Study on the Complexity Reduction of Observed Sequences Based on Different Sampling Methods: A Case of Wind Speed Data
Round 1
Reviewer 1 Report (Previous Reviewer 1)
The authors revised thoroughly the paper according to the comments and recommendations.
Author Response
Thank you very much for the comment.
Reviewer 2 Report (Previous Reviewer 2)
The manuscript has presented a study the complexity reduction method of observed time sequence based on wind speed data. It is an interesting study and presented in a comprehensive manner with detailed analysis. I would like to recommend the article for publication in its current state.Author Response
Dear reviewer,
Thank you very much for your approval.
Reviewer 3 Report (Previous Reviewer 3)
Many researches have confirmed
-->
Many studies have confirmed
Overall, I don't understand what the paper wants to insist.
"And the complexity of the observed wind speed is closer to the complexity of a random sequences, which indicates that the wind speed sequence is not easy to predict."
If the wind speed sequence is difficult to predict, what the authors write a paper for?
I doubt if it is worth while to publish a paper for what they have found.
Author Response
Dear reviewer, Thank you very much for the comment.
First of all, I am very sorry that you do not agree with the conclusion of this manuscript, which is calculated from the 10Hz observed wind speed data.
In fact, it is needed to study the complexity reduction method of observed wind speed for it’s difficult to predict. The further conclusion shows that the average method has lower complexity and better prediction ability than the original sequence. This provides an idea for prediction. In the longer-term prediction, we can use the average method to obtain the data with low time resolution, so as to obtain an accompanying prediction result, which can be used as the constraint condition of longer-term prediction.
Reviewer 4 Report (Previous Reviewer 4)
The study focuses on reducing the complexity of the observed time sequence based on wind speed data. The authors tested five sampling methods, including random method, average method, sequential method, max method and min method to obtain a new time sequence with low resolution from a high resolution time sequence. The subject is interesting and practical to industry practitioners. However, the contribution of the research is limited and it wasn't clear what its novelty was. The reviewer would like to suggest the following comments for authors to revise their work:
[1] Literature review is not extensive. Conducting an extensive literature review would significantly help readers understand what already existed in the literature and what the gap in research was.
[2] The novelty of the work must be better outlined and more emphasized. In the current version, it’s so hard to understand what's new in this research.
[3] Sampling methods seem very simple and straightforward, and they already have been reported in different papers in the literature. What’s new here?
[4] Section 4 is perhaps the most interesting part of this paper. It presents the results of testing the methods on sample dataset. However, it lacks rich discussion. The results could be discussed and compared with some other methods (studies) reported in the literature. Some KPIs could also be introduced for this purpose to make comparisons unbiased.
Author Response
Dear reviewer,
Thanks your valuable suggestion. Over the past two weeks, we improved the manuscript and hope it meet the requires. We reply the comments one by one as follows.
[1] Literature review is not extensive. Conducting an extensive literature review would significantly help readers understand what already existed in the literature and what the gap in research was.
REPLY: We revised the manuscript to improve section INTRODUCTION. More methods of wind speed prediction based on machine learning were added. Please give more details about the references if it still can’t meet your requires, and that would be very thankful.
[2] The novelty of the work must be better outlined and more emphasized. In the current version, it’s so hard to understand what's new in this research.
REPLY: We enhanced the INTRODUCTION and added more information.
[3] Sampling methods seem very simple and straightforward, and they already have been reported in different papers in the literature. What’s new here?
REPLY: Only the common data processing methods were used. The results show that the average method can reduce the complexity of the sequence and improve the prediction ability in a very short time interval (less than 1 minute). At present, the average forecast of many wind farms is 15 minutes (with time interval 15 minutes). If more precise forecast data can be obtained, it is undoubtedly beneficial for wind power prediction.
[4] Section 4 is perhaps the most interesting part of this paper. It presents the results of testing the methods on sample dataset. However, it lacks rich discussion. The results could be discussed and compared with some other methods (studies) reported in the literature. Some KPIs could also be introduced for this purpose to make comparisons unbiased.
REPLY: The results in the manuscript show that the complexity of the original observed wind speed sequence is very high, almost consistent with the complexity of the random number. However, for the time interval of 10s or longer, such as 1 minute, its complexity obtained by the average method is reduced. We have enhanced the discussion in this part as follows and hope that it meets the requirements.
