Electricity Consumption Prediction in an Electronic System Using Artificial Neural Networks
Round 1
Reviewer 1 Report
Comments on “Electricity Consumption Prediction in an Electronic System using Artificial Neural Networks”
Dear Authors,
The paper must be significantly improved. Please consider the following remarks:
Major comments:
(1) Abstract part: “The electricity consumption dataset is obtained from a cold storage facility, which generates data in hourly intervals. The data obtained is measured for a period of over 2 years.” Please explain: one electricity consumer? Please specify. Only one is not enough for examination several methods.
(2) Please improve abstract part. Please answer the following question: What were your main results?
(3) Please add Nomenclature part.
Minor comments:
(1) Line 66: Please improve the way of calling references
(2) Table 1. Is not relevant. Figure 1 please improve in map: days (horizontal axis), hour (vertical axis)
(3) Table 4. Please explain accuracy of results
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
REVIEWER COMMENTS FOR AUTHORS
(1st ROUND)
1. The abstract of the manuscript is well written. However, it can be improved. The authors are advised to add the percentage efficiency of each implemented algorithm and compare it with the proposed one which is giving the best results. Also mention the performance indices on which the performance of algorithms is evaluated.
2. The 2nd Paragraph (Lines 35 – 41) discusses quite important aspects of Electrical load forecasting, but unfortunately it lacks any of the reference cited. The authors are advised to add at least relevant 2 or 3 references.
3. The authors are advised to add some more reference in 3rd Paragraph (Line 42 – 54)
4. Lines No. (55 – 60) are poorly managed. There is no need of extra paragraph for it. The authors are advised to merge it with 2nd paragraph.
5. The authors are advised to give some reference of the report of any International Agency (e.g., International Energy Agency etc.) to prove their stance regarding the dire need of robust and high-performance forecasting methods for appropriate power Despatch and generation planning. This would make this article make this article worth reading.
6. Lines No. (70 – 72) are poorly managed. There is no need of extra paragraph for it. The authors are advised to merge it with previous paragraph.
7. Before discussing neural networks and deep-learning algorithms, the authors are advised to add some literature regarding some machine learning algorithms, such as K-Nearest neighbors, Bagged trees, Support vector machine, and other such hybrid machine learning algorithms. Also discuss the pros and cons of these algorithms in literature.
8. What is the reason of choosing deep-learning algorithms over machine learning ones, despite the fact that later ones are simpler, less complex, consumes less computational time and doesn’t require high performance computing?
9. The literature review lacks the discussion and literature regarding input selection for their proposed forecasting model.
10. The literature review is quiet week and seems incomplete and in-sufficient. There is a lot of room for improvement in it. The authors are advised to update their literature review and add some updated research work in the literature section.
11. The introduction section also doesn’t include the discussion regarding the factors / performance indices on which the performance of algorithms is evaluated.
12. Summarize the discussion of Section No. 1 Introduction in a nutshell in the form of bullets at the end of section. This would make this article make this article worth reading.
13. On what basis the authors have chosen the inputs for prediction model? There is no such analysis produced in the paper regarding the Selection of Input matrices. The authors are advised to perform various exploratory data analysis prior finalizing any input factor for prediction model.
14. Figure No. 1 represents the presence of seasonal patterns with the electrical load consumption profile throughout the day, the month, and the year. Therefore, from this observation, it is quite evident that climatic and metrological factors chiefly contribute towards electricity consumption. The authors are advised to add relevant climatic factors, such as temperature, humidity, dewpoint etc. in input matrix after proper exploratory data analysis.
15. The text present inside the figures 4 and figure 5 are quite short. It should be increased for proper visibility.
16. The methodology is well written and well presented.
17. The MAPE of the results is quite high enough, and results from 24% – 29%. This means that the forecasted results have 24-29% error in it, which is quite high enough. The authors are advised to re-calculate the error result. The results with such high in-accuracy are NOT ACCEPTABLE.
18. The authors have stressed upon the requirements and dire need of short-term electrical load forecasting (which ranges from 24 – 72hrs) in their literature and hypothesis. However, the results in Table 4 suggest that forecasting results of 720hrs are more likely accurate than 24hrs, which is quite unlikely. This means that the forecasting model developed by authors more likely supports and give better simulation results in medium-term electrical load forecasting rather than day-ahead short-term forecasting (STLF) and is not well suit for STLF problems.
19. Since the electrical load pattern, as suggested by the authors in Figure No. 01 have seasonality throughout the year; therefore, the authors are advised to separate the data month-wise or season-wise (which ever gives better results). The authors should perform separate training and testing (month-wise or season-wise), and so different models (month or season wise) would be developed for the same case study. For the more clarity authors, the reviewer has sorted few papers (mentioned below) for the better understanding of authors. The authors are advised to go through it and grasp the concept presented. Also, it would help them a lot in modifying their literature review.
· https://doi.org/10.3390/en14175510
· https://doi.org/10.1016/j.eswa.2022.117689
· https://doi.org/10.1049/joe.2017.0873
20. The authors are advised to propose the method in their result discussion and conclusion which gives better results in their case-study, and also give their recommendations regarding it.
21. The authors are advised to add some hybrid deep learning algorithms in their future work in conclusion section. This would enhance their motivation and the publicity of the article too.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Accept
Author Response
Thank you!
Reviewer 2 Report
REVIEWER COMMENTS FOR AUTHORS
(2nd ROUND)
1. The modified lines 46-56 seems to be copy pasted / plagiarized from some source. The line no. 49 states that “…electrical load and meteorological data of one city in Pakistan were considered ….”. Are the authors sure that they are using Pakistan’s data?
2. The line no. 52-53 also seems to be plagiarized. It states that “… where a novel hybrid deep learning-based Encoder-Decoder technique …”
3. The authors have addressed all my queries and the paper is all ok now, except the above two points.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
