Time-Series Deep Learning Models for Reservoir Scheduling Problems Based on LSTM and Wavelet Transformation

Round 1

Reviewer 1 Report

This study presents, WT-LSTM, a deep learning model to improve the scheduling operation of hydorpower plants. The methods part appropriately provide the process of model development and data collection. The results shows that WT-LSTM shows improved model performance compared to the traditional model.

There are minor comments that may improve the understanding of the readers.

 

1 ) The term BP neural network (line 13)look referring back propagation if it so the full name may be presented when it is first mentioned.

 

２）The font in figures such as Fig 6~Fig 11 look too small. Increasing the font would be appreciated by the readers

 

3) In lines 523-524 “In general, it is thought that the model's prediction accuracy is good when both the RSR and NSE values are more than 0.75.”

It seems that this comment may not necessary in the paper.

Although this is generally acceptable the critical criteria for acceptable model performance would be varied for study area and characteristics of models. For example, even though a model shows NSE 0.6 or 0.7 that model may provide useful information and it would be not always true that the model is not a good model.

Author Response

Thanks for your comments concerning our manuscript. We appreciate your encouragement and endorsement of our manuscript. Those comments are all valuable and helpful for revising and improving our paper. We have studied all comments carefully and have made conscientious correction.

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

1.     The manuscript is concerned with time series deep learning models for reservoir scheduling problems based on LSTM and wavelet transformation, which is interesting. It is relevant and within the scope of the journal.

2.     However, the manuscript, in its present form, contains several weaknesses. Adequate revisions to the following points should be undertaken in order to justify recommendation for publication.

3.     Full names should be shown for all abbreviations in their first occurrence in texts. For example, BP in p.1, RRMSE in p.1, RRSR in p.1, NNSE in p.1, ARIMA in p.3, etc.

4.     For readers to quickly catch the contribution in this work, it would be better to highlight major difficulties and challenges, and your original achievements to overcome them, in a clearer way in abstract and introduction.

5.     p.1 - time series deep learning model based on LSTM and wavelet transformation is adopted for reservoir scheduling problems. What are other feasible alternatives? What are the advantages of adopting this soft computing technique over others in this case? How will this affect the results? The authors should provide more details on this.

6.     p.3 - Xiluodu hydropower station is adopted as the case study. What are other feasible alternatives? What are the advantages of adopting this case study over others in this case? How will this affect the results? The authors should provide more details on this.

7.     p.3 - historical records of November 2012 to January 2021 are taken. Why are more recent data not included in the study? Is there any difficulty in obtaining more recent data? Are there any changes to the situation in recent years? What are its effects on the result?

8.     p.6 - Mallat algorithm is adopted to decompose and reconstruct the traffic sequence. What are other feasible alternatives? What are the advantages of adopting this algorithm over others in this case? How will this affect the results? The authors should provide more details on this.

9.     p.7 - an M-layer wavelet decomposition is adopted in this study. What are other feasible alternatives? What are the advantages of adopting this approach over others in this case? How will this affect the results? The authors should provide more details on this.

10.  p.7 - the structure as shown in Figure 3 is adopted for the proposed prediction model. What are other feasible alternatives? What are the advantages of adopting this structure over others in this case? How will this affect the results? The authors should provide more details on this.

11.  p.11 - model input factors as shown in Table 1 are adopted in the model. What are the other feasible alternatives? What are the advantages of adopting these factors over others in this case? How will this affect the results? More details should be furnished.

12.  p.11 - the ReLU function is adopted as the activation function. What are other feasible alternatives? What are the advantages of adopting this function over others in this case? How will this affect the results? The authors should provide more details on this.

13.  p.11 - a greedy search strategy is adopted in this study. What are other feasible alternatives? What are the advantages of adopting this strategy over others in this case? How will this affect the results? The authors should provide more details on this.

14.  p.14 - three evaluation metrics are adopted to compare the forecasting performance of the models. What are the other feasible alternatives? What are the advantages of adopting these evaluation criteria over others in this case? How will this affect the results? More details should be furnished.

15.  p.14 - the models as shown in Table 2 are adopted as benchmarks for comparison. What are the other feasible alternatives? What are the advantages of adopting these models over others in this case? How will this affect the results? More details should be furnished.

16.  Some key model parameters are not mentioned. The rationale on the choice of the set of parameters should be explained with more details. Have the authors experimented with other sets of values? What are the sensitivities of these parameters on the results?

17.  The discussion section in the present form is relatively weak and should be strengthened with more details and justifications.

18.  Some assumptions are stated in various sections. More justifications should be provided on these assumptions. Evaluation on how they will affect the results should be made.

19.  Moreover, the manuscript could be substantially improved by relying and citing more on recent literature about real-life applications of soft computing techniques in different fields such as the following. Discussions about result comparison and/or incorporation of those concepts in your works are encouraged:

●          Quilty, J., et al., “A maximal overlap discrete wavelet packet transform integrated approach for rainfall forecasting-A case study in the Awash River Basin (Ethiopia),” Environmental Modelling & Software 144: 105119 2021.

●          Wang, W.C., et al., “A comparison of BPNN, GMDH and ARIMA for monthly rainfall forecasting based on Wavelet Packet Decomposition,” Water 13 (20): 2871 2021.

●          Farajpanah, H., et al., “Ranking of hybrid wavelet-AI models by TOPSIS method for estimation of daily flow discharge,” Water Supply 20 (8): 3156-3171 2020.

20.  Some inconsistencies and minor errors that needed attention are:

●          Replace “…due to the no stationarity of reservoir operation time series data…” with “…due to the non-stationarity of reservoir operation time series data …” in line 91 of p.2

21.  In the conclusion section, the limitations of this study, suggested improvements of this work and future directions should be highlighted.

Author Response

Thanks for your comments concerning our manuscript. We appreciate your encouragement and endorsement of our manuscript. Those comments are all valuable and helpful for revising and improving our paper. We have studied all comments carefully and have made conscientious correction.

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The revised paper has addressed all my previous comments, and I suggest to ACCEPT the paper as it is now.