Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents a transformer inrush current identification scheme using a CNN model suitable for embedded devices, leveraging single-terminal electrical quantities (current and voltage) of the transformer. The approach demonstrates significant practical merits and considerable potential for engineering applicability. There are some problems as follows:

1. The abstract mentions a comparative analysis of the model's identification speed, but this aspect is not substantiated in the main text. If this aspect hasn't been adequately addressed or supported with empirical evidence, it would be appropriate to remove such claims to maintain accuracy.
2. The text above Figure 18 contains erroneous bracket notation. Please verify and correct it accordingly.
3. Regarding the current waveforms in Figure 8, please specify whether they are derived from simulation or actual field measurements. If obtained through simulation, evidence demonstrating the consistency between these waveforms and real engineering scenarios should be provided. Relevant validation should be supplemented accordingly.

Author Response

Thank you for your comments and suggestions on our manuscript. Our responses to the reviewer‘s comments are provided below. You can also find the responses in the attached file.

Comments 1. The abstract mentions a comparative analysis of the model's identification speed, but this aspect is not substantiated in the main text. If this aspect hasn't been adequately addressed or supported with empirical evidence, it would be appropriate to remove such claims to maintain accuracy.

Response 1：Thank you for your valuable suggestion. We have modified the inaccurate phrasing in the abstract to improve clarity and accuracy. The changes have been marked in red font in the text.

Comments 2. The text above Figure 18 contains erroneous bracket notation. Please verify and correct it accordingly.

Response 2：Thank you for your valuable comments. The relevant text has been amended accordingly. The changes have been marked in red font in the text.

Comments 3. Regarding the current waveforms in Figure 8, please specify whether they are derived from simulation or actual field measurements. If obtained through simulation, evidence demonstrating the consistency between these waveforms and real engineering scenarios should be provided. Relevant validation should be supplemented accordingly.

Response 3：Thank you for your valuable comments. We have now clarified in the text that the data in Figure 8 are all authentic field data, which reinforces the practical basis and credibility of our case study. The changes have been marked in red font in the text.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

-Summarize the detailed explanation of the different previous algorithms applied for the classification of magnetizing inrush current.

- The discrimination of inrush with the faulty condition needs more elaboration. For example, describe the difference between( or the superiority of)  the proposed algorithm and other methods proposed in the literature, such as the total harmonics distortion‐based solution proposed for discrimination of the inrush condition from fault and other abnormal conditions using the short time Fourier transform technique.

-Figures 14 up to 17 have to be drawn clearly (larger size)

-The collection of real‐time data is better to be used based on laboratory hardware
setup instead of PASCAD results. However, this may be considered in your future work.

Author Response

Thank you for your comments and suggestions on our manuscript. Our responses to the reviewer‘s comments are provided below. You can also find the responses in the attached file.

Comments 1. Summarize the detailed explanation of the different previous algorithms applied for the classification of magnetizing inrush current.

Response 1：Thank you for your valuable suggestion. In line with your suggestion, we have analyzed and summarized the advantages and limitations of both the traditional identification methods and the AI-based techniques to ensure our viewpoints are presented clearly and unambiguously. The changes have been marked in red font in the text.

Comments 2. The discrimination of inrush with the faulty condition needs more elaboration. For example, describe the difference between( or the superiority of)  the proposed algorithm and other methods proposed in the literature, such as the total harmonics distortion‐based solution proposed for discrimination of the inrush condition from fault and other abnormal conditions using the short time Fourier transform technique.

Response 2：Thank you for your valuable comments. As suggested, the paper has been expanded to clarify the distinctions and advantages of our proposed method over both traditional and other AI-based identification schemes, thereby further highlighting the innovativeness of our research. The changes have been marked in red font in the text.

Comments 3. Figures 14 up to 17 have to be drawn clearly (larger size)

Response 3：Thank you for your valuable comments. In accordance with your suggestion, we have increased the size of the figures to enhance their clarity. The changes have been marked in red font in the text.

Comments 4. The collection of real‐time data is better to be used based on laboratory hardware setup instead of PASCAD results. However, this may be considered in your future work.

