Data-Driven Capacity Modeling of 18650 Lithium-Ion Cells from Experimental Electrical Measurements

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper focuses on the capacity modeling of 18650 lithium - ion batteries. Overall, it is of high quality, with certain innovation and academic value. However, there are still some details that can be improved. A minor revision is recommended before acceptance.

1.In the process of selecting the dataset, the author excluded some cells due to missing critical information like temperature. However, considering the complexity of real - world applications, how would the author handle datasets with missing values in a more comprehensive and practical way in future research?

2.the author evaluated the effects of different scaling methods on the model performance. But in practical scenarios where the data distribution may change over time, how do the author plan to adapt the scaling method to ensure the stable performance of the model?

3.Although the author proposed using neural networks to predict the capacity of lithium - ion cells, and the results show a certain degree of accuracy, there are still some errors. Have the author considered combining other models or techniques to further improve the prediction accuracy, and if so, which ones? 

Comments on the Quality of English Language

The English in this manuscript is generally of a good standard and effectively conveys the complex scientific content.

Author Response

We sincerely thank the reviewer for the positive evaluation of our manuscript and for recognizing its innovation and academic value in the context of capacity modeling for 18650 lithium-ion batteries. We are also grateful for the constructive comments and suggestions, which we have carefully addressed to improve the overall clarity, rigor, and presentation of the work. Below, we provide a detailed response to each point raised, along with the corresponding revisions made in the manuscript.

Comment 1:

In the process of selecting the dataset, the author excluded some cells due to missing critical information like temperature. However, considering the complexity of real - world applications, how would the author handle datasets with missing values in a more comprehensive and practical way in future research?

Response 1:

We appreciate the reviewer’s thoughtful comment. As rightly pointed out, energy storage systems often operate under complex and uncertain real-world conditions. In future research, we intend to address missing data more comprehensively by exploring data-driven modeling techniques that incorporate uncertainty, such as probabilistic models or hybrid approaches combining machine learning with state estimation methods. These strategies will allow for more robust analysis even when dealing with incomplete datasets, thereby enhancing the practical relevance of the modeling framework. Please check lines 418 to 424.

Comment 2:

the author evaluated the effects of different scaling methods on the model performance. But in practical scenarios where the data distribution may change over time, how do the author plan to adapt the scaling method to ensure the stable performance of the model?

Response 2:

We thank the reviewer for this valuable and insightful comment. As noted, modeling and parameter estimation in energy storage systems, such as lithium-ion cells, are highly dependent on the proper treatment of input data. In this study, the application of scaling techniques served two main purposes: (1) addressing the presence of features with different magnitudes, which is a common challenge in data-driven modeling, and (2) capturing the distinct behaviors of specific signals. For practical scenarios, as the reviewer correctly points out, adapting scaling methods becomes essential in the presence of evolving data distributions. In future implementations, especially in embedded or real-time systems, we plan to incorporate adaptive scaling strategies that update transformation parameters dynamically. Such techniques not only help maintain model stability and predictive accuracy over time, but also contribute to computational efficiency, which is critical in real-world applications. Please check lines 467 to 472.

Comment 3:

Although the author proposed using neural networks to predict the capacity of lithium - ion cells, and the results show a certain degree of accuracy, there are still some errors. Have the author considered combining other models or techniques to further improve the prediction accuracy, and if so, which ones? 

Response 3:

We appreciate the suggestion. Indeed, data-driven models such as neural networks can be enhanced by integrating other modeling approaches or complementary techniques to improve predictive performance. While this study focused on neural network-based modeling, we acknowledge the potential of exploring alternative architectures—such as convolutional or recurrent networks—as well as hybrid strategies that combine machine learning models with physics-based or statistical methods. These approaches may offer improved accuracy, albeit at the cost of increased computational complexity. We consider this a valuable direction for future research and will mention it accordingly in the revised manuscript. Please check lines 478 to 482.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript proposes a data-driven approach 10 model capacity degradation in 18650 lithium-ion cells. A systematic search was conducted to identify publicly available experimental datasets reporting charge/discharge processes. The proposed model achieved a prediction error of 3.35% during training and 3.48% during validation, demonstrating the robustness and efficiency of the implemented method. However, some problems need to be addressed for improving the manuscript. The authors could consider the following comments.

