Auto TPI behaviour investigation #1739
Replies: 28 comments
|
Hello, I let @KipK (our Auto-TPI developer) answer your legit question. |
|
Hi, Here is why it behaves like this with valves: #1704 |
|
I see that the stars are aligned in the analysis of Vtherm's behavior :D I will have a look closely to the post and your reco. Thanks ! |
then maybe @eflye-veepeetech should avoid auto-TPI and set the coefs manually. What do you think ? |
|
@jmcollin78 II think if the min/max opening valve is set correctly, it should work "ok" , at least should find "working" coeff, but still, the TPI algo has flaws on such situation as the curve is still not linear. @caiusseverus made a tremendous work, it explains a lot of issues. For now I would advice to use the internal TVRZB regulation instead of TPI as it was enginered with the current hardware in mind. |
|
Very clear, i'll do my homework and read closely @caiusseverus post. maybe a clarification within the documentation could be helpful from both of you regarding TPI / AutoTPI regarding valves integration on Vhterm? |
|
However @eflye-veepeetech I'd like you to test auto-tpi after you have well configured the min/max opening and post the graph here to compare. |
|
other thing , if your boiler adjust water temperature with external T° sensor, then it's useless. |
For now no, i don't have any boiler adjust water temperature with external T° sensor. It could change in the future (i have a Frisquet boiler, there's active project around it especialy on hacf forum, with LoRA module and use external sensor on it) |
What do you think about @KipK ? |
|
I don't know it depends of your setup. Follows caius advice on how to set them:
For the max valve position, I guess you have to check when the heating curve get's linear. Means there's no power increase anymore. Or use an infrared thermometer or thermal cam to mesure heater t°. When t° doesn't increase anymore significantly means you've reached the max. You can also try caius blueprint he posted, but follow his intructions. Seems a better way. |
|
And are you sure the min max value you've puit are ok ? They need to be exactly the effective range. It looks like you've just choosen randomly 10-30 ain't it ? Also read the auto TPI doc as it seems you haven't follow at all the guidelines for a correct auto tpi learning session. I see no setpoint changes to gives Kint learning window. It's just always the same setpoint for days. Looking at the regulation anyway it seems the value it choosed are still ok as the temp curve seems to stick to the set point quiet good, so for me your power range is still not adjusted to your valve, but it still work with those low Kint |
|
ok i'll try to do a more accurate test than this one :) |
|
What about the heat rate ? I see 0.16 , is it real ? That's not a heater, that's a candle no ? :P |
|
What does it gives you when you use the service call calculate_capacity ? |
|
I will focus only on "Chambre parents" and try to get
climate.chambre_parents_vthermostat:
success: true
observed_capacity: 3.15
kext_compensation: 0.108
avg_delta_t: 10.8
samples_used: 11
samples_before_filter: 11
outliers_removed: 0
reliability: 40
min_power_threshold: 0.95
period: 10.7
max_capacity: 3.258 |
|
yes so it completely screwed up during the learning. :) However, AutoTPI is not advised for Valve system as there's too much parameters and knowledge that enter in the game. Much simpler and I'm pretty sure it works better. |
So in your opinion, TRZB + control of valve / TPI would be worst than using VT + TRVZ internal climate with or without self regulation ? ... you make me so sad... ^^ I have to admit that i rush on the control of valve when @jmcollin78 implemented it, i (wrongly?) imagined that VT with this feature would be by default better than embedded algo. I could go wrong, i will do the test ! |
|
the thing is the TRZB manufacturer knows the hardware valve specs embeded and their regulation is supposed to be engineered for that. If you manage to tweak your setup to have a linear power curve shape then it should be ok. But as you seen in Caius post, it's not that easy. Also you should add an automation to stop auto-tpi when the boiler is off, and start it few minutes after it started to heat ( when the water is supposed to be at correct temperature ). If water doesn't have always the same temp while auto tpi is running it's a waste of time, as power is never the same at same % |
How tight do you want the temperature regulation to be? I found the internal logic to at least be acceptable, providing you are sending the the temperature to the valve from an external sensor. I saw about 0.2-0.3º variation with the internal on/off control. If you set up my blueprint and determine the minimum valve opening accurately (manually adjust the valve to open 1% at a time until it starts to get warm), and use something like 25% and 85% for the valve opening and flow settings you should get much better response. You can still use the tpi algorithm, but I would follow KipK's advice about turning auto-learn on and off with the boiler. Alternatively, use manually set coefficients - start with internal at 0.5, and 0 for external. If you don't get overshooting, then gradually add some external coefficient - start at something like 0.002 and add 0.001 at a time until you get the room temperature settle at the set point. If you get oscillation reduce k_int a bit and if you get constant overshoot reduce k_ext slightly.
The thing is the tpi algorithm expects a linear response to its output. Radiator valves (usually) aren't linear at all, so if you can correct for that you should get better results. |
|
Yes that’s quite a big swing. It will vary from system to system, depending on things like room size compared to radiator size, how hot the radiators get etc. The TRVZB is getting proportional control in a firmware update but it’s not out for zigbee2mqtt until the end of march unfortunately. |
where is the source on this? both the firmware update and the z2m compatibility ? |
There is also at least one additional unknown for each valve, namely the setting of the valve bodies with integrated presetting which can be individually set up differently depending where radiator is in the water network depending of the pressure.
I'm also very curious about this info, you are well connected with Sonoff ;) |
|
They also add PID control on trvzb new fw. Edit: Here is an untested external converter made from the zigbee herdsman converter PR so you can try : Ota fw is here: https://github.com/CubeZ2mDeveloper/zigbee-OTA/raw/ff2899832662f27163663f77acd1f0fe190a32f6/images/Sonoff/trvzb_v1.4.4.ota |
|
OTA updates are starting to roll out |
|
Hello, I turn this into a discussion, because it is ... a discussion and not an issue as far I understand |





Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Hello everyone,
I applied AutoTPI on all my Vhterm when the feature was deployed, probably a bit mad but this is a very valuable feature in my opinion.
i would to share here one behaviour that i have difficulty to understand
First of all find below current Vtherm and auto_tpi conf for this room:
sensor.salon_vthermostat_auto_tpi_learning_state:
climate.salon_vthermostat:
#version
You will find below AutoTPI graph

As you may see, my graph started after VTherm last update, I did a TPI reset before starting back auto TPI. TPint was at 0.5 initialy and TPext at 0.05.
I don't understand why TPI fall very agressive on boot and after is very very slow to increase back, creating a situation of having hours of being below temperature setpoint because opening is very very small.
Do i miss something regarding config ? I'm very open to test any change on my config :)
Précision: overheat period here is because of a wood stove within the living room.
EDIT: Jan 29th at 16:00, notif received:
Etienne
All reactions