![Estimation methods of different batteries [19-23]](https://www.researchgate.net/profile/B-P-Patil/publication/341778248/figure/tbl2/AS:897246006632448@1590931526450/Estimation-methods-of-different-batteries-19-23_Q320.jpg)
Content uploaded by B P Patil
Author content
All content in this area was uploaded by B P Patil on May 31, 2020
Content may be subject to copyright.
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15053
Published by: The Mattingley Publishing Co., Inc.
A Review on AI based Predictive Battery
Management System for E-Mobility
Satyashil D. Nagarale, B. P. Patil
Asst. Professor, Dept of ETC, PCCOE, Pune
Principal, Army Institute of Technology, Pune
Article Info
Volume 83
Page Number: 15053 – 15064
Publication Issue:
May - June 2020
Article History
Article Received: 1May 2020
Revised: 11 May 2020
Accepted: 20 May 2020
Publication: 24May 2020
Abstract:
Abstract: The advancement in digitalization and availability of reliable sources of
information that provide credible data, Artificial Intelligence (AI) has emerged to
solve complex computational real life problems which was challenging earlier. The
Artificial Neural Networks (ANNs) play a very effective role in digital signal
processing. However, ANNs need rigorous main processors and high memory
bandwidth, and hence cannot provide expected levels of performance. As a result,
hardware accelerators such as Graphic Processing Units (GPUs), Field
Programmable Gate Arrays (FPGAs), and Application Specific Integrated Circuits
(ASICs) have been used for improving overall performance of AI based
applications. FPGAs are widely used for AI implementation as FPGAs have
features like high-speed acceleration, low power consumption which cannot be
done using central processors and GPUs. FPGAs are also a reprogrammable unlike
central processors, GPU and ASIC. In Electric-powered vehicles (E-Mobility),
Battery Management Systems (BMS) perform different operations for better use of
energy stored in lithium-ion batteries (LiBs). The LiBs are a non-linear
electrochemical system which is very complex and time-variant in nature. Because
of this nature, estimation of States like State of Charge (SoC), State of Health
(SoH) and Remaining Useful Life (RUL) is very difficult. This has motivated
researchers to design and develop different algorithms which will address the
challenges of LiBs states estimations. This paper intends to review AI based data-
driven approaches and hardware accelerators to predict the SoC, SoH and RUL of
the LiBs. The goal is to choose an appropriate algorithm to develop an advanced AI
based BMS that can precisely indicate the LiBs states which will be useful in E-
Mobility.
Keywords: Artificial Intelligence, Battery Management Systems, Electric-powered
automobiles, State of Charge, Remaining Useful Life, State of Health, Field
Programmable Gate Arrays, Graphic Processing Units, Application Specific
Integrated Circuits, and Li-ion batteries.
I. Introduction
In recent decades, E-Mobility has gained
importance due to environmental concerns and
the depletion of fossil fuels. Renewable energy
is preferred over fossil fuels due to
characteristics like renewability, non-
depletion, and it does not give rise to
environmental issues. LiBs have achieved
considerable attention in the E-Mobility
industry due to their safety, low maintenance,
longer life span, high efficiencies, nominal
voltage, reasonable cost, and ability to operate
in a large temperature range [1].
The availability of extensive credible data and
advancement in digital electronics that have
high computing power, have revived AI [2].
The accurate determination of the LiB states is
extremely challenging because of its high non-
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15054
Published by: The Mattingley Publishing Co., Inc.
linear, time-variant and complex
electrochemical system. AI based algorithms
can be used to estimate battery states
accurately independent of granular information
of electrochemical processes of batteries. AI
algorithms or Data driven approaches will be
useful for predicting battery states like SoC,
SoH and RUL. The SoC, SoH & RUL protect
the LiBs by providing visual signals or
warnings for any problems in the BMS. This
facilitates monitoring of the voltage, internal
resistance, current, temperatures, ageing and
life cycles of the LiBs [3, 4]. This paper
discusses a general working of BMS and
energy storage systems. This paper presents a
relative analysis of AI based data driven
estimation approaches of SoC, SoH and RUL
and evaluates various AI based approaches for
enhanced battery management. Finally, as part
of future approach, this review paper
recommends use of AI based BMS for E-
Mobility which will be beneficial for
automobile engineers, experts, and researchers.
II. Battery management system (BMS) for
E- Mobility
The past decades have witnessed an increase in
E-Mobility. Over 125 million electric vehicles
will be in use by 2030 according to the
projections [5]. BMS ensures a reliable,
efficient, and safe operation of the LiBs. BMS
is an electronic system that estimates the SoC,
SoH, and RUL; controls over-charge and
under-charge; controls voltage and
temperature, and ensures safety and fault
assessment of the cells. Due to the battery
being an electrochemical product, it poses an
uncertainty of performance in various
environmental and operational conditions.
