A Driving Cycle for a Fuel Cell Logistics Vehicle on a Fixed Route: Case of the Guangdong Province
Round 1
Reviewer 1 Report
- Line 34: It isn’t clear “… the 4 working conditions…”, they are mentioned in abstract but it is difficult to associate.
- Table 1: Some variables doesn’t have units; the title is “Simulation parameters”, why is it Simulation? They aren’t real parameters or do you use those variables for any simulation?
- Line 41: hcm doesn’t appear in Table 1.
- Lines 46-50: It isn’t clear the reason of paragraph because it is contradicting data selection.
- Lines 58-59: What is the meaning of consistent in “… the driving condition in each segment is consistent.”
- Paragraph 60: Huertas et al. (2019) define characteristic parameters: “They are parameters or variables that result from any combination of speed and time, such as mean speed, mean positive acceleration.” For this reason, I don’t suggest to assign throttle and brake pedal angle as characteristics parameters. (Huertas, J. I., Quirama, L. F., Giraldo, M., Diaz, J. (2019): Comparison of Three Methods for Constructing Real Driving Cycles. Energies, 12(4), 665. doi: 10.3390/en12040665. https://www.mdpi.com/1996-1073/12/4/665)
- Equation 1: What is the meaning of dsini? In paragraph above, It refers to “… a characteristic parameter vector…”, I suggest to change dsini for CPi.
- Equation 2: You doesn’t mention variables description.
- Figure 8: The points in legends are to small and it is difficult to appreciate colors.
- Lines 119-120: “…the velocity-acceleration probability distribution matrix…” is known in literature as SAPD (speed acceleration probability distribution), I suggest use this nomenclature.
- Line 130: You mentioned “42-day driving data …” but in this line refers only to 1 day, that idea reduces the scope of your work.
- Line 133: You never mentioned the acronyms “FCV and ICEV”, i.e. you always refer to fuel cell vehicle but not FCV.
- I suggest to improve discussion about different between driving cycles for fuel cell vehicles and for internal combustion engine vehicles, and differences between your method and other methods to produce driving cycles.
- Can you use your method to produce driving cycles for ICEV? If the reply is yes, you need to reconsider the scope of your work or compare the results to apply this method for FCV and ICEV.
Author Response
Point 1:
Line 34: It isn’t clear “… the 4 working conditions…”, they are mentioned in abstract but it is difficult to associate.
Respose 1:
We have added the descriptions about "4 working conditions" after it.
Point 2:
Table 1: Some variables doesn’t have units; the title is “Simulation parameters”, why is it Simulation? They aren’t real parameters or do you use those variables for any simulation?
Response2:
Units have been added. The word is misused and we have changed it into "vehicle parameters"
Point 3:
Line 41: hcm doesn’t appear in Table 1.
Response 3:
hcg in table 1 shall be hcm instead, we have modified.
Point 4:
Lines 46-50: It isn’t clear the reason of paragraph because it is contradicting data selection.
Response 4:
RDE is a widely used test. However, there is some difference between the fuel cell vehicles we study and the internal combustion engine vehicle, so we think it's not wise to totally follow the regulation. The paragraph has been changed as follow:
On the one hand, the chosen route consists to the regulation of the real drive emission(RDE) test, which stipulates that the journey's largest altitude should be smaller than 700m. On the other hand, the reason why this article doesn't use RDE test is that it stipulates the difference of the altitude of the whole journey should be less than 100m, however, the chosen route has an altitude's difference of about 300m, because fuel cell logistics vehicle usually runs between cities and the difference of altitude is large.
Point 5:
Lines 58-59: What is the meaning of consistent in “… the driving condition in each segment is consistent.”
Response 5:
Every segment is one kilometer long and it's assumed that the vehicles dynamic parameters, such as vehicle and acceleration, in one particular segment are basically the same.
