Can anyone explain this temperature profile to me?

Feedback & Discussion
1

This is way different than it was a few weeks ago. I have no idea what is going on here.

IMG_1750
IMG_17502868Ă—1320 281 KB

2

Lol looks like beestat had an aneurysm.

It will probably clean up as time goes on. Here’s a more normal example:

image
image700Ă—316 21.7 KB

Basically the curved lines are the “real data” where beestat samples the effectiveness of your heating, and the straight lines are just a “best fit” to make it easier to see what’s going on.

In your case, the curved lines are really noisy. Beestat does it’s best to remove that noise, but when everything is so excessively outside the norm, that becomes the norm. :slight_smile:

I would expect over time things will average out a little better, but you could also see a lot of this if you run a lot of other heat sources like space heaters, or maybe it’s a work area and you open a window for a cool breeze even if the heat is on, etc.

3

[quote=“Jon Z, post:2, topic:1375, username:ziebelje”]
Lol looks like beestat had an aneurysm.
[/quote] :joy:

With a possible +1.0F/h on Heat 1 @ 6F, why are you even using Heat 2, Aux 1, & Aux 2?

Your staging looks rather odd to me. Heat 1 off below 22? Heat 2 off below 25?

4

Thanks for the input. Those are not settings I have set myself.

I have a brand new BOSCH air-source heat pump paired with a new BOSCH gas furnace. The BOVD-36HDN1-M20 and the BGH96M080B3B. Ecobee interprets this as a two-stage heatpump (heat 1 and heat 2) and two-stage gas (aux 1 and aux 2).

Living in Denver, gas is actually cheaper to heat with per btu than electricity. Staging is set so that it switches to gas at lower temperatures due to cost.

1 Like
5

Heat 1 and Resist look normal to me; the others are unusual but if you look closely you’ll see that you have very little actual data for them, meaning beestat is fitting a line through a bunch of noise and comes out with nonsense, which is not entirely unexpected

Adding confidence intervals or quality of fit metrics would be interesting (at least to me), although it would likely make the graph even more confusing to read. Maybe hiding profiles with a quality of fit below a certain threshold would be a better approach?