BMS, therefore, controls and monitors the
states of the batteries at various levels like
module, cell and battery pack levels making it
a critical component in E-Mobility for
optimum operation of LiBs in varying driving
conditions. Relays, Sensors, and breakers are
incorporated in the BMS of E-Mobility to
ensure protection from under voltage, over
voltage, over-charge, under-charge, under-
current, over current, etc. [6]. The table below
showcases BMS components and their
operational details [7].
Table 1 – BMS Components and their Operational Details
BMS Components
Operational Details of Component
Measurement or
Data acquisition
Measures physical or electrical parameters such as current, voltage,
temperature, pressure. A computer converts them into digital values.
Measurement can be done with the help of sensors, hardware and software.
Battery Algorithm
Estimate SOC, SOH and RUL with reference to current, voltage, ageing, and
temperature.
Capability
Estimation
After battery states estimation, BMS has to find the maximum available charge
and discharge current at any instant in accordance by using an algorithm. The
output of this component is passes to the Electronic Control Unit (ECU) to
prevent over charge and discharge of battery.
Cell Balancing
Cell Balancing equals or balances voltage level of each cell.
Thermal
Management
Any divergence from an operational temperature range is addressed by
activating cooling or heating system, etc. which facilitates thermal stability.
Vehicle
Communications
(CAN and
Ethernet)
Communication with all battery components.
Safety
Safety and Reliability of LiB relates to the construction and installation of the
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15055
Published by: The Mattingley Publishing Co., Inc.
cells. Protects the pack from damaging events such as over-charge and under-
charge and proactively detects potentially dangerous conditions within the
battery pack.
The literature reports that BMS can be
designed using various approaches based on
the suitable operation for a particular
application, but most approaches focus on a
specific operation, such as SoC estimation,
SoH estimation, RUL estimation or Cell
Balancing [8-10].
BMS has three different types of design
structures viz. centralized, modular, and
distributed. Based on the objectives to be met,
developers use suitable structure like a
distributed one for scalability and portability
[11]. Microcontroller Unit (MCU) or FPGAs
are used as a hardware component in BMS.
LiBs are extensively equipped in both high-
power applications and low-power electronics
products, such as hybrid-motor engines,
electric cars, smart phones, tablet, laptops, etc.
The table below showcases Commercial BMS
used in E-Mobility [12-17].
Table 2 - Commercial BMS used in E-Mobility
Sr.
No.
BMS
Manufactures
Description
Key Features
1
SPIKE
SPIKE’s developed
advanced BMS with
highly efficient thermal
management.
Each module comes with an advanced Battery
Management System and is optionally
equipped with liquid cooling plates to
significantly improve lifetime, safety and
performance.
2
AMP
AMP provides the best-
in-class BMS for E-
Mobility.
Estimate SoC & SoH, Power Prediction,
temperature & Humidity Sensing, Lower
system cost, Highest battery utilization and
etc.
3
ION ENERGY
ION Energy’s battery
designs and the
ecosystem use a
customizable and
modular approach.
Configurable single board compact
architecture.It supports multiple battery
chemistries. SoC & SoH estimations based on
advanced algorithms. Easily Customizable
features like additional temperature
measurements, current sensor, higher
balancing current, heater, etc.
4
RENESAS
ELECTRONICS
Renesas' provide BMS
solutions are specifically
designed to meet the stiff
safety, performance
requirements and
reliability
Cell equalization and safety.
Designed for next generation E-Mobility
5
GREENECO
Green Eco products are
used in renewable and
energy efficient products.
Charge – Discharge Protection for each cell
Over Current Protection, Impendence and
Temperature measurements.
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15056
Published by: The Mattingley Publishing Co., Inc.
BMS for E-Mobility require efficient hardware
and software along with advanced battery state
estimation algorithms. The manufacturers have
not disclosed the advanced algorithms they
have used in BMS.
III. Energy Storage Device (ESD) for E-
Mobility
Currently, LiB technology is growing very fast
as the most reliable electrical power source for
E- Mobility. Various ESDs like LiB, VRFB,
NiCd, NaNICl, ZBFB, NaS and lead acid are
available. The table below lists the Different
types of ESD for E- Mobility with their
properties [18-19].