Point 6:
Paragraph 60: Huertas et al. (2019) define characteristic parameters: “They are parameters or variables that result from any combination of speed and time, such as mean speed, mean positive acceleration.” For this reason, I don’t suggest to assign throttle and brake pedal angle as characteristics parameters. (Huertas, J. I., Quirama, L. F., Giraldo, M., Diaz, J. (2019): Comparison of Three Methods for Constructing Real Driving Cycles. Energies, 12(4), 665. doi: 10.3390/en12040665. https://www.mdpi.com/1996-1073/12/4/665)
Response 6:
We firstly add the two pedal angle and try finding something innovative. However, the correlation analysis indicates that the two pedal angles are theoretically not suitable to used as characteristic parameters. So the throttle and brake pedal angle are eliminated from the parametersgroup.
Point 7:
Equation 1: What is the meaning of dsini? In paragraph above, It refers to “… a characteristic parameter vector…”, I suggest to change dsini for CPi.
Response 7:
Great suggestion and we have modified.
Point 8:
Equation 2: You doesn’t mention variables description.
Response 8:
We have added the descrition.
"in which p is the point to be clustered,miis the cluster center,|p−mi|^2 means the Euler distance between them, k is the number of cluster center."
Point 9:
Figure 8: The points in legends are to small and it is difficult to appreciate colors.
Response 9:
Modified.
Point 10:
Lines 119-120: “…the velocity-acceleration probability distribution matrix…” is known in literature as SAPD (speed acceleration probability distribution), I suggest use this nomenclature.
Response 10:
Modified.
Point 11:
Line 130: You mentioned “42-day driving data …” but in this line refers only to 1 day, that idea reduces the scope of your work.
Response 11:
Sorry, there is a mistake on writing. We use data of 42 days, and only show the one of them in consideration of the limited space. The route is over 90km long and data from one day is divied into about 90 segments, so the number of segments of 42-day data is over 4000.
Point 12:
Line 133: You never mentioned the acronyms “FCV and ICEV”, i.e. you always refer to fuel cell vehicle but not FCV.
Response 12:
We add the full name of FCV and ICEV the first time they appear.
Point 13:
I suggest to improve discussion about different between driving cycles for fuel cell vehicles and for internal combustion engine vehicles, and differences between your method and other methods to produce driving cycles.
Response 13:
"
Nowadays, there are not many fuel cell vehicles actually running on the roads of China, resulting in not much feedback on the development of fuel cell systems based on actual driving cycle. Therefore, the purpose of this article is mainly to focus on an existing structured road with fuel cell commercial logistics vehicles running every day, as a research to promote subsequent fuel cell system development, such as energy management research, control system optimization, and so on.
This study proposed a methodology to construct a representative driving cycle of fuel cell vehicles in Guangdong Province. After being pre-processed, the original data are divided in spatial dimension. Characteristic parameters are derived and K-means algorithm for cluster analysis is used. The results of the effectiveness analysis show that the typical driving cycle we constructed can cover the original data well and reflect the four road structures of highway, urban road, national highway and others.
In the process of constructing the driving cycle, a few prominent differences between fuel cell vehicles (FCV) and internal combustion engine vehicles (ICEV) are shown. For example, a fuel cell logistics vehicle have higher average velocity and smooth acceleration due the its cruising on a highway, which suggests that the occurrence possibility of highway conditions is the most. However, a fuel cell logistics vehicle has a much lower maximum speed no more than 80 km/h, which leads to a lower energy consumption when the cycle is used for vehicle power analysis. The facts above proves again the necessity of building an independent driving cycle for FCV. The validity analysis of the driving conditions has provided an outstanding evidence of the correctness of the driving cycle built in this paper, and the cycle can be taken as basis for the optimization research of energy management strategies of fuel cell powertrain system. As more and more enterprises pay much attention on the development and application of commercial vehicles that run on a typical route, it's possible for the researchers in those enterprises to analyze the energy consumption along the route following the procedure this paper puts forward, and plans wisely where a gas station should be built."
