Calibration type: PID?

[SwedishChef]

Hi,
I just started to use Better Thermostat because it could solve a lot issues with my Homematic mixed setup using old (non-IP) and new (IP) devices that can't be connected in RaspberryMatic.

As far as I understood the way how the target room temperature is tried to reached, I found on https://better-thermostat.org/calibration that the following formula is used:
trv_target_temp = target_set_temp - external_sensor_temp + internal_trv_temp

For me this seems to be a simple proportional controller. Therefor the target temperature will never be reached, what I observe now with my setup:
target temp: 22.5 °C
measured room temperature 23,5 °C
TRV target temp 24,5
Event though the room is already too warm the TRV is still increasing instead of decreasing the target temperature!

Shouldn't a Proportional-Integral(-Derivative) (PID) controller be used to reach the room target temperature as exact as possible?

Edit:
Before starting this topic I only searched Discussions for 'PID'. But now I found some PID related topics in Issues. But I still does not know if there is already a PID controller implemented or not. At least for me (as my charts show) it seems to be not.

[aberridg]

AFAIK it isn't. I am in a very similar position to you. I have used PID for fuel injection control and think it is ideally suited to this project. Another advantage is that a PID controller doesn't need to know anything about the actual state of the radiator valve. It can accurately hit a set point (provided it's tuned correctly) very quickly. Tuning is the challenge as one needs to tune each part of the PID algorithm. I was considering experimenting with PID control, once I understand the code. I am not (yet) a contributor to the project.

Edit: there is a decent PID algorithm written in Python that's already available. You just need to update it with current vs target each time either changes, and it outputs something that can be used to control a system.

[wtom]

Yes, i wanted to go that way as well. The problem will be the calibration as you said. And that not very easy. For example in my house, i have strong heater and weaker ones. Different room sizes. Some that get a lot of wind from outside, some having sun if the sun is shining.

Our goal is for the next heating season, so next winter. Having an AI ready, that learns itself from the behavior of the heating, so to say.

For this heating season, the main goal is to have a stable very that can be a non beta release.

[aberridg]

Haha... FYI I have implemented a very simple PID controller in my own system with BT. I am now experimenting with tuning. The biggest issue I've seen so far (without doing any tuning) is wide swings for the set point of the real TRV. It's trying to fix the error very quickly. However, I have temporarily fixed it by restricting the minimum/maximum values for the output of the PID algorithm to BT set temperature +/- 10 degrees C.

To my mind, PID is much simpler than calculating heating power, etc.. although perhaps I don't yet understand enough about heating power.

[SwedishChef]

@aberridg: could you show me (us) what part of code you changed to add PID controller. Many thanks!

[aberridg]

Sure - I've forked and created a new branch:

https://github.com/aberridg/better_thermostat/tree/pid_experiments

My current challenge is understanding the control flow of the BT code. I really want to call the PID controller far more frequently than I currently do, but I'm not sure how to achieve that. Right now I'm only calling it when anything changes. Maybe that's ok, but I'm not sure. PID controllers are designed to be called much more frequently as they're time based. I'm also tuning the PID controller right now - just the P term first.

Since I don't yet understand the control flow of the BT code, I have hacked in the PID controller in the easiest way possible just to test the theory. I think the control_queue logic in controlling.py is a bit flawed. Surely a queue should hold the "operations" that need to be performed, or the events that have been triggered, rather than the main climate (self) object that then decides what to do?

I've tried modifying the control_queue function to call the PID controller frequently, but almost any change to that function seems to break it - even logging. This is another reason why I suspect it's not implement properly.

[aberridg]

I've figured out how to update the PID controller with a set period, and integrated it all a bit better. Will commit to the same repo within the next day or so.

[wtom]

Anything new regarding the PID controller mode?

[aberridg]

Yes, exactly! I am in the same situation. I don't mind if things aren't perfect right now. Even now, BT is way better than what I had before. I am using it in a bedroom where the TRV is almost touching the radiator, and the radiator is underneath a window in a poorly insulated room. Before BT, the room would cycle rapidly between hot and cold. The TRV would be off, then detect the temperature drop, open, and detect too high a temperature almost immediately, and shut down again, and repeat the cycle! Now, the temperature is much more stable and comfortable.

…

On Thu, 22 Dec 2022, 19:22 Wolfgang Tom, ***@***.***> wrote: Yes, i wanted to go that way as well. The problem will be the calibration as you said. And that not very easy. For example in my house, i have strong heater and weaker ones. Different room sizes. Some that get a lot of wind from outside, some having sun if the sun is shining. Our goal is for the next heating season, so next winter. Having an AI ready, that learns itself from the behavior of the heating, so to say. For this heating season, the main goal is to have a stable very that can be a non beta release. — Reply to this email directly, view it on GitHub <#806 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHLJITILHPZF4LH4WJMOHHLWOSTAHANCNFSM6AAAAAATGXTAAM> . You are receiving this because you commented.Message ID: <KartoffelToby/better_thermostat/repo-discussions/806/comments/4478969@ github.com>

[ovizii]

You guys might want to have a look at this project, you might be able to get some inspiration from there: https://github.com/ScratMan/HASmartThermostat

[DawidMT]

Also this one looks pretty interesting:
https://github.com/dadge/awesome_thermostat

[aberridg]

Sorry guys! In the middle of a building project at the moment. I'll commit my latest changes to my fork and let you know. I had to move away from using it for a bit due to various reasons, but I have further things to try with it. Andrew

…

On Thu, 12 Jan 2023 at 09:59, ovizii ***@***.***> wrote: You guys might want to have a look at this project, you might be able to get some inspiration from there: https://github.com/ScratMan/HASmartThermostat — Reply to this email directly, view it on GitHub <#806 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHLJITP5ZKUJLYKHJU46C5LWR7IXFANCNFSM6AAAAAATGXTAAM> . You are receiving this because you were mentioned.Message ID: <KartoffelToby/better_thermostat/repo-discussions/806/comments/4665882@ github.com>

-- *Andrew Berridge - *TIBCO Data Science | *M*: +44 77 89764017 | *E:* ***@***.*** | www.tibco.com