Hierarchical Optimization Method for Energy Scheduling of Multiple Microgrids
Round 1
Reviewer 1 Report
Contributions of the paper are not clear, why should someone use the proposed method? how it is different from the state of the art?
Hierarchical optimization method for energy scheduling of multiple microgrids is an interesting topic and there are many reported results based on different methods. The authors have not benchmarked their studies with any of the existing studies. I would suggest doing so for improving the quality of the paper.
Ci is the cost degradation of the ESS, what were the values used for Ci, how practical the assumed values are?
Literature survey could be improved
Author Response
Since the formulas cannot be displayed, please refer to the uploaded file.
Responses to Reviewer #1:
Reviewer #1:
Comment 1: Contributions of the paper are not clear, why should someone use the proposed method? how it is different from the state of the art?
R1: Thank you for your suggestions. We reiterate the contributions of the paper in the last paragraph of the revised introduction. The hierarchical optimization method proposed in this paper is hour-ahead, which is more accurate than the current day-ahead based method which has been studied a lot. In addition, the rolling optimization method is adopted to solve the problem of energy storage dispatch in MG. The price incentive method based on Stackelberg game is used to encourage MGO to participate in MMG energy transaction, which is helpful to improve the economic benefits of MGs.
Comment 2: Hierarchical optimization method for energy scheduling of multiple microgrids is an interesting topic and there are many reported results based on different methods. The authors have not benchmarked their studies with any of the existing studies. I would suggest doing so for improving the quality of the paper.
R2: Compare with the hierarchical method in references [14], [16] and [17] in revised manuscript, a significant feature of the proposed hierarchical optimization method is that energy storage and load demands are dispatched hourly which makes the energy transaction in MMGs more reasonable than day-ahead optimization method, especially under the uncertainties of PV output and load demands. Besides, you can find these words in the introduction of revised manuscript.
Comment 3: Ci is the cost degradation of the ESS, what were the values used for Ci, how practical the assumed values are?
R3: Thank you for your suggestions. The degradation cost coefficient is 0.005, which can be estimated by the following formula.
(R-1)
(R-2)
where, is the maximum charging/discharging power given by battery manufacturer, is the number of effective charging/discharging cycles under maximum power, is the directly purchasing cost of the battery unit. In the paper, ,,.
Comment 4: Literature survey could be improved
R4: Thank you for your suggestions. We have modified the literature survey in the introduction and marked it red in the revised manuscript.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is well written and described each section thoroughly. Appendix helps the readers to understand the equations and results in detail.
Conclusion must be improved - highlight the important points.
Also while discussing about the figure - the abbreviation is not same everywhere. In some places it is stated as Figure and in some places as Fig. This must be changed and kept same throughout the paper.
Table 2 must be placed in the middle of the discussion of the results, instead of in the end.
Multiple microgrids hierarchical optimization method is discussed in 2 stages - rolling optimization, Stackelberg game theory. the economic efficiency is improved along with the PV utilization.
Author Response
Responses to Reviewer #2:
Reviewer #2:
Comment 1: The paper is well written and described each section thoroughly. Appendix helps the readers to understand the equations and results in detail.
R1: Thank you for your affirmation.
Comment 2: Conclusion must be improved - highlight the important points.
R2: Thank you for your suggestions. The conclusion is modified in revised manuscript and marked as red.
Comment 3: Also while discussing about the figure - the abbreviation is not same everywhere. In some places it is stated as Figure and in some places as Fig. This must be changed and kept same throughout the paper.
R3: Thank you for your comments. We have changed all ‘Fig.’ to ‘Figure’, and marked them red in the revised manuscript.
Comment 4: Table 2 must be placed in the middle of the discussion of the results, instead of in the end.
R4: Thank you for your suggestion. We have placed Table 2 in the middle of the results discussion, instead of in the end.