1. What is the conclusion drawn from the data distribution in Figure 7? Please provide a brief explanation.
2. “The results reveal that the charge/discharge current is correlated with the cell's terminal voltage, whereas these two variables do not exhibit a strong association with the cell temperature. Nonetheless, temperature plays a pivotal role in the capacity degradation process of lithium-ion cells, as it significantly accelerates the aging of the battery.”What is the logical relationship between these two sentences?
3. The dataset was standardized using MinMaxScaler, Standard Scaler, and RobustScaler, and the training-test split was set at 75%-25%, but in page 11, it is shown that this step is fundamental prior to splitting the dataset into training and testing sets, with a ratio of 75% for training and 25% for testing.Is this an interval or the ratio between the two?
4. “With the model's predictors defined for the charging and discharging of an 18650lithium-ion cell under a 1C2C profile” what is the 1C2C profile?
5. What is the “val_loss”in Figure12?
6. “To this end, key experimental variables such as voltage, current, and capacity were considered, obtained from CCCV charge and discharge profiles under different current rates.”What is the CCCV charge and discharge profiles?
7. The size of typefaces for all the figures are too small to read, which should be changed.
8. The title of ordinate in Figure 4 had better change to “percent”.
9. The paragraph between the line 61 and 66 is repeated.

Author Response

We sincerely thank the reviewer for their thoughtful and constructive evaluation of our manuscript. We are pleased to know that the data-driven approach, the systematic dataset selection, and the robustness of the proposed model were well received. We also appreciate the reviewer’s comments highlighting areas for improvement. We have carefully addressed each of the points raised and implemented the suggested revisions to enhance the clarity, accuracy, and overall quality of the manuscript. Our detailed responses to each comment are provided below.

Comment 1:

What is the conclusion drawn from the data distribution in Figure 7? Please provide a brief explanation.

Response 1:

Figure 7 presents the data distribution as part of the initial exploratory analysis necessary when developing data-driven models. These distributions indicate that the data were collected under controlled experimental conditions, capturing both resting states and charging/discharging processes. Furthermore, the consistency in capacity values throughout the cycles suggests that the dataset represents the early stages of cell degradation. Please check lines 230 to 234.

Comment 2

The results reveal that the charge/discharge current is correlated with the cell's terminal voltage, whereas these two variables do not exhibit a strong association with the cell temperature. Nonetheless, temperature plays a pivotal role in the capacity degradation process of lithium-ion cells, as it significantly accelerates the aging of the battery. What is the logical relationship between these two sentences?

Response 2:

The two sentences illustrate a contrast between statistical correlation and physical relevance. While the charge/discharge current and terminal voltage exhibit a measurable correlation within the dataset, temperature—although not strongly correlated with those variables—plays a fundamental physical role in lithium-ion cell degradation. This highlights that a variable can be critical to system behavior, such as aging, even if it does not show a strong statistical correlation with other measured parameters. Please check lines 248 – 252.

Comment 3:

The dataset was standardized using MinMaxScaler, Standard Scaler, and RobustScaler, and the training-test split was set at 75%-25%, but in page 11, it is shown that this step is fundamental prior to splitting the dataset into training and testing sets, with a ratio of 75% for training and 25% for testing. Is this an interval or the ratio between the two?

Response 3:

Thank you for your observation. In this study, the dataset was divided into training and testing subsets using a 75%-25% ratio, as commonly applied in machine learning practices. The standardization of features was performed prior to model training because the variables involved have different physical scales. Emphasis was placed on evaluating the impact of different scaling methods on model performance, while adhering to a proper data partitioning strategy required for training neural networks.

