International Journal of Power Electronics and Drive Systems (IJPEDS)

Vol. 15, No. 3, September 2024, pp. 1942~1950

ISSN: 2088-8694, DOI: 10.11591/ijpeds.v15.i3.pp1942-1950  1942

Dimensionality reduced deep learning-based state of health

estimation of Lithium-Ion batteries using standard dataset

Vimala Channapatna Srikantappa, Seshachalam Devarakonda

Department of Electronics and Communication Engineering, BMS College of Engineering, Bangalore, India

Article Info ABSTRACT

Article history: Lithium-Ion batteries are used in everyday DC equipment’s, electric vehicle
technology, and microgrid technology. The necessity to verify the battery's
Received Aug 23, 2023 state is crucial for the dependent apps to continue operating without
Revised Jan 11, 2024 interruption due to advancements in battery technology & adaption. This
Accepted Feb 7, 2024 study uses dimension decreases in input attributes along with deep learning
methods to determine the state of health of lithium-Ion batteries (LIB).
principal component analysis (PCA), a deep learning technique, is combined
Keywords: with recurrent neural networks (RNN) to reduce dimensionality. For the
purpose of evaluating the effectiveness of the dimensionality reduction used
Lithium-Ion battery in the data, the state of health (SOH) estimate using the RNN with and without
Mean square error PCA is compared. The use of PCA-powered RNNs using mean square error
Principal component analysis (MSE) as the loss function throughout the training and testing stages of state-
Recurrent neural network of-health (SOH) estimation showed great performance in terms of loss. This
State of health was seen during the training and testing processes' respective testing and
validation phases.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Vimala Channapatna Srikantappa
Department of Electronics and Communication Engineering, BMS College of Engineering
Bangalore, India
Email: vimalasrikantappa@gmail.com

1. INTRODUCTION
In recent years, lithium-Ion batteries or LIB have dominated the electrical storage sector. The
condition of the battery must be examined to ensure uninterrupted operation of the devices that rely on batteries
for power. The LIB is more prevalent on the market for electric vehicles. For either routine maintenance or
replacement, each application requires information on the battery's condition. By carefully calculating the state
of charge (SOC), that stands for the amount of energy stored in the batteries, while maintaining track of the
SOH, which represents the general condition and degree of degradation of the batteries, it is possible to
implement into practice efficient methods which improve the operational life expectancy of the batteries. The
single particle model is a simplified battery electrochemical model that is used in this study to track changes
in lithium-ion content with varied loaded currents and calculate the number of recyclable lithium-ions at
various aging degrees. In order to ensure both operational speed and optimization accuracy, the model's
parameters are determined using a hybrid technique called the hybrid coyote optimization technique and the
grey wolf optimizer. With mean absolute percentage errors for battery capacity below 2% and SOC prediction
below 1.8%, the suggested technique exhibits great accuracy and resilience. SOC forecast time is less than 2
seconds during federal urban driving cycles [1].
For the purpose of assessing the state of health (SOH) of batteries, the unscented particle filter (UPF)
algorithm integrates recent measurements to predict the battery health and expresses uncertainty. A UPF-based
estimator accomplishes accurate SOH evaluation with less than 5% maximum estimation error and exhibits

