Evohome - poor temperature regulation

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • erik
    Automated Home Guru
    • Feb 2015
    • 244

    Set point and actual = 17. Then override set point to something obscene like 23. After an hour, when the actual temperature gets to 19, then set it back to 17 again. You will find that not only will the temperature keep increasing due to the heat in the system, the large integral-error term that has build up will mean for the next 30 mins or so, it will actually keep calling for heat.
    Haven't noticed that yet (haven't tried it) and would be pretty bad if it's true. A setpoint change like that should instantly (give or take 5 minutes) override any integral etc. Just shut the valves and stop calling for heat immediatly (if there's no other zones calling for heat).

    Comment

    • greyhound1234
      Automated Home Jr Member
      • Mar 2015
      • 33

      With regard to the HR92 valves.

      As I said, I have found that if zones have only just reached a stable temperature (say after 30 min), and you introduce temperature switch points in one zone, it affects all the other zones as well.
      You get a increase in the other zones, sometimes by 1 - 1.5 degrees, and sometimes for several hours. The settling time of this system in my house seems to be about 1-2 hours under light load. (FYI in colder weather, it is much more stable.)

      Now when all zones have been stable for several hours at requested temperature, if I turn one zone up, the others will not heat much. Or at all, they will close their valves.
      But as I said if the zones have only just reached temperature recently (or are even slightly above), and I turn up one zone, all of the zones will tend to then overshoot as the radiator valves are still open.

      Typically, the valves are 80% open when < T-0.5. 30 % open at T-0.5. Sometimes they are still 10-20% open at T+0.5 as well, if it has just overshot.
      I believe this is because of the accumulated integral error again, the system still has low demand in there, which takes a little time to get rid of.
      After an hour or so at temperature, the valves can even be 0%-closed, at T+0.5. I have even seem them closed at exactly the setpoint.

      I tested this by looking at the bathroom valve when I gave it a big temperature step. Then at T 0.5 degrees higher than set point it was still 73% open! It got to 1.5 degrees over before thinking it would close.

      It takes a little while for for the accumulated integral error to be 'used up', hence you get demand overshoot and radiator valves open.

      One thing I will say, is that the thermal comfort I get from this system is far better than the old system - provided I leave it alone!
      Last edited by greyhound1234; 20 March 2015, 09:40 AM.

      Comment

      • greyhound1234
        Automated Home Jr Member
        • Mar 2015
        • 33

        A comment for Honeywell.
        PI controllers are designed to drive 'symmetric' systems - those which can be driven up or down. Heating systems are asymmetric, they can only be driven up, but come down by heat loss only. There seems to be no capability in the system to deal with integral overshoot at low load conditions.
        My recommendation would be to implement one or more of the following:
        - We know the system is not stable to temperature steps exceeding a few degrees. Implement temperature change requests in a linear or gradual fashion e.g. 1 degree per hour.
        - Do not start accumulating integral effects, until a zone is at Tset minus 1 degree. This will reduce overshoot when large deviations are requested.
        - Purge the accumulated integral term if the temperature is 0.5 degrees above setpoint and rising.

        If you do this, you will get less complaints from technically-minded people who like to fiddle with their heating system and set complex schedules...
        Last edited by greyhound1234; 20 March 2015, 09:45 AM.

        Comment

        • greyhound1234
          Automated Home Jr Member
          • Mar 2015
          • 33

          A question for Honeywell:
          The minimum run time setting (default 1 minute).
          Is this the minimum the controller will fire for, or is the telling the controller what the boiler overrun is (for the learning algorithm?).

          Just that I have noticed whilst set to 1 minute, sometimes the controller will call for heat for several minutes, but other times I have seen it click on and off in 10 seconds!
          If it is to tell the system, if you pulse for 10 seconds, you will run the heating for a minute and 10 seconds? It may need to know this to improve the learning.

          In which case, I need to set it to 2 minutes since that is what the pump overrun is on my boiler.

          Comment

          • erik
            Automated Home Guru
            • Feb 2015
            • 244

            thanks for all your detailed analysis/findings, greyhound.