“…Many methods have been used to predict wind speed, and performed well. Such as mesoscale numerical models, statistical methods, and some machine learning methods, etc. However, if the complexity of the sequence is reduced before the prediction, the wind speed observation ability can be further improved, which is completely different from the improvement of the prediction method. …”
“…This problem is common in the prediction of wind farms with a time resolution of 15 minutes. However, if the time interval is shorter (like 1 minute) with lower complexity, we can predict the maximum wind speed more accurately.”
“…In addition, with the high predictability of low resolution data, we can also make a prediction profile as the high-resolution adjoint function.”
More details were shown in the manuscript with modification trace.
Round 2
Reviewer 3 Report (Previous Reviewer 3)
At least, in American English, people do not use the term "researches". They use "research". Please google it.
Using sampling techniques, the complexity can be decreased.
Average method helps prediction.
Most remarks in the paper is way too obvious to publish.
No worthy of publication.
Author Response
Dear reviewer,
Thank you very much for your comment.
Our research on reducing the complexity of observation data to improve its predictability will be continued as always. And the work would be improved according to your valuable comments.
In the future, we hope to submit our new work to you for review.
Reviewer 4 Report (Previous Reviewer 4)
The authors have improved the manuscript by addressing some of the comments; however, the comment associated with the novelty (contribution) isn't yet addressed properly. The authors have to clearly state what the gap in the literature was, how their research is filling the gap, and what's new in their research that makes it different than published studies. This will help reviewers and readers better understand the contribution of this piece of research.
Author Response
Comment: The authors have improved the manuscript by addressing some of the comments; however, the comment associated with the novelty (contribution) isn't yet addressed properly. The authors have to clearly state what the gap in the literature was, how their research is filling the gap, and what's new in their research that makes it different than published studies. This will help reviewers and readers better understand the contribution of this piece of research.
Dear reviewer,
Thanks for your valueable comment. We improved the manuscript by adding more explanation about the aim of our work.
The previous researches about the predictioin methodes mainly focused on the prediction results, barely paid attention to the reduction of the complexity of the observation sequence itself.
We improved the introduction as follows:
â‘ The third paragraph of Section introduction: " However, the complexity of the sequence obtained by these sampling methods and whether it affects the predictability of the sequence are still inconclusive. It is necessary to figure out their impacts on the complexity and predictability of the sequence. "
â‘¡ The fourth paragraph of Section introduction: "However, these prediction methods are mainly focused on the prediction results. If the complexity of observation data itself can be reduced, the effectiveness of these forecasting methods will be improved. A couple of sampling methods were used to analyze how to reducte the data complexity."
â‘¢ Besides, we also modified the first paragraph of Section conclusion and discussion: "Most of the research on wind speed prediction focuses on how to improve the prediction method, and there is little research on reducing the complexity of the observation sequence itself. We used five methods to obtain a new time sequence from the observed time series with high resolution and studied their complexities which was expressed by the ApEn exponent. The main conclusions are as follows."
Please refer to the manuscript for more details.
Thank you again for your helpful suggestions. We hope this version can meet the requirements.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The paper titled "Study on Complexity Reduction of Observed Sequence-Based on Different Sampling Methods: A Case of Wind Speed Data" has some potential but it requires major improvement to meet the standard of this journal:
*) The abstract needs to be crisp and clear in the problem statement, prior research efforts, proposed approach, and contributions of the work. The abstract is vague and does not engage the reader to read further. Please also state the key results of the work.
*) contributions and justification of using the pivotal quantity approach must be explained.
*) Please enumerate the contributions of the work in the introduction, as of now it is unclear. Please do not state what is the paper, but contextualize how your work improves on prior work in the list of contributions.
*) The related work section needed to be added. Please include a table that summarizes the key papers on the topic and then describes how your work is different. Key prior studies have not been discussed. It is important that such works are covered for a thorough and comprehensive literature review required for archival articles.
*) In the experimental and discussion please include a table that summarizes the experimental hardware and software setup.
*) Please add a conclusion and discuss some future work in a new Section 5.
*) what is/are the benefits of your research to the readers of this journal.