Journal homepage: http://ijpeds.iaescore.com

Int J Pow Elec & Dri Syst ISSN: 2088-8694  1943

robustness across multiple lithium-ion battery types by collecting an online health indicator (HI) from
measurable parameters and using it in a state-space model [2]. In order to accurately estimate the SOC and
SOH of batteries in electric vehicles, a dual extended Kalman filter (EKF) and a fractional-order model (FOM)
are used [3]. In order to overcome the difficulty of the batteries' complex aging mechanism, a novel method
for precisely predicting the remaining useful life (RUL) and determining the state-of-health (SOH) of energy
storage systems using LIB has been developed [4]. This method establishes a support vector regression-based
battery SOH state-space model and makes use of a particle filter to estimate impedance degradation parameters.
In order to improve accuracy and explore the connection between internal parameters and battery states, a joint
extended Kalman filter-recursive-least squares method has been proposed [5] for state-of-charge estimation.
This method estimates the SOH of lithium-ion batteries. Particle swarm optimization-least square support
vector regression is used after parameter identification to produce reliable and accurate SOH estimation with
good generalizability. The suggested method for state-of-health determination has undergone experimental
tests on lithium iron phosphate batteries at various aging stages, and the results show that it is highly accurate
and suitable. The remaining usable life (RUL) of a battery, for instance, can be predicted using a long short-
term memory (LSTM) network based on capacity degradation [6]. A convolutional neural network and a long
short-term memory unit are combined to provide a deep learning method for online capacity estimate of
lithium-ion batteries.
The suggested algorithm achieves accurate capacity estimation with an absolute error of fewer than
0.021 Ampere-hour and 0.11 Ampere-hour for two battery types, enabling quick online capacity estimation.
Partial charging voltage and current data are used in the proposed algorithm without wide pre-processing,
leading to simpler training preparation and lower computational intensity [7]. Calculations that simulate
various battery types are used in model-based SOH estimates. Data-driven approaches train machine learning
algorithms using real-time data from these calculations. By contrasting the factors comprising advantages,
estimation error, and downsides of each estimating method, an examination of such methods is given in the
literature [8]. In order to calculate ECM parameters as well as battery SOC using dual time scales, a multi-
scale extended Kalman filter relying on the first-order equivalent-circuit model (ECM) is used [9]. This
emphasizes the significance of optimizing the excitation current for precise parameter and state estimation.
The study illustrates the effect of excitation current selection on estimation accuracy using Cramer-
Rao bound analysis while taking voltage noise, current amplitude, and frequency into account. It then offers
recommendations for creating battery current profiles that produce better SOC and SOH estimation
performance. In order to estimate the SOC and SOH of power batteries used in electric vehicles accurately, the
article [10] offers a hierarchical estimation model that takes into consideration the current rate. The proposed
method significantly improves SOC and SOH estimation accuracy through the use of a fractional-order model,
data-driven parameter identification, along with a multiscale dual extended Kalman filter (DEKF). It
outperforms conventional DEKF addresses by 35.8% to 36.5% for SOC estimation as well as 34.8% to 43.1%
for SOH estimation under various current conditions.
In order to improve the performance of battery management systems in LIB used in diverse
applications, the study [11] suggests a co-estimation method for SOC and SOH based on fractional-order
calculus. The plan comprises a dual fractional-order extended Kalman filter for simultaneous estimate of SOC
and SOH, as well as a fractional-order equivalent circuit model parameterized utilizing a hybrid genetic
algorithm/particle swarm optimization technique. The usefulness of the proposed approach in battery
management tasks is validated by experimental results showing that it can estimate SOC and SOH having
maximum steady-state errors of less than 1% and is resilient to battery aging. For estimating SOC and SOH,
there is significant interest in artificial intelligence and machine learning (ML), including feed forward neural
network (FNN), recurrent neural network (RNN), support vector machine (SVM) and radial basis function
(RBF) and Hamming networks. Comparative analysis of these approaches takes into account elements like
input and output quality, test conditions, battery types, as well as accuracy, placing emphasis on the significance
of multiple training iterations, similar network structures, and identical data for accurate comparisons between
estimation techniques [12]. Accurate capacity prediction of batteries in ESS is critical for safe operations, and
this paper offers a deep learning-based battery management system (BMS) that estimates battery health and
capacity using multiple channel charging profiles (MCCPs) [13]. Deep domain adversarial network (DDAN)
is a model that includes a deep feature generator, dense bidirectional gated recurrent unit, as well as
unsupervised feature alignment metrics to improve feature learning and knowledge transfer, and it has been
shown to be effective in SOH estimation on a battery dataset [14]. In real-time simulations and hardware tests,
a successful approach to forecast battery aging & estimating battery health.
The research project develops and implements mathematical models on a standalone hardware
platform to determine an optimal machine learning-based method for SOH and RUL estimation, taking into
account factors such as SOC, discharge voltage transfer features, internal resistance, and capacity. According
to experimental findings, although a long short-term memory neural network efficiently predicts battery RUL
with an accuracy of 10 cycles, a deep neural network predicts SOH with an acceptable error rate of 5%. The
Dimensionality reduced deep learning-based state of health … (Vimala Channapatna Srikantappa)
1944  ISSN: 2088-8694

