Data-Driven Coordinated Voltage Control Strategy for Distribution Networks with High Proportion of Renewable Energy Based on Voltage–Power Sensitivity
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
Reviewer Comments
Actually, the manuscript presents a well-structured and insightful contribution to the field. The topic is timely and relevant for the field, and the overall quality of the writing is commendable. The data are presented, and the findings offer valuable insights. However, a few areas would benefit from further refinement to enhance the overall rigor and impact of the work. My specific suggestions are as follows:
-
Introduction:
While the introduction provides a good overview of the research context, it would benefit from a more detailed discussion of the current state of the art. Please consider expanding this section to better position your study within the existing body of literature. The inclusion of more recent and relevant references would strengthen the rationale and demonstrate a thorough understanding of recent developments in the field. -
Methodology:
The methodology is generally sound, but certain procedures and parameters require clearer explanation. Providing additional details regarding sample preparation, experimental conditions, or analytical techniques will improve reproducibility and allow readers to more fully assess the validity of your approach. -
Results and Discussion:
The results are promising and well-illustrated. However, some findings would benefit from deeper analysis and comparison with previously published data. Please consider referencing more recent studies to contextualize your results and highlight the novelty and significance of your contributions. -
References:
Several references are outdated. I strongly recommend the inclusion of more recent publications (from the last 5 years) to support your claims and discussions throughout the manuscript. This will also help underline the relevance of your work in the current scientific landscape.
Overall Recommendation:
In sum, the manuscript is of high quality and merits publication after minor to moderate revisions. Addressing the points outlined above will significantly enhance the article's clarity, robustness, and scientific value.
Author Response
Data-Driven Voltage Coordinated Control Strategy for Distri-bution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity
Ziwei Cheng 1*, Lei Wang 1, Can Su 1, Runtao Zhang 1, Xiaocong Li 2, Bo Zhang 2
- State Grid Hebei Electric Power Research Institute, Shijiazhuang 050021, China
- Key Laboratory of Distributed Energy Storage and Micro-grid of Hebei Province
North China Electric Power University, Baoding 071003, China
Manuscript Number: sustainability-3577313
Initial Submitted: 25 Mar 2025.
Decision: 11 Apr 2025.
Dear Editors and Reviewers:
Thanks for your valuable comments concerning our manuscript entitled "Data-Driven Voltage Coordinated Control Strategy for Distribution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity " (ID: sustainability-3577313). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made correction which we hope meet with approval.
A detailed response to the associate editor’s and the reviewers’ comments/suggestions has addressed and reported in the following parts, in which the font type of the suggestions is bold and the corresponding response of the author is common font. We have tried to meet all the requests made by the reviewers.
In the revised version of the manuscript the modified parts have been highlighted with the red color fonts.
Responds to the reviewer’s comments:
Reviewer #1:
- Introduction: While the introduction provides a good overview of the research context, it would benefit from a more detailed discussion of the current state of the art. Please consider expanding this section to better position your study within the existing body of literature. The inclusion of more recent and relevant references would strengthen the rationale and demonstrate a thorough understanding of recent developments in the field.
Response: Relevant literature has been added to the text, and the details are shown:
- Methodology: The methodology is generally sound, but certain procedures and parameters require clearer explanation. Providing additional details regarding sample preparation, experimental conditions, or analytical techniques will improve reproducibility and allow readers to more fully assess the validity of your approach.
Response: The main innovation points of this paper mainly focus on two aspects: data-driven voltage-power sensitivity acquisition and voltage coordinated control of active distribution networks based on voltage-power sensitivity. The case verification part is also carried out for these two parts. The case verification is mainly conducted for the typical distribution systems of IEEE 33 nodes and IEEE 141 nodes. In order to verify the generality of the proposed strategy, two situations, namely, the consistent and different meteorological environments within the distribution network area, are respectively considered to verify the proposed data-driven voltage-power sensitivity acquisition method and the voltage regulation effect of reactive compensation/active power curtailment in the distribution network, which fully validates the effectiveness of the theoretical analysis and the proposed method.
According to the reviewers' comments, the configuration of the calculation examples and the details of the result analysis in the simulation verification part were supplemented.
- Results and Discussion: The results are promising and well-illustrated. However, some findings would benefit from deeper analysis and comparison with previously published data. Please consider referencing more recent studies to contextualize your results and highlight the novelty and significance of your contributions.
Response: The main innovation points of this paper mainly focus on two aspects: data-driven voltage-power sensitivity acquisition and voltage coordinated control of active distribution networks based on voltage-power sensitivity.
To more clearly illustrate the innovation points and significance of this paper, the author supplemented the latest relevant literature and conducted a comparative analysis from the aspects of method principles and implementation effects. In particular, the real-time performance and real-time effects of the traditional power voltage sensitivity calculation method and the distribution network voltage optimization regulation method were analyzed. The results show that the proposed method has obvious technical advantages.
