Analysis of Charging Infrastructure for Private, Battery Electric Passenger Cars: Optimizing Spatial Distribution Using a Genetic Algorithm
Round 1
Reviewer 1 Report
The paper provides a good and thorough research of charging infrastructure planning in Berlin. The paper could be published but several shortcomings have to be overcome first.
State of the art
I do not mind that the authors only use meta studies to review the state-of-the-art but their conclusions are a bit to quick. There are already a number of studies out there that use genetic algoritms in combination with traffic data. It has to be made more explicit what the research gap is. A quick search in google scholar already resulted in the following papers:
https://etrr.springeropen.com/articles/10.1007/s12544-017-0239-7
https://www.sciencedirect.com/science/article/pii/S0925231220316556
https://dl.acm.org/doi/abs/10.1145/3512290.3528859
https://www.sciencedirect.com/science/article/pii/S1569190X18300911
3.1.3 Agent detour
The choice for detour limits is made here, but why these choices are made is discussed in section 5. It is better to already do this in this section already. Please also not that more literature is already available on this topic. See e.g.
https://www.sciencedirect.com/science/article/pii/S0965856421001087
3.2 EV population
The authors choose to charge the vehicles only to 80% because of efficiency losses. This could be true for fast charging, but is not the case for AC charging points. As well this does not correspond to the results in which the majority of trips starts with 85% SoC.
3.3 Starting population
In lines 279-283 decisions about home charging availability are discussed. It is however unclear why 40% availability leads to 60% home charging in Berlin and 87% to 100% in Brandenburg.
4.2 Results of the final generation
Table 5: Please change Mean start/end SoC to clarify that this is about trips and not charging sessions.
Table 6: This is my biggest point for the review: There are very big variations in the bottem and top edge solutions. The top edge results also lead to way more charging stations and sessions, but agents with SoC 0% barely changes. Why are there so many more charging processes? What value do they constitute. I am a bit puzzled. I am not sure how the pareto optimisation worked. It seems to have over estimated the importance of mean detour
5.1 results
This section is very long. It is better to use sub-sections or more whitespaces to make it more readable.
Author Response
Dear reviewer,
thank you for your valuable feedback and for taking your time to read our paper. We tried to respond to your suggestions and remarks as best as possible.
The paper provides a good and thorough research of charging infrastructure planning in Berlin. The paper could be published but several shortcomings have to be overcome first.
State of the art
I do not mind that the authors only use meta studies to review the state-of-the-art but their conclusions are a bit to quick. There are already a number of studies out there that use genetic algoritms in combination with traffic data. It has to be made more explicit what the research gap is. A quick search in google scholar already resulted in the following papers:
https://etrr.springeropen.com/articles/10.1007/s12544-017-0239-7
https://www.sciencedirect.com/science/article/pii/S0925231220316556
https://dl.acm.org/doi/abs/10.1145/3512290.3528859
https://www.sciencedirect.com/science/article/pii/S1569190X18300911
Thank you for this feedback. We tried to keep the SoA section short but we see your point. We were aware of the recommended papers (and some more) and have included a detailed description of them as well as a delimitation of our work.
3.1.3 Agent detour
The choice for detour limits is made here, but why these choices are made is discussed in section 5. It is better to already do this in this section already. Please also not that more literature is already available on this topic. See e.g.
https://www.sciencedirect.com/science/article/pii/S0965856421001087
SrV Berlin 2018 Tabellenbericht p.62: mean distance walking trips 900m:
https://www.berlin.de/sen/uvk/verkehr/verkehrsdaten/zahlen-und-fakten/mobilitaet-in-staedten-srv-2018/
We included a short explanation on the values we assumed for the acceptable and maximum detour of the agents in section 3.1.3 around line 286.
3.2 EV population
The authors choose to charge the vehicles only to 80% because of efficiency losses. This could be true for fast charging, but is not the case for AC charging points. As well this does not correspond to the results in which the majority of trips starts with 85% SoC.
We see this as a conservative assumption and added a short explanation in section 3.2 around line 318. Please note that the mean start SoC means the mean SoC before the first trip of all agents and not the mean SoC at the beginning of each trip. We renamed the values accordingly within the affected tables and also within the text at several positions.
