Help with understanding Range Loss over time

[jonwitt]

Hi BMW EV and PHEV Community,

I’m Jon, a PhD candidate in Chemical Engineering specializing in electrochemistry at the University of Washington. I’m working with folks at a new company in Seattle, Recurrent, with a goal of helping electric vehicle (EV) buyers understand the battery health and range of used electric cars. We are developing machine learning algorithms to predict future battery life and range across all major EV models.

Here’s where I need your help. We’re assembling a group of current EV drivers/enthusiasts who are willing to share anonymous access to their battery range and state of charge data over time in a variety of weather and driving conditions. At the moment, over 2500 EV drivers and enthusiasts have signed up! While we have plenty of Tesla, Chevy and Nissan data, we need more BMWi3, i8, 5-series plug-in, 3-series plug-in, and X5 plug-in volunteers to share their vehicle data. To participate, you’ll need an active subscription to BMW Connected Drive (Connected Package or better) and the car must be a 2014 or later model for BMWi3, and 2017 or later for all the other BMWs mentioned above. It’s automatic, no work on your end.

Here’s the signup link and FAQ. A nice added bonus: you’ll get a free monthly battery analytics report as we build up our comparative data.

So if you drive an EV and really want everyone else to drive an EV in the future, please consider volunteering your data. Let me know if I can answer any questions about the program or the company.

Jon Witt
Linked In

 

[alohart]

I signed up our 2014 BEV hoping to learn how our actual range loss compares with other 2014-2016 i3's. I suspect that actual range loss for 60 Ah i3's could be due to battery cell degradation and/or the battery management system reducing the battery pack's usable capacity. I don't know how such a survey might distinguish between these two explanations, but maybe there's a way.

 

[Anonymous]

Art, how did you sign up? When I do it, I get a "No Compatible Vehicle Found" message, and reading their FAQ for that, it says that only 2017 and later are eligible for the study.

 

[Pahaska]

I signed up my 2019 and it went very quickly and easily. Just a few questions and input my username and password to allow them to link to my account.

 

[jonwitt]

I signed up our 2014 BEV hoping to learn how our actual range loss compares with other 2014-2016 i3's. I suspect that actual range loss for 60 Ah i3's could be due to battery cell degradation and/or the battery management system reducing the battery pack's usable capacity. I don't know how such a survey might distinguish between these two explanations, but maybe there's a way.

Thanks for signing up Art! I'm excited to hear what you think about your first report and I hope we can help with figuring out the difference between the two modes of range loss that you mention. What we suspect is that if anything, the i3 is likely opening up your usable capacity window as the car ages, but it is quite a slow process. At the same time, very real degradation of the battery is occuring. As the study grows, and more BMW's join, we will get a clearer picture over time. I really hope it will be helpful to all EV owners and future ones too! If you know of any others who may want to join then feel free to send them our way!

Don't hesitate to ask any other questions. When you get your first report, there will be an email attached to it that you can provide feedback to.

Best,
Jon Witt

 

[jonwitt]

Art, how did you sign up? When I do it, I get a "No Compatible Vehicle Found" message, and reading their FAQ for that, it says that only 2017 and later are eligible for the study.

Hi Mark,
Sorry about that. I should have mentioned the model year requirement in the original post. I edited it to reflect that. Thanks for the heads up. We are looking into ways to include folks that don't match the criteria for remote connection so keep an eye out for that!

Best,
Jon Witt

 

[Anonymous]

Hi Jon. Thanks for the info. Was just wondering how Art (above), who has a 2014 i3 managed to sign up?

Would definitely think that looking at EVs older than 2017 would help in understanding battery degradation. Hopefully us 'older' owners will eventually be included.

 

[alohart]

Was just wondering how Art (above), who has a 2014 i3 managed to sign up?

Would definitely think that looking at EVs older than 2017 would help in understanding battery degradation. Hopefully us 'older' owners will eventually be included.

I just ignored the 2017 and later "requirement" to see whether our 2014 would be accepted. I didn't think excluding earlier i3's made any sense, and apparently the Website agreed

 

[Anonymous]

I just ignored the 2017 and later "requirement" to see whether our 2014 would be accepted. I didn't think excluding earlier i3's made any sense, and apparently the Website agreed

You lucked out!

They must have changed their website software after you signed up. When you add in your connected drive info and it does the connection verification, it errors out with a "no compatible vehicle found" now if your car is older than 2017.

I agree, excluding earlier i3's makes absolutely no sense :ugeek: .

 

[CanisLupus]

I personally won't give my username and password out . I don't mean to say that they aren't legitimate but with systems getting hacked continuously I would not want access to my vehicle to get into the wrong hands .

 

[skeptic]

While I will not be participating, I do have a guess as to why they want 2017 and newer. If they want to track degradation over time they probably want to know when battery degradation happens. If the car is 4 years old and it's 20% degraded, did that happen in the first year or two or accelerate it's demise over the last 12 months... Just making up a completely fictitious example of course, but I can see where older cars are less useful to this study.

