Short-Term Load Forecasting Based on Spiking Neural P Systems
Round 1
Reviewer 1 Report
General comments
- The manuscript needs proofreading regarding grammar, punctuation, typesetting.
- The figures can be improved.
- The presentation and discussions ,conculsion about the results has to be improved.
Author Response
Response to Reviewer 1 Comments
Point 1: The manuscript needs proofreading regarding grammar, punctuation, typesetting.
Response 1: We thank the reviewer for comment. In the revision, we have carefully and meticulously proofread for grammar, punctuation, and typesetting.
Point 2: The figures can be improved.
Response 2: We thank the reviewer for comment. In the revision, we have reworked the experimental results graph to make it clearer.
Point 3: The presentation and discussions, conclusion about the results has to be improved.
Response 3: We thank the reviewer for comment. In the revision, we have redescribed the conclusion section and improved the presentation of the experimental results. Please see the revision.
Author Response File: Author Response.docx
Reviewer 2 Report
This study presents a fresh short-term load forecasting model, termed as LF-SNP model. The following points should be considered in the revised version:
1- The problem statement needs to be extra clarified.
2- Extra details should be given about the "structure and attribute of biological neurons".
3- I suggest adding a flowchart describing the logic of the proposed model (to be added in section 2).
4- It is hard to understand the results obtained in Figure 3. Could you please add more clarifications?
5- How can the presented model be beneficial to the industrial sector? I believe a clear answer should be added before the conclusion section.
6- The "conclusions" section is still not solid, and needs to be rewritten to highlight the importance of the developed model.
Overall, the paper offers good attempt. The authors need to consider the abovementioned comments.
Author Response
Response to Reviewer 2 Comments
Point 1: The problem statement needs to be extra clarified.
Response 1: We thank the reviewer for comment. In the revision, we have clarified the problem as follows:
“Although these models have shown good prediction performance, however, power system is a complex dynamic system, so load forecasting is still a challenging task. This study focuses on developing a new recurrent-like prediction model for load forecasting tasks.”
Please see the fourth paragraph in Introduction section in the revision.
Point 2: Extra details should be given about the "structure and attribute of biological neurons.
Response 2: In Definition 1, we define NSNP system, which is composed of m neurons. Each neuron is composed of a state unit and a nonlinear spiking rule. The nonlinear spiking rule describes dynamic characteristics. Equations (2) and (3) are mathematical model of NSNP neuron.
In the revision, we have improved the representation to describe the NSNP neuron model more clearly.
Point 3: I suggest adding a flowchart describing the logic of the proposed model (to be added in section 2).
Response 3: We thank the reviewer for comment. In the revision, we have added the flowchart in section 2). Please see Section 2 in the revision.
Point 4: It is hard to understand the results obtained in Figure 3. Could you please add more clarifications?
Response 4: We thank the reviewer for comment. In the previous description, we only gave a qualitative analysis, in the revision, we added a quantitative analysis. Please see section 3.3.1 in the revision.
Point 5: How can the presented model be beneficial to the industrial sector? I believe a clear answer should be added before the conclusion section.
Response 5: We thank the reviewer for comment. In the revision, we have added a paragraph as follows:
“As mentioned earlier, load forecasting plays an extremely important role in power system planning and operation. Accurate prediction of future loads in different time horizons will be able to considerably enhance the stable and efficient management and coordinated dispatching of power grids. Because of its advantages in structure and performance, the proposed LF-NSNP model must provide a potential choice for the power industry sector.”
Please see the last paragraph in Experiment section in the revision.
Point 6: The "conclusions" section is still not solid, and needs to be rewritten to highlight the importance of the developed model.
Response 6: We thank the reviewer for comment. In the revision, we have redescribed the conclusion section. Please see conclusions section in the revision.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper presents a new method for load forecasting. The authors claimed that a new method has been presented for an old method. The paper has some limited contributions and there are some questions as follows:
1- The contribution of paper is not presented well.
2- Most of the abbreviated words are not defined. (suggest: a nomenclature can be added to the paper).
3- It seems that a pre-preparation of data should be done before using in the classifier. (Discuss)
4- The classification structure (number of inputs, outputs, input layer and etc) is not presented.
Author Response
Response to Reviewer 3 Comments
Point 1: The contribution of paper is not presented well.
Response 1: We thank the reviewer for comment. In the revision, we added a paragraph to describe contribution of this paper. Please see the last second paragraph in Introduction section in the revision.
Point 2: Most of the abbreviated words are not defined. (suggest: a nomenclature can be added to the paper).
Response 2: We thank the reviewer for comment. In the revision, we have added a nomenclature before introduction. Please see Nomenclature section in the revision.
Point 3: It seems that a pre-preparation of data should be done before using in the classifier. (Discuss)
Response 3: The load forecasting is a regression problem. In training the regression model, input is load time series data and the corresponding out is also the same load time series. After training the model, the well-trained model is used to predict the future values.
In experiments, benchmark data set is used. Since load time series is used as input of model and the corresponding output, no pre-preparation is required in experiments.
Point 4: The classification structure (number of inputs, outputs, input layer and etc) is not presented.
Response 4: The proposed LF-NSNP model is a recurrent-like prediction model. As usual in recurrent neural networks (RNN), the LF-NSNP model does not have a fixed number of input units and output units. This is very different from multi-layer perceptron (MLP), convolutional neural network (CNN) and support vector machine (SVM).
Since the proposed LF-NSNP model is a recurrent-like model, it is particularly suitable for short-term load forecasting.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors claim The LF-SNP is very different from multi-layer perceptron (MLP), convolutional neural network (CNN) and support vector machine (SVM).why??, How many inputs and outputs ,the description is very confusing
Author Response
Response to Reviewer 1 Comments
Point 1: “The authors claim the LF-SNP is very different from multi-layer perceptron (MLP), convolutional neural network (CNN) and support vector machine (SVM). Why??, How many inputs and outputs, the description is very confusing”
Response 1: Power load time series belongs to sequence data, just like natural language data. Sequence data usually has a challenge: each sequence has a different length. As in English, each sentence may have a different length or contain a different number of words.
Usually, multi-layer perceptron (MLP), convolutional neural network (CNN) and support vector machine (SVM) require a fixed input length, that is, the number of input cells must be fixed, not variable. Although they can also process sequential data (after special processing), this input method is unnatural and may not capture the correlation between long and short term data. This is an important disadvantage of these models when dealing with sequential data. To overcome this shortcoming, recurrent neural networks (RNN) have been proposed. RNN is a recurrent-like model, which is very suitable for processing such variable length time series data, for example, power load time series and natural language data.
The proposed LF-NSNP model is a recurrent-like prediction model and it is implemented in the RNN framework. Therefore, as usual in RNN, the LF-NSNP model does not have a fixed number of input units and output units. Since the proposed LF-NSNP model is a recurrent-like model, it is particularly suitable for short-term load forecasting.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors answer well to all of my questions.
Author Response
We thank the reviewer for comment.