Study of Lithium-Ion Battery Charge State Estimation Based on BP Neural Network Fusion Optimized Sparrow Algorithm

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper presents the results of training neural networks for predicting a technologically important parameter for industry and not only for the battery charging degree. I did not find any significant problems or shortcomings, but there are several questions after answering which with appropriate additions in the text the article can be accepted

1. Please provide training and validation loss functions
2. Please add links to the datasets you use (UDSS. DST) or mark explicitly where the data on these tests were obtained from

Author Response

Comment 1:
Please provide training and validation loss functions.

Response 1:

Thank you for pointing this out. We agree with this comment. Therefore, we have provided the training, validation, and testing loss curve (MSE) in this response to demonstrate the model’s convergence behavior. The model reaches its best validation performance at epoch 87, with an MSE of 0.00011331, indicating effective learning and strong generalization capability.

The corresponding figure has been included in this response file for reference.

 

Comment 2:

Please add links to the datasets you use (UDSS, DST) or mark explicitly where the data on these tests were obtained from.

Response 2:

Thank you for pointing this out. We agree with this comment. Therefore, we clarify that the DST, UDDS, and FUDS data used in this study were obtained from battery testing experiments conducted in our laboratory under standard dynamic operating conditions. These datasets are internally generated and not derived from any publicly available sources.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper is about lithium-ion battery charge state estimation. However, there are several things to consider.

1. Please consider the scope of the journal, which is coatings. The current article is disconnected from the scope of the journal. The article is about machine learning.
2. I would like to suggest submitting to a different journal.

Author Response

Comment :

The paper is about lithium-ion battery charge state estimation. However, there are several things to consider.

1.Please consider the scope of the journal, which is coatings. The current article is disconnected from the scope of the journal. The article is about machine learning.

2.I would like to suggest submitting to a different journal.

 

Response :

Thank you for pointing this out. We agree with this comment and fully understand the importance of aligning our manuscript with the scope of the journal. While our research focuses on battery state-of-charge (SOC) estimation using machine learning methods, we respectfully note that Coatings has previously published work on related topics. For example, the article titled “A Joint Estimation Method Based on Kalman Filter of Battery State of Charge and State of Health” (Coatings 2022, 12(8), 1047) addresses battery state estimation and algorithmic modeling, which share methodological similarities with the present study.

We also believe that our work aligns with the journal’s interdisciplinary focus, particularly regarding the application of coatings in electrochemical energy systems. In lithium-ion power batteries used in electric vehicles, coatings are widely employed across key components to enhance electrochemical stability, thermal management, and operational safety. Electrode surface coatings mitigate interface degradation and improve cycle life; ceramic or polymer coatings on separators enhance thermal resistance and inhibit lithium dendrite growth; thermal interface and phase-change coatings support efficient heat dissipation and suppress thermal runaway; while protective layers on current collectors reduce corrosion and degradation under high-rate cycling. These technologies are fundamental to ensuring battery reliability under dynamic automotive load conditions, where accurate SOC estimation is critical for real-time performance evaluation and safety control. In this context, our machine-learning-based SOC prediction model provides a valuable computational tool to support the design and assessment of coating-enabled battery systems in electric vehicles.

In addition, recent publications in Coatings reflect the journal’s growing interest in data-driven modeling and simulation research in energy and materials science, including:

1.“Predicting UV-Vis Spectra of Benzothio/Dithiophene Polymers for Photodetectors by Machine-Learning-Assisted Computational Studies” (Coatings 2025, 15(5), 558);

2.“Application of Machine Learning in Terahertz-Based Nondestructive Testing of Thermal Barrier Coatings with High-Temperature Growth Stresses” (Coatings 2025, 15(1), 49);

3.“Machine-Learning-Driven Optimization of Cold Spray Process Parameters: Robust Inverse Analysis for Higher Deposition Efficiency” (Coatings 2025, 15(1), 12).