suggested method provides an ideal answer for estimating battery life by showcasing multiple machine learning
models on a real hardware platform [15]. Auto-regression nested sequence (ARNS), described in literature
[16], is a novel data-driven approach that effectively aggregates channel- and cycle-level information whereas
including relaxation impacts for peak prediction. When applied to NASA and CALCE datasets, ARNS
outperforms existing methods, especially during peak periods spanning multiple SOH states and cycles.
Indirect health indicators (IHIs), which reflect battery capacity loss, can be extracted from voltage, current, and
temperature curves throughout the charging and discharging processes, according to literature [17]. High
estimation accuracy is achieved by selecting the significant IHIs using (PCA) and using them as inputs for
SOH estimation using gaussian process regression (GPR). The development of a unique SOH [18] estimate
method is based on the battery pack's behavior in active charge balancing (ACB), and has a positive correlation
with SOH. We outperform state-of-the-art methods by leveraging this metric and other cell parameter to train
a Random Forest (RF) regression estimator, which leads to accurate SOH estimate with 1.94% accuracy for
capacity as well as 4.28% accuracy for resistance. A lithium-ion battery (LIB) online SOH monitoring
technique is presented in article [19] that makes use of the battery's charge-discharge characteristic in real time.
The proposed method demonstrates accurate real-time prediction of LIB SOH using NASA
experimental data, with strengthened accuracy due to the use of techniques such as complete ensemble
empirical mode decomposition with adaptive noise (CEEMDAN), logistic regression based on sliding time
window (LR-STW), Kalman filter (KF), and radial basis function neural network (RBFNN), as well as (PCA)
to analyze the relationship between the charge-discharge feature and SOH. 124 commercial lithium iron
phosphate/graphite cells with varied cycle lifetimes were cycled under fast-charging conditions, and a
comprehensive dataset was created [20]. The potential of combining intentional data generation with data-
driven modeling to comprehend complex dynamical systems is demonstrated by predictive models that achieve
a 9.1% test error for quantitatively predicting cycle life as well as a 4.9% test error for classifying cycle life
into two groups via using machine-learning tools to discharge voltage curves from early cycles. The changing
pattern of the charging current during the constant-voltage (CV) phase serves as the foundation for developing
an approach to detect battery SOH [21]. By establishing a measurable link between the normalized battery
capacity and the current's time constant, an accurate estimation of SOH becomes achievable. The time constant
of the CV charging current is acknowledged as a dependable indicator associated with battery aging. The
proposed approach may successfully show the SOH of several batteries under a 2.5% error bound, according
to experimental results.
The idea of effective battery capacitance is first mentioned in literature [22] as a way to forecast the
SOH of certain cells. A linear relationship among maximum effective capacitance and SOH is found and
validated using data from several Toyota Prius battery packs. The maximum effective capacitance is
determined through a voltage versus charge curve analysis and acts as a signal for end-of-life or catastrophic
failure of battery modules. A innovative method for predicting long-term remaining usable life (RUL) and
estimating short-term (SOH) online is presented in literature [23] by combining particle filtering (PF) and a
degradation model based on Brownian motion (BM). The review paper [24] discusses the viability and
economics of data-driven methods for estimating battery health in practical applications by utilizing
developments in "Big Data" analytics using statistical/computational tools. The authors classify these
techniques according to the models and algorithms that underlie them, talk about their benefits and drawbacks,
and explore the difficulties of managing the real-time battery health. This review intends to advance data-
driven battery health estimate and forecasting across all technology readiness levels by offering insights into
commercial technology decisions and research goals. An adaptable health estimation model is presented in the
literature [25], which demonstrates the effectiveness of a Bayesian non-parametric approach employing
Gaussian process regression for gauging capacity degradation across diverse usage scenarios. The author
effectively anticipates long-term capacity fading using this technique on the NASA randomized battery usage
dataset, attaining an optimal normalized root MSE of 4.3% alongside accurate computation of predictive
uncertainty. Various literatures that discuss on the different battery performance evaluation and different
chemical composition, novel methods for performance evaluation are considered [25]-[43].
Though the SOH implementation for batteries is used to build machine learning algorithms in the
prior research, dimensionality reduction of the input characteristics using techniques for feature extraction is
not addressed. In order to achieve the SOH estimation using the SVM implementation, this article uses the
PCA-based feature extraction method. The implementation process is covered in section 2. Section 3 includes
a description of the results and discussion, followed by the conclusion and references.

2. DIMENSIONALITY REDUCED RNN-BASED SOH ESTIMATION

The variables are prepared for PCA analysis to determine the total number of primary components to
be used for SOH estimate. The dimensionality reduction method utilized in machine learning applications is

Int J Pow Elec & Dri Syst, Vol. 15, No. 3, September 2024: 1942-1950
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1945

(PCA). By encapsulating every variable in a small number of variables, principal components (PC) can
condense a sizable number of variables. All variables are linearly combined to create the principal components
(PC), which are then generated. The SVM decision-making algorithm receives its input from the PCs.
A collection of experimental data called the NASA dataset for LIB can be used to calculate a battery’s
SOH. SOH stands for the battery's lifetime energy storage and delivery capacity. This dataset, which was
produced by the NASA Ames research center and the jet propulsion laboratory, contains measurements of
several lithium-ion battery properties made across a number of charge and discharge cycles. The dataset
contains data on the voltage, current, temperature, capacity, and impedance of the batteries. It also includes
information about the battery's structure and content, as well as the testing environment. Different testing
techniques, such as accelerated aging and real-world cycling, were used to gather the results. As shown in
Table 1, the dataset contains variables that were noticed during the charging and discharging cycles. The
experiment is done on the data acquisition test bed using the lithium-ion battery's charging and discharging
cycles. Rechargeable batteries in the 18650 size that are readily available on the market are employed. Table 2
lists the characteristics noticed for the impedance operation.
Models and algorithms are created using the dataset, which comprises all the variables gleaned from
the charge-discharge cycles, in order to anticipate battery performance and calculate the battery's remaining
usable life. LIB can be designed better and used more effectively in a variety of applications with the help of
these models. After performing a PCA-based dimensionality reduction on the dataset, the data-driven SOH
estimation is implemented using a recurrent neural network. To compare the outcomes of PCA extracted and
non-PCA implementation, model loss is obtained for both. Figure 1 shows the block diagram for the SOH
estimate using PCA and RNN.