3.3 Starting population
In lines 279-283 decisions about home charging availability are discussed. It is however unclear why 40% availability leads to 60% home charging in Berlin and 87% to 100% in Brandenburg.
There was a typo for the values regarding Berlin.
For Brandenburg it is assumed that every agent has a homecharger due to the high availability of private parking space and the fact that no chargers are placed in Brandenburg in this scenario. We also added a short explanation in section 3.3 around line 361.
4.2 Results of the final generation
Table 5: Please change Mean start/end SoC to clarify that this is about trips and not charging sessions.
Like mentioned before, we renamed those values.
Table 6: This is my biggest point for the review: There are very big variations in the bottem and top edge solutions. The top edge results also lead to way more charging stations and sessions, but agents with SoC 0% barely changes. Why are there so many more charging processes? What value do they constitute. I am a bit puzzled. I am not sure how the pareto optimisation worked. It seems to have over estimated the importance of mean detour
Due to the higher number of charging points more agents have access to charging infrastructure. That's why the number of charging processes increases (explained in 4.1). The agents with zero percent soc do not increase in the same way due to the fact that the scenario only lasts one day and their journeys over the day are not long enough to reach zero percent. This problem is also discussed in section 5.1. We also added an explanation regarding the comment at line 706.
5.1 results
This section is very long. It is better to use sub-sections or more whitespaces to make it more readable.
We added whitespace for this section.
Reviewer 2 Report
The author did a fantastic job in this work. However few elements are not clear to me. Hence they may consider those aspects which will make it wider.
First, they focused too much on the Berlin context whereas which is a global issue. Also, the author addressed the lack of data that makes this type of work poses a critical situation. How future readers can get data from your work?
Figures such as Figure 4 and 6 and also others needs a bit more clarity. Now it does not make any clear sense.
Overall take home message must be more clearer. In discussion section talking about method is not a good idea.
Author Response
Dear reviewer,
thank you for your kind words, the valuable feedback and for taking your time to read our paper. We tried to respond to your suggestions and remarks as best as possible.
The author did a fantastic job in this work. However few elements are not clear to me. Hence they may consider those aspects which will make it wider.
First, they focused too much on the Berlin context whereas which is a global issue. Also, the author addressed the lack of data that makes this type of work poses a critical situation. How future readers can get data from your work?
The introduced method of providing a suitable charging infrastructure scenario is tested with the example of berlin. With a similar microscopic traffic simulation it is also transferable to other cities or areas. The provided paper does not aim to provide final data about charging infrastructure. It tries to provide a methodology to generate suitable charging infrastructure based on data of an arbitrary agent-based microscopic traffic simulation which can be optimized by arbitrary criteria. We tried to be more clear about mentioning that within the abstract and the conclusion of the paper.
Figures such as Figure 4 and 6 and also others needs a bit more clarity. Now it does not make any clear sense.
We are not sure if we can follow your comment about those figures. The data points show the solutions(infrastructure scenarios) and their characteristics regarding the evaluation criteria. The color of the dots shows which Front of solutions a solution belongs to. The solutions in front zero are pareto optimal.
Overall take home message must be more clearer. In discussion section talking about method is not a good idea.
We do discuss our method in the discussion section in all of our papers. We tried to specify the goal of the paper and the outcome within the introduction and the conclusion.
Reviewer 3 Report
In the submitted paper a methodology for spatial distribution of charging infrastructure is evaluated by investigating a scenario with a market penetration of electric vehicles with onboard batteries of hundred percent. The aims of the work are the development of various charging infrastructure scenarios, either public or private, which are suitable to cover the charging demand.
In what follows some comments are reported which could be considered by the Authors as possible improvements.
The paper is a discussion on some scenarios. There no models, no clear theoretical work beyond some data. The approach seems trivial, since neglecting several aspects related to the inclusion of charger infrastructure in power networks. Also, the assumptions made are not clearly defined from a theoretical point of view.
The problem of charging infrastructure in case of massive deployment of electric vehicles is strictly related to the ability of the electrical network to allows larger share of power. However, this aspect is not considered among the constraints included in the submitted paper. This aspect should be clearly defined and discussed.
