Adaptive Edge–Cloud Framework for Real-Time Smart Grid Optimization with IIoT Analytics

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper proposes GridOpt, an adaptive edge-cloud framework for real-time smart grid optimization with IIoT analytics. It addresses challenges in latency, scalability, and security caused by integrating Distributed Energy Resources (DERs). GridOpt leverages edge computing for real-time state estimation, anomaly detection, and control using TDNN and ADMM, while offloading intensive tasks like predictive maintenance (via HTM models) to the cloud. The framework integrates homomorphic encryption, zero-knowledge proofs, a private blockchain with PPBFT consensus, and an ontology-based interoperability layer. Simulations show GridOpt achieves the lowest latency (76 ms), energy consumption (25 J), high scalability (>10 req/s), and low resource utilization (54%) compared to existing approaches. I believe this paper can be accept after minor revision.

1. The figures in the manuscript, particularly Figure 1, suffer from poor readability due to small font sizes and low visual clarity. All textual elements (labels, annotations, axis titles, etc.) should be significantly enlarged and optimized for legibility, especially considering that the architecture diagram is central to understanding the proposed GridOpt framework. Please revise all figures to ensure they meet publication-quality standards and are easily interpretable by readers.
2. The paper employs several AI models (e.g., TDNN and HTM) for predictive maintenance and real-time analytics, but it lacks essential details regarding their implementation. Specifically, the authors should provide model architectures, hyperparameter settings (e.g., learning rate, number of layers, window size, sparsity levels), training/validation/test data splits, and evaluation metrics (such as accuracy, precision, recall, F1-score, or RMSE). Without this information, it is difficult to assess the reproducibility, robustness, and actual performance gains of the proposed AI components. Please include these details in a dedicated subsection or table.
3. The manuscript requires thorough proofreading to eliminate typographical and grammatical errors. A notable example is in the Keywords section, where “Artificial Intelligenc” should be corrected to “Artificial Intelligence”. Such oversights undermine the professionalism and credibility of the work. The authors are advised to carefully review the entire text for similar inconsistencies, spelling mistakes, and formatting issues before resubmission.

Author Response

1. Summary
Thank you very much for taking the time to review this manuscript. Please find detailed responses below and the corrections highlighted in the re-submitted files.
2. Point-by-point response to Comments and Suggestions for Authors
Comment 1: The figures in the manuscript, particularly Figure 1, suffer from poor readability due to small font sizes and low visual clarity. All textual elements (labels, annotations, axis titles, etc.) should be significantly enlarged and optimized for legibility, especially considering that the architecture diagram is central to understanding the proposed GridOpt framework. Please revise all figures to ensure they meet publication-quality standards and are easily interpretable by readers.
Response 1: I sincerely thank the reviewer for this constructive and valuable comment. We agree that clear and well-readable figures are essential for effectively conveying the proposed GridOpt framework, particularly for architecture-centric illustrations that play a key role in understanding the methodology.
In response, all figures in the manuscript have been carefully revised to enhance visual clarity and readability. Specifically, the font sizes of all textual elements, including labels, annotations, legends, and axis titles, have been increased to ensure comfortable readability in both digital and printed formats. Figure 1, which illustrates the overall GridOpt architecture, has been redesigned with improved spacing, higher resolution, and clearer component boundaries to better highlight the functional relationships among system modules. Similar enhancements have been applied consistently across the remaining figures to ensure uniform presentation quality throughout the manuscript.
[In the revised manuscript, this change can be found – page number# 05 and Figure 2.]
Comments 2: The paper employs several AI models (e.g., TDNN and HTM) for predictive maintenance and real-time analytics, but it lacks essential details regarding their implementation. Specifically, the authors should provide model architectures, hyperparameter settings (e.g., learning rate, number of layers, window size, sparsity levels), training/validation/test data splits, and evaluation metrics (such as accuracy, precision, recall, F1-score, or RMSE). Without this information, it is difficult to assess the reproducibility, robustness, and actual performance gains of the proposed AI components. Please include these details in a dedicated subsection or table.
Response 2: I sincerely thank the reviewer for this insightful and technically important comment. We agree that providing implementation-level details of the employed AI models is essential for ensuring transparency, reproducibility, and a fair assessment of the reported performance gains.
In response, we have added a concise yet complete subsection that summarizes the architectural configurations, key hyperparameter settings, data partitioning strategy, and evaluation criteria used for the TDNN and HTM models. The intention was to improve clarity while keeping the revisions
minimal and aligned with the current scope of the manuscript. To further enhance readability, the core parameters are presented in a compact table format, allowing readers to quickly understand and replicate the experimental setup.
[In the revised manuscript, this change can be found – page number# 13, paragraph 5, section-“Simulation Outcomes” and line 331-33 and the page no#14, table2.]
3. Response to Comments on the Quality of English Language
Comment 03: The manuscript requires thorough proofreading to eliminate typographical and grammatical errors. A notable example is in the Keywords section, where “Artificial Intelligenc” should be corrected to “Artificial Intelligence”. Such oversights undermine the professionalism and credibility of the work. The authors are advised to carefully review the entire text for similar inconsistencies, spelling mistakes, and formatting issues before resubmission.
Response 03: Thank the reviewer for highlighting this important point regarding language quality and presentation. We fully agree that typographical and grammatical accuracy is essential for maintaining the professionalism and credibility of the manuscript.
In response, the entire manuscript has been carefully proofread to identify and correct spelling errors, grammatical inconsistencies, and minor formatting issues. The specific typo noted by the reviewer in the Keywords section has been corrected from “Artificial Intelligenc” to “Artificial Intelligence.” In addition, a thorough review was conducted across all sections to ensure consistent terminology, proper sentence structure, and uniform formatting throughout the paper.
[In the revised manuscript, this change can be found – page number# 01, “section-Keywords”, and line 16.]