Point 14:
Can you use your method to produce driving cycles for ICEV? If the reply is yes, you need to reconsider the scope of your work or compare the results to apply this method for FCV and ICEV.
Response 14:
Yes it can be used to produce driving cycles for ICEV. The method might be similar but the vehicles under studied are different. This paper makes some research of the differnce between the FCV and ICEV, such as the vehicle parameters and the routine's charactersic parameters, and produce a driving cycle as this paper indicates, and successfully used as a supportive data for an energy distribution strategy used for an FCV. It's like an instruction and show it's possible to produce such a driving cycle especially for FCV.
Reviewer 2 Report
## Abstract
- I would suggest to rework your abstract, based on the following questions:
- What is the overall topic of your paper?
- What is the motivation for your research?
- How does your dataset looks like
- Describe your overall methodology but, leave out details like the number of clusters. That's to specific for an Abstract.
- What are your main results, what is your contribution on that field?
## Introduction:
- What do you mean with mild working conditions? Is it true that commercial vehicles are operated with less stress than personal vehicles? Please provide a source for that statement.
- Line 35, wich working conditions do you mean with above? Can't find them.
- Can you specify, which type of vehicle you address? Is it a truck, ist it a van?
## Chapter State of the Art
- Missing entirely. For a Journal Article I would expect at least 1-2 pages where you describe the state of the art, and discuss the weaknesses and needs for further research on your topic. You did that already in parts in your Introduction, but that's rather short.
- You may consider this article and´the references included: https://doi.org/10.1109/itsc.2018.8569547
- Even if there may not exist driving cycles for Fuel Cell Commercial Vehicles, the methodology of other approaches on how to generate a driving cycle is still valid. Whats your difference there? Please discuss that in this chapter.
## Data aquisition and pre-processing
- Please provide a more detailed description how the data was collected. Vehicle Parameters seem to be rather unimportant, as vehicle dimensions won't primarliy affect the driving cycle.
- How many observations do you have, whats the sampling frequency, whats the accuracy of the sensor platform, which sensors do you use, do you have outages in the recording...
- Are you using, just a single route? If so, why exactly this route?
-
## Clustering of Vehicle Working Conditions Based on K-means
- Line 58 how do you ensure that the driving condition is consistent? Why 1 hour intervals?
- 3.2. Where do you get those parameters from? Please describe your data-set more in detail. Which features come from measures, which features are derived from other features.
- Provide a table of all characteristic parameters.
- Line 63/64, why?
- Figure 5 and 6 which features belong to which numbers?
## Discussion
- The discussion is to short, and does not list the main issues of this paper:
- The probe data-set contains only a single route. How can one obtain a standard driving cycle from that route?
- You mentioned above that the terrain of the route may be critical. How does this affect your results?
- How does your contribution improve the current state of the art?
- Which points need further research?
Author Response
Point 1: Abstract
- I would suggest to rework your abstract, based on the following questions:
- What is the overall topic of your paper?
- What is the motivation for your research?
- How does your dataset looks like
- Describe your overall methodology but, leave out details like the number of clusters. That's to specific for an Abstract.
- What are your main results, what is your contribution on that field?
Response 1:
The abstract has been modified as suggested.Below is the revised abstract.
Point2: Introduction
What do you mean with mild working conditions? Is it true that commercial vehicles are operated with less stress than personal vehicles? Please provide a source for that statement.
- Line 35, wich working conditions do you mean with above? Can't find them.
- Can you specify, which type of vehicle you address? Is it a truck, ist it a van?
Response 2:
- There is indeed something wrong with using the word mild in the article. We modify this description into a more structured road condition. What we want to express in this article is that most commercial vehicles are not as complicated as passenger cars because of their simple paths.
– The four famous standard driving cycles, such as JC10, FTP75 and ECE+NEDC, listed in line 21 are the 4 working conditions mentioned in line 35.