Table 3 - Types and Properties of ESD for E- Mobility
Energy
storage
device
Specific
Energy
Density
(Wh/L)
Power
density
(W/L)
Cell
Voltage
(V)
Life
Cycle
Nominal
Capacity
(%)
Round trip
Efficiency
(%)
Estimated
cost,
(USD/kWh)
LiB
200–400
1500-
10,000
4.3
10,000
95
96
200–1260
VRFB
25–33
1–2
1.4
13,000
100
70
315–1050
NiCd
60–150
80–600
1.3
2500
85
83
-
NaNiCl
160–275
150–270
-
3000
100
84
315–488
ZBFB
55–65
1–25
1.8
10,000
100
70
525–1680
NaS
140–300
140–180
2.08
5000
100
80
263–735
Lead-
acid
50–80
10–400
2.0
1500
50
82
105–475
Table 3 shows that VRFB has the highest life
cycle and LiB has the highest power densities
and specific energy as compared to the other
storage devices. The nominal voltage is
important point as it determines the quantity of
single cell and number of cells required in a
pack of battery for reliable and safe operations.
Round Trip efficiency is defined as the ratio of
energy recovered and energy input to ESS.
Compare to other Energy storage device LiB
has better round trip efficiency which plays an
important role in BMS to improve overall
performance. LiB is better choice because of
its power density, specific energy density, cell
voltage, life cycle, round trip efficiency and
cost. Although LiB is the best choice for E-
Mobility, the challenges of reduction in capital
cost and high life cycle need to be addressed.
BMS is an essential component of all LiB
packs. These battery packs can be classified
into Low Voltage (LV) or High Voltage (HV)
(UNECE 2013).
6
ORION
The Orion BMS,
specifically designed to
meet the difficult
requirements of
protecting and managing
packs of battery for E-
Mobility with automotive
grade quality.
BMS settings include: Over Charge and
Under Charge voltage, Over current limits.
SoC Estimation, Thermal Management along
with current sensor, communication unit and
etc.
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15057
Published by: The Mattingley Publishing Co., Inc.
Table 4: Classifications of LiB packs
Classifications of
LiB packs
Range
Usability
Low Voltage (LV)
30 – 1000 VAC
Light electric and hybrid vehicles, two and three wheelers.
High Voltage (HV)
60 – 1500 VDC
Heavy electric vehicles such as elevators, conveyors, trolleybuses and
etc.
2. SoC, SoH and RUL Estimation Methods
As pointed out earlier, accurate battery states
estimation is a very critical function of the
BMS design. Therefore, different literature
presents various methods to calculate the SoC,
SoH and RUL precisely. The SoC estimation
helps to optimize the battery performance and
extend battery lifespan.
SoC = (available capacity/rated capacity*) X
100
SoH = (maximum charge capacity /rated
capacity*) X 100
RUL = Total life cycles - Actual life cycles
(*Rated capacity is given by manufactures
which is constant through the battery lifetime)
Table 5 shows the estimation methods of
different batteries.
Table 5 - Estimation methods of different batteries [19-23]
Sr.
No.
Classification
of Estimation
Methods
Working
Principle
Methods
Key Limitations
1
Direct
Measurement
Physical
properties such
as impedance,
and
charge/discharge
current.
Internal
Resistance(IR),
Open-circuit voltage
(OCV),
Electromotive
force(EMF) &
Impedance
Spectroscopy (IS)
Unable to measure initial states.
Open loop, minute errors will build
up with time due to the integration
term.
Requires high precision sensors.
2
Book-
keeping
Input will be
Battery charge
and discharge
current
Coulomb Counting
(CC)
Open Loop, Sensitive to current
sensor precision.
Accuracy depends on quality of
sensors and initial LiB states.
Accuracy depends on battery
history, temperature, discharge
current, and life cycle of battery.
Not suggested for online BMS as
the battery needs long time resting.
3
Model-based
By developing
battery models
battery ageing is
evaluated but
these methods
involve
complicated
mathematical
Adaptive methods:
The Kalman filter;
Particle filter; &
Least square
Accuracy highly depends on quality
of the model. The algorithms have
high complexity, high
computational cost, and need
experimental tests to form different
datasets. The performance of the
KF variants is required on previous
knowledge and measurement noise
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15058
Published by: The Mattingley Publishing Co., Inc.
operations.
covariance which is causing
performance to degrade and leads
to poor convergence or slow
adaptation.
4
Data Driven
or AI or
Computer
Intelligence
Used Advanced
Algorithms.
FL, ANN, SVM,
RNN
Require high computational time
and storage size. Accuracy depends
on quantity of data and quality
training NN.
As compared to other methods, the data driven
approaches can successfully capture the
nonlinear relationship of battery parameters
and are more suitable to estimate SoC, SoH
and RUL. These methods work independent of
battery operating condition and complex
electrochemical process of LiBs. For battery
state estimation by the data driven FL, SVM,
ANN are accepted by few of researchers
accepted [24-27]. Based on the study of
various methods to estimate battery states, the
AI based data-driven methods generate more
accurate results and are simpler as compared to
the direct, model-based, and adaptive filter
methods.