Comment  4:

With the model's predictors defined for the charging and discharging of an 18650 lithium-ion cell under a 1C2C profile – what is the 1C2C profile?

Response 4:

This is an excellent question that allows us to clarify the experimental conditions from which the dataset was obtained. In this context, '1C2C profile' refers to the charge/discharge protocol: the cells were charged at 1C (2.4 A) and discharged at 2C (4.8 A), relative to their nominal capacity of 2400 mAh. These rates are standard descriptors in battery testing and denote the current applied during cycling. Please check lines 341-344.

Comment 5:

What is the 'val_loss' in Figure 12?

Response 5:

In Figure 12, the term 'val_loss' shown in the legend refers to the validation loss during the training process of the model. It indicates the model’s performance on the validation set and is a key metric used to monitor generalization during training. The new legend are Training error and Validation error. Please check lines 392 – 393.

Comment 6:

To this end, key experimental variables such as voltage, current, and capacity were considered, obtained from CCCV charge and discharge profiles under different current rates. What is the CCCV charge and discharge profiles?

Response 6:

Thank you for this opportunity to expand on the experimental setup. CCCV refers to a Constant Current–Constant Voltage profile, a common charging method for lithium-ion cells. It involves charging the cell first with a constant current (CC) until a specified voltage is reached, followed by maintaining that voltage constant (CV) while reducing the current. This approach ensures safe and efficient charging and is widely used in both research and commercial applications. We add a new image to explain the concept, please check Figure 4.

Comment 7:

The size of typefaces for all the figures are too small to read, which should be changed.

Response 7:

Thank you for your helpful suggestion. We will revise the font sizes in all figures to improve readability and enhance the overall quality and clarity of the manuscript.

Comment 8:

The title of ordinate in Figure 4 had better change to 'percent'.

Response 8: We appreciate your observation. The ordinate label in Figure 4 will be revised to 'percent' in the updated version of the manuscript to ensure clarity and consistency. Please check Figure 3 in the new version.

Comment 9:

The paragraph between line 61 and 66 is repeated.

Response 9:

Thank you for pointing out this formatting error. We acknowledge the duplication and will remove the repeated paragraph. In this place, we will incorporate additional content based on other reviewer suggestions to strengthen the introduction.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

In this manuscript, the authors proposed a data-driven approach to investigate the modeled capacity degradation of 18650 lithium-ion batteries, demonstrating the potential of data-driven models for accurately predicting lithium-ion battery degradation. This work demonstrates a certain level of innovation, but the dataset lacks comprehensiveness. Consequently, the reviewer recommends its publication in Sustainability after revisions.

1. The author mentions that temperature is a crucial factor affecting battery degradation but does not delve into performance variations under different temperature conditions. Please include capacity degradation data at different temperatures (e.g., 25°C, 40°C) to assess the impact of temperature on degradation prediction.
2. The experimental data in this study covers a maximum of 100 cycles, whereas the lifespan of batteries in practical applications typically spans hundreds to thousands of cycles. To better evaluate the long-term predictive capability of model, it would be beneficial to supplement data for a greater number of cycles (e.g., 500 or 1000 cycles).
3. This study is based solely on data from the LISHEN 2400mAh 18650 battery. To assess the generalization capability of model, could authors incorporate data from other brands or different chemical systems (e.g., NCM or LFP)?
4. This study primarily analyzes data from a single battery (Cell 52). Could authors include a table comparing key performance metrics (e.g., initial capacity, maximum/minimum capacity, degradation rate, etc.) across all selected batteries (e.g., 44, 48, 52, 57)? This would help evaluate the representativeness of the data and ensure the broader applicability of the research findings.
5. The references section has significant issues, including an insufficient number of citations, a large proportion of outdated references (not from the past three years), and inconsistencies in formatting. Please revise and update accordingly.

