Optimizing Photovoltaic System Diagnostics: Integrating Machine Learning and DBFLA for Advanced Fault Detection and Classification

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The study presents a promising hybrid optimization approach that significantly enhances PV fault detection accuracy through DBFLA-based machine learning optimization. The use of real and simulated datasets, ensemble learning, and comparative performance analysis makes the work scientifically rigorous. However, further research is needed to compare DBFLA with other metaheuristic methods, assess real-time deployment feasibility, and improve interpretability of the models. Despite these limitations, the study contributes valuable insights to the field of photovoltaic fault diagnostics and machine learning-based optimization. I suggest to pubblish the paper after a minor revision, following the sequent comments.

While the paper claims that DBFLA outperforms traditional optimization methods, there is no direct comparison with other established metaheuristic techniques such as Particle Swarm Optimization, Genetic Algorithms, Grey Wolf Optimizer. A benchmark study would strengthen the claims of superiority.

While the paper focuses on performance metrics, there is no discussion on the interpretability of the machine learning models. Methods like SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Model-agnostic Explanations) could be used to understand which features contribute most to fault classification.

The dataset is specific to one PV system configuration, which may limit the generalization of results. It would be beneficial to test the model across multiple PV topologies, including different panel types, inverter architecture, and environmental conditions.

The study mentions future integration into real-time PV monitoring, but there is no discussion of computational efficiency or real-time feasibility of DBFLA.

Author Response

 

Response to Reviewer Comments We sincerely appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful feedback has been invaluable in enhancing the clarity, depth, and rigor of our study. Your suggestions have helped us strengthen our work, particularly in the areas of benchmarking DBFLA, model interpretability, and real-time feasibility. We have carefully addressed all your comments, incorporating a comprehensive performance comparison with other metaheuristic algorithms, SHAP and LIME-based interpretability analysis, and a discussion on DBFLA’s computational efficiency for real-time deployment. Regarding the generalization concern related to the dataset, we provide a justification for our current scope while outlining potential future extensions. Below, we provide a point-by-point response detailing the revisions and justifications made in response to your feedback.

Comments 1: the paper claims that DBFLA outperforms traditional optimization methods, but there is no direct comparison with established techniques such as Particle Swarm Optimization (PSO), Genetic Algorithms (GA), and Grey Wolf Optimizer (GWO). A benchmark study would strengthen the claims of superiority.

Response 1: We have now included a benchmark comparison (Table 10), evaluating DBFLA against PSO, GA, and GWO based on key performance metrics, including accuracy, convergence speed, computational efficiency, and fault detection robustness. This comparative analysis demonstrates that DBFLA consistently outperforms other algorithms in accuracy (99.50%) and convergence speed while requiring lower parameter tuning complexity. This inclusion strengthens our claim that DBFLA is a superior optimization method for PV fault detection.

Comments 2: While the paper focuses on performance metrics, it lacks an explanation of which features contribute most to fault classification. The reviewer suggests using SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Model-Agnostic Explanations) for interpretability analysis.

Response 2: We have now integrated SHAP and LIME feature importance analysis in the manuscript.

• SHAP provides global and local interpretability, allowing us to analyze the impact of voltage, current, irradiance, and temperature on classification decisions.
• LIME focuses on local instance-based explanations, offering insights into why the classifier predicts specific faults.
• Figure 11 visually compares SHAP and LIME feature importance, demonstrating that voltage and current are the dominant features in PV fault classification.
• This analysis further reinforces DBFLA’s role in refining feature impact, ensuring robust fault detection performance.

Comments 3: The dataset is specific to one PV system configuration, which may limit the generalization of results. Testing across multiple PV topologies, including different panel types, inverter architectures, and environmental conditions, would enhance validity.

Response 3: While we acknowledge the importance of evaluating multiple PV configurations, this study focuses on a specific PV system as a case study to demonstrate the effectiveness of DBFLA-optimized machine learning models. The selection of a real PV dataset with simulated fault scenarios ensures practical relevance. Expanding the study to include multiple PV topologies would require significant additional datasets and computational resources, which are beyond the scope of this paper. However, we highlight this as an area for future work, where DBFLA can be tested on diverse PV configurations and environmental conditions.

