Simplified A-Diakoptics for Accelerating QSTS Simulations

Round 1

Reviewer 1 Report

Dear Authors,

This paper presented a simplified algorithm for implementing A-Diakoptics to perform QSTS simulations on large-scale networks using OpenDSS. This is a very interesting topic, the paper has been well-written and supported by simulation results. Following are my comments

• Please improve the introduction by providing more information regarding the existing methods (other than A-Diakoptics) in the literature on network partitioning techniques. 
• Several variables/parameters in the equations are not correctly mentioned in the text. For instance, IPC(E), I0, ZCC, ZTT etc. The equations have a subscript whereas they are missing in the text. 
• On page 8, it has been mentioned that the simulation fidelity is as good for the whole yearly simulation as for the previous cases. What previous cases are you referring to? 
• Figure 5 showed the improvement in solution times when the network is partitioned into different number zones. What is the improvement in simulation time compared to the earlier version of A-Diakoptics?
• If possible, It would be nice to compare the solution times of A-Diakoptics with any of the existing methods in the literature. 
• Circuit reduction techniques can also be employed to reduce the solution times. Should we use circuit partitioning methods to preserve the entire data of the network?

Author Response

Dear reviewer. Thanks for your kind feedback and comments, as well as the time dedicated for reviewing our paper. Our answers to the reviewer’s comments as follows:

Please improve the introduction by providing more information regarding the existing methods (other than A-Diakoptics) in the literature on network partitioning techniques.

Thanks for your comment and we agree, however, in previous publications on the same topic we have provided such literature review, the first time in 2015 and since then, in several publications. To address the reviewer’s inquiry, we have modified the paragraph starting at line 48 as follows:

“Other techniques link the computational burden to the circuit complexity, which is addressed by tearing the interconnected circuit into multiple sub-circuits. These new sub-circuits are solved separately using multicore computers to obtain the solution for the interconnected circuit. This approach is known as spatial parallelization, which addresses multiple techniques aiming to tear the interconnected model for its processing. These techniques are classified in two big groups depending on the power flow problem formulation: Bordered Block diagonal matrix (BBDM) [3] and piecewise methods, the last one is the one in which Diakoptics can be found, which is the base for the model presented in this paper called Diakoptics is based on actors (A-Diakoptics) [4, 5]. A more detailed description and bibliography about the differences between the model partitioning technique groups mentioned above can be found here [4, 6].”

We have also provided some extra references to complement this suggestion.

 

Several variables/parameters in the equations are not correctly mentioned in the text. For instance, IPC(E), I0, ZCC, ZTT etc. The equations have a subscript whereas they are missing in the text.

We have added a nomenclature table at the begging of the paper, thanks for your recommendation.

On page 8, it has been mentioned that the simulation fidelity is as good for the whole yearly simulation as for the previous cases. What previous cases are you referring to?

Thanks for this remark, as mentioned along the paper, the results presented in this paper are a part of a larger project, reason for which we probably mentioned the other studies and comparisons performed in other circuit models. These comparisons were not brought into this paper since they address other features and advantages obtained from spatial parallelization beyond the scope of this paper. To correct this issue, we have reformulated lines 297 and 298 as follows:

“As expected, the simulation fidelity comparison reports that the results obtained in the new implementation differ from the original by a negligible margin”

Figure 5 showed the improvement in solution times when the network is partitioned into different number zones. What is the improvement in simulation time compared to the earlier version of A-Diakoptics?

This is a very interesting question given the aiming of the previous version of A-Diakoptics. As presented in citations 4-6, the aim back then was to simulate the model in Real-Time for recreating electromechanical transients and protection testing. However, in this paper the method proposed is oriented to standard computing architectures focusing on fast simulation, which sets an important difference since as mentioned in 4-6, the implementation of A-Diakoptics was made using a Real-Time operation system. For this paper the operating system and computer architectures used are not dedicated. In this case is up to the reader to review the data presented in these publications (4-6), nevertheless, the conditions are different.

If possible, It would be nice to compare the solution times of A-Diakoptics with any of the existing methods in the literature.