Table 1. Data structure for charge, discharge operations

Attribute Attribute explanation Unit
Voltage_measured Battery terminal voltage Volts
Current_measured Battery output current Amps
Temperature_measured Battery temperature Degree Centigrade
Current_charge Current measured at the charger Amps
Voltage_charge Voltage measured at the charger Volts
Time Time vector for the cycle Secs
Capacity (Only for Discharge operation) Battery capacity (Ahr) for discharge till 2.7V Ahr

Figure 1. Dimensionality reduced SOH estimation using RNN

The block diagram shows how the specified CSV files are used to read the dataset. For the
implementation of SOH estimation, the attributes are divided into input and output. Following the application
of PCA, the data has been normalized utilizing the min-max scaler normalization method and trained on the
RNN network having the following structure. To carry out the dimensionality-reduced SOH estimation
utilizing RNN, Python-based code has been created. The MSE is used as the objective for convergence in the

Dimensionality reduced deep learning-based state of health … (Vimala Channapatna Srikantappa)

1946  ISSN: 2088-8694

training and testing of SOH estimation using RNN. To appreciate the effectiveness of the approach, the
comparative MSE convergence of the RNN without PCA and RNN with PCA is derived.

Table 2. Data structure for impedance operations

Attribute Attribute explanation Unit
Sense_current Current in sense branch Amps
Battery_current Current in battery branch Amps
Current_ratio The ratio of the above currents NIL
Battery_impedance Battery impedance (Ohms) computed from raw data Ohms
Rectified_impedance Calibrated and smoothed battery impedance (Ohms) Ohms
Re Estimated electrolyte resistance (Ohms) Ohms
Rct Estimated charge transfer resistance (Ohms) Ohms

3. RESULTS AND DISCUSSION

Figure 2 shows the box plots for the attributes: Figure 2(a): capacity, Figure 2(b): voltage measured,
Figure 2(c): current measured, Figure 2(d): temperature measured, Figure 2(e): current load, and Figure 2(f):
voltage load. Using a box plot, the input attributes are represented in order to determine their range of values.
Figure 2 illustrates the attributes of the battery charging procedure alongside the span of values encompassing
these characteristics.

(a) (b)

(c) (d)

(e) (d)

Figure 2. Box plots for the attributes of (a) capacity, (b) voltage measured, (c) current measured,
(d) temperature measured, (e) current load, and (f) voltage load

Int J Pow Elec & Dri Syst, Vol. 15, No. 3, September 2024: 1942-1950
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1947

To understand the battery's capacity profiling, the graph analyzes the data for SOH versus cycle.
Figures 3(a) and 3(b) show the SOH versus cycle profile and the capacity versus cycle, respectively. Figures
3(a) and 3(b) exhibit the instantaneous SOH and capacity values along with the average SOH and capacity
values. The Figure 3 clearly shows how the capacity and SOH have declined.
For efficient memory usage in SOH prediction training and testing, the attributes for SOH estimation
must be feature minimized. The correlation diagram provides information about how closely the attributes are
related; the higher the correlation, the less frequently an attribute is used. Although there is a higher correlation
among all the distinctive factors, the ambient temperature has a larger correlation. For PCA analysis, all seven
attributes are kept in place. Since PCA is used to extract variance from characteristics, it is calculated as a
linear combination of attributes, and as a result, the implementation of PCA shows a reduction in
dimensionality. Different percentage variances are contributed by the principal components (PCs). However,
the first and second PCs, PC1 and PC2, are the principal variations that account for the majority of the variance
in the characteristics. Figure 4(a) shows the PC graph, whereas Figure 4(b) shows the correlation graph between
the qualities. Table 3 shows the PCs and their proportional contribution to reflecting the variance of the
characteristics. It is evident from the chart that the top three or four PCs' contributions account for the majority
of the differences in all the parameters.
To create the model loss versus epoch graphs, the RNN technique is used in conjunction with the
original attributes and the extracted PCs, which may be all of them or only three or four of them. To compare
the training convergence and testing model loss vs the epochs, the model loss graph with the train and test
operation on the dataset is obtained. Model loss in machine learning is a metric indicating how effectively a
model can forecast the target variable given a specific set of input data. The difference between the projected
value and the actual value is used to calculate it. Figure 5 shows the model loss responses for the various
instances addressed. Figures 5(a) and 5(b) show that the model loss response is identical when the whole set
of characteristics and all PCs are used for the SOH estimation. While the model loss while employing 4 PCs
and 3 PCs is comparable, as shown in Figures 5(c) and 5(d), respectively.
When using 4 Pcs for the SOH estimation, the root mean square of 0.1379859619348809 was
achieved. When the characteristics are employed directly, the SOH estimation achieved using PCA and RNN
essentially yields equal values of model loss. As a result, when dimensionality is reduced using PCA, the model
loss values are close to those obtained when PCA is not used. As a result, as compared to an implementation
that does not use PCA, the SOH estimation approach uses less RAM.