- References: Several references are outdated. I strongly recommend the inclusion of more recent publications (from the last 5 years) to support your claims and discussions throughout the manuscript. This will also help underline the relevance of your work in the current scientific landscape.
Response: According to the opinions of the review experts, more recent references have been updated, and the details are shown:
Thank you again for your positive and constructive comments and suggestions on our manuscript.
We hope you will find our revised manuscript acceptable for publication.
Data-Driven Voltage Coordinated Control Strategy for Distri-bution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity
Ziwei Cheng 1*, Lei Wang 1, Can Su 1, Runtao Zhang 1, Xiaocong Li 2, Bo Zhang 2
- State Grid Hebei Electric Power Research Institute, Shijiazhuang 050021, China
- Key Laboratory of Distributed Energy Storage and Micro-grid of Hebei Province
North China Electric Power University, Baoding 071003, China
Manuscript Number: sustainability-3577313
Initial Submitted: 25 Mar 2025.
Decision: 11 Apr 2025.
Dear Editors and Reviewers:
Thanks for your valuable comments concerning our manuscript entitled "Data-Driven Voltage Coordinated Control Strategy for Distribution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity " (ID: sustainability-3577313). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made correction which we hope meet with approval.
A detailed response to the associate editor’s and the reviewers’ comments/suggestions has addressed and reported in the following parts, in which the font type of the suggestions is bold and the corresponding response of the author is common font. We have tried to meet all the requests made by the reviewers.
In the revised version of the manuscript the modified parts have been highlighted with the red color fonts.
Responds to the reviewer’s comments:
Reviewer #1:
- Introduction: While the introduction provides a good overview of the research context, it would benefit from a more detailed discussion of the current state of the art. Please consider expanding this section to better position your study within the existing body of literature. The inclusion of more recent and relevant references would strengthen the rationale and demonstrate a thorough understanding of recent developments in the field.
Response: Relevant literature has been added to the text, and the details are shown:
- Methodology: The methodology is generally sound, but certain procedures and parameters require clearer explanation. Providing additional details regarding sample preparation, experimental conditions, or analytical techniques will improve reproducibility and allow readers to more fully assess the validity of your approach.
Response: The main innovation points of this paper mainly focus on two aspects: data-driven voltage-power sensitivity acquisition and voltage coordinated control of active distribution networks based on voltage-power sensitivity. The case verification part is also carried out for these two parts. The case verification is mainly conducted for the typical distribution systems of IEEE 33 nodes and IEEE 141 nodes. In order to verify the generality of the proposed strategy, two situations, namely, the consistent and different meteorological environments within the distribution network area, are respectively considered to verify the proposed data-driven voltage-power sensitivity acquisition method and the voltage regulation effect of reactive compensation/active power curtailment in the distribution network, which fully validates the effectiveness of the theoretical analysis and the proposed method.
According to the reviewers' comments, the configuration of the calculation examples and the details of the result analysis in the simulation verification part were supplemented.
- Results and Discussion: The results are promising and well-illustrated. However, some findings would benefit from deeper analysis and comparison with previously published data. Please consider referencing more recent studies to contextualize your results and highlight the novelty and significance of your contributions.
Response: The main innovation points of this paper mainly focus on two aspects: data-driven voltage-power sensitivity acquisition and voltage coordinated control of active distribution networks based on voltage-power sensitivity.
To more clearly illustrate the innovation points and significance of this paper, the author supplemented the latest relevant literature and conducted a comparative analysis from the aspects of method principles and implementation effects. In particular, the real-time performance and real-time effects of the traditional power voltage sensitivity calculation method and the distribution network voltage optimization regulation method were analyzed. The results show that the proposed method has obvious technical advantages.
- References: Several references are outdated. I strongly recommend the inclusion of more recent publications (from the last 5 years) to support your claims and discussions throughout the manuscript. This will also help underline the relevance of your work in the current scientific landscape.
Response: According to the opinions of the review experts, more recent references have been updated, and the details are shown:
Thank you again for your positive and constructive comments and suggestions on our manuscript.
We hope you will find our revised manuscript acceptable for publication.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper looks at the problem of voltage regulation in local grids caused by high solar energy input. It uses a neural network to predict voltage behavior and applies a two-step control method: first adjusting reactive power, then reducing active power, to manage voltage quickly and accurately. The results are useful, but the manuscript needs several improvements before it can be considered.
Comments:
- The abstract is unclear and should be rewritten to better explain what the paper is about and what the main results are.
- Novelty issues: The paper should clearly explain what is new about this method compared to other AI-based voltage regulation methods.
- There should be a comparison with traditional voltage control methods, either in the introduction or discussion.
- The reason for using a two-step method (reactive power first, then active) should be explained clearly. Why this order? Are there situations where it might not work well?
