State Analysis of Grouped Smart Meters Driven by Interpretable Random Forest

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents an innovative approach for assessing smart meter group status using an explainable random forest model, enhanced by expert-guided grouping strategies and the SHAP interpretability framework. By classifying multi-source features into validation-, production-, and deployment-related categories, combined with production batch and distribution network grouping, it establishes a robust and practical framework for smart meter maintenance in power distribution systems. The study focuses on improving the efficiency and reliability of failure rate prediction, which is highly relevant to smart grid management.

To further enhance clarity and rigor, I recommend the following minor revisions:
(1) Provide explicit mathematical definitions or references for key statistical parameters (e.g., Theil index, kurtosis) in Section 3.2 to ensure clarity.
(2) Enrich the description of Figure 2 (violin plots) in Section 4.1 by detailing the normalization process and key insights from feature distributions.
(3) Briefly justify how the selected features were validated (e.g., via statistical analysis or expert consultation) in Section 3.2 to strengthen the methodology.

These modifications would improve the manuscript’s readability and impact.

Comments on the Quality of English Language

The English could be improved to more clearly express the research.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The explanations of the abbreviations appear too late. For example, SHAP is not explained until page 4.

Comments on the article:
The manner of description is imprecise. It was not explained what criteria were used to evaluate smart meters. The purpose of using diagnostic algorithms was not explained.
Smart meters operate in a feedback system. Information is transmitted bidirectionally. All diagnostic information is transferred from the meter to the billing server automatically and does not require any additional diagnostic algorithms.
Moreover, as we know, each recipient has an individual electricity usage profile, therefore it is impossible to remotely assess the accuracy of measurement and fairness of billing.
That is why I believe that the initial description in this article is insufficient. The results obtained are also unclear.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Mandatory statements like Author Contributions or Funding are missing. Please insert them inside the text according to the template.

Equations and relations have no numbers.

The section called Introduction is numbered as 0. Please renumber them starting from 1.

Is there any power range for the smart meters?

More future studies are required, but not specified.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have improved some aspects of this study. However, I must draw the authors' attention to one important issue:

It is necessary to specify precisely which quantities can be measured (in the introduction). Then, indicate which of them are observed by the authors.
It is best to describe or graphically present input and output information, and to present possible interferences.
It's also advisable to provide a more in-depth look at the features of the selected electricity meter. Expand the description to include the SMART-METERING system.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf