Surrogate DC Microgrid Models for Optimization of Charging Electric Vehicles under Partial Observability
Round 1
Reviewer 1 Report
No information neither on the DC microgrid nor on the EVs. The title poses a problem because the authors do not present any model and configuration of this DC micgorigrid.
How many source has been integrated into this DC microgrid for EVS charging applications?
No information regarding EV batteries in this study.
In parts 3.1 and 3.2, what are the problems to be solved and what are the cost functions? the formulation of the problems are absent.
The references are poor, theauthors can use the references below to improve their article.
1) More and Faster Energy Transfer Capability for Battery Chargers of Electric Vehicles, WAC 2021.
2) A state of the art of DC microgrids for electric vehicle charging, ISPCC 2017.
3) A smart cyber physical multi-source energy system for an electric vehicle prototype, Elsevier 2020.
Author Response
1. No information neither on the DC microgrid nor on the EVs. The title poses a problem because the authors do not present any model and configuration of this DC micgorigrid. How many source has been integrated into this DC microgrid for EVS charging applications? No information regarding EV batteries in this study.
Response: Since the paper considers focuses on charging planning under partial observability, battery properties like degradation are not relevant for the charging algorithm and therefore not reported. We have mentioned this argument in the paper, see lines 91-99.
2. In parts 3.1 and 3.2, what are the problems to be solved and what are the cost functions? the formulation of the problems are absent.
Response: The formulation of the optimization problems of the surrogate models are now included with full detail in sections 3.1 and 3.2.
3. The references are poor, theauthors can use the references below to improve their article.
1) More and Faster Energy Transfer Capability for Battery Chargers of Electric Vehicles, WAC 2021.
2) A state of the art of DC microgrids for electric vehicle charging, ISPCC 2017.
3) A smart cyber physical multi-source energy system for an electric vehicle prototype, Elsevier 2020.
Response: Indeed reference 2 above is relevant for this work. We’ve included it in line 93.
Reviewer 2 Report
The paper is well written. This paper studies the problem of EV charging planning under limited grid capacity and gives two methods to solve the limited-information problem by changing the grid topology. The contribution is, these solutions offer a trade-off between performance and quickness.
The shortcomings are,
- the methods did not be valided with real experiments.
- the figures in the paper are not Vector Graphics.
line 332, the reference is a journal or conference contribution? Please complete the information. Thank you.
Author Response
The paper is well written. This paper studies the problem of EV charging planning under limited grid capacity and gives two methods to solve the limited-information problem by changing the grid topology. The contribution is, these solutions offer a trade-off between performance and quickness. The shortcomings are,
1.The methods did not be valided with real experiments.
Response: It is indeed an important step to also validate the experiments in a real grid. However, we currently do not have access to such a grid. We have used a very detailed simulator which implements the physics as known to hold in a real grid. The experiments include one that uses the topology from the IEEE16 grid which has been used in the literature to study power flow in DC distribution grids (Li et al. [13]). To ensure realistic conditions, we have run experiments on a range of line capacities and demands, and also included some other (randomly generated) topologies. Therefore we are confident that the conclusions drawn from these simulation results are quite general and will also hold in practice. This discussion is included in the conclusion.
2. The figures in the paper are not Vector Graphics.
Response: We have converted the figures into the svg format.
3. line 332, the reference is a journal or conference contribution? Please complete the information. Thank you.
Response: We have completed the reference. Thank you for pointing that out.
Reviewer 3 Report
Dear author,
Please see attached.
Thank you
Comments for author File: 
Comments.pdf
Author Response
It should be noted that this is a scientific journal entry. Therefore, first-person pronouns such as “we” and “they”, should be avoided. Avoid starting sentences with a noun. For example, instead of “important feature” use “An important feature”. Sentences should begin with articles such as the, a, an, etc. unless for specific examples such as starting a sentence with plural nouns, for examples “Important features”.
Response: We have changed many uses of first-person pronouns and tried to improve the sentences.
This paper lacks consistency, for example, the use of present vs past tense and this paper vs we. The same tense should be used throughout the entirety of the paper with specific exceptions. As can be seen in Section 4, the authors switch to past tense. The graphs presented in the figures do not define the representation of the y-axis, with the exception of Figure 9. Again, not consistent. All graphs should be presented in the same manner. If a reader wishes to skim through an article or paper, figures should be easily readable. Finally, this paper includes extremely long sentences that often lead to more than two lines. This makes it difficult to follow the theme of the paragraph and leads to incoherency.
Response: We have fixed the tenses in the paper. All sections, except for Sections 4 and 5,, are now consistently using present tense. The experiments are described using past tense. We have also updated the figures by adding labels to the y-axis and shortened some of the sentences.
An abstract is usually composed of four main elements.
- Introduction
- Methodology
- Results
- Conclusion
The results derived from the methods are explained but the methods used are not clear. Also, conclusion is non-existent.
Response:We have modified the abstract.
List of per-line corrections.
Response: We have addressed the remarks you provided.
Reviewer 4 Report
In “Surrogate DC Microgrid Models for Optimization of Charging Electric Vehicles under Partial Observability”, the author developed a planning algorithm for optimal charging of electrical vehicles based on a weighted objective. The writing and organization of the paper are well done, but there are some critical information missing.