Comments on the Quality of English Language

None

Author Response

We sincerely thank the reviewer for their thoughtful evaluation of our manuscript and for recognizing the innovation and potential of the proposed data-driven approach for modeling capacity degradation in 18650 lithium-ion batteries. We appreciate your constructive feedback, particularly regarding the scope and comprehensiveness of the dataset, which we acknowledge as a relevant limitation. We have carefully addressed your comments and implemented the necessary revisions to improve the clarity, completeness, and scientific rigor of the manuscript. Detailed responses to each point are provided below.

Comment 1:

The author mentions that temperature is a crucial factor affecting battery degradation but does not delve into performance variations under different temperature conditions. Please include capacity degradation data at different temperatures (e.g., 25°C, 40°C) to assess the impact of temperature on degradation prediction.

Response 1:

We appreciate the reviewer’s valuable observation. Indeed, temperature significantly influences battery degradation both in practical operation and modeling scenarios. However, in real-world applications, temperature effects are typically managed through battery thermal management systems designed to ensure safe and reliable operating ranges. In the experimental setup, the available dataset was obtained using a LANHE CT2001B battery tester coupled with a GDBELL thermal chamber, strictly maintaining all tests at a controlled temperature of 25°C [citation 1][citation 2]. Therefore, our dataset does not currently include performance data at other temperatures. We will explicitly clarify this limitation in the revised manuscript to enhance transparency and avoid potential misunderstandings. Please check lines 258 a 261.

Comment 2:

The experimental data in this study covers a maximum of 100 cycles, whereas the lifespan of batteries in practical applications typically spans hundreds to thousands of cycles. To better evaluate the long-term predictive capability of the model, it would be beneficial to supplement data for a greater number of cycles (e.g., 500 or 1000 cycles).

Response 2:

We thank the reviewer for this insightful comment. Indeed, the dataset used in our study contains fewer cycles (up to 100 cycles), primarily because the cells and associated experiments were designed specifically to explore early-stage degradation phenomena. This design allowed us to observe notable capacity degradation within a relatively limited number of cycles, a phenomenon that, in other scenarios, might require a significantly higher cycle count to emerge clearly. Despite the limited number of cycles, the proposed model effectively captures early degradation behaviors, providing meaningful insights into initial battery capacity fade trends. We will highlight this consideration clearly in the revised manuscript to acknowledge and clarify this limitation.

Comment 3:

This study is based solely on data from the LISHEN 2400mAh 18650 battery. To assess the generalization capability of the model, could authors incorporate data from other brands or different chemical systems (e.g., NCM or LFP)?

Response 3:

We thank the reviewer for this important observation. One of the key contributions of this research lies in the selection, identification, and characterization of publicly available datasets provided by independent researchers, research laboratories, and industrial organizations. The chosen dataset was selected due to its relatively recent publication, the consistency in cell specifications (all batteries are of the same model and reference), and the fact that the nominal capacity (2400mAh) is representative of modern energy storage applications. While the current study focuses on the LISHEN 18650 cells, we agree that incorporating data from other brands or chemistries such as NCM or LFP would further enhance the generalizability of the model. We will emphasize this as a potential avenue for future research and clearly state this limitation in the revised manuscript.

Comment 4:

This study primarily analyzes data from a single battery (Cell 52). Could authors include a table comparing key performance metrics (e.g., initial capacity, maximum/minimum capacity, degradation rate, etc.) across all selected batteries (e.g., 44, 48, 52, 57)? This would help evaluate the representativeness of the data and ensure the broader applicability of the research findings.

Response 4:

We appreciate the reviewer’s constructive suggestion. In response, we will include a table summarizing the key performance metrics—such as initial capacity, maximum/minimum capacity, and degradation rate—of Cell 52, as well as the other selected cells (Cells 44, 48, and 57). Additionally, we will clarify in the revised manuscript that all tested cells share identical specifications in terms of form factor, model, and manufacturer reference. This addition will help provide a more comprehensive view of the dataset and support the representativeness and broader applicability of the findings. Please check 220 – 236.