– As a logistics function, the vehicles we use are mainly trucks
Point 3:Chapter State of the Art
- Missing entirely. For a Journal Article I would expect at least 1-2 pages where you describe the state of the art, and discuss the weaknesses and needs for further research on your topic. You did that already in parts in your Introduction, but that's rather short.
- You may consider this article and´the references included: https://doi.org/10.1109/itsc.2018.8569547
- Even if there may not exist driving cycles for Fuel Cell Commercial Vehicles, the methodology of other approaches on how to generate a driving cycle is still valid. Whats your difference there? Please discuss that in this chapter.
Response 3:
- To the first and second comment, we have added several references about the significance of the driving cycle of fuel cell to its R&D work.
– In the current tide of new energy for automobiles, there is no doubt that lithium-power battery is the most spotted power systyem. However, the shortcomings of lithium battery electric vehicles such as higher emissions throughout the full life cycle and long charging time give a segmented market for fuel cell vehicles. From the perspective of comprehensive policies and operating costs, commercial vehicles running on structured roads are the best entry point for these fuel cell engine suppliers. Nowadays, there are not many fuel cell vehicles actually running on the roads of China, resulting in not much feedback on the development of fuel cell systems based on actual driving cycle. Therefore, the purpose of this article is mainly to focus on an existing structured road with fuel cell commercial logistics vehicles running every day, as a research to promote subsequent fuel cell system development, such as energy management research, control system optimization, and so on.
It is true that the typical driving cycle extraction method used in this article has not changed much from the driving cycle of other power systems. Everyone follows such a pipelined process that data acquisition, determination of feature vectors, data pre-processing, and statistical methods are used to describe each feature quantity in the overall working condition. Finally, the typical driving cycles are determined, but those roads cannot be used well in the development of our existing fuel cell system. Therefore, on the eve of the start of this industry, we believe that it is very necessary to develop a typical driving cycle for fuel cell logistics vehicles in a specific area.
Point 4:Data aquisition and pre-processing
- Please provide a more detailed description how the data was collected. Vehicle Parameters seem to be rather unimportant, as vehicle dimensions won't primarliy affect the driving cycle.
- How many observations do you have, whats the sampling frequency, whats the accuracy of the sensor platform, which sensors do you use, do you have outages in the recording...
- Are you using, just a single route? If so, why exactly this route?
Response 4:
- For the method of DAQ, we use equipments provided by a third party. The specific solution involves the company's secrets and cannot be reflected in the paper. But we can briefly reveal in the comments that we arrange the sensors in the various subsystems of the vehicle, and then these data will be transferred to the database of our laboratory through LTE.
– As for the remotely received data, our sampling frequency is 1Hz. I cannot answer the question about the accuracy and type of the sensor. There have been interruptions, we also mentioned in the paper, the 42-day data is our relatively complete data after preliminary screening.
– The data in our database mainly comes from two cities, Shanghai and Guangdong. The reason for choosing this route is that the route includes relatively complete road features. In the future, as our vehicles run in more cities, the database will be updated in time. At that time, we can use the current method to obtain the typical driving cycle for that city.
Point 5:Clustering of Vehicle Working Conditions Based on K-means
- Line 58 how do you ensure that the driving condition is consistent? Why 1 hour intervals?
- 3.2. Where do you get those parameters from? Please describe your data-set more in detail. Which features come from measures, which features are derived from other features.
- Provide a table of all characteristic parameters.
- Line 63/64, why?
- Figure 5 and 6 which features belong to which numbers?
Response 5:
- Line 58 does not mention 1 hour. In theory, the finer the segements are divided, the better the continuity and consistency in each block. On the premise of collecting 42 days of valid data, this study left a total of about 4000 kilometers after preliminary screening of these data. We partition the data according to the spatial dimension, and consider the calculation accuracy and calculation speed, the division unit is set to 1 km
– changed as conducted, Considering space limitations, the characteristic parameter chart is omitted.