The paper scope is directed to present the improvement in computing time consumption when adapting a sequential power system simulator into parallel processing equivalent. The simulation time gains are documented in the paper to highlight the benefits when using A-Diakoptics. Also, we are providing the materials (circuit models and simulation engine -OpenDSS), which can be downloaded from the internet, for the users to perform their own comparisons if desired.

Please accept our apologies, but without having the same conditions (operating system, computing hardware) it doesn’t make sense for us to compare other documented cases, which will require to implement such methods in a local environment, which is out of the scope of this paper.

Circuit reduction techniques can also be employed to reduce the solution times. Should we use circuit partitioning methods to preserve the entire data of the network?

This is correct, and thanks for bringing up this topic. As mentioned in section 2.2, the results presented here are derived from the work performed at the SuNLaMP initiative, which was sponsored by the DOE from 2016 – 2018. Among the methods resulting from this research, we also implemented circuit reduction (https://sourceforge.net/p/electricdss/code/HEAD/tree/trunk/Version8/Distrib/Doc/Circuit%20Reduction%20for%20Version8.pdf).  In the conclusions of that work, circuit reduction was one of the techniques offering a good reduction in processing times. However, the partitioning methods offer a detailed description of the circuit preserving the detail contained within the original interconnected model. This is an interesting discussion that we didn’t bring to this paper because we considered it out of scope. However, if the reviewer considers it, we can think on talking about it as an additional topic. Thanks for your remarks, this has been a great conversation.

Reviewer 2 Report

The abstract should be rewritten by including more details of the existing challenges and the proposed solutions.

The equations should be written properly by including a description of all symbols used in the equation.

The acronyms/symbols used in the whole paper should be checked and described properly.

The conclusions should be rewritten by including more details and future directions.

 

Author Response

Dear reviewer. Thanks for your kind feedback and comments, as well as the time dedicated for reviewing our paper. Our answers to the reviewer’s comments as follows:

1. The abstract should be rewritten by including more details of the existing challenges and the proposed solutions.

Dear reviewer, thanks for your comments and for taking the time for reviewing our paper. I agree that the details on the needs and the study results should be widely addressed at the abstract, but considering the limited space of 200 words, we put a major effort on the introduction considering the observations made by other reviewers for better introducing the reader into this topic.

2. The equations should be written properly by including a description of all symbols used in the equation.

Thanks for your suggestion. For addressing this comment globally, we have added a nomenclature table at the beginning of the paper. We did a couple of changes to the abstract that we expect allow us to cover the reviewer’s comments. The changes can be seen in lines 17-22 as follows:

“This paper presents a simplified algorithm version of A-Diakoptics for modernizing sequential power simulation tools to use parallel processing. This simplification eliminates critical points found in previous versions of A-Diakoptics, improving the performance of the algorithm and facilitating its implementation to perform QSTS simulations. The performance of the new version of A-Diakoptics is evaluated as its implementation into EPRI’s open-source simulator OpenDSS, using standard computing architectures and is publicly available.”

3. The acronyms/symbols used in the whole paper should be checked and described properly.

Thanks for your suggestion. For addressing this comment globally, we have added a nomenclature table at the beginning of the paper.

4. The conclusions should be rewritten by including more details and future directions.

Thanks for this remark and yes, we have included a couple of paragraphs that we hope, will cover the reviewer’s comments:

“This implementation of A-Diakoptics was validated using yearly QSTS simulations in terms of fidelity and performance using as reference the interconnected model. The simulation results revealed the benefits of modernizing sequential power simulation tools into parallel processing, which is the standard for computing architectures nowadays. This paper introduces A-Diakoptics as a method for achieving such modernization and OpenDSS, given that is distributed as an open-source project through the internet, presents a practical example on how this modernization can be implemented.

A simulation test case was used for illustrating the computational time gains com-paring the sequential and parallel performance on a yearly QSTS. The computing time gains are evaluated for an incremental number of partitions, displaying, and discussing the benefits and limits of the algorithm. This test case is publicly available on the internet for the reader to evaluate if desired.”