IV. AI based Estimation Methods
i. Genetic Algorithm (GA)
GA is an optimization method and used to
estimate the model parameters of LiB system.
Zheng Chen used GA to measure SoH for
electric and hybrid electric vehicle applications
and also developed formula to calculate SoH
[28].
ii. Bacterial Foraging Algorithm (BFA)
This is a nature inspired optimization
algorithm. Which is based on Escherichia coli
bacteria’s social foraging behavior. This is
used to find the solution of engineering and
mathematical problems because of its high
efficiency and simplicity. To estimate the
unknown parameters of the LiB, BFA is also
utilized. [29].
iii. Particle Swarm Optimization (PSO)
Another algorithm which is based on nature-
inspired approach using the social behavior of
different species, such as birds or fishes,
interacting with each other or with the
surroundings is PSO. Main objective of this
method is data sharing in the group, where
each bird in the flock does not know the exact
location of the food, but they can track the
food site very easily through information
sharing method [30].
iv. iv. Fuzzy Logic (FL)
FL is another method like LiB which recognize
the unknown parameters of a highly complex
and nonlinear system. The working of FL has
four stages: fuzzification, fuzzy rule base,
inference engine, and defuzzification. To
evaluate the parameters of the SOC with
improved accuracy FL was also used [31].
However, FL requires large memory for
storage and time consuming.
v. v. Neural Network (NN)
The NN is composed of nodes or neurons
which is similar to human brains.
Fundamentally, NN uses a three-layer, input,
output and hidden layers containing neurons
with system specifications. The NN has self-
adaptability and self learning abilities. The
BSA and back propagation neural network
(BPNN) can be used to improve the
performance of NN [32]. The NN can be use to
estimate battery states efficiently.
vi. vi. Adaptive Neuro Fuzzy Inference System
(ANFIS)
An ANFIS is better form of the artificial NN,
which is based on the Takagi–Sugeno fuzzy
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15059
Published by: The Mattingley Publishing Co., Inc.
inference system. The ANFIS has the
advantages of FL and NN in a single
framework. For modeling, optimization, and
nonlinear mapping ANFIS is used as an
extraordinary tool. The comparative study
showed the better performance of ANFIS over
NN. [33].
vii. vii. Support Vector Machine (SVM)
The SVM is machine learning algorithm that
uses both classification or regression
challenges to change to a linear model from
nonlinear model. However, the complexity of
the SVM system is higher due to its complex
quadratic programming. SVM can be used to
estimate states of LiB [34].
viii. Multivariate Adaptive Regression Splines
(MARS)
Novel type of flexible regression analysis for
high dimensional data called as MARS was
suggested by Friedman in 1991. Similar with
data driven approach, no prior knowledge on
the form of the numerical operation is required
for MARS. The key advantages of MARS lie
in its capacity to capture the intrinsic
complicated data mapping in high-dimensional
data patterns and produce simpler, easier-to-
interpret models, and its ability to perform
analysis on parameter relative importance. The
important advantage of their suggested method
is that it can be implemented using low cost
microcontroller [35].
ix. ix. Recurrent neural network (RNN)
Current input decides the output of FNN. It is
inaccurate to use the FNN process for time
sequence problems such as battery states
estimation. The RNN is a type of an ANN that
add additional weights to the neurons to create
feedback loop in the NNs to maintain an
internal neuron. In traditional feed forward
neural network (FNN), all test cases are
considered to be independent whereas in RNN,
present state output depend on input as well as
output from previous state which more suitable
for LiB state estimation because battery states
not only depend on present state also depend
previous state. Also, number of inputs to FNN
is fixed and cannot be randomly changed,
which is not easy for battery states estimation
at any given time moment. This enables the
RNN to deal with sequential or time series
problem by remembering, storing, and
processing past complex signals. RNNs have
been mostly used in time series forecasting,
language or speech recognition, machine
translation, and system modeling. RNNs has
short-term memory problem. If a sequence is
long enough, they will have a hard time to
capture long-term sequential dependencies
[36]. To overcome these problems there are
commonly-used RNN such as Gated Recurrent
Unit (GRU) and Long-Short Term Memory
(LSTM) networks. GRU–RNN and LSTM
Networks both are popular ANN or self
learning networks have given excellent
solution to the problems in various domain
such as speech recognition, language
correction, robotic assistance, etc. GRU-RNN
and LSTM networks both are a type of RNN
which uses separate units in addition to RNN
structure. LSTM units consist of a 'memory
cell' that can maintain data in memory for long
period. A set of logical gates are used to
control while information enters the memory,
when it's output, and it's forgotten. This
architecture helps them to learn longer-term
dependencies. GRU-RNNs are similar to
LSTMs, but uses a simple structure. Due to use
of set of gates for control of flow of
information they don’t require separate
memory which reduces the count of gates.