Comments 4: The study mentions future real-time integration, but there is no discussion on computational efficiency or whether DBFLA is feasible for real-time PV monitoring.

Response 4: We have now incorporated a discussion on computational efficiency in the conclusions section.

• DBFLA achieves faster convergence compared to PSO, GA, and GWO, reducing the number of required iterations for optimization.
• Its low parameter tuning complexity minimizes computational overhead, making it suitable for real-time applications.
• Future work will focus on integrating DBFLA with real-time PV data acquisition and edge computing platforms, ensuring continuous fault detection with minimal latency.

 

 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Obtaining electricity through the conversion of primary renewable energy sources is an important alternative for reducing fuel consumption from finite primary sources, respectively for reducing pollution under current climate change conditions. Thus, the implementation on a larger scale of PV systems also requires the development of more efficient diagnostic methods. So, the topic addressed by the authors in this article is an actual and interesting topic for such applications.

The reviewer trusts the following comments can help the authors to improve the quality of their work:
1. Authors should review the text of the article, consult and respect the instructions presented in the “Instructions for Authors”, as follows:
References: References must be numbered in order of appearance in the text (including table captions and figure legends): The article does not respect this requirement and on the other hand references 1-7, 15, 17, 19 etc. do not appear as citations in the body of the article! All publications presented in the References section must be quoted in the body of the article.
All figures, schemes and tables should be inserted into the main text close to their first citation and must be numbered following their order of appearance (e.g., Figure 1, Scheme 1, Figure 2, Scheme 2, Table 1, etc.):
- Figure 6 appears twice: line 467, respectively 516.
- Figures 1, 7, 8, 9 are not referenced in the text.
- Tables 7, 8, 9 are also not referenced in the text of the article.
2. It is necessary to present a basic electrical diagram of the simulated and analyzed system – subparagraph 4.1.
3. In figure 6, two different quantities are represented in the same graphic: power and energy. As it is known, energy is the integral of power with respect to time (otherwise energy/time=power)! So a representation of energy in time could possibly be made by a bar diagram – energy/recording duration. On the other hand, in figure 6 the same order of magnitude appears regardless of the electrical quantities represented. “The state of charge of a battery is defined as the ratio between the available capacity and the reference capacity, which is the maximum capacity that can be withdrawn from the fully charged battery under reference conditions.”
Consequently, the conclusions based on the representations in figure 6 are questionable!

Comments on the Quality of English Language

No comments

Author Response

 

Response to Reviewer 1 Comments We sincerely appreciate the time and effort you have dedicated to reviewing our manuscript. Your insightful comments and constructive feedback have been invaluable in improving the clarity, accuracy, and quality of our work. Your expertise has helped us refine our methodology, enhance our visual representations, and ensure that our manuscript aligns with the journal’s formatting and citation requirements. We have carefully addressed all your comments, implementing the necessary revisions to strengthen our study. Below, we provide a point-by-point response detailing the changes we have made in accordance with your recommendations. Thank you again for your valuable insights and for contributing to the enhancement of our research.

Comment 1: References and Formatting Issues

• Issue: References were not numbered in order of appearance, and some references were listed in the bibliography but not cited in the text.
• Response: We have carefully reviewed the references and ensured that they are numbered in the correct order of appearance throughout the manuscript. We have also removed references that were not cited in the text and appropriately referenced all the relevant citations.
• Issue: Figures and tables were not referenced in the text, and some were duplicated.
• Response: We have revised the manuscript to ensure that all figures, tables, and schemes are inserted close to their first citation and are referenced in sequential order. Specifically:
• Figure 6 previously appeared twice (lines 467 and 516). This duplication has been removed.
• Figures 1, 7, 8, and 9 are now properly referenced within the text.
• Tables 7, 8, and 9 are now cited at appropriate points within the manuscript.

Comment 2: Basic Electrical Diagram of the Simulated System

• Issue: The manuscript lacked a fundamental electrical diagram illustrating the simulated PV system.
• Response: We have now included a detailed electric diagram of the simulated system in subsection 4.1. This diagram provides a clear visualization of the energy flow, system components, and their interconnections, ensuring a better understanding of the simulated PV system. Additionally, we have used distinct colors, varied node shapes, and labeled power flow arrows to enhance readability.