Also, adding the future directions and references to the work developed at the end of the conclusions:

“Additional to the methods presented in this paper, other using different parallelization techniques and approaches have been also investigated as the result of the Sunshot National Laboratory Multiyear Partnership Program (SuNLaMP). The findings, conclusions as well as a more extended explanation of these methods can be found at [13]. For future implementations and demonstrations on methods for accelerating simulations, we expect to keep using OpenDSS as the open-source project for serving as a proof of concept for this type of research, demonstrating viable techniques and methods for modernizing power simulation tools.”

Reviewer 3 Report

The article presents the A-Diakoptics method (Actor-based Diakoptics) that allows for the acceleration of flow analysis (QSTS) in distribution networks. The solutions proposed by the authors were implemented in software developed by EPRI (and other institutions) and referred to as OpenDSS. In fact, the article is complete in terms of structure, formulation of the problem, theoretical background, sample network analysed, and discussion of sample results.

However, I have doubts whether the group of readers who will understand the details appearing in it will be wide. The article is extremely condensed in terms of terminology, and its editorial level leave a lot to be desired. The article would be ready for publication, but requests major revision to be carefully prepared in an editorial sense. Authors should consider correcting the following points:

1. It is not known why subscripts of significant variables are not written as indexes but as small letters. This is not a coincidence but a trend that needs improvement. Examples: IPC (E), YII, YBUS, ZCC, ZTT, Ysystem.
2. Most of the variables used in the equations are complex numbers. This fact should be specially marked. And if not, at least the reader should be reminded of this fact.
3. The physical meaning of the E (basic analysis variable) is nowhere explained. You can guess (if you can) that this is a node voltage, but it has two parameters - module and angle and there is not a word about it.
4. Moreover, in Figure 1, voltage is marked with a different letter -V, is it really difficult to standardize it ?
5. The quality of the network diagram in Fig. 2 is poor, the main substation HV/MV location is not visible and described. It is not known whether there are dispersed sources and where. The meaning of the symbols (coloured arrows) should be given in the caption.
6. Figures 3 and 4 are of poor quality, even by using Excel spreadsheet, the figures can be better readable. Currently, they resemble the work of an impressionist painter
7. The list of literature is corrupt, 18 items in this area are not many. I remember work done in the nineties at Durham University.at least the elements indicated below out of respect for the reader.
8. Figures 3 and 4 are of poor quality, even by using Excel spreadsheet, the figures can be better readable. Currently, they resemble the work of an impressionist painter
9. The list of literature is rather short, 18 items in this area are not many. I remember work done on Diakoptics and parallel solution of load flow problem in the nineties at Durham University.

 

Author Response

Dear reviewer. Thanks for your kind feedback and comments, as well as the time dedicated for reviewing our paper. Our answers to the reviewer’s comments as follows:

1. However, I have doubts whether the group of readers who will understand the details appearing in it will be wide. The article is extremely condensed in terms of terminology, and its editorial level leave a lot to be desired. The article would be ready for publication, but requests major revision to be carefully prepared in an editorial sense. Authors should consider correcting the following points:

Dear reviewer, thanks for your comments and for taking the time for reading the paper. We agree, a topic such A-Diakoptics comprehends many applications and extensive background that is very difficult to condense in a single paper. However, since this is not the first time we write about this topic we tried to address the reader to our previous publications, which contain a detailed literature review as well as previous results. Also, by integrating the other reviewers’ comments we tried to address several holes that we failed to fill in the previous version of this paper. Those are mentioned in our responses to the other reviewers.

2. It is not known why subscripts of significant variables are not written as indexes but as small letters. This is not a coincidence but a trend that needs improvement. Examples: IPC (E), YII, YBUS, ZCC, ZTT, Ysystem.

Correct, we added a nomenclature table at the beginning of the paper.

3. Most of the variables used in the equations are complex numbers. This fact should be specially marked. And if not, at least the reader should be reminded of this fact.

This is highlighted when described variables in the nomenclature table.

4. The physical meaning of the E (basic analysis variable) is nowhere explained. You can guess (if you can) that this is a node voltage, but it has two parameters - module and angle and there is not a word about it.

Moreover, in Figure 1, voltage is marked with a different letter -V, is it really difficult to standardize it ?