Figure 10 is unclear and it is not clearly discussed. It is suggested to better present and clarify why those profiles have those shapes.
The quality of figures is generally low and it seems that are scanned from other documents. Please improve.
Fig. 1 seems taken from other works. If so, please correctly cite the font in the legend.
It is suggested to improve the literature survey by adding other relevant scientific papers.
In the references, documents should be reported together with all details which allow readers to find them.
Author Response
Dear reviewer,
thank you for your valuable feedback and for taking your time to read our paper. We tried to respond to your suggestions and remarks as best as possible.
In the submitted paper a methodology for spatial distribution of charging infrastructure is evaluated by investigating a scenario with a market penetration of electric vehicles with onboard batteries of hundred percent. The aims of the work are the development of various charging infrastructure scenarios, either public or private, which are suitable to cover the charging demand.
The goal which you describe here is indeed the aim of the method, but the paper aims more towards showing that the provided method is in principle able to do that. We tried to specify that in abstract, introduction and conclusion.
In what follows some comments are reported which could be considered by the Authors as possible improvements.
The paper is a discussion on some scenarios. There no models, no clear theoretical work beyond some data. The approach seems trivial, since neglecting several aspects related to the inclusion of charger infrastructure in power networks. Also, the assumptions made are not clearly defined from a theoretical point of view.
The problem of charging infrastructure in case of massive deployment of electric vehicles is strictly related to the ability of the electrical network to allows larger share of power. However, this aspect is not considered among the constraints included in the submitted paper. This aspect should be clearly defined and discussed.
To investigate the impact of charging infrastructure on the power grid much more data is necessary. We focus on the demand side which is also mentioned in the paper when choosing the evaluation criteria. Theoretically it is possible to include the usage of the power grid as a criterion if a suitable model is provided but this is not considered in this work. We also mention that most works regarding this topic are focussing on the demand side in our state of the art section. It is more likely that future work will produce data about necessary grid capacities in certain areas rather than directly consider the utilization of the available power grid due to the lack of data.
Figure 10 is unclear and it is not clearly discussed. It is suggested to better present and clarify why those profiles have those shapes.
Figure 10(now figure 12) is showing a heat map which contains informations about the number of started activities. It shows that there is a correlation between number of activities started and charging infrastructure placed.
The quality of figures is generally low and it seems that are scanned from other documents. Please improve.
We will inform the editors about that problem. In our pdf documents the resolution of the figures is much better compared to the figures in our pdf documents.
Fig. 1 seems taken from other works. If so, please correctly cite the font in the legend.
This is our own representation.
It is suggested to improve the literature survey by adding other relevant scientific papers.
We added some more relevant papers to our literature research.
In the references, documents should be reported together with all details which allow readers to find them.
Reviewer 4 Report
Comments for author File: 
Comments.pdf
Author Response
Dear reviewer,
thank you for your valuable feedback and for taking your time to read our paper. We tried to respond to your suggestions and remarks as best as possible.
We tried to specify the scope of the work within the abstract and the introduction and also the achievements and contributions of our work in the outcome section.
We also extended our literature research with different specific papers regarding our topic. We have read the two recommended papers with great interest. However, since the side of the energy network is not within the system boundary of this paper, the first paper cannot be related to this study. The second paper, however, represents an excellent investigation which we have included in the literature section and from which we have distinguished our own investigations. It is important to note that we explicitly exclude the power grid perspective in order to address the spatial placement and dimensioning of infrastructure. In future studies, this may well be brought together.
We added two flowcharts to make the method more understandable. See figure 1 and figure 5.
Furthermore we edited the discussion of the method and made the limitations more clear in this section and also in the conclusion
We informed the editors about the bad quality of the figures because in our pdf documents the quality is much better.
Round 2
Reviewer 2 Report
accept
Reviewer 3 Report
Thanks for replies.
I do not have other comments.
Reviewer 4 Report
I have seen that most of the suggestions I made in the first review have been corrected, but there are still some comments which are not been addressed.
However, on the bases of a major portion correction has been made, I accept the manuscript.
