[OpenTRV-dev] Feedback Mechanisms.

Mon Mar 25 09:38:35 GMT 2013

Re: overshoot and learning.

One way of tuning a control system is to apply an impulse or step to the system and measure the response. Assuming PID control, it would be interesting to see whether optimal tuning parameters could be obtained using this method; fully opening the radiator from cold and measuring the room temperature over time. Possible problems here would be the slow nature of the system and the number of external factors (doors/windows being opened, objects being relocated within the room). A correctly tuned PID loop should achieve fast response with no overshoot.

I wonder if the Honeywell replacement firmware project does this? Anyone familiar with it?


----------------original message----------------- 
To: "Closed list for developer discussions" EMAIL ADDRESS HIDDEN 
Date: Mon, 25 Mar 2013 07:09:07 +0000 

>A few quick thoughts:
>- On the topic of "central control": For my own home, I'd love central control of every 
>TRV parameter (so I can modify schedules and target temps from my Android phone / my wife's 
>iPad / the web / our HTPC). Also, there are homes where it might be possible for the system to 
>automatically learn the room occupancy schedule (hence the user wouldn't have to bother explicitly 
>stating a room-by-room heating schedule in order to get a comfortable, efficient home); but 
>this can only be done if central control of the TRVs is available. But I completely accept 
>that some folks don't want that.
>- I'd like to second Mike's statement that "Anywhere there is more than one possible UI 
>they MUST be in sync at all times. There is no technical reason why this can't happen."
>- On the topic of "smart" TRVs and overshoot. A little bit of overshoot doesn't bother me 
>as much as simply failing to reach the target temperature by the specified time because 
>the initial, "cold" temperature was too low. My ideal TRV would "learn" how long the room 
>takes to heat up and would modify the time it opens the valve to compensate for the starting 
>temperature. e.g. on the TRV just keep a rolling average for "minutes per degree C" for the past, say, 
>10 days.
>OpenTRV-dev mailing list

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opentrv.org.uk/pipermail/opentrv-dev/attachments/20130325/4bc39af5/attachment.html>

More information about the OpenTRV-dev mailing list