Reviewer 2 Report

Comments and Suggestions for Authors

This paper proposes a hybrid solution for the above challenges using edge and cloud capabilities in the form of GridOpt is proposed. The proposed system uses edge computing capabilities for real-time analysis and cloud capabilities for processing large volumes of data. The proposed system incorporates various security measures such as homomorphic encryption and blockchain consensus mechanisms coupled with an interoperability framework that provides seamless integration of heterogeneous devices.

 

The wording of this article is generally clear and easy to understand, but the author still has some issues that need to be addressed:

 

1. In the AI-Driven Predictive Maintenancemodule, the functional division between the HTM model and the TDNN model is not clearly distinguished: both are used for state prediction and anomaly detection. It is necessary to explain why the dual model design is adopted and how it works together in practical applications (such as parallel computing, complementary verification, etc.) to avoid the problem of functional redundancy.
2. The experimental setting is not clear. The specific parameters of the experimental scenario have not been specified. It is recommended to provide a detailed description of the specific parameter settings for the experimental scenario to ensure reproducibility of the experiment.
3. Suggest checking the writing details.Some formulas in Section 3 are not numbered, it is recommended to add numbering.  There is a spelling error in the keyword 'Artificial Intelligent' (should be 'Artificial Intelligence')
4. Inadequate experimental comparisons . The experimental section compared the proposed method with baselines such as ECCGrid, EdgeApp, and JOintCS, but lacked comparison work from the past three years. It is recommended to add the latest methods from the past three years for comparison.
5. The existing work section did not review the latest works published in 2024 and 2025. Moreover, this section should follows a certain line to introduce the related works and presents the novelty of this paper compared to these existing works, which is very important.

 

In general, there are still some flaws in this paper that need to be added or modified, and the suggestion is "Authors should prepare a minor revision".