– Since the two characteristic parameters of maximum acceleration and minimum acceleration are cumulative characteristics, they are not statistically significant.
– Excuse me, I don't understand what you mean.
Point 6:Discussion
- The discussion is to short, and does not list the main issues of this paper:
- The probe data-set contains only a single route. How can one obtain a standard driving cycle from that route?
- You mentioned above that the terrain of the route may be critical. How does this affect your results?
- How does your contribution improve the current state of the art?
- Which points need further research?
Response 6:
- The discussion has been modified
– At present, the number of vehicles running on the actual road is limited. We analyzed that this road contains a variety of road features, so we chose it.
– First of all, we listed the terrain data to prove that the road we chose is effective. Secondly, the impact of terrain on the fuel cell system is mainly reflected in the intake pressure of the air compressor, because the power level of fuel cell system we use at this stage is not very large, taking into account the cost and policy factors, if we choose a cross very high and very low altitude routes, then the controller design of the fuel cell system will have great challenges, the actual output power will be greatly reduced.
– Point4 and point5:This part of our research is mainly for the follow-up energy management and other subsystem controller development to provide reference, our research results can be used in the follow-up development of typical working conditions in other cities.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Point 1
- OK!
Point 2
- OK!
Point 3
- Can't check the added references, they are missing in this version.
- All those things you mentioned in the comments could be part of a discussion of the current state of the art.
- I leave that up to the decision of the Editor. IMHO, a journal Article should contain a section where you clearly work out the the present state of the art, discuss the findings of others, and state why your approach is needed to add a contribution to the science community.
Point 4
- This may be a miss understanding. I'm not that much interested in the specific solution. But from a methodological perspective more information, on how and which data was collected would be necessary here:
- Mentioning, that you use internal sensors and transmit those values over LTE would be enough of explanation, but must be provided. More information on which parameters you gather is mandatory!
- Just some thoughts (list them in a table for increased readability):
- How many vehicles did you observe,
- how many trips,
- how many km in total,
- how many drivers,
- which sensor parameters (GNSS, ACC, GYRO )
- Don't get me wrong, but vehicle parameters and the terrain profile of the route don't match to the heading ("Data acquisition and preprocessing"), if thats not the point you wanna address, use something like "experimental setup", "scenario definition"...
- Figure 4,5: Did not noticed that before: Typo in y-Axis label - must be (km/h).
- If there have been interruptions, just write a short explanation. Also mention shortly which cleaning techniques you applied.
- "The data in our database mainly comes from two cities, Shanghai and Guangdong. The reason for choosing this route is that the route includes relatively complete road features. In the future, as our vehicles run in more cities, the database will be updated in time. At that time, we can use the current method to obtain the typical driving cycle for that city." Why don't you give that explanation in the paper? - It's completely fine!
- One thought on Figure 3: It would be interesting to see, how much the different observations repeat in their patterns. Figure 3 and Figure 4 nearly show the same thing. It may improve the paper if you provide an analysis on that in Figure 3. (optional)
Point 5
- Sorry I made a typo, I meant 1 km sections.
- Still don't get why 1 km and why not 2 km or 500 m. It looks like a "guess". Thats not a problem, but please report that in the paper. Without explanation, one might ask why exactly 1 km is chosen.
- Don't understand your answer about space restrictions: "No Space Constraints, No Extra Space or Color Charges No restriction on the length of the papers, number of figures or colors" https://www.mdpi.com/journal/wevj/wevj_flyer.pdf
- You did a PCA to find the explaining features of your dataset. I assume Figure 6 shows the influence of the original features on your PCs? How is the mapping of your x-axis-labels to your original features. (e.g. 1=velocity ...)
Point 6
- OK!
Author Response
Point 3
-Can't check the added references, they are missing in this version.
- All those things you mentioned in the comments could be part of a discussion of the current state of the art.
