Uncertain Interval Forecasting for Combined Electricity-Heat-Cooling-Gas Loads in the Integrated Energy System Based on Multi-Task Learning and Multi-Kernel Extreme Learning Machine

Round 1

Reviewer 1 Report

• What is the size of the test set?
• How many historical data does the forecasting method use in the inputs? Better describe the input variables.
• Why aren't the same days used in MTL (figures 11 and 12) and STL (figures 13 and 14) methods? 
• In table 2, the PICP and PINAW indices are calculated for a single day, the data for a longer period should also be presented. 

Author Response

1. What is the size of the test set?

Thank you very much for your comment. As we mentioned in Lines 670-673 in the manuscript：‘The hourly data of different variables from January 2018 to June 2018 are selected as the training sample, and the hourly data of different variables of July 2018 is treated as the test sample for the established uncertain interval forecasting model of the industrial park’s combined electricity-heat-cooling-gas loads.’ Therefore, the test set is the hourly data of different variables from July 1, 2018 to July 31, 2018.

2. How many historical data does the forecasting method use in the inputs? Better describe the input variables.

Thank you very much for your comment. The historical data input into the forecasting model is the hourly data of different variables from January 1, 2018 to June 30, 2018. The input variables contain temperature, humidity, electricity price, heat energy price, gas price, per capita GDP, historical electric load, historical heat load, historical cooling load, and historical gas load. We have added the detailed description of the input variables in Section 3.1 marked in red color. Please check the added contents in the manuscript.

3. Why aren't the same days used in MTL (figures 11 and 12) and STL (figures 13 and 14) methods?

Thank you very much for your comment. In order to keep consistent, we modified the results illustrated in Figures 11, 12, 13, and 14 using the forecasting results of the same day. In Section 4.1, we compare the forecasting results obtained by STL based method using the same holiday and workday as those obtained by MTL based method to demonstrate the superiority forecasting performance of the MTL based model.

4. In table 2, the PICP and PINAW indices are calculated for a single day, the data for a longer period should also be presented.

Thank you very much for your comment. We have added the data for a long period in table 2. Please check the added contents marked in red color in the manuscript.

Reviewer 2 Report

The topic is interesting and it is adapt to this journal. The collaboration among several faculties is useful and I think that there is a great work behind the presentation of this work. The manuscript and the research in it are well structured. However, in my opinion, the paper is sometimes difficult to follow and more information is required on some issues. My comments:

 

-Clarify better the innovation of this work in the abstract and in the main text.

-Read articles to understand the structure of Mathematics. The following structure would be preferable based on the Energies Microsoft Word template file: 1. Introduction (1.1, 1.2, 1.3.), 2. Materials and Methods (2.1, 2.2., 2.3.), 3. Results (3.1, 3.2, 3.3), 4. Discussion (4.1, 4.2, 4.3), 5. Conclusions.

-Please search references to the equations. Equations should always be accurately and clearly referenced.

-Please add more information's about the model validation.

-Extend the conclusion with more general usability. What are the benefits of the results in a global context? Please explain this better in the manuscript.

-At the end of the study need to create a nomenclature table with units.

Author Response

The topic is interesting and it is adapt to this journal. The collaboration among several faculties is useful and I think that there is a great work behind the presentation of this work. The manuscript and the research in it are well structured. However, in my opinion, the paper is sometimes difficult to follow and more information is required on some issues. My comments:

1. Clarify better the innovation of this work in the abstract and in the main text.

Thank you very much for your comment. We have revised the abstract to better clarify the innovation of this work and revised Section 1.4 to better describe our contributions. Please check the revised contents in red color in the manuscript.

2. Read articles to understand the structure of Mathematics. The following structure would be preferable based on the Energies Microsoft Word template file: 1. Introduction (1.1, 1.2, 1.3.), 2. Materials and Methods (2.1, 2.2., 2.3.), 3. Results (3.1, 3.2, 3.3), 4. Discussion (4.1, 4.2, 4.3), 5. Conclusions.