(a) (b)

Figure 3. Profiling of SOH and capacity: (a) SOH versus cycle and (b) capacity versus cycle

Table 3. PC contribution
PC number Eigen value Percentage
0 4.172993 59.614191
1 1.350306 19.290089
2 0.910773 13.011049
3 0.385525 5.507497
4 0.138061 1.972299
5 0.026477 0.378236
6 0.015865 0.226640

Dimensionality reduced deep learning-based state of health … (Vimala Channapatna Srikantappa)

1948  ISSN: 2088-8694

(a) (b)

Figure 4. PCA implementation: (a) correlation diagram and (b) principal components (PCs)

(a) (b)

(c) (d)

Figure 5. Model loss response: (a) without PCA, (b) with all PCs, (c) with 3 PCs, and (d) with 4 PCs

4. CONCLUSION
The NASA dataset, a common dataset, is utilized in the machine learning context for SOH estimation.
The PCA is used to apply the dimensionality reduction technique to the input attributes. Using the PCs acquired
via PCA, the RNN from the Deep Learning technique is utilized to estimate the SOH. The model losses were
provided by the implementation to be quite close to those when all attributes are utilized. Even when

Int J Pow Elec & Dri Syst, Vol. 15, No. 3, September 2024: 1942-1950
Int J Pow Elec & Dri Syst ISSN: 2088-8694  1949

dimensionality reduction using PCA is used, the performance of the SOH estimation is still in a good range.
The algorithm's RMSE results show that the PCA with RNN algorithm has increased memory economy while
preserving performance equivalence to the non-PCA implementation.