Author Response

1. Summary
Thank you very much for taking the time to review this manuscript. Please find detailed responses below and the corrections highlighted in the re-submitted files.
2. Point-by-point response to Comments and Suggestions for Authors
Comments 1: In the AI-Driven Predictive Maintenancemodule, the functional division between the HTM model and the TDNN model is not clearly distinguished: both are used for state prediction and anomaly detection. It is necessary to explain why the dual model design is adopted and how it works together in practical applications (such as parallel computing, complementary verification, etc.) to avoid the problem of functional redundancy.
Response 1: Thanks the reviewer for this thoughtful and technically relevant comment. We agree that clearly distinguishing the roles of the HTM and TDNN models is important to avoid any perception of functional overlap and to better explain the rationale behind the dual-model design adopted in the GridOpt framework.
In response, we clarify that the two models serve complementary and non-redundant functions within the predictive maintenance module. The TDNN is employed for short-term temporal forecasting of system states using fixed-window historical data, whereas the HTM model is responsible for continuous online learning and anomaly detection from streaming data. The HTM operates in real time and adapts incrementally to evolving system patterns, while the TDNN provides stable predictive outputs based on trained temporal dependencies. In practice, both models operate in parallel, and their outputs are used in a complementary manner to enhance reliability and reduce false detections.
[In the revised manuscript, this change can be found – page number# 13, paragraph 5, and line 331-333.]
Comments 2: The experimental setting is not clear. The specific parameters of the experimental scenario have not been specified. It is recommended to provide a detailed description of the specific parameter settings for the experimental scenario to ensure reproducibility of the experiment.
Response 2: I sincerely thank the reviewer for this constructive comment highlighting the importance of clearly reporting the experimental setting. We fully agree that explicit specification of experimental parameters is essential for reproducibility and transparency.
In response, the experimental setup has been clarified and consolidated in the Simulation Outcomes section. In addition to the descriptive explanation already provided, we have introduced a dedicated parameter table (Table 02) that summarizes all key simulation and experimental settings, including the simulation platform, dataset, grid components, AI model configurations, data split strategy, evaluation metrics, and comparative methods. This table
provides a compact and clear overview of the experimental scenario, allowing readers to easily replicate the setup.
[In the revised manuscript, this change can be found – page number# 14, paragraph 1, Table#02.]
Comment 03: Suggest checking the writing details. Some formulas in Section 3 are not numbered, it is recommended to add numbering. There is a spelling error in the keyword 'Artificial Intelligent' (should be 'Artificial Intelligence')
Response 03: I sincerely thank the reviewer for their careful attention to the writing details and presentation quality of the manuscript. We agree that consistent equation numbering and correct terminology are essential for clarity and proper referencing.
In response, all mathematical expressions in Section 3 have been carefully reviewed and are now consistently numbered to facilitate easy cross-referencing. The only exceptions are basic formulas appearing within theorem statements and their corresponding proofs, which follow standard presentation practice and therefore remain unnumbered. In addition, the spelling error in the Keywords section has been corrected from “Artificial Intelligent” to “Artificial Intelligence.” A brief language check was also conducted to ensure no similar issues persist.
[In the revised manuscript, this change can be found – page number# 01, paragraph 2, and line 16.]
Comment 04: Inadequate experimental comparisons . The experimental section compared the proposed method with baselines such as ECCGrid, EdgeApp, and JOintCS, but lacked comparison work from the past three years. It is recommended to add the latest methods from the past three years for comparison.
Response 04: Thanks the reviewer for this comment regarding the selection of comparative methods in the experimental evaluation. We appreciate the emphasis on including recent studies to strengthen the relevance of experimental comparisons.
In response, we would like to clarify that the baseline approaches selected in this study—ECCGrid, EdgeApp, and JOintCS—were chosen deliberately due to their strong methodological alignment and functional similarity with the proposed GridOpt framework. These approaches address closely related aspects of edge-assisted smart grid optimization, distributed intelligence, and IIoT-enabled energy management, which makes them more suitable for a fair and meaningful comparison than several recently published methods that focus on different problem formulations or system assumptions.
Although some works have appeared in the past three years, many of these either target narrower subproblems, rely on centralized architectures, or assume infrastructure capabilities that differ significantly from the scope and constraints considered in GridOpt. Including such methods would therefore risk introducing biased or inconclusive comparisons. For this reason, we retained the current comparison set to ensure consistency, relevance, and experimental fairness.
Comment 05: The existing work section did not review the latest works published in 2024 and 2025. Moreover, this section should follows a certain line to introduce the related works and presents the novelty of this paper compared to these existing works, which is very important.
Response 05: I sincerely thank the reviewer for this valuable comment concerning the coverage and organization of the related work. We fully agree that incorporating recent studies and clearly positioning the novelty of the proposed work are essential for strengthening the contribution of the manuscript.
In response, the Related Work section has been revised to include a discussion of recently published studies that are closely aligned with the scope of this work. These additions help establish a clearer research trajectory, moving from existing edge- and AI-enabled smart grid solutions toward the proposed GridOpt framework. The revised section now follows a more structured flow, highlighting the
limitations of current approaches and explicitly clarifying how GridOpt differs from and advances beyond existing methods in terms of adaptive optimization, distributed intelligence, and system-level integration.
[In the revised manuscript, this change can be found – page number# 04, paragraph1, and line 97-104.]