Comment 5:

The references section has significant issues, including an insufficient number of citations, a large proportion of outdated references (not from the past three years), and inconsistencies in formatting. Please revise and update accordingly.

Response 5:

Following the reviewer’s recommendations, the references section will be revised in both content and formatting to ensure it meets the journal’s requirements. Regarding the formatting, it has been generated automatically using the Microsoft Word template, the Mendeley Cite reference manager, and the citation style corresponding to the journal.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The paper presents interesting work on the regression of capacity of the 18650. However, it needs major revision to improve the quality.

First, the introduction is not well written. Data-driven algorithms for regression of capacity of the 18650 should be listed, analyzed, and compared to support the selection of MLP for the purpose. A table can be helpful.

Second, I believe figures 3and 4 are not needed for the paper, please remove them for clarity.

Thirdly, for the algorithm research, it can be very helpful to add the regression result data for different number of neurons even for single hidden year. As the authors states that 10 neurons and ReLU performance very well. Please add result from 5-20 neruons to see the trends. Besides, results from Sigmoid or other activation function should be add for comparison.

Fourthly, the author employ the public dataset for analysis. Why don't you compare this work with other work using the same dataset. I think it is a must to show the results of the others on the same dataset to show the superiority of this one.

By the way, I think the contribution from line 449 to 465 can be moved to the end of introduction part.

Author Response

We sincerely thank the reviewer for taking the time to evaluate our manuscript and for acknowledging the relevance of our work on capacity regression of 18650 lithium-ion cells. We appreciate your recommendation for major revisions and have carefully addressed all comments to improve the scientific quality, clarity, and overall presentation of the manuscript. The changes made in response to your suggestions are detailed below, and we trust they contribute to strengthening the rigor and impact of the work.

Comment 1

First, the introduction is not well written. Data-driven algorithms for regression of capacity of the 18650 should be listed, analyzed, and compared to support the selection of MLP for the purpose. A table can be helpful.

Response: We appreciate this important observation. In response, we will revise the introduction to include a comprehensive list, analysis, and comparison (in tabular format) of related studies that have applied data-driven algorithms for capacity modeling of lithium-ion cells. This addition will strengthen the justification for selecting MLP and will also help broaden the reference base, aligning with the comment of another reviewer. Please check lines 87-88.

Comment 2

Second, I believe figures 3 and 4 are not needed for the paper, please remove them for clarity.

Response: In line with your comment, and considering remarks from other reviewers regarding these figures, we propose to remove Figure 3 and retain Figure 4 with modifications. Additionally, we will include a brief paragraph clarifying which research centers and institutions have contributed to the publication of experimental and operational data on lithium-ion cells. Figure 3 was removed and Figure 4 was retained at the request of one of the reviewers. You can find the paragraph suggest in lines 139-149.

Comment 3

Thirdly, for the algorithm research, it can be very helpful to add the regression result data for different number of neurons even for single hidden layer. As the authors state that 10 neurons and ReLU perform very well. Please add result from 5–20 neurons to see the trends. Besides, results from Sigmoid or other activation functions should be added for comparison.

Response: We appreciate this valuable suggestion, which will help enhance the quality and depth of the study. Several tests were conducted to identify the best configuration for the neural network model, balancing accuracy and computational efficiency for real-world deployment. Following your recommendation, we will include a comparative analysis of different network configurations, including variations in the number of neurons (from 5 to 20) and activation functions available in the Keras-TensorFlow library. A table with the experiments conducted has been added, along with computational performance metrics and visualizations that illustrate the model’s performance. Please refer to lines 467–533.

Comment 4

Fourthly, the author employed the public dataset for analysis. Why don't you compare this work with other work using the same dataset? I think it is a must to show the results of the others on the same dataset to show the superiority of this one.