- How is voltage sensitivity estimated? Is this method reliable in changing network conditions?
- For PV Penetration Levels: Does the method work well under different levels of solar power (low, medium, high)?
- Would the method still work well in more complex real-world grids?
- The paper should compare this method with standard voltage control tools like tap changers, capacitor banks, or droop control.
- How fast can the system react to sudden changes in solar power? Is it quick enough for real-time use?
- Can the method handle larger or more complex networks? Is it efficient?
- The paper should mention any weaknesses of the method and suggest possible improvements.
- It would help to explain cases where the method might not work well, like during reverse power flow or unstable reactive demand. (Optional)
- Figure captions should explain clearly what each figure is showing.
- As for Language: the paper has many grammar mistakes and should be carefully proofread.
Author Response
Data-Driven Voltage Coordinated Control Strategy for Distri-bution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity
Ziwei Cheng 1*, Lei Wang 1, Can Su 1, Runtao Zhang 1, Xiaocong Li 2, Bo Zhang 2
- State Grid Hebei Electric Power Research Institute, Shijiazhuang 050021, China
- Key Laboratory of Distributed Energy Storage and Micro-grid of Hebei Province
North China Electric Power University, Baoding 071003, China
Manuscript Number: sustainability-3577313
Initial Submitted: 25 Mar 2025.
Decision: 11 Apr 2025.
Dear Editors and Reviewers:
Thanks for your valuable comments concerning our manuscript entitled "Data-Driven Voltage Coordinated Control Strategy for Distribution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity " (ID: sustainability-3577313). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made correction which we hope meet with approval.
A detailed response to the associate editor’s and the reviewers’ comments/suggestions has addressed and reported in the following parts, in which the font type of the suggestions is bold and the corresponding response of the author is common font. We have tried to meet all the requests made by the reviewers.
In the revised version of the manuscript the modified parts have been highlighted with the red color fonts.
Responds to the reviewer’s comments:
Reviewer #2:
- The abstract is unclear and should be rewritten to better explain what the paper is about and what the main results are.
Response: The main research contents and achievements of this paper are concentrated in two aspects: the data-driven method for obtaining power voltage sensitivity and the cooperative control of active distribution network voltage based on the two-stage mode of reactive power compensation/active power reduction.
Based on the reviewers' comments, the author reorganized the abstract section, highlighting the main research content and innovation points of the paper.
- Novelty issues: The paper should clearly explain what is new about this method compared to other AI-based voltage regulation methods.
Response: As noted by the reviewers, several scholars have investigated the use of artificial intelligence (AI) methods, such as data-driven approaches, to achieve voltage optimization and regulation in active distribution networks. These studies primarily focus on two aspects: power flow calculation in active distribution networks and the formulation and solution of optimization models for these networks[1-3]. Nevertheless, data-driven methods for voltage optimization and regulation in active distribution networks encounter challenges such as complex environmental interactions, substantial consumption of training time and resources, difficulties in real-time power flow data updates, limited generalization capabilities of optimization models, and ambiguous optimization decision-making mechanisms. These issues significantly hinder the practical feasibility of data-driven voltage optimization and regulation methods in engineering applications.
In contrast, the method proposed in this paper is grounded in the fundamental control mode of photovoltaic participation in voltage optimization and regulation within distribution networks. This approach offers strong physical interpretability, a well-defined voltage regulation control mechanism, and leverages data-driven power voltage sensitivity calculations to markedly enhance the efficiency of the overall voltage regulation process, thereby facilitating its implementation in engineering contexts.
[1] LIU Haotian, WU Wenchuan. Two-stage deep reinforcement learning for inverter-based volt-var control in active distribution networks[J]. IEEE Transactions on Smart Grid, 2021, 12(3): 2037-2047.
[2] SHIN M, CHOI D H, KIM J. Cooperative management for PV/ESS-enabled electric vehicle charging stations:a multiagent deep reinforcement learning approach[J]. IEEE Transactions on Industrial Informatics, 2020, 16(5): 3493-3503.
[3] ZHANG Ying, WANG Xinan, WANG Jianhui, et al. Deep reinforcement learning based volt-var optimization in smart distribution systems[J]. IEEE Transactions on Smart Grid, 2021, 12(1): 361-371.
- There should be a comparison with traditional voltage control methods, either in the introduction or discussion.
Response: Among the traditional reactive power control equipment in the distribution network, OLTC, capacitor bank, and CB have the characteristics of large capacity, low cost, and significant steady-state regulation effect. However, the response speed of equipment such as switching capacitor banks and disconnectors is slow, which is insufficient to cope with the high-frequency and rapid fluctuations of the voltage in the distribution network with a high proportion of renewable energy access. Reactive power compensation devices such as SVC and SVG have a fast regulation response speed, but the equipment investment cost is relatively high and they are not suitable for large-scale application. The continuous increase in the penetration rate of distributed photovoltaic power in distribution networks not only poses challenges to the safe and stable operation of distribution networks, but also endows distributed photovoltaic power sources with great potential to participate in the voltage regulation of distribution networks. They have the characteristics of rapid and flexible response, continuous and controllable regulation, and no need for additional investment. The IEEE 1547.8 working group advocates making full use of distributed photovoltaic power sources to achieve reactive power and voltage regulation.
