fxstein
Well-Known Member
- First Name
- Oliver
- Joined
- Sep 17, 2024
- Threads
- 9
- Messages
- 107
- Reaction score
- 202
- Location
- California
- Vehicles
- 2023 R1T Quad Adventure
- Thread starter
- #1
Lots of things going on with Rivian vehicles under the cover. Also significant changes happening from firmware to firmware update. Seeing lots of discussions about phantom drain, charge rates, range.
As a data analytics person and former data scientist (and electrical engineer) I like data.
My moto has always been show me the data!
Individual observations or point in time snapshots can often be very misleading. Gut feel can lead to wrong conclusions.
Luckily there is a ton of data being collected by Rivian for their own App, their development and to integrate with 3rd parties.
We all can have access to a lot of that data and apply advanced analytics.
Here is a little initial how to, that explains one way to leverage AI to help you understand what is really going on with your Rivian.
First and foremost you need to collect your Rivian's data from Rivian. Without too much detail the easiest way to do that is to install Home Assistant on a small or tiny computer at home and add the Rivian API integration. Happy to answer specific questions, might even do a little demo first it down the road.
You have seen me post simple screenshots of dashboards and data from it in various threads.
For example the 2024.51.01 update on Phantom drain, SoC and battery capacity. Every single one of these charts allows you to download its data
Once downloaded, name and save that file so you can find it again and even tag it with Rivian or Home Assistant Tags:
Once saved head on over to ChatGPT and upload that file:
I like to tell it what it is that I sent it. ChatGPT is getting really good at reading these random files and a little extra info from you can help analyze that data quickly:
I the rename things (simply tell ChatGPT how you want to name them) for ease of use and simply plot the data
Now the fun starts after some filtering for bad data I let it plot just the core of the data (in this case I also told it to use a second Y axis so SoC % and kWh are both scale independently)
As I know there is something funky going on with SoC since firmware .51 I create a simple calculation to test a theory of mine: how about I use the SoC and capacity numbers Rivian reports to calculate the amount of Stored Energy in the battery pack at every point in time:
And the result:
Look at that. The amount of stored energy is pretty much a straight line (this is a parked vehicle draining away in Shipping Mode). So I would like to create a good average of that drain by asking for:
And the result:
Finally I want to know the formula of that linear regression:
After some further steps and dealing with a mistake (yes - ChatGPT and any AI will make horrendous mistakes at times - in this case its first attempt on the daily energy consumption came back with 25kWh per day! - Most of the time you simply call it our and tell it that that is wrong and it will go back and find its own mistake and fix it) by ChatGPT I end up getting the correct answer:
It’s a wrap:
Now this is just a super simple and primitive example. Skipped a few steps dealing with bad data and incorrect things, but was able to do all of it on my iPhone in less than 15 min.
With this setup I can analyze and compare any behavior, drain, charger sessions within minutes and know exactly what is going on.
Hope that gives some inspirations what is possible with data and AI these days.
As a data analytics person and former data scientist (and electrical engineer) I like data.
My moto has always been show me the data!
Individual observations or point in time snapshots can often be very misleading. Gut feel can lead to wrong conclusions.
Luckily there is a ton of data being collected by Rivian for their own App, their development and to integrate with 3rd parties.
We all can have access to a lot of that data and apply advanced analytics.
Here is a little initial how to, that explains one way to leverage AI to help you understand what is really going on with your Rivian.
First and foremost you need to collect your Rivian's data from Rivian. Without too much detail the easiest way to do that is to install Home Assistant on a small or tiny computer at home and add the Rivian API integration. Happy to answer specific questions, might even do a little demo first it down the road.
You have seen me post simple screenshots of dashboards and data from it in various threads.
For example the 2024.51.01 update on Phantom drain, SoC and battery capacity. Every single one of these charts allows you to download its data
Once downloaded, name and save that file so you can find it again and even tag it with Rivian or Home Assistant Tags:
Once saved head on over to ChatGPT and upload that file:
I like to tell it what it is that I sent it. ChatGPT is getting really good at reading these random files and a little extra info from you can help analyze that data quickly:
I the rename things (simply tell ChatGPT how you want to name them) for ease of use and simply plot the data
Now the fun starts after some filtering for bad data I let it plot just the core of the data (in this case I also told it to use a second Y axis so SoC % and kWh are both scale independently)
As I know there is something funky going on with SoC since firmware .51 I create a simple calculation to test a theory of mine: how about I use the SoC and capacity numbers Rivian reports to calculate the amount of Stored Energy in the battery pack at every point in time:
And the result:
Look at that. The amount of stored energy is pretty much a straight line (this is a parked vehicle draining away in Shipping Mode). So I would like to create a good average of that drain by asking for:
And the result:
Finally I want to know the formula of that linear regression:
After some further steps and dealing with a mistake (yes - ChatGPT and any AI will make horrendous mistakes at times - in this case its first attempt on the daily energy consumption came back with 25kWh per day! - Most of the time you simply call it our and tell it that that is wrong and it will go back and find its own mistake and fix it) by ChatGPT I end up getting the correct answer:
It’s a wrap:
Now this is just a super simple and primitive example. Skipped a few steps dealing with bad data and incorrect things, but was able to do all of it on my iPhone in less than 15 min.
With this setup I can analyze and compare any behavior, drain, charger sessions within minutes and know exactly what is going on.
Hope that gives some inspirations what is possible with data and AI these days.
Sponsored
Last edited: