Photovoltaic Energy All-Day and Intra-Day Forecasting Using Node by Node Developed Polynomial Networks Forming PDE Models Based on the L-Transformation
Round 1
Reviewer 1 Report
-
The authors presented a PV power forecasting using 2 node by node developed Polynomial Networks forming. The applied methods were presented clearly and discussed. However, the significance of the results was not discussed and analysed in relation to the PV power forecast. The mathematical (statistical) methods took a large portion of the article. This is a general method and has been discussed widely in previous publications. It is crucial that the authors:
- Make clear the importance of carrying this research and highlight the research questions.
- Give a proper discussion and the end of the article. Present a critical analysis of the results and their interpretation in the context of existing scientific knowledge about PV power forecast.
Author Response
- The authors presented a PV power forecasting using 2 node by node developed Polynomial Networks forming. The applied methods were presented clearly and discussed. However, the significance of the results was not discussed and analysed in relation to the PV power forecast. The mathematical (statistical) methods took a large portion of the article. This is a general method and has been discussed widely in previous publications. It is crucial that the authors:
Evaluation and Discussion Sec. were extended by comparison of the proposed all-day and intra-day prediction schemes. The results were analysed and evaluated in more detail. The D-PNN mathematical description in Sec.3 was reduced and referred. - Make clear the importance of carrying this research and highlight the research questions.
All-day one-step and intra-day multi-step statistical approaches in the proposed PVP prediction schemes (Sec.4) are compared and evaluated using several types of machine-learning and regression methods. A novel neuro-computing technique D-PNN with significant improvements in its model formation and optimization algorithms, designed by the author, is introduced and applied.
- Give a proper discussion at the end of the article. Present a critical analysis of the results and their interpretation in the context of existing scientific knowledge about PV power forecast.
The results and findings were analyzed in detail and interpreted in consideration of the published state of knowledge (Sec.6 Evaluation and 7 Discussion).
Author Response File: 
Author Response.docx
Reviewer 2 Report
In this paper, the author proposed a new AI-driven method: D-PNN in predilecting photo-voltaic energy. It is an interesting topic, and the author described the method well. However, I do have some major concerns as below:
1) The literature part greatly lack reference. The author made several statements without supporting references.
2) The author talked about the disadvantages of NWP model. However, I didn't see a comparison between D-PNN and NWP model results. I suggest the author also add NWP forecasting into the comparison.
Line 10. Do you mean "local cloud"?
Line 24 -44. Lack references.
Line 27. Expand "NWP".
Line 252. "PVP, solar and wind 10 min." Incomplete sentence. And do you mean the data was collected every 10 mins?
Table1. You might need to add weather station names to columns 2 and 3. Otherwise, it is quite confusing.
Line 285. Expand "RBF".
Figure 8. There is no SMTL and RNN in the figure.
Figure 9 and 10, MAE only can tell the averaged absolute value. It will be interesting to see if one method is consistently over(under) predicted.
Line 389 to 390. "Statistical prediction models can be all-over flawed (15 May, Fig.4 and Tab.11)." The sample size (one location, one day) is too limited to draw such a conclusion.
Author Response
In this paper, the author proposed a new AI-driven method: D-PNN in predilcting photo-voltaic energy. It is an interesting topic, and the author described the method well. However, I do have some major concerns as below:
1) The literature part greatly lacks reference. The author made several statements without supporting references.
The reference list was extended with resent publications supporting the related statements.
2) The author talked about the disadvantages of NWP model. However, I didn't see a comparison between D-PNN and NWP model results. I suggest the author also add NWP forecasting into the comparison.
NWP data are not available for this study, so that only the statistical approaches and model results are compared.
Line 10. Do you mean "local cloud"? Corrected.
Line 24 -44. Lack references. Referred.
Line 27. Expand "NWP”. The abbreviation is explained in Abstract.
Line 252. "PVP, solar and wind 10 min." Incomplete sentence. And do you mean the data was collected every 10 mins?
Yes, 10min. series, corrected.
Table1. You might need to add weather station names to columns 2 and 3. Otherwise, it is quite confusing.
Weather station-names are placed in col.2 and 3.
Line 285. Expand "RBF". The abbreviation explained.
Figure 8. There is no SMTL and RNN in the figure. RNN was applied only in the intra-day PVP predictions (Fig.7, Tab.2), its daily 24-hour results are not good. GPR is the selected SMLT method (Fig.8).