Thank you very much for your comment and suggestion. We have removed the contents of original Section 2 to Section 1.2 to analyze the complicated coupling relations among the various energy subsystems in the IES, hence we can verify that it is necessary to forecast the combined electricity-heat-cooling-gas loads in the IES instead of forecasting each load separately. Then we revised the sub-title of Section 2 into Materials and Methods and the sub-title of Section 3 into Results. Please check the revisions in the manuscript.

3. Please search references to the equations. Equations should always be accurately and clearly referenced.

Thank you very much for your comment and suggestion. We have added some references for the equations referred to related literatures. For the ISSA method, the basic SSA is referred to the literature published by an Australian scholar Mirjalili, which we have cited the reference in the manuscript and the equations of SSA are all cited from this literature. The ISSA is improved by introducing the dynamic inertia weight and chaotic local searching mechanism into the basic SSA intelligent optimization algorithm, so as to improve the optimization searching speed of the SSA and avoid falling into local optimum easily, which is the innovation of this research. Please check the added contents marked in red color in the manuscript.

4. Please add more information's about the model validation.

Thank you very much for your comment and suggestion. For model validation, we firstly compare the established Bootstrap-ISSA-MKELM model based on the MTL method with the Bootstrap-ISSA-MKELM model based on the STL method to verify the superior prediction performance of the established model by discussing the values of PICP and PINAW, as well as the computing time. Then the ELM method (MTL-Bootstrap-ELM), the ELM method optimized by the SSA (MTL-Bootstrap-SSA-ELM), Poly single kernel extreme learning method optimized by the ISSA (MTL-Bootstrap-ISSA-pELM), RBF single kernel extreme learning method optimized by the ISSA (MTL-Bootstrap-ISSA-rELM), and the MKELM method optimized by the SSA (MTL-Bootstrap-SSA-MKELM) are selected as the comparison models to verify the superiority of the ISSA and the MKELM method combined in the proposed model through comparing the forecasting results. After the complicated comparison analysis, we can verify the superiority and feasibility of the established model in uncertain interval forecasting of combined electricity-heat-cooling-gas loads in the IES. We have also revised and added some contents in Section 2.1 and Section 2.2 to describe the necessity of combining these methods. Please check the added and revised contents marked in red color in the manuscript.

5. Extend the conclusion with more general usability. What are the benefits of the results in a global context? Please explain this better in the manuscript.

Thank you very much for your comment. The primary significance of this research is establishing an uncertain interval forecasting model with high accuracy for combined electricity-heat-cooling-gas loads in the IES. Based on the analysis of the complex coupling relationship among the electric energy subsystem, heat energy subsystem, cooling energy subsystem, and gas energy subsystem, we know that there exists strong coupling relationship among various sub-energy systems. Therefore, it is necessary to forecast the combined electricity-heat-cooling-gas loads instead of predicting single energy load. And the accurate prediction of combined electricity-heat-cooling-gas loads on the demand side can provide important reference and decision-making basis for multiple energy planning and optimal scheduling, demand side management, as well as safe and stable operation of multi-energy systems in the IES of the industrial park. Hence, establishing an uncertain interval forecasting model with high accuracy for combined electricity-heat-cooling-gas loads in the IES is of great significance. Through analyzing the forecasting results and comparison analysis on several models, we prove the superior forecasting performance of the established model, therefore, it can be extensively employed in uncertain interval forecasting for combined electricity-heat-cooling-gas loads in the IES, so as to provide important reference and decision-making basis for multiple energy planning and optimal scheduling, demand side management, as well as safe and stable operation of multi-energy systems in the IES. We have added some contents in Section 5 to clarify the significance of this research. Please check the added contents marked in red color in Section 5 in the manuscript.

6. At the end of the study need to create a nomenclature table with units.

Thank you very much for your comment and suggestion. We have added a nomenclature table with units at the end of the study in Appendix. Please check the added contents marked in red color in Appendix in the manuscript.

Reviewer 3 Report

The manuscript proposes an original methodology and model for the forecasting of the combined electricity-heat- cooling-gas loads in the integrated energy system. It is a very important issue within the energy market, therefore predicting a (narrow as possible) interval of future loads is of great interest.

The paper is well prepared. Here I have some comments.