GRU-RNN and can measure SoC, SoH and
RUL more accurately by experimenting battery
historical data factors such as voltage, current
and at different temperature levels, hence does
not require information about battery complex
chemical reactions, internal chemistry, and
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15060
Published by: The Mattingley Publishing Co., Inc.
model parameters. FPGAs are best suitable to
implement on hardware compared to CPU,
ASIC and GPU. Further betterment can be
done on the GRU-RNN architecture’s by
experimenting or implementing various
activation function used in GRU to improve
accuracy [37].
Table 6 - Summary of the AI Algorithms available for LiBs states estimation.
AI
Algorithms
Advantages
Limitations
Applicable
in E-Mobility
GA
Robust to noisy environment
and accuracy level is high.
Time consuming and more
computational complexity.
Yes
BFA
Gives moderate accuracy for
dynamic current.
High Latency.
No
PSO
Moderate accuracy in
measuring the SOC online
High Latency.
Yes
FL
Moderate accuracy in different
conditions
Requires large data storage,
computational complexity and
Expensive data processing units.
Yes
NN
Most suitable methods for state
estimation in E-Mobility and
works in any condition.
Requires large data storage and
depend on training of NN.
Yes
ANFIS
Most suitable methods for state
estimation in E-Mobility and
works in any condition.
Requires large data storage and
depend on training of NN.
Yes
SVM
Suitable for non linear model
High Latency.
Yes
MARS
Moderate accuracy
Number of experiments are
needed to validate this technique
are more. Allowed only for
certain input variables
No
RNN
No prior knowledge of
battery’s internal parameters or
chemical process is required.
Suitable for non linear model.
Requires large data storage and
depend on training of NN.
Gradient Vanishing Problem.
Yes
GRU-RNN/
LSTM
Prior information of battery’s
internal parameters or chemical
process is not required and is
suitable for nonlinear model.
Reason of gradient decay over
layers is use of sigmoid or tangent
as an activation function.
Accuracy depends on activation
functions.
Yes
Based on the above summary, GRU-RNN can
give much better accuracy. The problem of
long term dependencies in the RNN is
overcome and is suitable for nonlinear model.
To extend this work, battery lifetime
experiments must be conducted to obtain
battery aging data for evaluating accuracy.
Future work will include prototype hardware
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15061
Published by: The Mattingley Publishing Co., Inc.
implementation for the BMS by using FPGA
based accelerators.
V. FPGA Accelerators
FPGAs offer high computing capabilities, re-
configurability, low power consumption,
concurrency, and customization of architecture
for particular applications which make it a very
suitable platform for hardware acceleration.
FPGAs are being used for implementing AI or
machine learning algorithms to improve
overall performance. Hardware accelerators
such as ASIC, FPGA, and GPU have been
used to betterment of performance of AI based
applications. AI based applications generally
use GPUs as hardware accelerator for
enhancing training and classification process.
This is because GPUs support complex
mathematical operations and provide high
memory bandwidth. The problem with GPU
accelerators is that these are power hungry.
Hence, GPU used in AI applications on cloud
service, large servers, or in E-Mobility will
consume more power. Whereas, FPGA have
comparatively limited on-chip memory, I/O
bandwidths and computing resources, FPGA
offers better results in terms of low power
consumption, faster data transfer and
flexibility than GPU. The throughput of AI
application can be enhanced by using ASIC as
hardware accelerator. But, ASICs are one time
programmable, requires long development
time and high design cost. Present EDA tools
supports general purpose programming
languages like C or C++ along with hardware
description languages (HDLs) which resolves
the programming concerns and shortens the
development time [38-43].
VI. Motivation of the research
The government of India intends a 30%
penetration in E-Mobility market and is
encouraging production and use of (Hybrid &)
Electric Vehicles. In some industries, the
government had compelled adoption of E-
Mobility. Use of E-Mobility provides health
benefits by reducing air pollution, noise
pollution, and water pollution caused by
gasoline or oil spills from petrol/diesel
vehicles. E-mobility is also much safer in a
possible collision and is less like to roll over.
E-mobility is powered by renewable energy
sources and reduces a country’s fuel exports.
[44-45] International Data Corporation
predicts spends on AI and Machine Learning
(ML) will grow from to $57.6 billion by 2021.
Revenue from AI-based software will reach
$105.8 billion by 2025 is predicted by Tractica
[46].