Comment 5: Multiple microgrids hierarchical optimization method is discussed in 2 stages - rolling optimization, Stackelberg game theory. the economic efficiency is improved along with the PV utilization.
R5: Thank you for your comments. The improving PV utilization is an optimization result of the proposed method, not the optimization goal of the method. We adopt the rolling optimization method to ensure the reuse of the excess PV energy and reduce the energy purchase cost of each individual MG. In addition, we adopt Stackelberg game based electricity pricing strategy to promote energy sharing among MGs hourly, so as to improving the economic benefits of the MMG system.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper elaborates on a hierarchical optimization method for energy scheduling of multiple microgrids in a distribution network.
Although the work covers a timely topic, there is still room for improvement. My concerns and suggestion are listed as follows:
· Some related literature in this subject area is missed. How do authors differentiate their model from previously published works such as:
https://doi.org/10.3390/su11020301
https://doi.org/10.1016/j.ijepes.2018.09.031
DOI: 10.1049/iet-gtd.2017.2061
Authors should explicitly state their contributions in light of the existing literature and compare the effectiveness of their hierarchical method against others such as bi-level models.
· Why is the length of the rolling optimization considered 8 hours? Selection of such length must be justified.
· Table 1: Why is the TOU pricing scheme just applied to the selling process and not the buying?
· Fig.8: why does MG3 have the biggest share in load adjustment? How are the responsive loads managed in the two-stage programming model? Who is responsible to perform load adjustment (MG operator or DSO)? Which type of demand response program has been utilized in this study?
· As understood from figure 9, internal market formation can greatly affect the MGs’ utilities. Authors are encouraged to make a sensitivity analysis over the bidding strategies adopted by the market operator!
Author Response
Since the formulas cannot be displayed, please refer to the uploaded file.
Responses to Reviewer #3:
Reviewer #3:
The paper elaborates on a hierarchical optimization method for energy scheduling of multiple microgrids in a distribution network. Although the work covers a timely topic, there is still room for improvement. My concerns and suggestion are listed as follows:
Comment 1: Some related literature in this subject area is missed. How do authors differentiate their model from previously published works such as:
https://doi.org/10.3390/su11020301
https://doi.org/10.1016/j.ijepes.2018.09.031
DOI: 10.1049/iet-gtd.2017.2061
Authors should explicitly state their contributions in light of the existing literature and compare the effectiveness of their hierarchical method against others such as bi-level models.
R1: Thank you for your comments. We have carefully studied the above literatures and make be a part of literature survey. Besides, the contributions of the paper are reiterated in the last paragraph of the revised introduction and marked as red.
Comment 2: Why is the length of the rolling optimization considered 8 hours? Selection of such length must be justified.
R2: Thank you for your comment. One important input of the rolling optimization is TOU pricing scheme, and these prices can be divide into peak period, flat period and valley period. The rolling length of 8 hours ensures that there are at least 2 different price periods in a rolling optimization, which improves the rationality of charging /discharging behavior of ESS in dealing with TOU pricing scheme.
Comment 3: Table 1: Why is the TOU pricing scheme just applied to the selling process and not the buying?
R3: Thank you for your comment. The TOU pricing scheme has been implemented by the electricity companies for selling in most parts of China. However, the feed-tariff of distributed PV energy is usually set by the electricity price for local coal-fire units, which is a constant value. In addition, you can find these prices through the following links: http://www.95598.cn/person/index.shtml
Comment 4: Fig.8: why does MG3 have the biggest share in load adjustment? How are the responsive loads managed in the two-stage programming model? Who is responsible to perform load adjustment (MG operator or DSO)? Which type of demand response program has been utilized in this study?
R4: Thank you for your comment. There are four questions in this comment, which can be numbered 1, 2, 3, 4 in order, and we will respond them in the order of 2, 3, 4, 1.
R4-2: The responsive load managed only in the second stage. By receiving the optimal prices from EMO, the MGOs can calculate the optimal load demands according to the formulas (22)-(27).