These examples demonstrate the journal’s openness to predictive modeling and algorithmic approaches that enhance coating performance, reliability, and system-level integration.

No textual changes were made to the manuscript in response to this comment. We sincerely appreciate the reviewer’s feedback and respectfully leave the final decision regarding the manuscript’s scope suitability to the editorial team.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have conducted a research and study on the lithium-ion battery charge state estimation based on BP neural network fusion optimized sparrow algorithm.

The main objective of the research has been to develop and propose a method to predict the state of charge (SOC - state of discharge) of lithium batteries much better and more accurately. These batteries are currently the most widely used, especially those with Lithium Iron Phosphate (LFP) technology. Applications range from photovoltaic solar installations for self-consumption to various applications in electric vehicles, etc... Currently, most lithium batteries (almost all of them) are equipped with the so-called BMS (Battery Management System), which controls voltage, current, temperature, SOC, etc.

The authors state that they have used a classic three-layer BP (Backpropagation) neural network, which at first glance does not seem very innovative since this type of network has been known and used for decades in numerous engineering applications. What is innovative is that they have also applied and combined a Sparrow Search Algorithm (SSA) to improve the model's ability to search for optimal values.

My first recommendation is to review the introduction section so that it is better explained and more concise. Instead of describing methods and algorithms in detail, it would be appropriate to explain the problem, its relevance, and the existing research gaps in a more detailed and clear manner. And how this research conducted by the authors addresses this need for improvement. To what extent is it important to estimate the SOC of a battery as accurately as possible? This should be better explained by delving into the potential problems in different application fields if it is not done accurately.

In this sense, there are references to multiple techniques for SOC (State of Charge) prediction, but there is no clear comparative analysis of why the BP Neural Network optimized with ISSA was selected over other methods. A comparative table highlighting advantages and disadvantages would be useful.

The methodology employed by the authors is well-structured, but from my point of view, it would be better to include a flowchart or visual diagram illustrating the steps of the ISSA-BP model to enhance clarity and understanding of the process. Similarly, the results section, presented with many tables and figures, would require further discussion and analysis regarding why certain methods (e.g., ISSA-BP) work better under certain test conditions than others. For instance, the inclusion of a more robust comparative statistical analysis, such as statistical tests to demonstrate the significance of the improvements, would be beneficial.

The conclusions section is well written according to the results obtained during the research. However, I recommend including and naming, on the one hand, the potential limitations of the study and, on the other hand, possible future work. This would provide more clarity on how the results could be further extended or improved.

Author Response

Comment 1:

The introduction should be revised to make it clearer and more concise. It currently contains too much detail about the methods and algorithms. The focus should be on the background, the significance of the research topic, current limitations in the field, and the motivation for the proposed approach. In particular, the importance of accurate SOC estimation and the consequences of inaccurate predictions in real-world applications should be emphasized.

Response 1:

Thank you for pointing this out. We agree with this comment. Accordingly, we have revised the Introduction section to reduce the level of detail concerning algorithms and methodologies. The updated version places greater emphasis on the research background and motivation, the significance of accurate SOC estimation in ensuring battery safety and system performance, and the limitations of existing methods. In addition, we have included a discussion on the potential risks and challenges posed by inaccurate SOC predictions in various practical applications.

This revision can be found on Pages 1–2 of the revised manuscript. The modified text is highlighted using blue Palatino Linotype font to ensure visibility.

 

 

Comment 2:

Although several SOC estimation methods are mentioned in the manuscript, the rationale for selecting the ISSA-BP combination is not sufficiently discussed. It is recommended to add a comparative table summarizing the advantages and disadvantages of different SOC estimation methods to better highlight the advantages of ISSA-BP.

Response 2:

Thank you for pointing this out. We agree with this comment. Therefore, we have added a new section (Section 3.3) in the revised manuscript, which provides a detailed comparison of commonly used optimization algorithms, including PSO, GA, GWO, and SSA. The section summarizes their respective advantages and disadvantages, and clarifies the motivation for selecting the Improved Sparrow Search Algorithm (ISSA).

