Analyze your Rivian's data with ChatGPT AI

[agame32]

100%. I'm constantly establishing a history of being decent and polite to AI.

If we don’t carry basic social norms to our increasing AI interactions we run the risk of a future state where AI-natives and others forget to transfer their AI norms to human interactions. Be polite to your AIs and teach your children to do the same.

Also, excellent work, OP!

Sponsored

 

[R1Thor]

OK, I'm now about 90% convinced I need to adopt Home Assistant...

 

[edman007]

In this simple example and in another thread already explained: Since firmware .51 the State of Charge (SoC) % can no longer be trusted in the first 2-4 days of parking a Rivian. It will say you gained charge or the truck did not discharge at all until day 3 when it starts catching up with reality. For real it is discharging at about 1kWh per day, every day when doing nothing but sitting there (even in shipping mode)

That's not what's going on, it's temperature compensated now (I don't think it was before), so now the SoC depends on battery temp, and you can't compare two numbers without the associated same battery temp.

Here is mine, note the change seemed to happen around .47 (I have a bit of a data gap there due to the two factor change)

 

[fxstein]

That's not what's going on, it's temperature compensated now (I don't think it was before), so now the SoC depends on battery temp, and you can't compare two numbers without the associated same battery temp.

Here is mine, note the change seemed to happen around .47 (I have a bit of a data gap there due to the two factor change)

Sorry but it’s not mainly temp related. I explicitly checked for that and took all measurements within a garage at steady state temps. While some temp compensation got introduced in .47, .51 did something entirely new. If you let the car sit at constant temp for days you can see that same behavior.

 

[edman007]

Sorry but it’s not mainly temp related. I explicitly checked for that and took all measurements within a garage at steady state temps. While some temp compensation got introduced in .47, .51 did something entirely new. If you let the car sit at constant temp for days you can see that same behavior.

My checks say it's pack temp, that really low dip on mine is when it was left outside and the pack hit 23°F, the following night it was 15°F, but I charged it so the pack was 55° and that's why it bounced right up to 129kWh or so. Charging heats the battery and will bring the number up. Things like parking a cold vehicle in a warm garage will also warm the pack and raise it, but much slower

It will also drift 1kWh or so on its own sitting, but it did that before the update too.

 

[fxstein]

My checks say it's pack temp, that really low dip on mine is when it was left outside and the pack hit 23°F, the following night it was 15°F, but I charged it so the pack was 55° and that's why it bounced right up to 129kWh or so. Charging heats the battery and will bring the number up. Things like parking a cold vehicle in a warm garage will also warm the pack and raise it, but much slower

It will also drift 1kWh or so on its own sitting, but it did that before the update too.

Not arguing that there is temp compensation going on. BUT that would never be a reason for SoC to climb in the first 24h of parking a Rivian, or for it to not loose any charge on day 2.
I am all for temp compensation, but it is at least done incorrectly when it comes to SoC or intentionally to make it appear there is less vampire drain than in reality.
This analysis simply shows that the car drains at about 1kWh per day even if the SoC number suggests it went up or stayed flat.
And for some of us in warm climates, we simply don't have the temperature differences. I can show you data for this truck parked for 40 days at near constant 61 degrees and the capacity numbers doing wild jumps when .47 rolled in.

 

[edman007]

Not arguing that there is temp compensation going on. BUT that would never be a reason for SoC to climb in the first 24h of parking a Rivian, or for it to not loose any charge on day 2.

I think it's attempting to show the extractable energy at the current pack temp. So if it's cold a full charge might be 120kWh. You park it in a warm garage and 24 hours later is stores 130kWh. So even if it vampire drained 1kWh, you'd only see 129kWh, which would be more miles.

Note, the number I plotted is the SoC, the capacity at full charge. So it doesn't vary due to vampire drain. It's normal for the SoC to go up. The displayed percentage probably shouldn't go up when

And I generally agree, the way they implemented it, it's shows confusing numbers, it's temperature compensating too much, if sitting for a few hours is enough to get an extra kWh of SoC, that's going to show as vampire charging. That's not what's actually happening, and if you took it for a road trip it would have done that anyways. So that extra kWh was always available if you assumed normal discharging. I'm not sure I agree with doing it the way they have it right now. My issue is it changes the "range on a full charge" which isn't true because driving and charging warms the battery, so it's basicallly extra conservative now

 

[Donald Stanfield]

Threads like this make me feel like a massive dumbass.

 

[logan]

Yes fully aware of that integration. Right now it is limited to simple interactions and not advanced data analysis. Over time this could (probably will) change. But given the privacy focus of HA, developers are hesitant to open up your personal HA data store to AI in general.

I think that is only because they give the LLM only a json with a snapshot of current sensors. That's why I was thinking if someone added tools for the LLM to retrieve historical metrics that it could then do more advanced data analysis. I suspect it is more a matter of no one getting around to implementing the tools support.

 

[Jerret]

If we don’t carry basic social norms to our increasing AI interactions we run the risk of a future state where AI-natives and others forget to transfer their AI norms to human interactions. Be polite to your AIs and teach your children to do the same.

Also, excellent work, OP!

Seen a lot of anecdotes that you’ll also get better replies if you’re polite, at least early on in the GPT3 era.

Sponsored