Tested and reliable datasets played vital role in
data driven approach for accurate prediction of
battery states estimation. Below are a few
public testing datasets, refer [47-49] for more
details on data sets.
Public Testing Datasets
Panasonic 18650PF
Samsung 18650-20R
Prognostic Center of Excellence (PCoE) of NASA
Ames Research Center
Department of
Mechanical Engineering
at McMaster University,
Ontario, Canada.
Center for Advanced
Life Cycle Engineering
at University of
Maryland
Data repository focuses exclusively on data
sets that can be used for development of
prognostic algorithms (prognostic data sets)
3. Challenge in BMS and Possible solutions AI based BMSs are in a developing stage. The
hardware required for BMS are very costly and
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15062
Published by: The Mattingley Publishing Co., Inc.
impractical. Whereas the testing of BMS
should be ideally performed in real operating
conditions of the roads, most of the research is
theoretical and conducted in laboratories. The
limitation is that AI based BMS do not take
into account effect of factors like vibration,
temperature variation, snow, rain, etc. To build
end-user trust, the difference between real
operating conditions and laboratory tests
should be viewed by next research.
Researchers must direct efforts to design an
efficient predictive hardware prototype
algorithm to improve speed, power, and
accuracy of the BMS.
VII. Conclusion
The key take-away from this review is data
driven estimation techniques suitable for
different battery types, and accurate estimation
of SoC, SoH and RUL. On the other hand, the
availability of testing datasets and robust
processing of data in real-time is a limitation.
The concern related to real-time processing
can be minimized with the advancement of the
computational tools for implementing AI based
predictive algorithms in BMS for E-mobility.
The outcomes of this review will contribute to
the future of E-Mobility as engineers and
experts develop an efficient SoC, SoH and
RUL estimation method/algorithm. The AI
based data driven proposed approach should be
chosen over other existing BMS because of
low cost, power consumption, high speed,
reprogrammable logic, and high on-chip
memory storage. The battery states
determination algorithms, developed with the
help of data driven approach, are highly
accurate and are a strong basis for practical
implementation. It will be fascinating to
experience the impact of this technology
applied in automobile, healthcare,
transportation, and robotics industry.
VIII. References
[1] Hannan MA, Lipu MSH, Hussain A, Mohamed
A. A review of lithium-ion battery state of
charge estimation and management system in
electric vehicle applications: challenges and
recommendations. Renew Sust. Energ Rev
2017; 78:834–54.
[2] Chen, J.; Ouyang, Q.; Xu, C.; Su, H. “Neural
network-based state of charge observer design
for lithium-ion batteries”, IEEE Trans. Control
Syst. Technol. 2017, PP, 1–9.
[3] Neeta Khare, Rekha Govil, and Surendra
Kumar Mittal, “A Process of determining State
of Charge and State of Health of a Battery”,
Indian Patent, 813/KOL/2005.
[4] Waag, W.; Fleischer, C.; Sauer, D.U. Critical
review of the methods for monitoring of
lithium-ion batteries in electric and hybrid
vehicles. J. Power Sources 2014, 258, 321–
339.
[5] https://www.cnbc.com/2018/05/30/electric-
vehicles-will-grow-from-3-million-to-125-
million-by-2030-iea.html [accessed on
22/03/2019]
[6] Tang, X.; Liu, B.; Gao, F. State of charge
estimation of lifepo4 battery based on a gain-
classifier observer. Energy Procedia 2017, 105,
2071–2076.
[7] Cheng, K.W.E.; Divakar, B.P.; Wu, H.; Ding,
K.; Ho, H.F. Battery-management system
(BMS) and SOC development for electric
vehicles. IEEE Trans. Veh. Technol. 2011, 60,
76–88.
[8] Cai, Y.;Wang, Q.; Qi,W. D-ukf based state of
health estimation for 18650 type lithium
battery. In Proceedings of the 2016 IEEE
International Conference onMechatronics and
Automation, Harbin, China, 7–10 August
2016, pp. 754–758.
[9] Chaoui, H.; Gualous, H. Adaptive state of
charge estimation of lithium-ion batteries with
parameter and thermal uncertainties. IEEE
Trans. Control Syst. Technol. 2017, 25, 752–
759.
[10] Beir, M.D.; Calado, M.d.R.A.; Pombo, J.A.N.;
Mariano, S.J.P.S. Balancing management
system for improving li-ion batteries capacity
usage and lifespan. In Proceedings of the 2016
IEEE 16th International Conference on
Environment and Electrical Engineering
(EEEIC), Florence, Italy, 7–10 June 2016; pp.
1–6.
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15063
Published by: The Mattingley Publishing Co., Inc.