- The reason for using a two-step method (reactive power first, then active) should be explained clearly. Why this order? Are there situations where it might not work well?
Response: In this paper, the regulation sequence of "reactive power first and then active power" is adopted, with the aim of reducing PV active power output and avoiding curtailment of PV and maximizing the reactive voltage regulation capacity of PV inverters. Potential limitations: This paper only starts from the aspect of the lowest active power regulation, and regulates the overall voltage level of the system, and if the reactive/active power regulation cost is considered, the control strategy proposed in this paper may have the situation that the regulation cost is not the optimal solution.
- How is voltage sensitivity estimated? Is this method reliable in changing network conditions?
Response: The voltage sensitivity proposed in this paper is firstly based on the inversion operation of the Jacobian matrix in the power flow equation, and the sensitivity data of photovoltaic nodes to other nodes are partially selected as the neural network data input set for relational mapping, so as to obtain the estimated voltage sensitivity mapping relationship model. It is worth noting that since the voltage sensitivity value fluctuates very little and is completely dependent on the grid structure, it can be approximated that this prediction relationship is still applicable without changing the system network structure.
In order to verify the validity and adaptability of the proposed voltage-power sensitivity estimation method, the typical distribution systems of IEEE33 nodes and 141 nodes are respectively verified and analyzed in this paper.
- For PV Penetration Levels: Does the method work well under different levels of solar power (low, medium, high)?
Response: As highlighted by the reviewers, to validate the effectiveness of the proposed control strategy, comparisons must be conducted from multiple perspectives, including various operating conditions and photovoltaic integration scenarios. Consequently, in this paper, the IEEE 33-node and 141-node typical distribution systems are respectively analyzed under different photovoltaic penetration rates for verification purposes. The photovoltaic penetration rates corresponding to these systems are randomly selected to ensure the generality of the examples. Owing to the length constraints of the paper, the impact of varying photovoltaic penetration rates on distribution systems of the same type is not further elaborated upon.
- Would the method still work well in more complex real-world grids?
Response: As the reviewers pointed out, the correctness verification part of the strategy proposed in this paper is carried out using typical IEEE examples. To further illustrate the effectiveness of the proposed method, its application effect in the actual power grid should be verified. However, it should be noted that the power system has extremely high requirements for power supply safety and reliability, and it is very difficult to conduct system-level experimental research in the power system. The conventional verification schemes for system-level research in this field are all carried out for typical IEEE systems, which is a recognized approach. The suggestions of the reviewers have practical engineering significance. The research team where the author belongs will also expand the experimental verification ideas in the next research work, strive to seek the experimental environment of the actual power grid, and strive to apply the proposed strategy in engineering as soon as possible.
- The paper should compare this method with standard voltage control tools like tap changers, capacitor banks, or droop control.
Response: There is mechanical wear and tear in the tap-changer, the overall voltage control accuracy is low, and the node voltage level may vary greatly; The capacitor bank cannot smoothly output, the response speed is slightly slow, and the voltage leveling cannot be completed by reactive power adjustment alone when the voltage exceeds the limit severely. The droop control relies on the voltage out-of-limit horizontal segmentation to adjust the voltage, and there is still a situation that the reactive power adjustment is too large.
- How fast can the system react to sudden changes in solar power? Is it quick enough for real-time use?
Response: The primary focus of this article is the rapid and precise implementation of voltage optimization and regulation in distribution networks, with an emphasis on the methodologies employed for such optimization and regulation. Typically, the Maximum Power Point Tracking (MPPT) system of a photovoltaic power supply adjusts its operating point at intervals ranging from 10ms to 100ms. However, the monitoring and perception of the actual operational status of the distribution network are not aligned with the time resolution of the photovoltaic MPPT system. The Supervisory Control and Data Acquisition (SCADA) system of the distribution network operates with sampling intervals at the minute level. From the perspective of voltage optimization and regulation requirements for real-world distribution networks, the time resolution for optimization and regulation does not need to reach the millisecond level. Consequently, the proposed control strategy effectively aligns with the sampling and regulation time resolution requirements of practical distribution networks.
- Can the method handle larger or more complex networks? Is it efficient?