Reviewer 3 Report

Comments and Suggestions for Authors

There are following observations about this research:

1. The abstract should be re-written to illustrate the noble contribution effectively.
2. The introduction section is too short, pls add scenario diagram and effectively define the noble contributions.
3. Literature review section is also too short, there must have one table for comparison of published works with current work.
4. Pls re-design the figure 1 it is not clearly reflecting the case.
5. Pls add the complexity analysis of the algorithm 1 & 2.
6. In the proposed framework section, pls add the detailed flow chart highlighting the flow of research.
7. Pls revisit the equation number 21.
8. Pls add rationale for the statement "Strong convexity of the primal quadratic and full row rank of G imply strong monotonicity and Lipschitz continuity of the KKT map."
9. Pls explain the configuration parameters for result section.
10. What is the limitations of this research.
11. Pls revisit the equation number 28.
12. The conclusion section also needs revision for highlighting the summary of finding with precise facts.
13. There are typo and grammar correction needed in the manuscript.

Comments on the Quality of English Language

There are typo and grammar correction needed in the manuscript for some paragraphs.

Author Response

1. Summary
Thank you very much for taking the time to review this manuscript. Please find detailed responses below and the corrections highlighted in the re-submitted files.
2. Point-by-point response to Comments and Suggestions for Authors
Comments 1: The abstract should be re-written to illustrate the noble contribution effectively.
Response 1: I sincerely thank the reviewer for this valuable suggestion regarding the clarity and impact of the abstract. I agree that the abstract plays a critical role in conveying the core contribution and significance of the proposed work.
In response, the abstract has been carefully rewritten to more clearly and concisely highlight the main motivation, core contributions, and key outcomes of the proposed GridOpt framework. The revised abstract now emphasizes the novelty of the approach, its role in AI-driven optimization for smart grids, and the practical benefits demonstrated through simulation results. The wording has been refined to improve coherence and technical focus, while maintaining consistency with the content of the manuscript.
[In the revised manuscript, this change can be found – page number# 01, paragraph 1, and line 2-14.]
Comments 2: The introduction section is too short, pls add scenario diagram and effectively define the noble contributions.
Response 2: Thanks the reviewer for this constructive suggestion. The Introduction section has now been expanded to better establish the problem context and improve clarity. A scenario diagram has been added to illustrate the operational flow between DER units, edge nodes, and the cloud environment, providing a clearer understanding of how the proposed system functions in a practical smart grid setting. Additionally, the key contributions of the paper are now explicitly defined in a separate bullet-style at the end of the Introduction. These revisions clarify the motivation, highlight the novelty of the work, and ensure that the introduction presents the research contributions more effectively.
[In the revised manuscript, this change can be found – page number# 02, Figure. 1 And line 49-54.]
Comments 3: Literature review section is also too short, there must have one table for comparison of published works with current work.
Response 3: I sincerely thank the reviewer for this helpful recommendation. I agree that extending the Literature Review and presenting a structured comparison table will provide clearer context and highlight how the proposed work differs from prior studies.
In response, the Literature Review section has been expanded to include additional recent works that align with edge intelligence, smart grid optimization, and secure AI-driven energy management. To address the reviewer’s request, a comparison table (Table 01) has been added to summarize key characteristics of the existing approaches and contrast them with the proposed GridOpt framework. The
table briefly compares architecture type, optimization focus, security mechanisms, scalability considerations, and noted gaps that GridOpt resolves. This enhancement provides a clearer progression from prior work to the novelty of the current contribution.
[In the revised manuscript, this change can be found – page number# 04, Table# 01, and line 104-105.]
Comments 4: Pls re-design the figure 1 it is not clearly reflecting the case.
Response 4: Thanks the reviewer for noting the clarity issue with Figure 1. The figure has now been redesigned and updated to more clearly represent the operational case and structural flow of the proposed framework. The revised version improves visual hierarchy, enlarges key components, clarifies data routing between DER units, edge nodes, and the cloud layer, and removes visual clutter. The updated figure now presents the system stages and interactions in a more readable and case-accurate manner without altering the technical meaning of the architecture.
[In the revised manuscript, this change can be found – page number# 05, Figure. 2.]
Comments 5: Pls add the complexity analysis of the algorithm 1 & 2.
Response 5: Thank the reviewer for pointing out the need to include the complexity analysis of Algorithm 1 and Algorithm 2. We agree that reporting the computational complexity helps to clarify the efficiency and scalability of the proposed framework.