[11] Feng, Y.; Cao, Z.; Shen,W.; Yu, X.; Han, F.;
Chen, R.;Wu, J. Intelligent battery
management for electric and hybrid electric
vehicles: A survey. In Proceedings of the 2016
IEEE International Conference on Industrial
Technology (ICIT), Taipei, Taiwan, 14–17
March 2016; pp. 1436–1441.
[12] https://spike.global/battery-
module/14s35p/[Last accessed on 22/03/2019]
[13] https://www.automotivepower.com/products/b
attery-management
systems.html?gclid=Cj0KCQjwmpb0BRCBA
RIsAG7y4za5peLhh6QPI4DECn_v5WAnQ8S
XvZh71L6AdPlre2KZjyt4Chqm9lAaAnF3EA
Lw_wcB[Last accessed on 22/03/2019]
[14] https://www.ionenergy.co/platform/battery-
management-systems/fs-ct/[Last accessed on
22/03/2019]
[15] https://www.renesas.com/in/en/[Last accessed
on 22/03/2019]
[16] http://greenecoindia.com/datasheet.html[Last
accessed on 22/03/2019]
[17] https://www.orionbms.com/[Last accessed on
22/03/2019]
[18] Wang, Q.; Jiang, B.; Li, B.; Yan, Y. A critical
review of thermal management models and
solutions of lithium-ion batteries for the
development of pure electric vehicles. Renew.
Sustain. Energy Rev. 2016, 64, 106–128
[19] Muhammad Umair Ali , Amad Zafar , Sarvar
Hussain Nengroo , Sadam Hussain
,Muhammad Junaid Alvi and Hee-Je Kim ,
“Towards a Smarter Battery Management
System for Electric Vehicle Applications: A
Critical Review of Lithium-Ion Battery State of
Charge Estimation”, Energies 2019, 12, 446;
doi:10.3390/en12030446
[20] P. Shen, M. Ouyang, L. Lu, J. Li, and X. Feng,
``The co-estimation of state of charge, state of
health, and state of function for lithium-ion
batteries in electric vehicles,'' IEEE Trans.
Veh. Technol., vol. 67, no. 1, pp. 92-103, Jan.
2018.
[21] Lipu, M.H.; Hannan, M.; Hussain, A.; Hoque,
M.; Ker, P.J.; Saad, M.; Ayob, A. A review of
state of health and remaining useful life
estimation methods for lithium-ion battery in
electric vehicles: Challenges and
recommendations. J. Clean. Prod. 2018, 205,
115–133.
[22] Juan Pablo Rivera-Barrera,Nicolás Muñoz-
Galeano and Henry Omar Sarmiento-
Maldonado,“SoC Estimation for Lithium-ion
Batteries: Review and Future Challenges”
[23] Huang, M.; Kumar, M. Electrochemical
Model-Based Aging Characterization of
Lithium-Ion Battery Cell in Electrified
Vehicles. In Proceedings of the ASME 2018
Dynamic Systems and Control Conference
(DSCC 2018), Atlanta, GA, USA, 30
September–3 October 2018; Volume 3, pp. 1–
10.
[24] Bijender Kumar, Neeta Khare ,P.K.
Chaturvedi, “FPGA Design Scheme for Battery
SoC & SoH Algorithms for Advanced BMS”,
[25] https://www.sciencedirect.com/topics/engineeri
ng/battery-management-system [Last accessed
on 22/03/2019]
[26] https://www.maximintegrated.com/en/design/te
chnicaldocuments/appnotes/6/6638.html?s_kw
cid=AL!8732!3!422714005632!b!!g!!&gclid=
Cj0KCQjw9tbzBRDVARIsAMBplx9aLoxNnt
xvYXLCBnUsksm4IMuls98z8ptrV9Xdm5IX
XYmRGdBhuPIaAuo0EALw_wcB [Last
accessed on 22/03/2019]
[27] https://www.grandviewresearch.com/industry-
analysis/battery-management-system-bms-
market [Last accessed on 22/03/2019]
[28] Chen, Z.; Mi, C.C.; Fu, Y.; Xu, J.; Gong, X.
Online battery state of health estimation based
on Genetic Algorithm for electric and hybrid
vehicle applications. J. Power Sources 2013,
240, 184–192.
[29] Ma, Y.; Ru, J.; Yin, M.; Chen, H.; Zheng, W.,
“Electrochemical modeling and parameter
identification based on bacterial foraging
optimization algorithm for lithium-ion
batteries”. J. Appl. Electrochem. 2016, 46,
1119–1131.
[30] Ali, M.U.; Habib, B.; Iqbal, M. Fixed head
short term hydro thermal scheduling using
improved particle swarm optimization.