Then again, the entire study seems a bit pointless to me. Battery chemistry is changing rapidly and companies are spending hundreds of millions (billions??) in research looking for the next big breakthrough. Solid state lithium batteries appear to be less than 5 years from being in production cars. Tesla has their new higher power density, longer lasting, faster charging battery design going into cars in a year or two (China first for whatever reason). GM announced their new Ultium batteries earlier this year. Toyota is promising a prototype w/solid state batteries in 2021 with some serious ICE killing specs (if you believe them). Nissan says by 2028 they will have their own solid state batteries in their production EVs. Everyone is doing something... By the time this study has enough useful data to to model and predict battery life, it will be obsolete.

 

[alohart]

Then again, the entire study seems a bit pointless to me.

The stated point is to provide battery pack health information for purchasers of used EV's, kind of like a CarFax report but for the battery pack. Therefore, the age of a used EV wouldn't matter.

 

[skeptic]

Then again, the entire study seems a bit pointless to me.

The stated point is to provide battery pack health information for purchasers of used EV's, kind of like a CarFax report but for the battery pack. Therefore, the age of a used EV wouldn't matter.

That's part of the stated purpose, it's the next bit I was referring to: "We are developing machine learning algorithms to predict future battery life and range across all major EV models."

If this is a decades long study then maybe once battery tech has stabilized. For now, battery tech is changing too rapidly for a study like this to have much real world value. By the time you have collected enough data (I'm guessing minimum 5 years worth of same year same model same battery size, but probably closer to 10 to be truly useful), that data will not be applicable to new vehicles. Use a bit of guestimation and machine AI to loosen up the restrictions a bit and you still need several years of data and by the time you have your 5+ years of data the battery tech has changed. Chemistry, density, cooling systems, charging options, software...

I get what they want to do and I agree that it would be useful if you were used car shopping and wanted to know projected battery life. I just think it's too early, stuff changing too quick to be of long term use. If you believe what car manufacturers and companies like QuantumScape are saying, we are somewhere between 4 and 10 years from current EV battery tech (and this data for them) being obsolete.

 

[Anonymous]

I think the "fuzziness" of all this comes from the stated purpose not really being the stated purpose. The purpose is in the first sentence of the request:

I’m Jon, a PhD candidate in Chemical Engineering specializing in electrochemistry at the University of Washington.

This is a graduate project, with the goal of earning a PhD. The small data set, 2017-2020 is likely because this is a proof of concept to be defended as part of a PhD dissertation.

To actually be 'real world' this kind of AI system would really have to be manufacturer-supported, with every EV automatically sending battery data in for long term tracking and analysis of battery 'health'.

 

[jonwitt]

I just ignored the 2017 and later "requirement" to see whether our 2014 would be accepted. I didn't think excluding earlier i3's made any sense, and apparently the Website agreed

You lucked out!

They must have changed their website software after you signed up. When you add in your connected drive info and it does the connection verification, it errors out with a "no compatible vehicle found" now if your car is older than 2017.

I agree, excluding earlier i3's makes absolutely no sense :ugeek: .

MKH and Art,
I think I have figured out what is going on here. Long story short, I made a mistake for specifically BMW i3s. BMWi3 should be compatible for 2014 and later models (2015 and on). It is the rest of the BMW models that don't have the right hardware and software to be accessed by our partner SmartCar unless they are year 2017 or later (2018 and on). Here is the rest of the explanation.

OEMs can only see data for their car brands, and no one elses. Given that we are not an OEM, we had to find a service that allowed us to access certain data points from any given make and model. The closest thing we found was SmartCar, a company with an API that can connect to a variaty of vehicles. I consulted with them today about the 2014 BMW i3 delimma and it turns out, we made a mistake for BMWi3's in our FAQ. 2014 BMWi3 and later models (2015 and on) are compatible with SmartCars software.

We are limited by what SmartCar can access. More technically speaking the necessary hardware telematics need to be installed in the car. Often times this limits what years of certain brands can be included in the study. We wish we could include all EV and PHEVs, but that isn't possible at the moment. We do include most of the most popular ones at the moment though. The second requirement is that the remote telematics is enabled by having a subscription to your OEMs remote service. Basically, the hardware and enabled software package needs to be present in order for you vehicle to connect. If you are having trouble connecting with your 2014 BMW i3, send an email to [email protected]. I can connect you with our trouble shooting partners at SmartCar via that email address. Building our own API to access vehicle data would have really slowed us down, so it was convenient and needed that SmartCar exist for this endeavor.

Sorry about the confusion and the mistake! I hope this explanation helps.

Best,
Jon Witt

 

[jonwitt]

While I will not be participating, I do have a guess as to why they want 2017 and newer. If they want to track degradation over time they probably want to know when battery degradation happens. If the car is 4 years old and it's 20% degraded, did that happen in the first year or two or accelerate it's demise over the last 12 months... Just making up a completely fictitious example of course, but I can see where older cars are less useful to this study.