Response: This is an important and insightful comment. When using publicly available datasets, it is customary to cite the original authors, which we have done in this study to preserve research integrity and acknowledge prior work. Although each study may follow a different modeling approach, we agree with your suggestion and will incorporate a comparison with other studies that have used the same dataset to highlight the contributions and performance of our proposed model. Please refer to lines  536 – 556.

Comment 5

By the way, I think the contribution from line 449 to 465 can be moved to the end of the introduction part.

Response: Thank you for your suggestion. We will revise the structure of the manuscript and move the mentioned contribution to the end of the introduction section to improve the overall clarity and flow of the article.

Author Response File: Author Response.pdf

Reviewer 5 Report

Comments and Suggestions for Authors

This paper proposed a data-driven approach to model capacity degradation in 18650 lithium-ion cells. The model performance has been studied to demonstrate robustness and efficiency. The method and results are vague and lack strong and logical support. This manuscript could be rejected in this journal.

1. The literature review covered the current battery degradation mechanisms. As the proposed method is for capacity modelling, the previous capacity prediction methods should also be summarized briefly in this section. The knowledge gap should be further justified to identify the significance of this work.

2. Some figures are hard to read. Please enlarge the text to be read easily, i.e., Figure 1, Figures 5-7, Figure 13, etc.

3. For Section 2.3, many details for the model parameters are missing. Please add more information on the determination of the parameters of neural networks. i.e., why are 10 neurons selected? How did the author choose the activation function? etc.

4. For Figure 10, why is the specific cycle selected? Please add more justifications.

5. For Figure 13(d), why does it show a noticeable difference with (a)-(c), since all the predictions are positive?

Author Response

We sincerely thank the reviewer for their time and critical evaluation of our manuscript. We acknowledge the concerns raised regarding the clarity and support for the proposed method and results. In response, we have made significant efforts to improve the manuscript by clarifying the methodology, reinforcing the rationale behind the modeling approach, and providing additional experimental evidence and visualizations to support the results. We respectfully hope that the revisions implemented adequately address your concerns and contribute to strengthening the scientific soundness and clarity of the study. Detailed responses to each of your comments are provided below.

Comment 1

The literature review covered the current battery degradation mechanisms. As the proposed method is for capacity modelling, the previous capacity prediction methods should also be summarized briefly in this section. The knowledge gap should be further justified to identify the significance of this work.

Response: We appreciate this insightful suggestion, which helped us identify an opportunity to improve the quality and clarity of the manuscript. In response to your comment, and in alignment with similar feedback from another reviewer, we have added a table that summarizes, analyzes, and compares existing studies related to lithium-ion cell parameter modeling with a specific focus on capacity prediction. Furthermore, a dedicated paragraph has been included to highlight the relevance and contribution of our work. A table presenting similar works has been included (line 87).

Comment 2

Some figures are hard to read. Please enlarge the text to be read easily, i.e., Figure 1, Figures 5–7, Figure 13, etc.

Response: Thank you for your observation. In response, we have revised the font sizes across all figures mentioned to ensure better readability and to enhance the overall presentation quality of the manuscript.

Comment 3

For Section 2.3, many details for the model parameters are missing. Please add more information on the determination of the parameters of neural networks. i.e., why are 10 neurons selected? How did the author choose the activation function? etc.

Response: We thank the reviewer for this valuable comment. Additional details have been incorporated in Section 2.3 to clarify the selection of model parameters. The number of neurons (10) and the ReLU activation function were chosen based on preliminary tests comparing different network configurations, with the goal of balancing model accuracy and computational efficiency. The model was implemented using Keras and TensorFlow libraries, and the experimentation was conducted using a dedicated virtual environment via Jupyter Notebook configured specifically for this study. Please refer to lines 465 to 533.

Comment 4

For Figure 10, why is the specific cycle selected? Please add more justifications.

Response: Figure 10 presents data from a single cycle to illustrate that each cycle includes both charging and discharging phases, which is central to the modeling challenge. This specific representation was selected to simplify visualization and enhance understanding. Additional explanation has been added to the revised manuscript to justify the figure’s inclusion and relevance.