- I leave that up to the decision of the Editor. IMHO, a journal Article should contain a section where you clearly work out the the present state of the art, discuss the findings of others, and state why your approach is needed to add a contribution to the science community
Response 3
- I’m very sorry that I don’t know why this happened. The pdf file I compiled on the online latex editing website Overleaf incluses all the references. I can clearly remember that I directly packaged the entire latex file and uploaded it, but I did find that all the references in the pdf file downloaded from the submission system are gone. Let me confirm this.
– To comment 2 and 3, thank you very much for your valuable comments, I have added the research significance to the article.
Point 4
-This may be a miss understanding. I'm not that much interested in the specific solution. But from a methodological perspective more information, on how and which data was collected would be necessary here:
- Mentioning, that you use internal sensors and transmit those values over LTE would be enough of explanation, but must be provided. More information on which parameters you gather is mandatory!
- Just some thoughts (list them in a table for increased readability):
- How many vehicles did you observe,
- how many trips,
- how many km in total,
- how many drivers,
- which sensor parameters (GNSS, ACC, GYRO )
- Don't get me wrong, but vehicle parameters and the terrain profile of the route don't match to the heading ("Data acquisition and preprocessing"), if thats not the point you wanna address, use something like "experimental setup", "scenario definition"...
- Figure 4,5: Did not noticed that before: Typo in y-Axis label - must be (km/h).
- If there have been interruptions, just write a short explanation. Also mention shortly which cleaning techniques you applied.
- "The data in our database mainly comes from two cities, Shanghai and Guangdong. The reason for choosing this route is that the route includes relatively complete road features. In the future, as our vehicles run in more cities, the database will be updated in time. At that time, we can use the current method to obtain the typical driving cycle for that city." Why don't you give that explanation in the paper? - It's completely fine!
- One thought on Figure 3: It would be interesting to see, how much the different observations repeat in their patterns. Figure 3 and Figure 4 nearly show the same thing. It may improve the paper if you provide an analysis on that in Figure 3. (optional)
Response 4
- Para1~9: Thank you very much for your guidance. I have deleted the original form that introduced vehicle parameters and replaced the form of information related to the remote signal acquisition platform in this position. From the perspective of the theme of this article, introducing the introduction of the platform is much more important to introducing the introduction of the vehicle parameters. I very much agree with your point of view.
– I have changed the mistake in Fig. 4, thank you very much.
– I have added a paragraph to shortly introduce the content of data preprocessing.
– I have mentioned the significance of our research in the end of abstract and introduction, so I decided not to write about it here.
– In the original paper, the figure 3 shows the original data, and figure 4 shows the preprocessed data ae well as devided into segements, now I have deleted the figure 3.
Point 5
- Sorry I made a typo, I meant 1 km sections.
- Still don't get why 1 km and why not 2 km or 500 m. It looks like a "guess". Thats not a problem, but please report that in the paper. Without explanation, one might ask why exactly 1 km is chosen.
- Don't understand your answer about space restrictions: "No Space Constraints, No Extra Space or Color Charges No restriction on the length of the papers, number of figures or colors" https://www.mdpi.com/journal/wevj/wevj_flyer.pdf
- You did a PCA to find the explaining features of your dataset. I assume Figure 6 shows the influence of the original features on your PCs? How is the mapping of your x-axis-labels to your original features. (e.g. 1=velocity ...)
Response 5
- Thank you for your valuable suggestions. I have indicated in the article that considering the accuracy and complexity of the calculation at Line 122, we choose 1km as a calculation unit, because we have a total of more than 4000 km for statistics.
– In terms of scientific rigor, it is more convincing to add a segment unit according to 2km and according to 500m, and compare their results with 1km. Unfortunately, we did not take this into consideration when we did this work, so we will consider it later when we do our work.
– Thank you very much for your suggestions, I have listed them as a table and put them in the article.
– Thank you for your previous suggestions. I have added an additional list of feature parameters and marked them with serial numbers in the table. The x-axis label in Figure 6 corresponds to the serial number in the table, which I have added a note in the text.