REFERENCES
[1] X. Sun, Q. Chen, L. Zheng, and J. Yang, “Joint Estimation of State-of-Health and State-of-Charge for Lithium-Ion Battery Based
on Electrochemical Model Optimized by Neural Network,” IEEE Journal of Emerging and Selected Topics in Industrial
Electronics, vol. 4, no. 1, pp. 168–177, 2022, doi: 10.1109/jestie.2022.3148031.
[2] D. Liu, X. Yin, Y. Song, W. Liu, and Y. Peng, “An on-line state of health estimation of lithium-ion battery using unscented particle
filter,” IEEE Access, vol. 6, pp. 40990–41001, 2018, doi: 10.1109/ACCESS.2018.2854224.
[3] L. Ling and Y. Wei, “State-of-Charge and State-of-Health Estimation for Lithium-Ion Batteries Based on Dual Fractional-Order
Extended Kalman Filter and Online Parameter Identification,” IEEE Access, vol. 9, pp. 47588–47602, 2021, doi:
10.1109/ACCESS.2021.3068813.
[4] J. Wei, G. Dong, and Z. Chen, “Remaining Useful Life Prediction and State of Health Diagnosis for Lithium-Ion Batteries Using
Particle Filter and Support Vector Regression,” IEEE Transactions on Industrial Electronics, vol. 65, no. 7, pp. 5634–5643, 2018,
doi: 10.1109/TIE.2017.2782224.
[5] D. Yang, Y. Wang, R. Pan, R. Chen, and Z. Chen, “State-of-health estimation for the lithium-ion battery based on support vector
regression,” Applied Energy, vol. 227, pp. 273–283, 2018, doi: 10.1016/j.apenergy.2017.08.096.
[6] K. Park, Y. Choi, W. J. Choi, H. Y. Ryu, and H. Kim, “LSTM-Based Battery Remaining Useful Life Prediction with Multi-Channel
Charging Profiles,” IEEE Access, vol. 8, pp. 20786–20798, 2020, doi: 10.1109/ACCESS.2020.2968939.
[7] Q. Xue, J. Li, Z. Chen, Y. Zhang, Y. Liu, and J. Shen, “Online Capacity Estimation of Lithium-Ion Batteries Based on Deep
Convolutional Time Memory Network and Partial Charging Profiles,” IEEE Transactions on Vehicular Technology, vol. 72, no. 1,
pp. 444–457, 2023, doi: 10.1109/TVT.2022.3205439.
[8] D. N. T. How, M. A. Hannan, M. S. Hossain Lipu, and P. J. Ker, “State of Charge Estimation for Lithium-Ion Batteries Using Model-
Based and Data-Driven Methods: A Review,” IEEE Access, vol. 7, pp. 136116–136136, 2019, doi: 10.1109/ACCESS.2019.2942213.
[9] Z. Song, X. Wu, X. Li, J. Sun, H. F. Hofmann, and J. Hou, “Current Profile Optimization for Combined State of Charge and State
of Health Estimation of Lithium Ion Battery Based on Cramer-Rao Bound Analysis,” IEEE Transactions on Power Electronics,
vol. 34, no. 7, pp. 7067–7078, 2019, doi: 10.1109/TPEL.2018.2877294.
[10] P. Xu, X. Hu, B. Liu, T. Ouyang, and N. Chen, “Hierarchical Estimation Model of State-of-Charge and State-of-Health for Power
Batteries Considering Current Rate,” IEEE Transactions on Industrial Informatics, vol. 18, no. 9, pp. 6150–6159, 2022, doi:
10.1109/TII.2021.3131725.
[11] X. Hu, H. Yuan, C. Zou, Z. Li, and L. Zhang, “Co-Estimation of State of Charge and State of Health for Lithium-Ion Batteries
Based on Fractional-Order Calculus,” IEEE Transactions on Vehicular Technology, vol. 67, no. 11, pp. 10319–10329, 2018, doi:
10.1109/TVT.2018.2865664.
[12] C. Vidal, P. Malysz, P. Kollmeyer, and A. Emadi, “Machine Learning Applied to Electrified Vehicle Battery State of Charge and
State of Health Estimation: State-of-the-Art,” IEEE Access, vol. 8, pp. 52796–52814, 2020, doi: 10.1109/ACCESS.2020.2980961.
[13] N. Khan, F. U. M. Ullah, Afnan, A. Ullah, M. Y. Lee, and S. W. Baik, “Batteries State of Health Estimation via Efficient Neural
Networks with Multiple Channel Charging Profiles,” IEEE Access, vol. 9, pp. 7797–7813, 2021, doi: 10.1109/ACCESS.2020.3047732.
[14] Z. Ye and J. Yu, “State-of-Health Estimation for Lithium-Ion Batteries Using Domain Adversarial Transfer Learning,” IEEE
