A Comprehensive Review on Machine Learning Techniques for Forecasting Wind Flow Pattern

Round 1

Reviewer 1 Report

1. The manuscript's discussion of the architectural stage of machine learning approaches appears to be misleading and demonstrates a lack of knowledge and experience in the field of machine learning. It is crucial for the authors to revisit this section and provide accurate and well-supported information to maintain the credibility of their work.

2. The form of the manuscript seems to aim for a scientific style, but it is too laconic in certain areas, such as the workings of CNN architecture, neural networks-based approaches, and hybrid models. A more detailed and comprehensive explanation of these concepts is necessary to ensure the manuscript provides a thorough understanding for readers.

The introduction is satisfactory, but the literature review requires significant improvement. Given the above concerns, I recommend that the authors revise and rewrite the manuscript before resubmitting it for consideration.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

It is advisable in the introduction before presenting the structure of the article to include which are the main contributions of the research.

In line 109 the authors use the autoregressive model (ARM) as an acronym and for the moving average model (MAM), it is recommended to use the acronyms that are normally used for them. In the case of autoregressive (AR) and moving average (MA).

In Table 1 some acronyms have not been described such as SKE, RKE, and LK.

In Table 1, Table 2, Table 3, and Table 4, it is necessary to align the content of the application and outcome columns, because in some cases it is not clear where each line starts.

It is recommended to use the same column names for the tables where the approaches are summarized.

It is recommended that the title of subsection 2.2.2 be Deep Learning Approaches since the previous subsection also discussed approaches based on neural networks, which can confuse the reader into thinking that they are the same.

Figure 2 should be better explained since if the reader is not familiar with the different architectures of deep learning algorithms, they could not understand the figure.

Figure 4 should be described before presenting it, the same for Table 5, which is presented without being explained.

After reading the article, some questions arise are:

* Certainly, it was indicated that they focused on univariate models but within the investigation, they did not find multivariate models.

* If these models were found, what variables would be the ones that could contribute to improving the forecast of the wind flow patterns?

* Various forecast horizons were presented, but the volume of data needed to make an adequate forecast was not discussed.

Check typing errors on lines 124, 134, 138, 141, and 265.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

It is well prepared review article that gives us an overview of current thinking on the topic. The paper presents a review of different wind flow prediction techniques and their challenges across various domains where many forecasting models were introduced. Each model has their own features and limitations. The content of all the manuscript sections will be of greater interest to the journal's readers. The paper is therefore appropriate for an International Journal publication.

Firstly, it has been presented the list the conventional experimental methods used in various applications as well as its challenges.  In next section overview of intelligent computing techniques such as machine learning and deep learning-based wind prediction approaches used in structural health monitoring have been explained in detail. In section 4 it has been discusses the research gap and recommendations for future enhancements.

Minor editing of English language required.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

The current manuscript presents a comprehensive review on machine learning techniques for forecasting wind flow pattern. The paper is overall well-written and comprehensively compares different approaches.

There are two things that I'm concerned during the review. The first one is the way how the authors organize the Section 2.2. It confuses me how the authors distinguish the machine learning approaches in 2.2.1 and the neural network approaches in 2.2.2. I understand that the authors may refer the ``machine learning'' approaches here as more basic machine learning techniques, such as SVR, Random Forests, while neural network based approaches refers to more sophisticated machine learning models. 

The other one is how the authors distinguish the statistical approaches with intelligent approaches. In my opinion, many statistical-based approaches, such as Gaussian process, even the random forest shares origins with statistical models. As far as I'm concerned, statistical approaches are also part of the intelligent ones. 

Other than that, I think the paper is well-organized and I recommend its publication if the authors could address my concerned. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 5 Report

In this work, the authors present a review on available modelling frameworks for forecasting wind. The review examined four categories of available models with detailed discussion on their application, performance, and general characteristics. This is capped off with a gap analysis and recommendations for further enhancement. The work is interesting and can be published considering several comments:

The introduction is well written with sufficient presentation of the governing issue related to wind and importance of wind forecasting, however, fails to report similar works done on reviewing wind forecasting models, or other reviews on application of machine learning for forecasting weather related properties, which are abundant in the literature. Also, the authors need to clearly present the scope of the review and intended benefit from this work.

Line 73, the information is repeated. Some other syntax and grammatical errors have to be adjusted throughout the manuscript.

The mathematical framework for the different types of models is missing. Small subsections on each type of modelling category are needed for reference for the reader.

Demonstration of merits and demerits for each type of model and their category are muddy and not clearly defined and must be explicitly articulated.

Graphical breakdown for each modelling category with model subtypes will be helpful, something like a branched tree.

Initial section on the physics of wind and relevant properties required for wind forecasting are needed.

The language has to be further polished and avoid repetition for similar syntax, which was seen throughout the manuscript. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I am satisfied with the answers provided by the authors.