In response, we have added a brief computational complexity analysis for both algorithms in the methodology section, immediately after the description of Algorithm 2. The added text explains the dominant operations and summarizes the asymptotic time complexity in terms of the main problem parameters, while also outlining the memory requirements. This addition is concise and does not change the design or operation of the proposed methods; it only documents their computational behavior.
[In the revised manuscript, this change can be found – page number# 9-10, section#3.3.1, and line 210-221.]
Comments 6: In the proposed framework section, pls add the detailed flow chart highlighting the flow of research.
Response 6: Thanks the reviewer for this helpful suggestion. The framework illustration in the proposed framework section has now been updated to include a more detailed flow chart that clearly represents the step-by-step progression of the research flow, including data acquisition, edge-level processing, cloud coordination, decision integration, and feedback control. The revised figure provides a clearer visualization of the operational sequence and improves the overall understanding of the proposed system. Now the proposed figure 2 is redesigned and now highlight the flow.
[In the revised manuscript, this change can be found – page number# 5and Fig.2.]
Comments 7: Pls revisit the equation number 21.
Response 7: I thank the reviewer for noting the issue with Equation (21). The equation has now been revisited and corrected for consistency with the surrounding formulation and notation. The updated form has been set in the manuscript, and no changes to the technical results were required.
[In the revised manuscript, this change can be found – page number# 8 and line 174-178.]
Comments 8: Pls add rationale for the statement "Strong convexity of the primal quadratic and full row rank of G imply strong monotonicity and Lipschitz continuity of the KKT map."
Response 8: I thank the reviewer for noting the issue with Equation (21). The equation has now been revisited and corrected for consistency with the surrounding formulation and notation. The updated form has been set in the manuscript, and no changes to the technical results were required.
[In the revised manuscript, this change can be found – page number# 8 and line 174-178.]
Comments 9: Pls explain the configuration parameters for result section.
Response 9: Thanks the reviewer for this helpful observation regarding the clarity of configuration parameters in the results section. In response, the required configuration details have now been clearly presented through the inclusion of Table 02, which summarizes the simulation environment, model settings, network parameters, and evaluation metrics used throughout the experiments. This table provides a structured view of the parameters applied in the result section, ensuring transparency and reproducibility without altering the existing experimental outcomes.
[In the revised manuscript, this change can be found – page number# 14 and table 02.]
Comments 10: What is the limitations of this research.
Response 10: Thanks to the reviewer for this important remark. To address this point, a brief limitations statement has now been included at the end of the conclusion section. The added text acknowledges that the current study is evaluated in a controlled simulation environment, does not yet include hardware-level deployment, and assumes stable communication conditions between edge and cloud components. These limitations do not affect the validity of the current results but indicate areas for future extension.
[In the revised manuscript, this change can be found – page number# 16-17 and line 412-415.]
Comments 11: Pls revisit the equation number 28.
Response 11: Thanks to the reviewer for highlighting the issue with Equation (28). The equation has now been revisited and adjusted for consistency with the preceding formulation and notation. The updated expression is correctly set in the manuscript, and no changes to the technical results or interpretations were required.
[In the revised manuscript, this change can be found – page number# 9 and line 198.]
Comments 12: The conclusion section also needs revision for highlighting the summary of finding with precise facts.
Response 12: Thanks to the reviewer for this valuable suggestion. The conclusion section has now been revised to present a clearer summary of the study’s outcome with more precise and factual statements. The updated conclusion briefly restates the objective, highlights the core performance findings (latency, energy consumption, scalability, and resource utilization), and ties them directly to the presented evaluation results. This revision improves clarity while maintaining technical accuracy and does not alter any reported results.
[In the revised manuscript, this change can be found – page number# 16-17 and line 401-415.]
3. Response to Comments on the Quality of English Language
Comment 13: There are typo and grammar correction needed in the manuscript.
Response 13: Thanks to the reviewer for pointing out the need for typographical and grammatical corrections. In response, the full manuscript has been carefully proofread to correct spelling inconsistencies, sentence structure issues, and minor formatting errors. These revisions focus solely on improving clarity and language quality and do not alter the technical content, results, or interpretations presented in the paper.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Authors have addressed the queries, the manuscript may be accepted for publication.