Response: As the reviewers pointed out, the correctness verification part of the strategy proposed in this paper is carried out using typical IEEE examples. To further illustrate the effectiveness of the proposed method, its application effect in the actual power grid should be verified. However, it should be noted that the power system has extremely high requirements for power supply safety and reliability, and it is very difficult to conduct system-level experimental research in the power system. The conventional verification schemes for system-level research in this field are all carried out for typical IEEE systems, which is a recognized approach. The suggestions of the reviewers have practical engineering significance. The research team where the author belongs will also expand the experimental verification ideas in the next research work, strive to seek the experimental environment of the actual power grid, and strive to apply the proposed strategy in engineering as soon as possible.
- The paper should mention any weaknesses of the method and suggest possible improvements.
Response: If the strategy is only regulated from the power adjustment amount, because the reactive/active power adjustment cost is different, a variety of different control costs are introduced, and the control cost may not be the lowest when adjusting according to this strategy, that is, the problem of the lowest control cost and the lowest power adjustment cannot be solved at the same time.
- It would help to explain cases where the method might not work well, like during reverse power flow or unstable reactive demand.
Response: Since the capacity of the photovoltaic inverter is fixed, if there is only a photovoltaic power supply in the system and no other adjustment unit, when the photovoltaic active power output reaches the limit, the voltage is seriously exceeded, and the photovoltaic unit with large power adjustable capacity happens to be located in the position of low voltage sensitivity, that is, the distribution of the adjustable unit is unbalanced, which is easy to lead to the voltage exceeding the limit after reactive/active power regulation. In fact, when there is the uneven distribution of the above-mentioned photovoltaic regulated units, the voltage over-limit level will also be reduced accordingly, and it is unknown which is stronger or weaker than the specific over-limit level and the corresponding regulation ability.
- Figure captions should explain clearly what each figure is showing.
Response: Based on the opinions of the reviewers, the author specifically examined the textual explanations corresponding to all the figures in the text and provided supplementary explanations for the parts that were not clearly expressed.
- As for Language: the paper has many grammar mistakes and should be carefully proofread.
Response: We apologize for the inconvenience caused to the reviewers by some grammatical errors and clerical mistakes in the text. In the revised draft, the author carefully checked the grammatical expressions and corrected the typos. In the revised draft, the author carefully examined the grammatical expressions and corrected the typos, hoping to meet the relevant requirements of the reviewers and the journal.
Thank you again for your positive and constructive comments and suggestions on our manuscript.
We hope you will find our revised manuscript acceptable for publication.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors(Lines 12-26)
The abstract clearly states objectives but lacks details on novelty.
Explicitly clarify how the proposed method differs or advances beyond current literature to underline the novelty.
(Lines 30-95)
The literature review adequately covers methods but lacks critical comparison in terms of practical applicability or complexity.
Enhance the introduction by briefly summarizing the weaknesses or limitations of each referenced method, clearly setting the stage for the proposed approach.
(Lines 96-136)
Approximate sensitivity calculation method limitations not sufficiently detailed.
Clearly describe scenarios or types of networks where approximation might cause critical errors, and specify the magnitude of potential cumulative errors.
(Lines 137-161)
The choice of the neural network architecture (e.g., hidden layers, number of neurons) and parameters are not explicitly justified.
Provide a more explicit rationale or reference for choosing neural network parameters and architecture, possibly through comparative analysis or cross-validation results.
(Lines 163-244)
Lack of clear criteria or algorithmic flowchart detailing PV node selection for control.
Include an algorithmic flowchart or pseudocode summarizing the node and PV selection procedures clearly, facilitating replication and implementation.
(Lines 245-361)
The simulation scenarios and assumptions (e.g., fixed threshold for power factor at lines 249 and 334-335) are not fully justified or contextualized.
Clearly justify or discuss the impact of chosen assumptions (e.g., fixed threshold for power factor) on generalizability and robustness of results.
(Lines 245-361)
Lack of sensitivity or robustness analysis in the presented simulations.
Include sensitivity analysis or discuss robustness explicitly, showing how variations in PV capacity, load distribution, and other parameters affect the control effectiveness.
(Lines 296-311)
Maximum prediction errors reported (~9.55%-9.93%) are noted but implications are insufficiently discussed.
Explicitly discuss the practical implications of the prediction error margins identified. Clarify in which operational scenarios these errors could significantly impact performance.
(Lines 356-381)
Comparative analysis of sensitivity calculation methods does not quantify or visually depict advantages/disadvantages.
Enhance clarity by adding comparative graphical results (e.g., bar charts or radar plots) highlighting explicit differences in performance (computation time, power adjustments) among the three methods tested.
(Lines 382-401)
Conclusions are somewhat generic and fail to summarize concrete numerical outcomes or improvements achieved by the proposed method.
Summarize key numerical improvements (percentages of power savings, speed-up achieved in calculation) demonstrated by the proposed approach compared to traditional methods.
(Lines 397-401)
Future research directions are too general, lacking specificity or connection to current limitations.