Comment 5

For Figure 13(d), why does it show a noticeable difference with (a)–(c), since all the predictions are positive?

Response: In Figure 13, regression results are shown for different cycles within the dataset. The difference observed in Figure 13(d) compared to Figures 13(a)–(c) was due to an error in the graphical representation that affected axis scaling. We have corrected this by retraining the model and generating a new image that aligns with the visual format and results presented in the rest of the subfigures.

Author Response File: Author Response.pdf

Reviewer 6 Report

Comments and Suggestions for Authors

1-  A few typographical errors exist in the manuscript that need revision. The repeated fourth paragraph of the Introduction should be deleted to enhance the text's clarity. A typing error in Subsection 2.3 needs correction by changing the term “108650 lithium-ion cell” to “18650 lithium-ion cell.”

2-  Some issues or suggestions related to the figures and tables are as follows:
- Figure 2: Clearly label battery types (e.g., confirm Lithium-Cobalt-Oxide, LCO) and replace “Forma” with “Form.” Ensure color consistency between Figures 2a and 2b.
- A comparison of Figure 4 shows an actual discharge profile percentage that matches 24% instead of the recorded 23% while also determining the correct reporting accuracy of percentages in discharged profiles. Confirm if the research only includes Li-ion battery studies since the identified number of papers (33) appears low. 
- An explicit x-axis label should be included in Figure 6. 
- Figures 7 and 13 require enlarged axis labels and numbers to improve clarity. 
- Figure 9 needs to indicate whether Compile executes alongside Fit or follows it sequentially in the workflow. Figure 10 requires the use of labels “(a)” and “(b)” to identify the sub-figures alongside clarification about their distinct characteristics in the caption.
- Figure 11 reveals that MinMaxScaler consistently produces low MSE values but only delivers the minimum MAE during cycles 10 and 15 despite the text indicating opposite results. Modify the text description to present the findings correctly.
- Figure 12: Authors should clarify in the caption that the left plot displays a magnified view of epochs 5–40 from the right plot to aid reader interpretation.
- Table 3: Verify the listed epochs. The text mentions five values, although only four appear in the provided table. 

3- I recommend that the authors discuss how these results might generalize to other battery chemistries or use cases, noting any constraints of the dataset.

4- I suggest the authors consider adding a brief comparison against other modeling approaches (both traditional and ML-based) to strengthen the discussion section and reinforce the scientific value of their research.

Author Response

We sincerely thank the reviewer for their thorough and detailed evaluation of our manuscript. Your comments have been invaluable in identifying areas for improvement, particularly regarding typographical accuracy, figure and table clarity, and the overall presentation of results. We have carefully addressed all the points raised, including corrections to the introduction and subsections, revisions to figure labels and formatting, clarification of workflows, and adjustments to textual descriptions where inconsistencies were identified. Additionally, we have expanded the discussion to address the generalizability of the results to other battery chemistries and have included a brief comparison with alternative modeling approaches to enhance the scientific context of our study. We are confident that these revisions have significantly improved the quality and clarity of the manuscript. Detailed responses and justifications for each item are provided below.

Comment 1

A few typographical errors exist in the manuscript that need revision. The repeated fourth paragraph of the Introduction should be deleted to enhance the text's clarity. A typing error in Subsection 2.3 needs correction by changing the term “108650 lithium-ion cell” to “18650 lithium-ion cell.”

Response: We appreciate the reviewer’s attention to detail and valuable feedback aimed at improving the quality and clarity of the manuscript. All the identified errors have been corrected in the revised version, including the removal of the repeated paragraph in the Introduction and the correction of the typographical error in Subsection 2.3.

