Multi-Site Wind Speed Prediction Based on Graph Embedding and Cyclic Graph Isomorphism Network (GIN-GRU)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presents a learning-based integrated multi-site wind speed prediction model for wind farms.

Please include numerical results in the abstract. Also, what was the comparative model, and how was higher prediction accuracy obtained?

Ensure that all the keywords are included in the abstract text.

 The introduction is clear and concrete and considers relevant previous work in the field.

Please check Eq. (1); Gini must depend on i; the Eq 1 appears to miss this index.

In line 185, please include the dimension of Matrix R.

Figure 2 appears twice on pages 6 and 7. Also, check figure 3.

After the title of Section 3, there should be a paragraph introducing the content of this section. Also, please ensure that the format and presentation of this section are carefully reviewed, as they appear to be sloppily written.

Include references supporting all you wrote about error metrics in section 3.1.3.

Improve the quality of Figure 7.

Lines 449 and 485 are repeated; please improve how the information is written.

Figures 8 and 9 lack the y-axis label. Also, the information presented is unclear, and the quality of those figures must be improved.

Please explain in detail how the author reached this conclusion: on line 504: “However, the competing models result in more significant prediction errors during wind speed fluctuations.”

 

Do Tables 5 and 6 both present information on error metrics at 10 m? The results are different; please explain.

Comments on the Quality of English Language

I recommend reviewing the paper by a professional editor or using language editing services, especially if English is not the authors' first language.

 

Ensure a smooth transition between sections headers and paragraphs.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The journal readers are aware of global energy problems; thus, they should avoid using general statements that are well-known to the public.

Verify that every reference is cited correctly.

The paper presents many sequential mistakes; the method is not fully described and cannot be reproduced. 

In line 129, the Gini impurity formula must be included.

Achronims must be fully declared and not included in the headings without a full description.

In line 159, the structure of the data set must be fully described, including a description of the measuring process, frequency of measurement, data structure, etc.

In line 164, the method cannot be reproduced. Many terms are not fully described.

In line 178. it is unclear how matrix Xb is obtained.

In eq. 5, how is D calculated or determined?

How is matrix Xf obtained?

Eq. 13 cannot be reproduced. Many terms have not been defined.

Fig. 1 does not correspond to the description included in the text.

Fig. 3 is out of format and not fully explained.

Fig. 9 must include where the "other methods" were obtained.

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

In their article, the authors propose a new model for predicting wind speed, which takes into account the spatial features and geographical structure of the location of the power plant. The authors of the study propose a combined data analysis mechanism using random forest and principal component analysis (PCA-RF), which allows you to preserve the main parameters and extract important information by reducing the dimensionality of residual features. Additionally, the model builds graph networks by integrating graph embedding techniques with the Mahalanobis metric to synthesize correlation information between features from different locations (however, it would be nice to calculate its concordance coefficients. This is not an observation, it is a thought). This approach effectively integrates adjacent data and captures complex interactions between different locations. Finally, the graph isomorphism network (GIN) learns the internal relationships in graph networks, and the group recurrent unit (GRU) combines these relationships with temporal correlations, thus the research carried out by the authors can solve wind speed forecasting problems. Experiments conducted on wind farm data off the coast of California in 2019 showed that the proposed model has higher prediction accuracy compared to the comparison model. Overall, the paper represents a significant contribution to the field of wind speed forecasting and can be useful for improving the efficiency of wind power generation.

Notes:

1. Line 57. Error in abbreviation.

2. Not all variables in the formulas are disclosed.

3. Line 197, figure 1. Needs explanation after the figure.

4. line 295-299. Error in drawings. The formula in the figure is not readable.

Conclusion. The article is written in detail and logically. No significant comments. There are several design flaws. This is easy to fix. I recommend it for printing with minor modifications.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have satisfactorily addressed all of my previous comments. They have clarified all figures, mathematical expressions, and the information that was unclear in the initial version. I believe that the manuscript is now suitable for publication. However, I still recommend a general English revision to ensure linguistic accuracy and fluency. Additionally, enhancing the text size and quality of all figures would improve the overall readability and presentation of the manuscript.

Comments on the Quality of English Language

I still recommend a general English revision to ensure linguistic accuracy and fluency