014 (ITEC Asia-Pacific),
[73] J. Chen, Yu Zhang, W. Li, W. Cheng, Q. Zhu, State of charge estimation for Beijing, China, 2014, 31 August–3 September.
lithium-ion batteries using gated recurrent unit recurrent neural network and [100] W. Wang, H.S.H. Chung, J. Zhang, Near-real-time parameter estimation of an
adaptive Kalman filter, J. Energy Storage 55 (2022), 105396, https://doi.org/ electrical battery model with multiple time constants and SoC-dependent
10.1016/j.est.2022.105396. Part AISSN 2352-152X. capacitance, IEEE Trans. Power Electron. 29 (2014) 5905–5920.
[74] C.C. Lee, P. Hu, C.Y. Li, S.H. Wang, State of charge estimation of the lithium-ion [101] H.H. Afshari, M. Attari, R. Ahmed, M. Farag, S. Habibi, Modeling,
battery based on neural network in electric vehicles, in: Proceedings of the IEEE parameterization, and state of charge estimation of li-ion cells using a circuit
International Symposium on Product Compliance Engineering - Asia (ISPCE- model, in: Proceedings of the Transportation Electrification Conference and Expo
ASIA), Guangzhou, Guangdong Province, China, 2022, pp. 1–3, https://doi.org/ (ITEC), Dearborn, MI, USA, 2016, 27–29 June.
10.1109/ISPCE-ASIA57917.2022.9971063. [102] Z. Yu, L. Xiao, H. Li, X. Zhu, R. Huai, Model parameter identification for lithium
[75] O. Ali, M.K. Ishak, F. Memon, M.S.M. Asaari, Estimation of battery state-of-charge batteries using the coevolutionary particle swarm optimization method, IEEE
using feedforward neural networks, in: Proceedings of the 19th International Trans. Ind. Electron. 64 (2017) 5690–5700.
Conference on Electrical Engineering/Electronics, Computer, [103] X. Hu, H. Yuan, C. Zou, Z. Li, L. Zhang, Co-estimation of state of charge and state
Telecommunications, and Information Technology (ECTI-CON), Prachuap Khiri of health for lithium-ion batteries based on fractional-order calculus, IEEE Trans.
Veh. Technol. 67 (2018) 10319–10329.

13
A. Singh et al. e-Prime - Advances in Electrical Engineering, Electronics and Energy 6 (2023) 100328

[104] H. Aung, K.S. Low, J.J. Soon, State-of-charge estimation using particle swarm Dr. Kirti Pal is an Associate Professor in Electrical Engineering
optimization with inverse barrier constraint in a nanosatellite, in: Proceedings of Department, School of Engineering, Greater Noida, India. She
the Industrial Electronics and Applications (ICIEA), Auckland, New Zealand, has total 14 year of teaching experience. Dr. Kirti holds a PhD
2015, 15–17 June. degree in Electrical Engineering, 2013 from RGTU, Bhopal, M.
[105] M. Ye, H. Guo, B. Cao, A model-based adaptive state of charge estimator for a E. degree from MITS Gwalior, 2006, BE degree from L.N.C.T.
lithium-ion battery using an improved adaptive particle filter, Appl. Energy 190 Bhopal, 2004. She has published many research and conference
(2017) 740–748. paper in various journal and conferences. She has guided
[106] Y. Kishor, C.H. Kamesh Rao, R.N. Patel, L.K. Sahu, An architectural and control project and dissertation work of many B.Tech and M.tech
overview of DC-microgrid for sustainable remote electrification, Singh, V.K., Students. Currently she is having 3 PhD students under his
Bhoi, A.K., Saxena, A., Zobaa, A.F., Biswal, S.. Renewable Energy and Future guidence. She has delivered many keynote and expert talks in
Power Systems. Energy Systems in Electrical Engineering, Springer, Singapore, national/international conferences/workshop.
2021, https://doi.org/10.1007/978-981-33-6753-1_11. Her research areas are Restructuring of Power System,
[107] A. Sreeram, Y. Kishor, C.H. Kamesh Rao, R.N. Patel, Modeling and simulation of Power System Analysis and optimization, Soft Computing
SoC-based BMS for stand-alone solar PV-fed DC microgrids. Enabling Techniques, Renewable Energy systems and Electric Vehicle.
Methodologies for Renewable and Sustainable Energy, CRC Press, 2023, pp. 1–31.
Edition1st EditioneBook ISBN9781003272717.
[108] Y. Kishor, R.N. Patel, PV-battery fed two-stage non-isolated buck-boost converter
Dr. Chandra Bhan Vishwakarma is an Associate Professor in
for low-voltage DC-microgrid, in: Proceedings of the IEEE 2nd International
Electrical Engineering Department, School of Engineering, H.B.
Conference on Smart Technologies for Power, Energy and Control (STPEC),
T.I Kanpur, India. He has 20 years of teaching experience. Dr.
Bilaspur, Chhattisgarh, India, 2021, pp. 1–6, https://doi.org/10.1109/
CB Vishwakarma holds a Ph.D degree in Electrical Engineering,
STPEC52385.2021.9718690.
2010 from IIT Roorkee, MTech degree from IIT Roorkee in
2002, B.Tech degree from IIT Roorkee in 1998. He has pub­
lished many research and conference paper in various journal
Abhishek Singh is a Ph.D scholar in Electrical Engineering and conferences. She has guided project and dissertation work
Department, School of Engineering, Greater Noida, India. He of many B.Tech and M.tech Students.
has total 12 years of teaching experience. Abhishek holds a His-areas of Research Interest: • Model Order Reduction of
MTech degree in Signal Processing from A.I.T Delhi in 2012, linear and nonlinear dynamic systems. • Controller Design:
BTech degree from RGEC Meerut in 2005. His-research area is PID, Fuzzy, Neuro-Fuzzy, MPC, PLC, IMC, etc. • Intelligent
Electric Vehicle. Controls • Artificial Intelligence and Machine Learning • En­
gineering Optimizations • Fractional Order Modelling and
Control

14