R4-3: Load demand of each MG is calculated and performed by MG operator in this paper.
R4-4: As mentioned above, MGO responds the optimal prices by adjusting load demand. Therefore, the demand response in this study is price-motivated.
R4-1: The magnitude of load adjustment mainly depends on the benefit coefficient and internal prices in (2), and the role of energy trading. The benefit coefficient is calculated based on the predicted net load of MG.
(R-3)
where is the forecast load demand under grid prices, and is the forecast PV energy.
In Figure 8, all the load adjustments are in the range of maximum schedulable load, which indicates the optimal load can be expressed by
(R-4)
Therefore, the load adjustment can be expressed by
(R-5)
Formula (R-3) shows that MGO is seller when . For the given internal buying price, the load adjustment increases with the decrease of forecast load . Besides, MGO is buyer when . For the given internal buying price, the load adjustment increases with the increase of forecast load . Coincidentally, the load demand of MG 3 in time slots 8-11 and that of MG 3 in time slots 13-15 coincide with the above two situations respectively. Therefore, the load adjustment of MG3 in Figure 8 is the biggest.
Comment 5: As understood from figure 9, internal market formation can greatly affect the MGs’ utilities. Authors are encouraged to make a sensitivity analysis over the bidding strategies adopted by the market operator!
R5: Thank you for your comment. Formulas (26) and (27) show that if , then , and if , then . Therefore, we mainly discuss the case of and .
For MG who is a buyer, if , the optimal load demand can be expressed as
(R-6)
The corresponding utility is rewritten as
(R-7)
Therefore, the derivative of benefit with respect to price
(R-8)
For MG who is a seller, if , the optimal load demand can be expressed as
(R-9)
The corresponding utility is rewritten as
(R-9)
Therefore, the derivative of benefit with respect to price
(R-10)
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
I only have a small suggestion, add a comparison table by referring to methods in ref [14], [16-17] and show the improvement. This would clearly indicate that the method has been appropriately benchmarked with the most recent research work.
Author Response
Responses to Reviewer #1:
Reviewer #1:
Comment 1: I only have a small suggestion, add a comparison table by referring to methods in ref [14], [16-17] and show the improvement. This would clearly indicate that the method has been appropriately benchmarked with the most recent research work.
R1: Thank you for your suggestions. We have added the utility comparisons with the method proposed in ref [16] and [17].
To further illustrate the advantages of proposed hour-ahead optimization over day-ahead optimization. The methods in references [16] and [17] are applied to the proposed case, and the cost optimization problem of references [16] and [17] are transformed into the utility optimization problem to make a comparative analysis of the results.. In addition, the day-ahead stochastic prediction errors of PV and load demand are set to 10% and 12%, while the hour-ahead stochastic prediction errors are set to 5% and 6%, respectively. The results are shown in Table 3.
Table 3. Comparisons with other methods.
The results show that EMO and MGO can achieve more benefits by using the optimization method proposed in this paper.
You can find these words in lines 344-253 of the revised manuscript.
Author Response File: Author Response.pdf
Reviewer 3 Report
Most of the comments have been addressed suitably. Regarding my last comment, authors are encouraged to embed their calculations in another Appendix (such as Appendix C) while providing a figure in the result section showing that how the proposed utility function would change with prices and other parameters sated in (R-10).
Author Response
Responses to Reviewer #3:
Reviewer #3:
Comment 1: Most of the comments have been addressed suitably. Regarding my last comment, authors are encouraged to embed their calculations in another Appendix (such as Appendix C) while providing a figure in the result section showing that how the proposed utility function would change with prices and other parameters sated in (R-10).
R1: Thank you for your comments. We have added the calculations in Appendix C. Besides, we added two figures (Figure13 and Figure14) to illustrate the utilities of MGs response to the internal electricity prices, and you can find them in Appendix C of revised manuscript.
Author Response File: Author Response.pdf