Thanks for your remark. We have updated the figure to make the equations consistent along the paper. Also, we have added notes at the nomenclature table specifying that the current and voltage values are given as complex numbers.

5. The quality of the network diagram in Fig. 2 is poor, the main substation HV/MV location is not visible and described. It is not known whether there are dispersed sources and where. The meaning of the symbols (coloured arrows) should be given in the caption.

Figure 2 has been improved to provide a better lookout of the model under discussion.

6. Figures 3 and 4 are of poor quality, even by using Excel spreadsheet, the figures can be better readable. Currently, they resemble the work of an impressionist painter

Thanks for this comment, we agree, and we appreciate the reviewer’s description of what turns out to be a messy error graph. To solve this issue, we have updated the graph using a block-and-whiskers graph, which is more organized and delivers the same type of feedback.

7. The list of literature is corrupt, 18 items in this area are not many. I remember work done in the nineties at Durham University.at least the elements indicated below out of respect for the reader.

Thanks for this comment and we feel identified with the reviewer on this observation. However, given that we have previous publications presenting the features and advances on A-Diakoptics implementations, our aim was to redirect the reader to those works, where an extensive literature review and cases was made, instead of copy-paste the same references and content in this work. Trying to solve the reviewer concerns, we have added the following text in line 65 aiming to redirect the reader in case a more extensive literature review is required:

“In previous publications, the authors of this paper made an extensive literature review on Diakoptics as well as provided a populated list of references for the reader to review [4-6].”

Reviewer 4 Report

The paper lacks novelty and contributions. Employing an OpenDSS software to evaluate the performance of the new version of A-Diakoptics is not a contribution. The first half of the paper is purely theoretical and results are not just apt that may suit well for a conference paper and not a reputed journal like Energies. More work needs to be done.

Author Response

Dear reviewer. Thanks for your kind feedback and comments, as well as the time dedicated for reviewing our paper. Our answers to the reviewer’s comments as follows:

1. The paper lacks novelty and contributions. Employing an OpenDSS software to evaluate the performance of the new version of A-Diakoptics is not a contribution. The first half of the paper is purely theoretical and results are not just apt that may suit well for a conference paper and not a reputed journal like Energies. More work needs to be done.

Dear reviewer, thanks for your constructive comments and for your time dedicated for reading our paper. However, we would like to highlight several features about the work presented here to address your concerns:

1. As mentioned at section 2.2, this work and the results here presented are derived from the Sunshot National Laboratory Multiyear Partnership Program (SuNLaMP), a DOE sponsored initiative from 2016 to 2018, that later was enhanced and updated with the improvements here presented.
2. The work presented is not just a simulation done in OpenDSS. Instead, as described across the paper it is a demonstration on the benefits of updating sequential power system simulation tools. The proof of concept is provided to the reader through its direct implementation in an open-source initiative materialized in OpenDSS. So, the methods here described theoretically have a practical implementation that is available for the public.
3. Also, the test cases here presented have been provided for the public so the reader can validate and use the results of this paper.
4. Parallel processing is a research field in power systems simulation that is widely used nowadays. The content presented in this paper by introducing the reader to spatial parallelization, also mentions other works as well as cites current and recent research projects dealing with advancing simulation in the power systems domain.

These are just a few remarks that we would like the reviewer to consider. We hope that by mentioning these points to the reviewer as well as the feedback obtained from the other reviewers, reviewer 4 could reconsider his position respect to the paper content and scope.

Round 2

Reviewer 2 Report

The paper can be considered for publication after careful and serious proofreading.

 

Reviewer 3 Report

The authors took into account the comments presented in the review in a significant way. The quality of the article is better now. In a scientific sense, I still rate it as "average", but believe it can be published as presented. The authors claim that the bibliography is sufficient, I have a different view in this respect - 19 items, including 4 of Mr Montenegro and 2 sources from 60 years ago, are not enough in today's information era. But this is a question of the authors' responsibility. More specifically, the better terms for quantities E and I_PC are vectors rather than arrays.

Reviewer 4 Report

The authors have given a satisfactory explanation. The paper may be considered now. Thanks for clarifying things in detail.