- Have the authors considered using real city grid topology for optimization?
- In the definition of the optimization problem, the authors tried to optimize J over a duration of t, as in Eqn.3. However, in the application of the model in the POEVCP problem, the optimality was only defined at an exact time step. Does the local optimality guarantee total optimality? Why?
- As the authors stated, it is better to plan on the fly as the partially observability of the EV’s on the grid. So why using random guessing algorithm in particular? How does it compare to many other optimization techniques such as particle swarm, genetic algorithm, etc.
- Have the authors considered the charging nodes on the actual grid may be owned by different companies, also, to my personal experiences, the charging stations near downtown or far away at the residential areas may drastically vary in the pricing, could the weighting coefficients in the J formula account for those?
- Please elaborate on the definition or selection of the u_k and c_g weighting coefficients’ values in detail, as they govern the final results.
- As a paper solely focused on numerical optimization, in common practice, the authors should clearly state the values of the upper and lower bounds of the variables, in stead of just a variable name. In addition, convergence criteria are missing from the manuscript. Last but not the least, what are the choices of initial conditions?
- Please demonstrated the optimized parameters, for at least a sample grid.
Author Response
In “Surrogate DC Microgrid Models for Optimization of Charging Electric Vehicles under Partial Observability”, the author developed a planning algorithm for optimal charging of electrical vehicles based on a weighted objective. The writing and organization of the paper are well done, but there are some critical information missing.
1. Have the authors considered using real city grid topology for optimization?
Response: It is indeed an important step to also validate the experiments using real city grid topologies. However, we could not find an open source dataset suitable for this purpose. In our work we used the IEEE16 grid which has been used in the literature to study power flow in DC distribution grids (Li et al. [13]). To ensure realistic conditions, we have run experiments on a range of line capacities and demands, and also included some other (randomly generated) topologies. Therefore we are confident that the conclusions drawn from these simulation results are quite general and will hold for grids with different topologies and/or capacities. We have included this argument in the conclusion.
2. In the definition of the optimization problem, the authors tried to optimize J over a duration of t, as in Eqn.3. However, in the application of the model in the POEVCP problem, the optimality was only defined at an exact time step. Does the local optimality guarantee total optimality? Why?
Response: In our study, we define the planning and execution framework to ensure feasibility of the charging schedules. At each timestep, the planner solves the PO-EVCP problem with a multi-timestep objective. The part of the solution corresponding to the current timestep is then used as an upper bound in the Executor problem, which is a single-timestep OPF problem.
We updated Section 3.3 by expanding the description of the framework and added Algorithm 1 to better illustrate the interaction between the planner and the executor.
3. As the authors stated, it is better to plan on the fly as the partially observability of the EV’s on the grid. So why using random guessing algorithm in particular? How does it compare to many other optimization techniques such as particle swarm, genetic algorithm, etc.
Response: We use the random guessing algorithm only to resolve the uncertainty of the PO-EVCP problem, and the actual solution is performed by the means of classical optimization. Essentially, the random guessing algorithm samples an assignment of EVs to the loads within each cable. Then, this assignment is used to solve the deterministic EVCP problem. While other optimization techniques can be used to solve the optimization problem, we have not considered them, since the classical optimization methods seem to be a better choice for the constrained optimization problem.
We improved the description of the algorithm and renamed it to blind guessing, in order to avoid potential confusions. The updated description can be found at Sections .
4. Have the authors considered the chargng nodes on the actual grid may be owned by different companies, also, to my personal experiences, the charging stations near downtown or far away at the residential areas may drastically vary in the pricing, could the weighting coefficients in the J formula account for those?
Response: We have not considered that in our study. However, different pricing schemes in different nodes can be incorporated in the utility and cost coefficients uk and cg. We believe that this will not affect any of the suggested methods and conclusions will still hold. We have included this argument in the paper, see lines 122-128.
5. Please elaborate on the definition or selection of the u_k and c_g weighting coefficients’ values in detail, as they govern the final results.
Response: It is indeed important to mention that the utility coefficients u_k were an order of magnitude larger than cost coefficients c_k. That made the objective function monotonically increasing. We included a brief discussion about this topic in the paper, see lines 293-297.
6. As a paper solely focused on numerical optimization, in common practice, the authors should clearly state the values of the upper and lower bounds of the variables, in stead of just a variable name. In addition, convergence criteria are missing from the manuscript. Last but not the least, what are the choices of initial conditions?
Response: We have included upper and lower bounds of the variables in Section 4, see lines 278-283. Concerning initial conditions and convergence criteria, we have used the default MOSEK solver with manually set tolerances. We have included this information in the manuscript, see lines 302-304.
7. Please demonstrated the optimized parameters, for at least a sample grid.
Response: We have included tables with decision variables obtained by different models in the appendix.
Round 2
Reviewer 1 Report
The article does not present any DC microgrid model. The article is poorly prepared and organized. The title of this article does show neither DC microgrid nor EV charging. The references are poor.
Reviewer 4 Report
In the revised version of “Surrogate DC Microgrid Models for Optimization of Charging Electric Vehicles under Partial Observability”, the authors have made substantial improvements over the previous edition, the paper is now more readable and logical. I’m satisfied with the authors' answers to my previous questions. The paper can be accepted as is.