Comment 3: Issues with Figure 6 (Power and Energy Representation)

• Issue: Figure 6 incorrectly combined power (instantaneous) and energy (cumulative) in the same plot, making the conclusions questionable.
• Response: We have completely revised Figure 6 to ensure a proper representation of power and energy.
• Corrections Implemented:
• PV Power Output (kW) is now represented as a line graph to reflect its instantaneous nature.
• Battery State of Charge (SOC %) is now represented as a bar chart, instead of being incorrectly plotted as a continuous function.
• The order of magnitude for each variable has been corrected, ensuring that energy and power are visually distinct.
• The figure legend and axis labels have been refined to clearly distinguish between power and energy parameters.
• Issue: The definition of battery state of charge (SOC) was not properly applied in the figure.
• Response: We have revised the explanation in the text to clarify that the SOC is the ratio of available capacity to reference capacity and ensured that our graphical representation follows this definition accurately.

Final Remarks

All suggested modifications have been incorporated into the revised manuscript. We sincerely appreciate the reviewer’s constructive feedback, which has significantly contributed to the improvement of our work. Thank you for your valuable time and consideration.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

I think the article deals with current issues. Machine learning, artificial neural networks and fault detection of PV systems are important for the development and operation of photovoltaic power plants. The novelty is not high but acceptable. However, there are many inaccuracies in the article, see my comments below. I can recommend the article for publication after a major revision.

Comments:

- Introduction - I recommend softening the claim „As a clean energy alternative to fossil fuels, PV systems play a critical role in reducing environmental impact and addressing global energy needs.“ I agree with the authors that photovoltaic power plants have an important place in the energy mix, but their importance should not be overestimated. There are disadvantages as well. The main disadvantage is the low efficiency of energy conversion and a short service life. Although the manufacturers indicate a lifespan of around 25 years, the real lifespan is around 12 years and in locations with extreme climatic conditions it is below 10 years. There are problems with recycling, with the release of heavy metals into the soil, instability in the distribution network and so on.

- Tables 5, 6 - It is not good if one table is divided on two different pages.

- There are two different figures and both figures are marked like Figure 6 (Lines 467, 515).

- Figure 6 (line 467) – I think, there is not „Power/Energy (kW/kWh)“, there should be „Power (kW), Energy (kWh)“. I do not understand the first blue peak during the night and the oscillations of PV power output during the night. The oscillations even go into negative values.

- Figure 6 (line 515) and Figure 7 – the formate should be uniform, (units %).

- Table 10 – The table caption should be on the same page like the table.

- The references are a bit confusing. I don't mind that they are not listed sequentially in the text. But some references are not mentioned in the text (1, 2, 3, 4, 5, 6, 7, 15, 26, 29, 30). On the contrary, references 41, 44 are mentioned in the text (line 71, tab.1, tab.10), but they are not listed in the references. There are 34 references in total.

- On the other hand, I think, the article doi:10.31545/intagr/192173 also deals with the detection of PV panel defects. It could be inspiration for authors and it could be mentioned in the references.

Author Response

 

Response to Reviewer 2 Comments We sincerely appreciate your thoughtful review of our manuscript. Your detailed feedback has provided valuable insights that have helped us enhance the clarity, accuracy, and overall quality of our work. We have carefully considered each of your comments and have implemented the necessary revisions to address all concerns. We particularly appreciate your suggestions regarding the introduction, figure formatting, table placements, and reference consistency, all of which have been corrected in the revised manuscript. Your recommendation regarding the additional reference has also been taken into account, and we have incorporated it into our discussion. Below, we provide a point-by-point response detailing the improvements made in response to your comments. Thank you once again for your constructive feedback, which has significantly strengthened our study.