Define more specific future work directions. Suggest particular areas (e.g., advanced AI models, real-time hardware implementation, cybersecurity implications) that can build on identified limitations or current study results.
Author Response
Data-Driven Voltage Coordinated Control Strategy for Distri-bution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity
Ziwei Cheng 1*, Lei Wang 1, Can Su 1, Runtao Zhang 1, Xiaocong Li 2, Bo Zhang 2
- State Grid Hebei Electric Power Research Institute, Shijiazhuang 050021, China
- Key Laboratory of Distributed Energy Storage and Micro-grid of Hebei Province
North China Electric Power University, Baoding 071003, China
Manuscript Number: sustainability-3577313
Initial Submitted: 25 Mar 2025.
Decision: 11 Apr 2025.
Dear Editors and Reviewers:
Thanks for your valuable comments concerning our manuscript entitled "Data-Driven Voltage Coordinated Control Strategy for Distribution Networks with High Proportion of Renewable Energy Based on Voltage-Power Sensitivity " (ID: sustainability-3577313). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied the comments carefully and have made correction which we hope meet with approval.
A detailed response to the associate editor’s and the reviewers’ comments/suggestions has addressed and reported in the following parts, in which the font type of the suggestions is bold and the corresponding response of the author is common font. We have tried to meet all the requests made by the reviewers.
In the revised version of the manuscript the modified parts have been highlighted with the red color fonts.
Responds to the reviewer’s comments:
Reviewer #3:
- Lines 12-26 The abstract clearly states objectives but lacks details on novelty. Explicitly clarify how the proposed method differs or advances beyond current literature to underline the novelty.
Response: In this paper, compared with the traditional control methods, the reactive/active sequence control of neural network + voltage sensitivity is combined: firstly, machine learning is used to improve the efficiency and accuracy of voltage sensitivity; Second, the sequential control method is combined to reduce the power adjustment of the traditional control strategy, which has certain advantages in solving speed and power adjustment.
- Lines 30-95 The literature review adequately covers methods but lacks critical comparison in terms of practical applicability or complexity. Enhance the introduction by briefly summarizing the weaknesses or limitations of each referenced method, clearly setting the stage for the proposed approach.
Response: According to the reviewers' comments, the introduction has been revised to include more precise expressions concerning the practical applicability and complexity of distribution network voltage regulation. Additionally, the latest references have been incorporated, and the content has been systematically organized and summarized from the perspectives of voltage power sensitivity acquisition and distribution network voltage regulation.
- Lines 96-136 Approximate sensitivity calculation method limitations not sufficiently detailed. Clearly describe scenarios or types of networks where approximation might cause critical errors, and specify the magnitude of potential cumulative errors.
Response: The approximate sensitivity calculation depends on the network topology, resistance, and reactance values. Based on the power flow equation, the standard voltage sensitivity is calculated based on the inversion of the Jacobian matrix, and in fact, the voltage sensitivity of each node fluctuates slightly but changes slightly each time it is calculated. Since the approximate sensitivity is calculated by using the fixed impedance value for each calculation, there will be errors in the voltage sensitivity compared with the standard inversion method during a single adjustment, and the errors will accumulate and amplify when multiple photovoltaic regulation units are repeatedly adjusted, especially with the increase of photovoltaic adjustable capacity, this difference will be more obvious.
- Lines 137-161 The choice of the neural network architecture (e.g., hidden layers, number of neurons) and parameters are not explicitly justified. Provide a more explicit rationale or reference for choosing neural network parameters and architecture, possibly through comparative analysis or cross-validation results.
Response: According to the loss function, the backpropagation algorithm is used to calculate the gradient of each weight and bias parameter to the loss, and then update the network parameters accordingly. In the process of error backpropagation, according to the gradient of the loss function, the error is backpropagated from the output layer to the input layer, and the gradient of each layer is calculated. According to the gradient descent algorithm, the backpropagation is used to calculate the weights and bias parameters in the gradient update network to reduce the loss, and the expression is shown as:
where λ is the learning rate, which is generally defined as ∈(0,1).
- Lines 163-244 Lack of clear criteria or algorithmic flowchart detailing PV node selection for control. Include an algorithmic flowchart or pseudocode summarizing the node and PV selection procedures clearly, facilitating replication and implementation.
Response: The authors have updated the references to add references from 2025. In this paper, the node and photovoltaic selection methods are briefly stated, as shown in Figure ***. In this example, all node voltages are arranged in descending order, and the node with the deepest degree of over-limit is selected as the target node. Calculate the voltage-power sensitivity of each photovoltaic power supply connected to the distribution network, arrange all the voltage and power sensitivities in descending order, and select the photovoltaic power source with the largest value for control.
- Lines 245-361 The simulation scenarios and assumptions (e.g., fixed threshold for power factor at lines 249 and 334-335) are not fully justified or contextualized. Clearly justify or discuss the impact of chosen assumptions (e.g., fixed threshold for power factor) on generalizability and robustness of results.