Nucleus (Islamabad) 2015, 52, 107–114.
[31] Singh, P.; Vinjamuri, R.; Wang, X.; Reisner,
D. Design and implementation of a fuzzy
logic-based state of charge meter for li-ion
batteries used in portable defibrillators. J.
Power Sources 2006, 162, 829–836.
[32] Hannan, M.A.; Lipu, M.S.H.; Hussain, A.;
Saad, M.H.; Ayob, A. Neural network
approach for estimating state of charge of
lithium-ion battery using backtracking search
algorithm. IEEE Access 2018, 6, 10069–
10079.
[33] 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
May – June 2020
ISSN: 0193-4120 Page No. 15053 – 15064
15064
Published by: The Mattingley Publishing Co., Inc.
vehicle) traction batteries. Energy 2015, 80,
350–360.
[34] Anton, J.A.; Nieto, P.G.; Viejo, C.B.; Vilan,
J.V. Support vector machines used to estimate
the battery state of charge. IEEE Trans. Power
Electron. 2013, 28, 5919–5926.
[35] Antón, J.C.Á.; Nieto, P.J.G.; de Cos Juez, F.J.;
Lasheras, F.S.; Viejo, C.B.; Gutiérrez, N.R.
Battery state-of charge estimator using the
mars technique. IEEE Trans. Power Electron.
2013, 28, 3798–3805.
[36] Esteva, A.; Kuprel, B.; Novoa, R.A.; Ko, J.;
Swetter, S.M.; Blau, H.M.; Thrun, S.
Dermatologist-level classification of skin
cancer with deep neural networks. Nature
2017, 542, 115–118.
[37] Zhang, Y.; Xiong, R.; He, H.; Pecht, M.G.
Long short-term memory recurrent neural
network for remaining useful life prediction of
lithium-ion batteries. IEEE Trans. Veh.
Technol. 2018, 67, 5695–5705.
[38] A. Shawahna, S. M. Sait & A. El-Maleh,
“FPGA-Based Accelerators of Deep Learning
Networks for Learning and Classification: A
Review”, IEEE Access, vol. 7, pp. 7823-7859,
2019
[39] A. Vasudevan, A. Anderson, and D. Gregg,
“Parallel multi channel convolution using
general matrix multiplication,” in Application-
specific Systems, Architectures and Processors
(ASAP), 2017 IEEE 28th International
Conference on. IEEE, 2017, pp. 19–24.
[40] J. Misra and I. Saha, “Artificial neural
networks in hardware: A survey of two decades
of progress,” Neurocomputing, vol. 74, no. 1-3,
pp. 239–255, 2010.
[41] L. Du, Y. Du, Y. Li, J. Su, Y.-C. Kuan, C.-C.
Liu, and M.-C. F. Chang, “A reconfigurable
streaming deep convolutional neural network
accelerator for internet of things,” IEEE
Transactions on Circuits and Systems Regular
Papers, vol. 65, no. 1, pp. 198–208, 2018.
[42] A. Putnam, A. M. Caulfield, E. S. Chung, D.
Chiou, K. Constantinides, J. Demme, H.
Esmaeilzadeh, J. Fowers, G. P. Gopal, J. Gray
et al., “A reconfigurable fabric for accelerating
large-scale datacenter services,” ACM
SIGARCH Computer Architecture News, vol.
42, no. 3, pp. 13–24, 2014. 3, 13
[43] Y. Liang, K. Rupnow, Y. Li, D. Min, M. N.
Do, and D. Chen, “High-level synthesis:
productivity, performance, and software
constraints,” Journal of Electrical and
Computer Engineering, vol. 2012, p. 1, 2012.
[44] https://www.weforum.org/agenda/2019/10/how
-can-india-transition-to-electric-vehicles-heres-
a-roadmap [accessed on 9th April, 2020]
[45] https://www.ergon.com.au/network/smarter-
energy/electric-vehicles/benefits-of-electric-
vehicles [accessed on 9th April, 2020]
[46] https://www.capgemini.com/consultingde/wpc
ontent/uploads/sites/32/2017/09/artificialintelli
gence-report.pdf [accessed on 20th Feb 2020]
[47] https://www2.deloitte.com/content/dam/Deloitt
e/global/Images/infographics/technologymedia
telecommunications/gxdeloitte-tmt-2018-
predictions-full-report.pdf [accessed on 20th
Feb 2020]
[48] https://semiengineering.com/using-fpgas-for-
ai/ [accessed on 20th Feb 2020]
[49] https://www.apriorit.com/dev-blog/586-fpgas-
for-ai [accessed on 20th Feb 2020]