1. Figure 1 is mentioned twice: at line 252 to illustrate the energy flow within IES and at line 656 as an industrial park of China. Please clarify what the figure represents and change accordingly
2. Within the methodology (chapter 3) the authors only gave general mathematical formulations, it would be better to provide more details with regard to the load forecast model.
3. Besides the comparison of the results obtained by using different models, the comparison with actual data (load) should be given, to validate the model.
4. Language and format used in the manuscript can be improved.

Author Response

The manuscript proposes an original methodology and model for the forecasting of the combined electricity-heat- cooling-gas loads in the integrated energy system. It is a very important issue within the energy market, therefore predicting a (narrow as possible) interval of future loads is of great interest. The paper is well prepared. Here I have some comments.

1. Figure 1 is mentioned twice: at line 252 to illustrate the energy flow within IES and at line 656 as an industrial park of China. Please clarify what the figure represents and change accordingly.

Thank you very much for your comment. For Figure 1 mentioned at Line 252, it illustrates the basic energy structure in the IES. For Figure 1 mentioned at Line 656, it illustrates the energy supply and demand structure of the selected IES. In other words, the IES of the selected industrial park is composed of electric energy subsystem, heat energy subsystem, cooling energy subsystem, and gas energy subsystem, including CCHP unit, Photovoltaic unit, electric boiler, gas boiler, electric refrigerator, absorption refrigerator, other energy generation and conversion equipment, as well as electric energy storage, heat storage and other energy storage equipment, just shown as Figure 1.

2. Within the methodology (chapter 3) the authors only gave general mathematical formulations, it would be better to provide more details with regard to the load forecast model.

Thank you very much for your comment. In Sections 2.1, 2.2, 2.3, and 2.4, we introduced the primary methods used in establishing the combined electricity-heat- cooling-gas loads prediction model. Then in Section 2.5, we described the established uncertain interval forecasting model for the combined electricity-heat-cooling-gas loads in the IES in details. We have also added some contents in Section 2 marked in red color about the model validation. Please check these contents in the manuscript.

3. Besides the comparison of the results obtained by using different models, the comparison with actual data (load) should be given, to validate the model.

Thank you very much for your comment. We have compared the forecasting results with the actual data in Section 3.2 before conducting the comparison analysis of different models in Section 4. The prediction interval coverage probability (PICP) and the prediction interval normalized averaged width (PINAW) are selected as the evaluating indicators to validate the established forecasting model. The PICP represents the probability that the forecast interval covers the actual electricity-heat-cooling-gas loads. The larger the PICP, the higher the reliability of the interval obtained by the prediction model is, which can better avoid the electricity-heat-cooling-gas loads from exceeding the upper and lower limits of the prediction interval. The PINAW demonstrates the uncertainty degree of the prediction interval of the model. The narrower the width, the lower the uncertainty of the prediction interval and the better the accuracy of the model. In Section 3.2, through analyzing the results of PICP and PINAW, the established model forecasting performance has been validated. In order to further verify the effectiveness of the established model in uncertain interval forecasting for combined electricity-heat-cooling-gas loads in the IES, we also compare the PICP and PINAW values obtained by various comparison models, and the comparison analysis verified the effectiveness and feasibility of the established uncertain interval forecasting model for combined electricity-heat-cooling-gas loads in the IES.

4. Language and format used in the manuscript can be improved.

Thank you very much for your comment and suggestion. We have modified the language and format in this manuscript. Please check the revised contents marked in red color in Lines 32, 45, 47, 58, 81, 100, 101, 139, 151, 157, 189, 203, 211, 214, 216, 230-231, 345, 370, 393-394, 400, 401, 419-422, 437, 440, 575, 588, 590, 599, 821, and 822. Additionally, we have removed the contents of original Section 2 to Section 1.2 to analyze the complicated coupling relations among the various energy subsystems in the IES, hence we can verify that it is necessary to forecast the combined electricity-heat-cooling-gas loads in the IES instead of forecasting each load separately. Then we revised the sub-title of Section 2 into Materials and Methods and the sub-title of Section 3 into Results according to the Mathematics Microsoft Word template file. Please check all the revisions in the manuscript.

Round 2

Reviewer 1 Report

No comment.