            I think this one:
            There seems to be no capability in the system to deal with integral overshoot at low load conditions.
            is probably the most important.

            afaik, the minimum run time setting is suposed to be the minimum time that the relay sends an ON signal to the boiler. So 10 seconds should not happen. I've seen this happening as well sometimes though, but not very often. I think only after setpoint changes.

            Comment

            • orange
              Automated Home Guru
              • Dec 2014
              • 149

              Originally posted by erik View Post
              thanks for all your detailed analysis/findings, greyhound.

              I think this one: is probably the most important.

              afaik, the minimum run time setting is suposed to be the minimum time that the relay sends an ON signal to the boiler. So 10 seconds should not happen. I've seen this happening as well sometimes though, but not very often. I think only after setpoint changes.
              as I've mentioned previously I also see this all the time too

              Comment

              • greyhound1234
                Automated Home Jr Member
                • Mar 2015
                • 33

                I'll post my summary that is under 'positive experiences' thread here as well to round this off.
                ---
                Initially, I was shocked at the lack of control, and overshooting behaviour I encountered. I was very concerned and assumed it was faulty.
                So naturally I did loads of fiddling with it to work out how well it held a certain temperature, what the behaviour of other zones was etc. This didn't help.

                One thing I found particularly surprising coming from a simpler thermostat, that it would be possible to get heat demand when all zones are above the set point. But this is possible with a proportional integral controller (i.e. TPI) if subject to a couple of large set point perturbations. As my feel for how the system behaves has improved, I have altered how I use the system and things are better.

                Some general observations
                - the more you change the temperature, the worse the control gets.
                - If you change the set point up by more than a few degrees, you will get an overshoot before it settles down.
                - For a given set of zone temperatures, it takes a few hours for the system to settle down, each zone to find the correct valve setting, and the boiler demand to stabilise.
                - Other zones will then overshoot again if one zone is increased again in this 2 hour 'settling period'. This will be most evident in rooms that are small and therefore heat up faster.
                - After the settling period of a zone, it won't then increase when other zones are set higher.
                - Once this happens, the control is very good.
                - It is better to schedule in a series of stepped rises that one big step. It is also better for thermal comfort to use slightly lower set point but held for longer, compared to a big increase for a short time.

                So basically, the more you fiddle the worse it gets. If you are perturbing the equilibrium the algorithm will take a few hours to settle down. During this period, you may get overshooting. You have to be 'gentle' with it.

                I have made a few suggestions for how Honeywell could avoid this effect (caused by a build up of the 'integral' part of the P-I algorithm), but it is easily avoided if you don't make huge whacking changes in set-point.

                For example, now after a few hours all zones are in equilibrium and I can achieve lounge - 18.0 as set, bedroom 16.0 as set. Perfect.
                Also comfort is way better than the old thermostat, which keeps the Wife happy!

                Energy use I will report back when I have a year of data. I average 4000 kWh/yr gas usually. But getting control of the upstairs was the main selling point for me, which has been achieved.

                Comment

                • erik
                  Automated Home Guru
                  • Feb 2015
                  • 244

                  Are you saying it's going to take a few hours of settling down EVERY DAY (or actually even every time you have a 'big' change in setpoint?) before it achieves temperature as set?

                  Comment

                  • greyhound1234
                    Automated Home Jr Member
                    • Mar 2015
                    • 33

                    Basically, yes. But it depends on your definition of 'achieve'. Can get to the temperatures easily enough, but to be stable at those temperatures a few hours.

                    E.g. set 18.0 from 16.
                    For me, time to pass 18.0 (to 18.5 or 19.0 sometimes) = 30 minutes.
                    Time to settle back to 18.0 = 30 minutes. (give or take, depends on room and overshoot)
                    Time to become stable at that set point = 30-60 minutes.

                    With this system, it is not just achieving the set point but achieving stability at the set point. From my observations, that takes a little while. But once achieved, it is held quite nicely, even if you ramp up other zones again. The longer the zone is at temperature, the more stable it seems to be as well.
                    So long unchanging schedules are best.