Comment 1: Introduction – Overstatement of PV Importance

• Issue: The introduction overemphasized the role of PV systems while neglecting their limitations (e.g., low efficiency, short service life, recycling challenges, and network instability).
• Response: We have softened the claim and revised the introduction to acknowledge both the advantages and challenges of PV systems.
• The revised text now highlights the limitations of PV technology, including efficiency constraints, degradation rates, and recycling concerns.
• We have added discussion on the real-world lifespan of PV panels (12–15 years on average, shorter in extreme climates) and potential grid stability issues.
• This balanced approach ensures that the introduction accurately reflects both the potential and the challenges of PV adoption.

Comment 2: Tables 5 and 6 - Split Across Pages

• Issue: Tables 5 and 6 were divided across two different pages, making them difficult to read.
• Response: We have reformatted these tables to ensure that each table appears on a single page, improving readability.

Comment 3: Duplicate Figure 6 (Lines 467, 515)

• Issue: Two different figures were mistakenly labeled as Figure 6, leading to confusion.
• Response: We have renumbered the figures to ensure a sequential and correct order, eliminating the duplicate Figure 6 issue.

Comment 4: Figure 6 (Power and Energy Representation Errors)

• Issue: The original Figure 6 incorrectly combined power and energy in the same plot using inappropriate labeling ("Power/Energy (kW/kWh)") and contained unexpected oscillations and negative values at night.
• Response:
• We have corrected the y-axis labels to clearly differentiate Power (kW) and Energy (kWh).
• The battery charge is now represented as a bar chart instead of a misleading continuous function.
• The PV power output curve has been adjusted to remove oscillations at night and prevent incorrect negative values.
• The explanation of Figure 6 in the text has been revised to correctly define battery state of charge (SOC) and ensure accuracy.

Comment 5: Formatting Consistency for Figures 6 and 7 (Units in %)

• Issue: The formatting of Figures 6 and 7 was inconsistent, particularly regarding unit representation (e.g., percentages).
• Response: We have standardized the formatting of Figures 6 and 7, ensuring:
• All percentages are consistently formatted as % across both figures.
• Labeling, font size, and axis scaling are uniform for readability.

Comment 6: Table 10 - Caption Placement

• Issue: The caption for Table 10 was placed separately from the table, affecting readability.
• Response: We have adjusted the formatting so that the caption now appears directly above the table, following the journal's guidelines.

Comment 7: Reference Issues

• Issue 1: Some references (1, 2, 3, 4, 5, 6, 7, 15, 26, 29, 30) were listed in the bibliography but not cited in the text.
• Response: We have ensured that all references are properly cited within the body of the manuscript or, if unnecessary, removed them from the reference list.
• Issue 2: Some references (41, 44) were cited in the text but not listed in the references section.
• Response: We have added missing references to the bibliography, ensuring that all cited works are properly documented.
• Issue 3: The total number of references was 34, but the inconsistencies made the reference section appear confusing.
• Response: We have now restructured the reference list, verified all citations, and numbered them correctly to ensure clarity and compliance with journal formatting requirements.

Comment 8: Additional Recommended Citation

• Issue: The reviewer suggested adding the reference doi:10.31545/intagr/192173, which also discusses PV panel defect detection.
• Response: We have reviewed the recommended article and found it relevant to our study.
• We have cited this reference in our literature review.
• This addition strengthens our study by providing further context on PV life span.

Final Remarks

All suggested modifications have been carefully implemented in the revised manuscript.
We appreciate the reviewer’s detailed comments, which have greatly improved the clarity, accuracy, and presentation of our research.

Thank you again for your valuable insights and for helping us refine our work.

 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The changes made, in accordance with the comments of the reviewers, led to an improvement of the article in terms of its quality of presentation and intelligibility.

Author Response

Thank you for your valuable feedback. We appreciate your recognition of the improvements made to the article based on the reviewers' comments. Your insightful suggestions have greatly contributed to enhancing the quality and clarity of our work. If there are any further recommendations, we would be happy to address them.

Reviewer 3 Report

Comments and Suggestions for Authors

I think, my comments were accepted and the article was improved. I have no additional comments. I can recommend the article for publication in present form.

Author Response

Thank you for your positive feedback and for taking the time to review our work. We appreciate your thoughtful comments and are glad that the improvements align with your expectations. Your recommendation for publication is greatly valued.