A Novel Machine Learning-Based Approach for Fault Detection and Location in Low-Voltage DC Microgrids
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn the paper the fault detection method based on CS and RT and the location method based on LSTM in the DC microgrid were studied. The methods were not descripted clearly, and seem lack of innovations.
1. The dataset was generated by using the MATLAB/Simulink environment. What are the values of n and m in Eq.(3) and Eq.(4). How to create the faulty signal?
2. CS was used to process the dataset. What are the values of the matrixes P and H? How about the performances of CS in the processing? Detailed examples of CS are recommended to be presented.
3. The Regression Tree and LSTM model should be trained by applying the dataset. Please introduce the training dataset and test dataset. How about the loss functions and the changing situations during training?
4. How to define the accuracy to locate the faults? What are the units of the parameters in the vertical coordinates in Fig.7 to Fig.15?
5. Some typos were found. For example, in Line 187 (5) should be (1). In Figure 5, there are some mistakes in LSTM model.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis manuscript takes DC microgrid as the research object, and conducts relevant research on fault detection and fault location of DC microgrid. The fault detection of DC microgrid is completed by combining compressed sensing and decision tree, the fault location is completed by using long short-term memory network, and the proposed method is verified by using the signal generated by the simulation model.This manuscript is somewhat innovative, but it also has certain problems. The main problems are as follows:
(1) Abstract is a bit lengthy and needs to be condensed, and the introduction of the relevant building models is not necessary.
(2) The Introduction section lacks an introduction to machine learning-based fault detection and fault location.
(3) The descriptions of equations (2), (3), (4), (5), and (6) in Section 2.2.1 are not clear, and corresponding introductions need to be added, and it is recommended to use pictures to introduce the theory of compressed sensing.
(4) Figure 3 shows that there is no real or imaginary part of the current signal, so please add an introduction to the signal transition process.
(5) The manuscript lacks an introduction to some of the symbols that appear in the formula, such as equations (9)-(12), (17), (18), etc.
(6) Equation (17) in Section 2.3.2 has cases where the formula does not correspond to the description.
(7) In the manuscript, there are cases in which the legend obscures the content of the picture in pictures (7)-(12), and the ordinate of pictures (13)-(15) lacks units.
(8) The Conclusions section is lengthy, and it is suggested that it should be condensed and highlighted as innovative in this manuscript.
There are also some grammatical problems in this manuscript, and the author is advised to read them carefully and correct them.
Comments on the Quality of English LanguageThere are some grammatical errors in this manuscript, but the language is generally fluent and there are no major problems.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThere are a few typographical errors that can be corrected with a careful proof-reading - for instance line 121 - "...microgrid. proposed..." should be "...microgrid. The proposed..."
Line 108 - a citation is needed for CS and RT.
Line 165 - "mathematical basics" -> "mathematical foundations"
Line 170 - "...swiftly decay..." -> "...decay swiftly..."
Line 195 - in addition to square of the difference, the absolute value of the difference may also be used (and the authors do describe the use of MAE, along with MSE, later in the paper). It will be helpful to mention the absolute value of difference here, too.
Equation 8: does the error greater than exactly zero, or some value that is very close to zero? The measurement of electrical current and the representation of its signal in data can have some noise, so a margin of error in measurement may needed to be used for determining whether the error is acceptable, in place of using the difference between actual and predicted value of the imaginary part of the signal to be exactly zero.
Comments on the Quality of English Language
Apart from a few errors, identified above, no noticeable errors that affect the readability of the manuscript were found.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for Authors1.In Figure 5,the structure of LSTM model is not the same with Figure 5 of Ref.50, and not consist with Eq.(17). Please explain the reasons.
2. Sections 2.2.1 and 2.3.2 are recommended to be simplified since the paper has no improvements of the theory of CS and LSTM.
3. In Sections 3, the performances of the proposed the fault detection and location method was validated by using the data generated through the MATLAB/Simulink simulation. Can authors provide the experiment results when processing the data obtain from the real environment?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have made some changes to the manuscript based on the questions raised, and the current problems are as follows:
In the revision of the literature review part, some literature on fault detection and fault location based on machine learning has been added, but the length is still small, and the shortcomings of the current research status are not fully summarized,As a result, readers still have insufficient understanding of the research status of fault detection and fault location based on machine learning, and the title of the manuscript is fault detection and location based on machine learning, which should be mainly described in the literature review part.
Comments on the Quality of English LanguageThe author's English writing level has basically reached the journal standard.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