                    I therefore have altered my scheduling so that rather than allowing living room to drop to 15 whilst at work, and bringing it up to 18 at 5pm, I only let it drop to 16.5 whilst at work and bring it up to 18 at like 2pm, so system is stable when I get home.
                    Slightly more energy use one would think, but 2 hours I don't think is a big issue, since it does take a couple of hours for the thermal mass of a room to get to the air temperature anyway, which affects thermal comfort quite a bit.

                    Comment

                    • erik
                      Automated Home Guru
                      • Feb 2015
                      • 244

                      Thanks for explaining.

                      Doesn't seem like very smart behavior though. And in my case, it's not hapenning for some reason. Because even after running an entire day on 1 setpoint, temperature is still going up and down a lot. If we lived closed together, we could do some comparing. Too bad

                      Comment

                      • greyhound1234
                        Automated Home Jr Member
                        • Mar 2015
                        • 33

                        Originally posted by erik View Post
                        Thanks for explaining.

                        Doesn't seem like very smart behavior though. And in my case, it's not hapenning for some reason. Because even after running an entire day on 1 setpoint, temperature is still going up and down a lot. If we lived closed together, we could do some comparing. Too bad
                        If you set it the same temperature in zoned mode, and it is still unstable as you have said, then it is not behaving itself.
                        Today is a 'low load' day, fairly warm here in UK, I have a well insulated modern house. My readings have been constant for half the day at: lounge 18.0 (set 18.0); bedroom 1 and 2 16.0 (set 16.0); bathroom 17.0 (set 16.0). The boiler comes on once an hour or so. There is no unstable behaviour.
                        Given that I have noticed an improvement since resetting everything, that is all I can suggest. Failing that Honeywell should sort it out, fix/replace/refund whatever.

                        You have already reset-rebound before I see. I can tell you exactly what I did...
                        - Factory reset on the controller, after confirming yes to reset, it says all settings have been reset / bindings cleared (tick). I didn't press the tick else it would return to the main screen in thermostat mode (it is a pre-bound unit). I took the batteries out at that point instead. Left them out for 5 minutes.
                        - I then went round each HR92, I unbound it (hold 15 sec in binding mode). I did a 'reset' on it from the menu. I then took the batteries out, and left them out for 5 minutes.
                        - I then put batteries in and reassembled all the HR92s.
                        - I put the batteries back in the Evohome controller.
                        - It comes up that I do not have any bindings, and do I want installer or guided configuration. Chose guided configuration.
                        - I added 1 zone in guided configuration (living room).
                        - When binding, I put the HR92 in binding mode first. Then when Evohome says do I want to use the internal sensor, press no. Then press bind on HR92. Then I waited 30 seconds before pressing the green tick. Do I want to add more rads? No.
                        - I repeated for other zones using the setup menu (add zone).
                        - Then program a constant temperature in each zone during the day, a lower temperature at night, leave for a few days so it can learn the rooms before programming anything more complex.

                        This is what I did.

                        Comment

                        • erik
                          Automated Home Guru
                          • Feb 2015
                          • 244

                          just to be sure, you're also steering the boiler with a BDR91? And you are seeing it going ON only once per hour? That's truely amazing

                          Thanks for your explanation. Did you reset the BDR91 as well? (not listed)

                          Comment

                          • greyhound1234
                            Automated Home Jr Member
                            • Mar 2015
                            • 33

                            It is a bdr91. I didn't reset bdr91, but it did get rebound as part of guided configuration. Good luck.

                            Comment

                            • erik
                              Automated Home Guru
                              • Feb 2015
                              • 244

                              it is a bdr91 and it's only going on once per hour? Wow, I'd pay to see that

                              and HR92 totally set to default? No full stroke/torque? (sorry for asking so many questions )
                              Last edited by erik; 20 March 2015, 07:30 PM.

                              Comment

                              • greyhound1234
                                Automated Home Jr Member
                                • Mar 2015
                                • 33

                                Yes, it is nice to have it behaving. No other changes. Normal stroke. 1 minute min cycle, 6 per hour. Happy to help.

                                Comment

                                Working...
                                X