Page 2 of 7

Re: Advanced algorithms (PID and Prediction) computation

Posted: 30 Dec 2015, 08:03
by Dr.Fix
It seems like PID is not considering at all outside temperature...

For having 20.5 degrees at 17:30 it started about at 14:30 last week:
23-dec.png
23-dec.png (15.94 KiB) Viewed 689 times
This week the outside temperature dropped a lot, and I expected the thermostat to automatically anticipate further IN REAL TIME, since it KNOWS (or should know) the outside temperature via local weather data. But it doesn't do it:
29-dec.png
29-dec.png (20.58 KiB) Viewed 689 times
Where is the smartness of the whole thing? I expected the anticipation to take in consideration the global situation and adapt QUICKLY.
I know that it will "learn", but what if next week temperatures raise again: it will relay on this week low ones and turn-on too much early?

Re: Advanced algorithms (PID and Prediction) computation

Posted: 30 Dec 2015, 19:00
by JanMetDePet
I have had a bunch of "self learning thermostats" and they ALL produced what you are describing: they need time (expirience) to learn. If the PID is adjusted to soon you will get chaos.

Re: Advanced algorithms (PID and Prediction) computation

Posted: 31 Dec 2015, 16:23
by cmugur
@Dr. Fix

Brieuc stated in the first post of this topic:
Brieuc_Netatmo wrote:
NB: the PID is automatically disabled for all setpoints lower or equal to 14°C. For these setpoints, the maximum duration of Anticipation is also modified: it is about 3 hours maximum for orders above 14°C, and up to 8 hours for lower or equal to 14°C setpoints.

Re: Advanced algorithms (PID and Prediction) computation

Posted: 11 Jan 2016, 15:43
by rnucci
i don't get the point of limiting to 3 hours the max duration of anticipation...
can Brieuc explain why?

Re: Advanced algorithms (PID and Prediction) computation

Posted: 12 Jan 2016, 10:32
by TypeC
rnucci wrote:i don't get the point of limiting to 3 hours the max duration of anticipation...
can Brieuc explain why?
I'm not Brieuc (clearly) but at the risk of adding speculation rather than fact to the discussion, I think it was because some people's homes were taking ages to warm up and the heating was coming on in the middle of the night. I tried the anticipation feature for about a week when it was very cold last year, but our bedroom (which, with the door shut, heats up a lot quicker than the rest of the house) was getting uncomfortably warm at about 3:30am! I switched it off after that.

A setting rather than a hard limit might be better... but then some people want different things at different times of the day (for example: no heating at 3am, live with the cooler house in the morning; definitely want a warm house returning home at 5pm) so it's quite difficult to fit everyone's needs. A killer feature for me would be a script interpreter on the server so you could write a little bit of JavaScript (or whatever) to attach to a schedule and provide whatever custom mechanisms you could dream up for whatever time of the day or week required it.

Re: Advanced algorithms (PID and Prediction) computation

Posted: 12 Jan 2016, 11:23
by Dr.Fix
The solution would be: for each scheduled period of time set the max pre-heating time.
So you could set not to pre-heat before the morning schedule but to pre-heat max 3-4 hours before evening one.

Since I needee more than 3 hours for getting house warm, I created another interval before the evening one with an intermediate temperature of about 1 hour, so the pre-heating starts 4 hours before the time I need.

Re: Advanced algorithms (PID and Heating Prediction) computa

Posted: 21 Jan 2016, 09:58
by marks14
I too am a bit unhappy with the predictive aspect of the thermostat which now wants to start heating my house about 4 hours before I get up in the morning.

I installed the thermometer when our weather was quite mild - around 10 to 15 degrees. Now we are down to 0 temperatures at night. That is why I asked a question (so far unanswered) about deleting the data. I assume that if I do this then the thermostat will "re-learn" that it requires less time to heat up - is that right?

Incidentally, since I don't want my boiler to wake me up at 4 am in the morning, I have taken the advice elsewhere on the forum and set the original controller switch on my boiler to switch the boiler off at night until 6 am. At least this way I can force the boiler off in the middle of the night.

Re: Advanced algorithms (PID and Heating Prediction) computa

Posted: 21 Jan 2016, 17:57
by Giorgio71
Hello.

Please, is it possible to activate my PID? I have only this option (soglia Isteresi):

Thank you.

Giorgio

s/n J0a22a2

Re: Advanced algorithms (PID and Heating Prediction) computa

Posted: 22 Jan 2016, 10:08
by Brieuc_Netatmo
Hello,
Please wait for the PID to be ready, you've installed your Thermostat the 15th of January...

Re: Advanced algorithms (PID and Heating Prediction) computa

Posted: 22 Jan 2016, 10:25
by Giorgio71
Ok, thank you.