Response 4：Thank you for your valuable comments. As you suggested, our current focus is on demonstrating the effectiveness of the proposed approach. Our team’s next step is to implement this approach on protective devices and carry out demonstration applications in substations, thereby further enhancing its practicality.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

I would like to point out the following items for the authors to consider :

1) Although the residual flux is of extreme importance in r determining the maximum amplitude of the inrush current, the authors seem to neglect its influence, although a mention of it is made in lines 169 and 170;  This is a crucial item , since it changes everything. 

2) Circuit of Figure 4 shows some mistakes : 

   2.a ) R2 and L2 are wrongly located ;

    2.b) It seems L1 and L2 are leakage inductances, whereas the is no sense in considering these parameters for a no load transformer, since they make sense only if  there are current in both primary and secondary windings.

 

3) Graph of Fig. 9 is very strange and hard to comprehend;

4) While graphs of Figs. 1 and 2 are very representative of the inrush current, those of Figs. 8, 12 and 13 seem to be too much "mild". Is there any reason ?

Thank you.

Author Response

Thank you for your comments and suggestions on our manuscript. Our responses to the reviewer‘s comments are provided below. You can also find the responses in the attached file.

Comments 1. Although the residual flux is of extreme importance in r determining the maximum amplitude of the inrush current, the authors seem to neglect its influence, although a mention of it is made in lines 169 and 170;  This is a crucial item , since it changes everything.

Response 1：Thank you for your valuable comments. We thank the reviewer for this important observation. We fully agree that the residual flux in the transformer core is a critical factor as it directly influences the core saturation level and the subsequent characteristics of the inrush current. To address this, the dataset constructed in Chapter 5 has been designed to include inrush current data under various residual flux conditions. Our proposed method has been validated to effectively identify inrush currents even with these variations. Furthermore, to underscore the comprehensiveness of our evaluation, we have clarified the text in the dataset section to explicitly state that it encompasses a wide range of operating conditions, including different residual flux states. The changes have been marked in red font in the text.

Comments 2. Circuit of Figure 4 shows some mistakes :

   2.a) R₂ and L₂ are wrongly located ;

   2.b) It seems L₁ and L₂ are leakage inductances, whereas the is no sense in considering these parameters for a no load transformer, since they make sense only if  there are current in both primary and secondary windings.

Response 2：Thank you for your valuable comments. As noted by the reviewer, a parameter was incorrectly labeled in the original figure due to a drafting error. We have corrected this and updated the figure. This revision improves the clarity of our results and the overall rigor of the paper. we acknowledge that leakage inductance is only defined in the presence of current. Accordingly, during no-load energization, it should only be represented on the side where the current flows (the switching-on side). We have updated the figure accordingly to ensure this physical principle is accurately depicted, strengthening the paper's technical foundation. The changes have been marked in red font in the text.

Comments 3. Graph of Fig. 9 is very strange and hard to comprehend

Response 3：Thank you for your valuable comments. Here we provide an explanation of the calculation process for Figure 9. The figure shows the ratio of the second harmonic to the fundamental component in Phase B's differential current, as calculated by the protection relay in a real-world engineering scenario. These values were computed at different time points following a transformer energization. Due to the decaying nature of the inrush current, the second harmonic content in the differential current exhibits significant non-constant fluctuations. This leads to the frequent variations observed in Figure 9, where the ratio oscillates above and below the 15% threshold. Crucially, when this ratio dips below 15%, it may lead to an erroneous operation of the differential protection.

Comments 4. While graphs of Figs. 1 and 2 are very representative of the inrush current, those of Figs. 8, 12 and 13 seem to be too much "mild". Is there any reason ?

Response 4：Thank you for your valuable comments.

The distinction in saturation depth is indeed central to our argument. Figures 1 and 2 depict classic, severe inrush, while Figure 8 shows a more challenging, mild saturation case from the field that caused a protection maloperation.

Furthermore, we acknowledge that the original waveforms in Figures 12 and 13 were unclear. Their purpose was purely to define the input features for our scheme, not to represent a specific physical event. Following the reviewer's valid point, we have replaced these waveforms with real engineering data to make their illustrative purpose clear and to prevent any misinterpretation.

The changes have been marked in red font in the text.

Author Response File: Author Response.docx