Transactions on Power Electronics, vol. 37, no. 3, pp. 3528–3543, 2022, doi: 10.1109/TPEL.2021.3117788.
[15] P. Venugopal et al., “Analysis of Optimal Machine Learning Approach for Battery Life Estimation of Li-Ion Cell,” IEEE Access,
vol. 9, pp. 159616–159626, 2021, doi: 10.1109/ACCESS.2021.3130994.
[16] T. Pham et al., “ARNS: A Data-Driven Approach for SoH Estimation of Lithium-Ion Battery Using Nested Sequence Models With
Considering Relaxation Effect,” IEEE Access, vol. 10, pp. 117067–117083, 2022, doi: 10.1109/ACCESS.2022.3217478.
[17] D. Y. Reddy, B. Routh, A. Patra, and S. Mukhopadhyay, “Gaussian Process Regression based State of Health Estimation of Lithium-
Ion Batteries using Indirect Battery Health Indicators,” 2021, doi: 10.1109/ICPHM51084.2021.9486519.
[18] A. Lamprecht, M. Riesterer, and S. Steinhorst, “Random Forest Regression of Charge Balancing Data: A State of Health Estimation
Method for Electric Vehicle Batteries,” 2020 International Conference on Omni-Layer Intelligent Systems, COINS 2020, 2020, doi:
10.1109/COINS49042.2020.9191421.
[19] L. Mao, H. Hu, J. Chen, J. Zhao, K. Qu, and L. Jiang, “Online State-of-Health Estimation Method for Lithium-Ion Battery Based
on CEEMDAN for Feature Analysis and RBF Neural Network,” IEEE Journal of Emerging and Selected Topics in Power
Electronics, vol. 11, no. 1, pp. 187–200, 2023, doi: 10.1109/JESTPE.2021.3106708.
[20] K. A. Severson et al., “Data-driven prediction of battery cycle life before capacity degradation,” Nature Energy, vol. 4, no. 5, pp.
383–391, 2019, doi: 10.1038/s41560-019-0356-8.
[21] J. Yang, B. Xia, W. Huang, Y. Fu, and C. Mi, “Online state-of-health estimation for lithium-ion batteries using constant-voltage
charging current analysis,” Applied Energy, vol. 212, pp. 1589–1600, 2018, doi: 10.1016/j.apenergy.2018.01.010.
[22] P. Leijen, D. A. Steyn-Ross, and N. Kularatna, “Use of Effective Capacitance Variation as a Measure of State-of-Health in a Series-
Connected Automotive Battery Pack,” IEEE Transactions on Vehicular Technology, vol. 67, no. 3, pp. 1961–1968, 2018, doi:
10.1109/TVT.2017.2733002.
[23] G. Dong, Z. Chen, J. Wei, and Q. Ling, “Battery health prognosis using brownian motion modeling and particle filtering,” IEEE
Transactions on Industrial Electronics, vol. 65, no. 11, pp. 8646–8655, 2018, doi: 10.1109/TIE.2018.2813964.
[24] Y. Li et al., “Data-driven health estimation and lifetime prediction of lithium-ion batteries: A review,” Renewable and Sustainable
Energy Reviews, vol. 113, 2019, doi: 10.1016/j.rser.2019.109254.
[25] R. R. Richardson, M. A. Osborne, and D. A. Howey, “Battery health prediction under generalized conditions using a Gaussian
process transition model,” Journal of Energy Storage, vol. 23, pp. 320–328, 2019, doi: 10.1016/j.est.2019.03.022.
[26] K. Omiloli, A. Awelewa, I. Samuel, O. Obiazi, and J. Katende, “State of charge estimation based on a modified extended Kalman
filter,” International Journal of Electrical and Computer Engineering, vol. 13, no. 5, pp. 5054–5065, 2023, doi:
10.11591/ijece.v13i5.pp5054-5065.
[27] Y. Boujoudar et al., “Fuzzy logic-based controller of the bidirectional direct current to direct current converter in microgrid,” International
Journal of Electrical and Computer Engineering, vol. 13, no. 5, pp. 4789–4797, 2023, doi: 10.11591/ijece.v13i5.pp4789-4797.
[28] N. K. Trung and N. T. Diep, “Online parameter identification for equivalent circuit model of lithium-ion battery,” Indonesian
Journal of Electrical Engineering and Computer Science, vol. 31, no. 1, pp. 151–159, 2023, doi: 10.11591/ijeecs.v31.i1.pp151-159.