Comment 2

Some issues or suggestions related to the figures and tables are as follows:
- Figure 2: Clearly label battery types (e.g., confirm Lithium-Cobalt-Oxide, LCO) and replace “Forma” with “Form.” Ensure color consistency between Figures 2a and 2b. (The correction has been made as suggested.)
- A comparison of Figure 4 shows an actual discharge profile percentage that matches 24% instead of the recorded 23% while also determining the correct reporting accuracy of percentages in discharged profiles. Confirm if the research only includes Li-ion battery studies since the identified number of papers (33) appears low. (The correction has been made as suggested.)
- An explicit x-axis label should be included in Figure 6. (The correction has been made as suggested.)
- Figures 7 and 13 require enlarged axis labels and numbers to improve clarity. (The correction has been made as suggested.)
- Figure 9 needs to indicate whether Compile executes alongside Fit or follows it sequentially in the workflow. Figure 10 requires the use of labels “(a)” and “(b)” to identify the sub-figures alongside clarification about their distinct characteristics in the caption. (The correction has been made as suggested.)
- Figure 11 reveals that MinMaxScaler consistently produces low MSE values but only delivers the minimum MAE during cycles 10 and 15 despite the text indicating opposite results. Modify the text description to present the findings correctly. (The correction has been made as suggested, please refer to lines 441 to 453.)
- Figure 12: Authors should clarify in the caption that the left plot displays a magnified view of epochs 5–40 from the right plot to aid reader interpretation. (The correction has been made as suggested, please refer to lines 459 to 471.)
- Table 3: Verify the listed epochs. The text mentions five values, although only four appear in the provided table. (The correction has been made as suggested.)

Response 2: We are grateful for the reviewer's thorough evaluation and constructive suggestions. Items 1 and 3–7 have been carefully reviewed and addressed in the revised version of the manuscript. Clarifications, corrections, and enhancements were implemented accordingly to improve clarity and accuracy. However, we could not establish a direct link between item 2 and the figure referenced. All applicable revisions have been incorporated into the updated manuscript.

Comment 3

I recommend that the authors discuss how these results might generalize to other battery chemistries or use cases, noting any constraints of the dataset.

Response 3: We appreciate this valuable comment. In the revised version, we have expanded the discussion section to reflect on the generalizability of the proposed model to other battery chemistries and use cases. Given the characteristics and limitations of the dataset used in this study, this topic has also been highlighted as a key direction for future research. Please refer to lines 642 to 654.

Comment 4

I suggest the authors consider adding a brief comparison against other modeling approaches (both traditional and ML-based) to strengthen the discussion section and reinforce the scientific value of their research.

Response 4: Following this insightful suggestion, and in alignment with other reviewers’ feedback, we have included a comparative analysis of the proposed model against both traditional and machine learning-based approaches in the discussion section. This addition enhances the scientific contribution and contextualizes our results within the broader research landscape. Table 1 has been included.

Author Response File: Author Response.pdf

Round 2

Reviewer 4 Report

Comments and Suggestions for Authors

It can be accepted in the present form. The author has addressed all my concerns. I believe that it is good enough.

Author Response

We sincerely appreciate your positive feedback and the time you dedicated to reviewing our manuscript once again. We are pleased to know that the revisions have adequately addressed your concerns and that the manuscript is now considered acceptable.

Reviewer 5 Report

Comments and Suggestions for Authors

The authors addressed all my previous comments and greatly improved the manuscript. A small issue is that the number of figure captions should be revised since the current one has some duplicate numbers, i.e., Line 189 vs. Line 291; Line 229 vs. Line 293. It is recommended that the manuscript be received after minor revisions.

Author Response

We sincerely appreciate your positive feedback and thoughtful recommendations. We have carefully addressed the issue regarding the duplicate figure caption numbers that you kindly pointed out (specifically Lines 189 vs. 291 and Lines 229 vs. 293). The captions have been revised accordingly to eliminate any duplications.

The suggested adjustments, as well as the minor issues identified, have been fully resolved and corrected throughout the manuscript. We thank you again for your valuable contributions, which have significantly helped improve the quality of our work.