Response: In this paper, the focus is on the descending regulation of voltage sensitivity under neural network, and the proposed control strategy is a fast voltage regulation method. In the proposed control strategy, the fixed power factor threshold is selected only to illustrate the necessity and superiority of using photovoltaic sequential active power reduction in the case of voltage depth exceeding the limit and insufficient reactive power regulation margin. Taking the power factor threshold of the PV adjustable unit as the control variable, compared with other active control methods, the active power reduction of the proposed control strategy is still the lowest.
- Lines 245-361 Lack of sensitivity or robustness analysis in the presented simulations. Include sensitivity analysis or discuss robustness explicitly, showing how variations in PV capacity, load distribution, and other parameters affect the control effectiveness.
Response: According to the opinions of the review experts, the paper supplements the comparative analysis of the voltage regulation process of the distribution network corresponding to the proposed strategy under two conditions: the same photovoltaic capacity and different photovoltaic capacities.
- Lines 296-311 Maximum prediction errors reported (9.55%-9.93%) are noted but implications are insufficiently discussed. Explicitly discuss the practical implications of the prediction error margins identified. Clarify in which operational scenarios these errors could significantly impact performance.
Response: Taking the system voltage depth limit as an example, if the adjustable capacity of multiple PV units is large, and all of them are PV that should be operated, and the voltage sensitivity error is positive/negative, the increase of multiple cumulative regulation power adjustment is obvious.
- Lines 356-381 Comparative analysis of sensitivity calculation methods does not quantify or visually depict advantages/disadvantages. Enhance clarity by adding comparative graphical results (e.g., bar charts or radar plots) highlighting explicit differences in performance (computation time, power adjustments) among the three methods tested.
Response: In the actual calculation, in terms of speed, the approximate sensitivity can be directly called because it only needs to be calculated once in advance and the subsequent adjustment can be called, so there is no need to repeat the calculation; However, the inversion method of the current Jacobi matrix needs to calculate the matrix and find the inversion every time, and the solution speed becomes slower with the increase of the matrix dimension, so the calculation speed is the lowest. The proposed control method uses the neural network to obtain the mapping relationship, and the dynamic photovoltaic adjustable unit variables and load variables are input and calculated, and then the solution is fast and the solution speed is moderate. The strategic advantages in terms of power adjustment are still illustrated.
- Lines 382-401 Conclusions are somewhat generic and fail to summarize concrete numerical outcomes or improvements achieved by the proposed method. Summarize key numerical improvements (percentages of power savings, speed-up achieved in calculation) demonstrated by the proposed approach compared to traditional methods.
Response: The control strategy proposed in this paper focuses on the rapid regulation of the out-of-limit voltage and the minimum power regulation requirements of the distribution network, and the node voltage only needs to be controlled to the threshold range to meet the requirements, and the voltage percentage is not analyzed in depth.
- Lines 397-401 Future research directions are too general, lacking specificity or connection to current limitations. Define more specific future work directions. Suggest particular areas (e.g., advanced AI models, real-time hardware implementation, cybersecurity implications) that can build on identified limitations or current study results.
Response: According to the opinions of the review experts, supplementary elaboration was made on the final outlook part of the conclusion.
Thank you again for your positive and constructive comments and suggestions on our manuscript.
We hope you will find our revised manuscript acceptable for publication.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsThe revised manuscript has been substantially improved and is now suitable for publication.
Author Response
Comments:The revised manuscript has been substantially improved and is now suitable for publication.
Response:Thank you to the reviewers for your hard work during the process of manuscript review and revision.
Reviewer 3 Report
Comments and Suggestions for AuthorsNeural network architecture justification
Please provide a more explicit justification for the selection of the neural network architecture (e.g., number of hidden layers, neurons per layer, activation functions). If available, include results from hyperparameter tuning, cross-validation, or comparative experiments demonstrating that the selected configuration offers the best trade-off between accuracy and computational efficiency.
You may also consider referencing recent work on AI models applied to photovoltaic systems, e.g.: https://www.nature.com/articles/s41467-025-56227-9
Lack of pseudocode or flowchart for PV node selection
To enhance reproducibility, we strongly recommend including a flowchart or pseudocode describing the PV node and control point selection logic. The current verbal description is helpful, but a visual or algorithmic representation would significantly clarify the implementation steps.
You could consider using a modular approach with clear decision criteria and thresholds. This is especially important if future studies aim to simulate different PV technologies (e.g., silicon-based vs. perovskite-based modules) with distinct behaviors under partial shading or variable irradiance.
No graphical comparison between methods
Please include comparative visualizations (e.g., bar charts, radar plots, or performance tables) to highlight the differences in computational cost, speed, and power adjustment efficiency among the three tested methods. Quantitative graphical summaries would provide a clearer view of the trade-offs and performance gains achieved by your proposed approach.
