Wind Power Short-Term Forecasting Method Based on LSTM and Multiple Error Correction
Round 1
Reviewer 1 Report
1) revise the abstract make it more clear and show contribution. The grammar errors and repeating sentences should be corrected.
2) Organization of the paper is not adequate. As of it is all is within each other. Make intro a seperate section. Make literature review section 2. Make Method section 3, Make empirical analysis section 4.
3) Delete ..... in Eq. 4.
4) Descriptive statistics table is not given. Is should be added.
5) Section 6 should be inside of Section 5, Empirical results section.
6) Table 1's name is RMSE values, inside the table it also says RMSE %. The equations for RMSE are for RMSE not for percentage calculation, I assume that inside the table, RMSE values are reported.
The RMSE values are very high, raising suspicions about the effectiveness of the model. Why is it? Know that RMSE is under the influence of unit of data. Is it the reason? Which trnasformations are the data went through, considering these, still the high RMSE's should be lowered.
7) For confirmatory analysis, RMSE is not enough. Present another statistics based on errors such as MAE or MAPE.
8) Overall, English grammar should be revised. Many errors in text. Example: By comparing M3 and M4 curves with M1 curves, we can see that. The affine 437 optimization model and the EFCM-LSTM model proposed in this paper can effectively 438 reduce the root-mean-square error of prediction results in data correction or error
9) Baseline model results should be seperately presented with their RMSE's and MAE or MAPES. Then, percentage reduction after your model should be presented. It will be more clear this way.
10) Contribution of the paper in terms of sustainability should be discussed and emphasized.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
With the increase in wind power production, accurate short-term wind power forecasting is becoming pivotal in the safe operation of power systems and the efficient scheduling of wind power generation. The submitted manuscript presents a prediction model that uses multiple error corrections and combines long-term and short-term memory to improve the accuracy of short-term wind power prediction. A multi-error correction method using a Long Short-Term Memory (LSTM) model for short-term wind power prediction is presented. The model includes a modified affine wind power correction model based on transfer learning, a self-moving window LSTM prediction model using an improved particle swarm optimization algorithm, and error feedback prediction to further correct initial prediction errors.
The simulation results show that the proposed method improves prediction accuracy through transfer learning and has good universality as it does not need to introduce other auxiliary methods. The proposed method improves upon traditional LSTM models by utilizing a self-moving window for more accurate predictions and uses an affine optimization model for data screening and a learning transfer technique for data correction. Additionally, the particle swarm optimization algorithm is utilized to optimize the parameters of the LSTM model, leading to an increase in prediction accuracy. However, following major and minor amendments are recommended prior to publication.
Major Revisions
1- Abstract: The abstract should be revised to reflect the significance of the proposed work. The details of the experiment and parameters should be moved to methodology section. Repetitive phrases (line 16 & 17) can be avoided. Also, the term used “learning transfer” seems to be incorrect.
2- Introduction: The review of literature should be updated with recent publications related to wind energy forecasting using LSTM, such as:
a. Forecasting Hydrogen Production from Wind Energy in a Suburban Environment Using Machine Learning
https://doi.org/10.3390/en15238901
b. Near real-time wind speed forecast model with bidirectional LSTM networks
https://doi.org/10.1016/j.renene.2022.12.123
c. A novel ensemble system for short-term wind speed forecasting based on Two-stage Attention-Based Recurrent Neural Network https://doi.org/10.1016/j.renene.2022.12.120
3- Page 3 Line 132-134
The authors have collected large amount of data but in the sentences, they mentioned ‘wind speed and direction of 10m, 30m, 50m and 70m’ which is unclear, as the wind direction is in degrees and wind speed is in meter per seconds.
4- Data Collection
The values collected for wind speed were on a single point or multiple points, please explain the data collection method briefly.
How was the seasonal variation captured during the training?
Page 11 Line 371: The data set namely 960 samples include only wind speed or other parameters. Also please provide details on the data set, its acquisition and duration.
5- Page 4 Line 147-157
The correlation figure is displaying the correlations between multiple parameters used as input for the predictive model however the paragraph should be rewritten to highlight, the implication of Figure 1.
6- Page 13 Line 425
What is meant by ten forecast results? Is the model trained 10 time? Is the input data same every time in each iteration and for all 5 models? Please explain the reasons of variation in the RSME% as illustrated in table 1.
7- Validation
Multiple samples should be randomly selected for validation to check the precision of the proposed model.
8- What are the main inputs of the model?
Minor Revisions
1- Page 1
Formatting error in the introduction section.
2- Citations
Citations should be in standard form and consistent.
3- Page 2 Line 117
The term used “learning transfer” seems to be incorrect which should be replaced by ‘transfer learning’.
4- Avoid excessive usage of ordinal adverbs
In the article there is excessive usage of the words “Firstly” and “Finally” which can be replaced with other words without changing the sentence structure.
5- Page 3 Line 134
On which criteria is the time step interval for the collection of data is set to 15 minutes.
6- Page 3 Line 139-141
Rewrite the sentence to clarify its meaning.
7- Page 8,9 Section 4.2
The steps displayed in the flow chart are described in detail under section 4.2 on page 8 and hence the figure 7 on page 9 is unnecessary as it is a sequential process.
8- Page 9 Line 301
Please correct the subscript “PPI” as “PI”.
9- Figure 8
Figure 8 should be revised with an improved version and a better caption.
10- Figure 10, 11, 12
The legends should be revised in proper format and should be presented in English. It appears that the top portion of the figure is missing. What are the units on X- axis in figure 10, how can no. of samples have a unit? Similarly, in Figures 11 and 12 have units. Please improve figures and captions. Why does the proposed model have high fluctuations in Fig 10 as compared to the other models?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear Authors, the comments and critiques that I gave are carefully corrected. Thank you. However, there is one exception. In the previous round I noted you should add descriptives table. I see that explanation table is added at Table 1. What I asked for was, the descriptive statistics for the dataset. These are Mean, Min., Max., standard deviation, skewness and kurtosis for the data used in the analysis. This table is very important. Because, the statistics in the upcoming sections that you are going to use are (some, not all) under the influence of the units of the data. If you show the descriptive table at the beginning of the empirical section, the problem will be solved. To keep the article size constant, I suggest extending Table 1 with this respect.
For an example paper that utilized a new neural networks model for forecast improvement and for a similar descriptive statistics table, please see, https://www.sciencedirect.com/science/article/abs/pii/S0957417408006477
I assume that this is a minor last correction and the paper is ready afterwards. I am satisfied with the corrections already made in the previous round. As a result, my decision is minor, which can be done by the authors easily. If the Editor choses to send the paper again, I will gladly take a final look for this table correction.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
The authors have made appropriate changes to the manuscript to address the concerns of this reviewer. The manuscript is therefore recommended for publication.
Author Response
The expert teacher thinks that the article has been properly revised and does not give his latest opinion. Many thanks to the expert teacher's ardent guidance and criticism. Sincerely wish the teacher happy every day, health, and happiness.