Author Response
Thank you very much. We improve our manuscript according to your suggestive comment and reply them one by one as the appendix. We hope that the manuscript is modified to meet the standard of journal.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript has discussed the impact of the sampling approach on the complexity of the dataset. The manuscript is interesting, however, it requires more discussion on the proposed study. The authors should consider the following comments while revising the manuscript:
- The authors have come to a conclusion based on their observation of a single dataset. It is advised to consider multiple datasets as well as a cross-validation approach to generalize the conclusion. Cross-validation or different sampling strategies can be found here: https://www.mdpi.com/1996-1073/13/10/2578
- The motivations and benefits of the proposed study are not well discussed in the manuscript.
- The qualities of figures in terms of resolutions and aesthetics is poor. It is advised to revise them.
Author Response
Thank you very much. We improve our manuscript according to your suggestive comment and reply them one by one as the appendix. We hope that the manuscript is modified to meet the standard of journal.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors used approximate entropy to measure the complexity of the wind speed data. And they used some sampling methods to sample data from the original data. I don't find anything new and anything worthy of publication in terms of methods.
Also, for the experiments, nothing is novel either. They applied ApEn and sampling methods to the data.
I don't find any related work in the paper. Is the authors' work the first study in the area?
In terms of results, nothing is novel to me.
"It indicates that the average wind speed is easy to predict, while the maximum and minimum wind speeds are as difficult to predict as the wind speed." Usually, average is easy to predict, but maximum and minimum is difficult to predict. Especially, sample mean tend to normal distribution (CLT).
"it is necessary to observe the wind speed with high time resolution first, and then use the average method to obtain the low-time-resolution wind speed". This can be one effective tip, but it is too marginal for publication.
No novelty in methods. Too small contribution.
Author Response
Thank you very much. We improve our manuscript according to your suggestive comment and reply them one by one as the appendix. We hope that the manuscript is modified to meet the standard of journal.
Author Response File: Author Response.pdf
Reviewer 4 Report
atmosphere-1691161: Study on complexity reduction of observed sequence based on different sampling methods: A case of wind speed data
This manuscript studied methods to reduce the complexity of the meteorological wind speed data obtained by instrument based on five sampling methods, namely, the random method, average method, sequential method, max method, and min method. The authors used the approximate entropy exponent to measure the complexity of the time series data. The subject of research is worth studying and can be also useful for industrial applications. However, the manuscript deals with some drawbacks which require authors’ further attention before it can further proceed.
Abstract should be improved. The main novelty of the work should be better outlined to motivate readers to continue reading the research. The introduction is not well developed. The problem statement is incomplete. Literature review is not done in an extensive way. Some papers are just cited without briefing their main contribution and mentioning how the current paper is different than the literature. What the gaps in the literature were?
Sections 2 introduces the approximate entropy method, the sampling methods, and the observed wind speed data. The first two subsections provide basic information which could be found in some other references. Therefore it’s suggested to authors to shorten subsections 2.1 and 2.2 and instead mainly focus on the wind speed data and expand the subsection 2.3.
Section 3 is the most interesting part of the paper. However it needs to be (i) enriched in terms of contribution and/or (ii) to be further expanded by including more verification (test) and validation results. Some comparisons with the results obtained from the already existing methods in the literature could be insightful to readers. The performance measures can be expanded as well. Discussion of results should be expanded too. Some tables/figures are left with little discussions. What were the limitations of this research? Further details on the dataset must be provided.
Author Response
Thank you very much. We improve our manuscript according to your suggestive comment and reply them one by one as the appendix. We hope that the manuscript is modified to meet the standard of journal.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors revised the paper according to the comments.
Reviewer 2 Report
The updates in the manuscript are satisfactory.
Reviewer 3 Report
The authors argue that they are interested in "reducing the time sequence complexity is the key to improve predictability". They argue that there are very "few studies for reducing the time sequence complexity", so they do not provide much related work. That is not true.
First, there are dimensionality reduction techniques for temporal data which eventually reduce complexity. Secondly, there are many sampling methods that reduce temporal data size and eventually reduces the complexity.
So, the authors argument, "there are few studies" seems to be just an excuse for no effort for exploring related work seriously and conducting detailed quantitative comparison.
Reviewer 4 Report
The authors have applied the comments and this version is in a better shape than the former version. Nevertheless, the discussion side of the research is not yet strong. The authors would need to enrich their discussions in the next version of the manuscript.