If space allows, it would also be useful to compare simulation results using different types of PV modules, for example perovskite-based panels, which may behave differently under dynamic control strategies. A good starting point could be: https://www.mdpi.com/2673-9941/3/3/20
Conclusions lack numerical results
Please revise the conclusion section to include concrete numerical outcomes from your simulations. For instance, specify the percentage reduction in power adjustment achieved by your method compared to traditional control strategies, as well as any reduction in computation time.
Quantifying improvements will not only reinforce the practical impact of your method but also help contextualize its benefits when applied to emerging PV technologies like perovskite solar panels, which are known for their rapid power output fluctuations and require fast-response control mechanisms.
Author Response
Reviewer #3:
Comment 1:Please provide a more explicit justification for the selection of the neural network architecture (e.g., number of hidden layers, neurons per layer, activation functions). If available, include results from hyperparameter tuning, cross-validation, or comparative experiments demonstrating that the selected configuration offers the best trade-off between accuracy and computational efficiency.
You may also consider referencing recent work on AI models applied to photovoltaic systems, e.g.: https://www.nature.com/articles/s41467-025-56227-9
Response: Based on the opinions of the review experts, relevant contents have been supplemented in the revised version.
Multiple experiments were conducted using Bayesian optimization[1-4], and the optimal parameters and structure of the BP neural network were determined as follows. Among them, the number of iterations epoch is 1000, the optimizer selects Tainlm, the learning rate is 0.001, the dynamic increase and decrease factors mu_dec=0.1, mu_inc=10, and the maximum allowable value of the learning rate mu_max=10 are set.
[1] Masum, H. Shahriar, H. Haddad, et al., "Bayesian Hyperparameter Optimization for Deep Neural Network-Based Network Intrusion Detection," 2021 IEEE International Conference on Big Data (Big Data), Orlando, FL, USA, 2021, pp. 5413-5419.
[2] C. Tung, B. Patel and P. T. Lin, "Multi-Fidelity Design Optimization (MFDO) for Fully Connected Deep Neural Network (FCDNN)," 2024 20th IEEE/ASME International Conference on Mechatronic and Embedded Systems and Applications (MESA), Genova, Italy, 2024.
[3] E. Puentes G., C. J. Barrios H. and P. O. A. Navaux, "Hyperparameter Optimization for Convolutional Neural Networks with Genetic Algorithms and Bayesian Optimization," 2022 IEEE Latin American Conference on Computational Intelligence (LA-CCI), Montevideo, Uruguay, 2022, pp. 1-5.
[4] Monzen, F. Stroebl, H. Palm and C. M. Hackl, "Multi-objective Hyperparameter Optimization of Artificial Neural Networks for Optimal Feedforward Torque Control of Synchronous Machines," in IEEE Open Journal of the Industrial Electronics Society, vol. 5, pp. 41-53, 2024.
Coment 2: To enhance reproducibility, we strongly recommend including a flowchart or pseudocode describing the PV node and control point selection logic. The current verbal description is helpful, but a visual or algorithmic representation would significantly clarify the implementation steps.
You could consider using a modular approach with clear decision criteria and thresholds. This is especially important if future studies aim to simulate different PV technologies (e.g., silicon-based vs. perovskite-based modules) with distinct behaviors under partial shading or variable irradiance.
Response: According to the opinions of the review experts, a process for selecting the regulation sequence of the controlled PV power source has been added. For details, please refer to the revised version.
Coment 3: Please include comparative visualizations (e.g., bar charts, radar plots, or performance tables) to highlight the differences in computational cost, speed, and power adjustment efficiency among the three tested methods. Quantitative graphical summaries would provide a clearer view of the trade-offs and performance gains achieved by your proposed approach.
If space allows, it would also be useful to compare simulation results using different types of PV modules, for example perovskite-based panels, which may behave differently under dynamic control strategies. A good starting point could be: https://www.mdpi.com/2673-9941/3/3/20
Response: Based on the opinions of the review experts, a comparison chart of the results of different methods in terms of pressure regulation effect and calculation time was supplemented.
Comment 4: Please revise the conclusion section to include concrete numerical outcomes from your simulations. For instance, specify the percentage reduction in power adjustment achieved by your method compared to traditional control strategies, as well as any reduction in computation time.
Quantifying improvements will not only reinforce the practical impact of your method but also help contextualize its benefits when applied to emerging PV technologies like perovskite solar panels, which are known for their rapid power output fluctuations and require fast-response control mechanisms.
Response: According to the opinions of the review experts, the conclusion section has supplemented the relevant descriptions of the numerical comparison results of the proposed strategy and the traditional method in terms of regulation time and regulation effect. For details, please refer to the revised version.
Author Response File: Author Response.docx