SSA was selected for its strong global search ability, fast convergence, and minimal control parameters. However, due to its sensitivity to initial population quality and tendency to fall into local optima, we enhanced it by integrating Tent chaotic mapping, sine–cosine mechanisms, and firefly disturbance strategies. These improvements result in the ISSA, which significantly improves optimization performance and robustness in training the BP neural network for SOC estimation.

This addition can be found in Section 3.3, Page 7 of the revised manuscript. The modified content is highlighted using blue Palatino Linotype font to enhance visibility.

 

 

Comment 3:

It is suggested to add a process flow diagram of the ISSA-BP model to improve clarity.

Response 3:

Thank you for pointing this out. We agree with this comment. In fact, we have already included a complete process flow diagram of the ISSA-BP model in the original manuscript. This diagram has been presented as Figure 4 on page 8, and it clearly illustrates the entire modeling and optimization process, including data preprocessing, SSA initialization, hybrid strategy integration (Tent chaotic mapping, sine-cosine algorithm, and firefly disturbance), and the final training of the BP neural network. We respectfully hope that this figure already meets the reviewer's expectations regarding clarity and completeness of the model structure.

 

 

Comment 4:

The results section would benefit from more detailed discussion and analysis explaining why certain methods (e.g., ISSA-BP) perform better under specific test conditions. Additionally, a more robust comparative statistical analysis, such as significance testing, would help to support the improvements observed.

Response 4:

Thank you for pointing this out. We agree with this comment. Therefore, we have expanded the analysis in the results section to provide a more in-depth explanation of the performance advantages of ISSA-BP under different driving cycles. Specifically, we discussed the model’s behavior across the DST, UDDS, and FUDS conditions using the reported MAE, MSE, and RMSE metrics in Tables 2–4. The revised section analyzes how the improved population diversity from Tent chaotic mapping, enhanced global-local search balance via sine–cosine mechanisms, and local refinement through firefly disturbance enable ISSA-BP to better adapt to the nonlinear and time-varying load characteristics of each test condition.

Moreover, we emphasized that the improvements of ISSA-BP over SSA-BP and traditional BP are consistent across all three scenarios, indicating not only better convergence accuracy but also stronger robustness. The discussion is based directly on the reported experimental results, without relying on further inference, and demonstrates that ISSA-BP achieves superior generalization across different dynamic environments.

This change can be found in the revised manuscript in Section 4.2, Pages 13–14, highlighted in blue Palatino Linotype font.

 

Comment 5:

The conclusions section is well written according to the results obtained during the research. However, I recommend including and naming, on the one hand, the potential limitations of the study and, on the other hand, possible future work. This would provide more clarity on how the results could be further extended or improved.

Response 5:

Thank you for pointing this out. We agree with this comment. Accordingly, we have revised the conclusion section to explicitly address the study's limitations, including the reliance on simulation data, limited generalizability to other battery chemistries, and the lack of validation in real-time embedded systems. We have also added a forward-looking statement outlining future work, such as applying the ISSA-BP model to real-world datasets, integrating it into battery management hardware, and extending it to joint SOC and SOH estimation tasks.

The revised conclusion can be found in Section 5 (Conclusion), Page 14 of the revised manuscript. To enhance visibility, the modified content is highlighted using blue Palatino Linotype font.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have significantly improved and expanded the paper, and together with the graphics, the article becomes much clearer and more coherent.

Author Response

Comment:

The authors have significantly improved and expanded the paper, and together with the graphics, the article becomes much clearer and more coherent.

 

Response:

We sincerely thank the reviewer for the positive evaluation and kind recognition of our efforts to improve the manuscript’s clarity, coherence, and overall presentation. We truly appreciate your valuable feedback throughout the revision process.

Author Response File: Author Response.pdf