EVE Pilot data is from a study of a mid-sized electric vehicle for a period of 3 years. The dataset includes data points such as:
Speed, distance covered, and other drivetrain data variables;
State of charge, State of health and other battery characteristics; as well as
external temperature variables.
The data can be cited as follows:
Osonuga, Seun; Wurtz, Frederic; Delinchant, Benoit, 2023, “The EVE Pilot: Usage Data from an Electric Car in France”, https://doi.org/10.57745/5O6QIH, Recherche Data Gouv, V2
Please feel free to make comments and questions below in English or French.
N’hésite pas à faire des commentaires et à poser des questions ci-dessous en français ou en anglais.
After processing and analysing the Voltage and Current of Battery Data of the whole battery pack for Power and then Energy calculation for each cycle, I noticed that there’re 2 points to take note for the dataset:
The missing data after joining HV LBC Current and HV LBC Voltage with 1-second aggregation apeears in between recordings of Voltage and Current
Moreover, the detailed SoC in charging periods was not fully recorded and can vaguely recognized with starting point and ending point of SoC. However, the ending point was not the actual time the SoC reached it maximum charge because the driver can unplug the charger later the battery was fully charged. This caused the multiple times the voltage dropped to zero after naturally losing power and fully charged again, which interrupts the real charging period.
The solution for this problem is detecting the first time voltage being dropped to 0 and replace the ending time correspondingly.
For comment 1, thanks this is well noted. As the data was from a logging device, only one variable is collected at a certain moment. We thought it best to leave it up to the user to address the forward filling
And for comment 2, yes effectively SoC is not collected at all during charging. The data that best corresponds to SoC during charging is “Available Discharge Energy”. However this variable starts to have some weird values when the battery voltage and current start going to zero close to the very end of the charge cycle as shown below.
Drop of the SOH of the battery after a BMS reprogramming:
For our information, a reprogramming of the BMS of the battery was made by the manufacturer of the car the 26th of april 2021. Before this reprogramming, the SOH decreased under 75 %. After the programming the SOH restarted at 98%
This figure gives the evolution of the SOH from oktober 20 to march 25:
Thank you for your reply. I think the way data was collected has no issues, I think I can manage it to deal with missing data.
In terms of the “Available Discharge Energy” dataset, I suppose the weird voltage values dropping to zeros are exactly the phenomenon I described in the second section. In particular, I think after being fully charged, the voltage was dropped, leaving for natural energy loss. After reaching a certain threshold, the voltage was raised again to charge the battery again. As the user did not always unplug the charger after full charge, this pattern can occur multiple times and cause many voltage drops as you can see at the end of charging periods.
Thank you for the informative reply! It turns out to be a system change that causes the increase in SoH.
I conducted small analysis around this phenomenon on 2 sub-sets divided for before and after the Gap of SoH change. Using the ratio of Energy over DoD, I witness the lower SoC gained after activating cells than before for the same amount of Energy in both charge or discharge periods. The lower SoC gain for the same Energy results in the lower ratio of Energy/DoD, which makes me think about the case when the backup parts of the battery might be turned on. Therefore, percentage curve lower than 75% before this Gap might supposedly be divided by pure 22kWh of the battery instead of the full capacity at that time which was lower than 22kWh with lower cells activated.
Another point I noticed is that all the data was missing from August 2022 to October 2023 apart from SoH estimation. I downloaded the Current, Voltage, SoC, … from the website and I think this missing part was not fully uploaded.
Thank you for the replay: can you remember what DoD means ?
As far as I know, when the battery starts to be used for this car, the useable capacity of the batterie is 22 kWh, but the real capacity of the battery is higher: at least 24 kWh (may be 26 kWh, I am not sure, I try to find this real value). So if I understand you well, when the BMS has been reprogrammed, it results by using a higher useable capacity (in other words, use of the reserve of capacity): is that correct ? If yes, are you able to evaluate what is the new value of allowed capacity used ?
DoD in the context means depth of discharge/charge, or Delta SoC, the SoC level difference from the start to end of the cycle. The ratio I used is to check the how much energy was charged or discharged into the battery for each unit of SoC in percentage.
Definitely that was what I meant. As you can see in the first graph, for the same 1% of SoC, it takes more energy to charge after the Gap than before. This means it’s very likely that the capacity was added to the battery after 26th of April 2021.
To quantify this added amount. I implemented SoH estimation using linear Intepolation on Charge and Discharge cycles. The SoH calculated in kWh showed that initially Battery’s capacity was around 16kWh, then it increased to nearly 20kWh. Our assumption is that all 192 cells of the battery have not been fully used. How do you think about his analysis?
I am working on a topic that develop a hybird-AI model for estimation of battery degradation in this EV context. However, the data is quite limited with only nearly 2-year duration, while the first year suffered from reprogramming which causes a non-seamless SoH curve.
Having a wider duration of measurement will significantly facilitate my work because covering more of battery’s lifespan data will make the degradation curve clearer. Therefore, it will be highly appreciated from my end to have access to further years of the data.
I am looking forward to hearing your response. Thank you for your support!
Kind regards,
Dat