Then again, the entire study seems a bit pointless to me. Battery chemistry is changing rapidly and companies are spending hundreds of millions (billions??) in research looking for the next big breakthrough. Solid state lithium batteries appear to be less than 5 years from being in production cars. Tesla has their new higher power density, longer lasting, faster charging battery design going into cars in a year or two (China first for whatever reason). GM announced their new Ultium batteries earlier this year. Toyota is promising a prototype w/solid state batteries in 2021 with some serious ICE killing specs (if you believe them). Nissan says by 2028 they will have their own solid state batteries in their production EVs. Everyone is doing something... By the time this study has enough useful data to to model and predict battery life, it will be obsolete.

Hi,
Sorry about the confusion. I made a mistake in my post and we had a documentation error regarding BMWi3s on our FAQ. Turns out we can accept 2014 and later BMWi3s (2014 to present). The short explanation is that we are limited by the telematics that car manufactuers put into their own vehicles. For most BMWs we can only access 2017 and later models. For BMWi3, turns out we can access 2014 and later. We do want as many vehicles as possible, currently we have 2500. Older cars are quite useful for our study because they tell us where newer cars may be going, if there hadn't been much changes in the future model year vehicles. Gathering snap shots of vehicles at different ages helps us pin down their calendar aging.

Once those solid state batteries come out, we will be really excited to ask them to join the research fleet. While those vehicles are making their rounds among new car buyers, there will still be plenty of used vehicles being sold. We hope we can help those used EV buyers out since those cars should be cheaper and more accessible to lots and lots of folks, thus accelerating EV adoption. Only time will tell I suppose.

I appreciate your critical eye, and such feedback is taken seriously and always examined to see how we can improve. I'm hoping that we can scale up fast enough to be useful. The study is housed within Recurrent Auto, a start up in Seattle, that has academic and national lab workers and support working on it (I'm one of them). So far we are making an impact at a few used car dealerships so far. For example, you can see a few reports on nearly all the Teslas of this dealerships website: https://www.idrive1.net/used-cars-carrollton-tx. We are working on useful BMW ones too for used car buyers. We also produce reports personalized for folks who join the fleet.

Hope this offers some clarification and let me know if you have any other questions.

Thanks,
Jon Witt

 

[jonwitt]

I think the "fuzziness" of all this comes from the stated purpose not really being the stated purpose. The purpose is in the first sentence of the request:

I’m Jon, a PhD candidate in Chemical Engineering specializing in electrochemistry at the University of Washington.

This is a graduate project, with the goal of earning a PhD. The small data set, 2017-2020 is likely because this is a proof of concept to be defended as part of a PhD dissertation.

To actually be 'real world' this kind of AI system would really have to be manufacturer-supported, with every EV automatically sending battery data in for long term tracking and analysis of battery 'health'.

Hi MKH,
We hope that one day manufacturers will join us! I also think that there is some value to third party validation as well. The study is a partnership between a start up, academic workers (like myself), and national lab workers. This personally won't go into my PhD dissertation. We will need more BMWi3s, but we have made lots of headway with Leafs, Teslas, Bolts and Volts. Examples for Teslas can be seen at https://www.idrive1.net/used-cars-carrollton-tx. The models used to make the reports for their used Teslas came from eager Tesla enthusiasts.

This one is a specific example: https://www.idrive1.net/vehicle-details/used-2015-tesla-model-s-85d-carrollton-tx-id-39263363 , just click on the Recurrent rating.

We are really exploring this question of how much data is needed to be useful? How much, what quality, and quantity of data is needed to answer certain questions?

I hope I was able to reduce some of the fuzziness. Sorry about that. Let me know if you have any other questions.

 

[jonwitt]

I personally won't give my username and password out . I don't mean to say that they aren't legitimate but with systems getting hacked continuously I would not want access to my vehicle to get into the wrong hands .

I definitely understand your concern around this.

Recurrent may have an opportunity later in the 2021 to participate manually in the research where you don’t need a subscription. I'm happy to keep this forum in the loop as well as post cool updates.

Best,
Jon Witt

 

[Anonymous]

Sorry about the confusion. I made a mistake in my post and we had a documentation error regarding BMWi3s on our FAQ. Turns out we can accept 2014 and later BMWi3s (2014 to present).

Thanks for the updated and more complete info you provided - interesting stuff. However, your sign-in software still wont accept any BMW i3 earlier than 2017.

 

[3pete]

To participate, you’ll need an active subscription to BMW Connected Drive

Does Connected Drive actually allow you to view battery capacity/ state of health directly? Do the other brands APIs allow this? I see a lot of references to range as the metric that is collected, not an actual State Of Health number in % or kWh.

What I'm getting at is: how do you differentiate someone with a lead-foot vs. a hyper-miler? If it's all based on range and SOC %, the driving style is going to skew the estimate one way or the other. Is that just left up to the machine learning to solve?

Interesting project either way!