Figure 9 and 10, MAE only can tell the averaged absolute value. It will be interesting to see if one method is consistently over (under) predicted.
Fig.8 and Fig.9 (orig. Fig. 9 and 10) present positive or negative trends in the daily avg. PVP prediction errors. Fig.10 and Fig.11 show the daily avg. R^2 accumulated in each prediction hour.
Line 389 to 390. "Statistical prediction models can be all-over flawed (15 May, Fig.4 and Tab.11)." The sample size (one location, one day) is too limited to draw such a conclusion.
A drastic change in weather is evident on the 4th prediction day (Fig.6). The statistical models are unable to predict in days with totally different patterns, compared to the trained and tested ones in previous days.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper is too long and too confused; the discussione is very short, only 12 lines as well as the conclusions, it is not possible! The authors must evidence the novelty and shorten the experimental part, show te results and definitly improve the discussion. Figures are too much, it is difficult to follow the meaning: please move some figs to the Suppler Mat or transformation part of them in tables. The figure quality os very low and difficult to understand. Further, the English must be revised. Finally, please follow the authors' guidelines for the reference. There is a lot of work to do before the paper could be considered for acceptance.
Author Response
The paper is too long and too confused; the discussion is very short, only 12 lines as well as the conclusions, it is not possible! The authors must evidence the novelty and shorten the experimental part, show the results and definitely improve the discussion. Figures are too much, it is difficult to follow the meaning: please move some figs to the Suppler Mat or transformation part of them in tables. The figure quality is very low and difficult to understand. Further, the English must be revised. Finally, please follow the authors' guidelines for the reference. There is a lot of work to do before the paper could be considered for acceptance.
The manuscript length and number of figures was reduced. Discussion and Experiments evaluation Sec. were extended. The novelty is highlighted in Abstract and Introduction Sec. (see the response to the rev.1) The original excel graphs are included below (the standard quality is adequate). Tables summarize the results presented in Figures.
The English was revised several times with an academic language writing tool. The reference style “Vancouver” was used (Chicago is incorrect).
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
I thank the author for making the edits and addressing my questions. The revised version is much clearer than the first version. However, I still have some suggestions as follows:
Figure 1 is missing.
Line 171 to 178. The figure overlaps the text.
Figure 2 and Figure 3, The quality of these two figures is poor. Please redo them.
Table 1, the format of the commas in this table is not consistent, and the number of digits after the decimal point.
Figure 5. Please keep the legend and figure on one page.
Line 383. why do you make the "of" bold?
Line 268. Change 'Tab 2' to 'Table 2'.
Line 391 to 392. "A fixed testing error can be determined, according to the past failure cases with relation to NWP data, to assess the optimal model." I don't get what you mean by this sentence. In your response, you were saying the NWP data is not available.
Author Response
I thank the author for making the edits and addressing my questions. The revised version is much clearer than the first version. However, I still have some suggestions as follows:
Figure 1 is missing.
Fig.1 is placed on page 4, the figure numbers are correctly ordered and referred to.
Line 171 to 178. The figure overlaps the text.
The Fig.1 position was corrected, there is no figure between the lines.
Figure 2 and Figure 3, The quality of these two figures is poor. Please redo them.
Fig.2-3 are Word designed in the adequate standard quality to demonstrate the 2 training procedures. Their size was enlarged to be more legible. The format can be easily changed by the editorial (if necessary).
Table 1, the format of the commas in this table is not consistent, and the number of digits after the decimal point.
Corrected.
Figure 5. Please keep the legend and figure on one page.
The legend is placed together with the map in one picture.
Line 383. why do you make the "of" bold?
I cannot find it in the line.
Line 268. Change 'Tab 2' to 'Table 2'.
Changed.
Line 391 to 392. "A fixed testing error can be determined, according to the past failure cases with relation to NWP data, to assess the optimal model." I don't get what you mean by this sentence. In your response, you were saying the NWP data is not available.
The encountered problems are discussed, and their possible ways of solutions are sketched out (if available NWP data) in Discussion Sec.7.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors improved the text. I have just a remark regarding the ref: the authors should follow the guidelines.
Author Response
The reference style was corrected according to the MDPI author guidelines.
Author Response File: Author Response.docx