Dimensionality reduced deep learning-based state of health … (Vimala Channapatna Srikantappa)

1950  ISSN: 2088-8694

[29] I. Boumedra, A. Diani, K. El Khadiri, A. Tahiri, M. O. Jamil, and H. Qjidaa, “High efficiency multi power source control constant
current/constant voltage charger lithium-ion battery based on the buck converter,” International Journal of Electrical and Computer
Engineering, vol. 13, no. 1, pp. 207–217, 2023, doi: 10.11591/ijece.v13i1.pp207-217.
[30] Z. E. Dallalbashi, S. Alhayalir, M. J. Mnati, and A. A. A. Alhayali, “Low-cost battery monitoring circuit for a photovoltaic system
based on LoRa/LoRaWAN network,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 29, no. 2, pp. 669–
677, 2023, doi: 10.11591/ijeecs.v29.i2.pp669-677.
[31] C. O. Omeje and C. U. Eya, “A comparative braking scheme in auto-electric drive systems with permanent magnet synchronous machine,”
International Journal of Applied Power Engineering, vol. 11, no. 4, pp. 251–263, 2022, doi: 10.11591/ijape.v11.i4.pp251-263.
[32] R. Ranom, R. S. Bacho, and S. N. A. S. A. Jamal, “The effect of electrolyte parameter variation upon the performance of lithium
iron phosphate (LiFePO4),” Indonesian Journal of Electrical Engineering and Computer Science, vol. 28, no. 1, pp. 58–66, 2022,
doi: 10.11591/ijeecs.v28.i1.pp58-66.
[33] H. Arshada, S. A. Zulkiflia, and M. H. Khan, “Adaptive filter algorithms for state of charge estimation methods: A comprehensive
review,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 26, no. 3, pp. 1360–1367, 2022, doi:
10.11591/ijeecs.v26.i3.pp1360-1367.
[34] I. Chaoufi, O. Abdelkhalek, and B. Gasbaoui, “State of charge estimation of lithium-ion batteries using adaptive neuro fuzzy inference
system,” IAES International Journal of Artificial Intelligence, vol. 11, no. 2, pp. 473–484, 2022, doi: 10.11591/ijai.v11.i2.pp473-484.
[35] C. N. Van and T. N. Vinh, “State of charge estimation for lithium-ion batteries connected in series using two sigma point Kalman
filters,” International Journal of Electrical and Computer Engineering, vol. 12, no. 2, pp. 1334–1349, 2022, doi:
10.11591/ijece.v12i2.pp1334-1349.
[36] K. Yamaguchi and K. Iida, “Auto tuning of frequency on wireless power transfer for an electric vehicle,” International Journal of
Electrical and Computer Engineering, vol. 12, no. 2, pp. 1147–1152, 2022, doi: 10.11591/ijece.v12i2.pp1147-1152.
[37] M. Lagraoui, A. Nejmi, H. Rayhane, and A. Taouni, “Estimation of lithium-ion battery state-of-charge using an extended kalman
filter,” Bulletin of Electrical Engineering and Informatics, vol. 10, no. 4, pp. 1759–1768, 2021, doi: 10.11591/eei.v10i4.3082.
[38] J. Khalfi, N. Boumaaz, A. Soulmani, and E. M. Laadissi, “An electric circuit model for a lithium-ion battery cell based on automotive
drive cycles measurements,” International Journal of Electrical and Computer Engineering, vol. 11, no. 4, pp. 2798–2810, 2021,
doi: 10.11591/ijece.v11i4.pp2798-2810.
[39] M. A. Nazri et al., “Fabrication and characterization of printed zinc batteries,” Bulletin of Electrical Engineering and Informatics,
vol. 10, no. 3, pp. 1173–1182, 2021, doi: 10.11591/eei.v10i3.2858.
[40] M. Elmarghichi, M. Bouzi, and N. Ettalabi, “Online parameter estimation of a lithium-ion battery based on sunflower optimization
algorithm,” Bulletin of Electrical Engineering and Informatics, vol. 10, no. 3, pp. 1505–1513, 2021, doi: 10.11591/eei.v10i3.2637.
[41] S. R. Salkuti, “Electrochemical batteries for smart grid applications,” International Journal of Electrical and Computer
Engineering, vol. 11, no. 3, pp. 1849–1856, 2021, doi: 10.11591/ijece.v11i3.pp1849-1856.
[42] A. M. Alsabari, M. K. Hassan, A. Cs, and R. Zafira, “Modeling and validation of lithium-ion battery with initial state of charge
estimation,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 21, no. 3, pp. 1317–1331, 2021, doi:
10.11591/ijeecs.v21.i3.pp1317-1331.
[43] B. V. Rajanna and M. K. Kumar, “Comparison of one and two time constant models for lithium ion battery,” International Journal
of Electrical and Computer Engineering, vol. 10, no. 1, pp. 670–680, 2020, doi: 10.11591/ijece.v10i1.pp670-680.

BIOGRAPHIES OF AUTHORS

Vimala Channapatna Srikantappa has completed her graduation in BE electrical

and electronics communication engineering from K.V.G College of Engineering, Mangalore
University. MBA a distant education from Sikkim Manipal university and M.Tech. in power system
from Ghousia college of Engineering, Ramanagram. At present a research scholar at BMS college
of Engineering, Bangalore, India and also working as an Assistant professor at Department of
Electrical and Electronics Engineering, SJB Institute of Technology, Bangalore. Member of MIE
(M- 612799). She has 9+ years of Industry Experience in abroad as well as in India and 9+ years of
teaching experience. Her area of expertise is in power systems, electrical machines, and battery
management systems in EV. She can be contacted at email: vimalasrikantappa@gmail.com.

Seshachalam Devarakonda is a academic dean and professor at Department of

Electronics and Communication Engineering, BMS College of Engineering, Bangalore, India. He
holds a Ph.D. degree in Electrical Engineering-Control Engineering and Power Electronics, Motilal
Nehru National Institute of Technology, Allahabad, U.P. His research interests are model order
reduction, power electronics, non-linear control engineering, and battery management systems in
EHV. He has played a major role in establishing a Centre of excellence along with Skill sector
council of India (ESSCI) in VLSI design. Member of several board of studies (BOS) such as
Dayananda Sagar College of Engineering, New horizon College of Engineering, Dr. AIT Bangalore,
Kongu college of Engineering Erode, Adhiamman College of Engineering Hosur and proposed
innovative curricula. He also played a key role in starting a third PG program, M.Tech. in VLSI
Design and Embedded systems during 2014, which is successfully running in the department now.
He is a member of evaluation committee for National Board of Accreditation and AICTE. He has
received funds under MODROBS scheme from AICTE & has conducted several workshops on
outcome-based education (OBE). A member of professional societies such as IEEE and ISTE. Has
supervised Four Ph.D. scholars and several M.Tech. students and now guiding three research
scholars. He can be contacted at email: dschalam.ece@bmsce.ac.in.

Int J Pow Elec & Dri Syst, Vol. 15, No. 3, September 2024: 1942-1950