Tweet
Hi All,
I see one other post about load scaling last month but will start a new thread here to hopefully draw more attention to the problem I saw overnight.
I've run with load scaling set to "partial" for years and while it can occasionally scale a zone too much I've found the benefits generally outweigh this occasional glitch.
It's not really documented exactly how load scaling works but the basic idea seems to be to observe how much heat demand a room really needs over time and then apply an invisible scaling factor to the heat demand for the zone before combining it with heat demands from other zones.
For example if it has decided that a load scaling factor of 4 is appropriate for my bathroom, the radiator calling for 100% heat would be scaled down to 25%.
In addition to this, load scaling performs a kind of "ramping" when there is a sudden large change in heat demand - for example a zone being scheduled from off to on. Without load scaling this causes the boiler demand to instantly go to 100% and can result in temperature overshoots in all over zones that are heating.
Load scaling largely avoids this by not going directly to 100%, but scaling down the heat demand request and then gradually creeping it up over around 30 minutes, which gives other zones time to react and avoid large overshoots - this is the feature of load scaling I like and why I use it.
However the heuristics which calculate the load scaling factor for a zone have always been a bit flakey IMHO and can go way too far. I've seen scaling factors of 10 or more applied at certain times and IMHO it should really be clipped at a factor of around 4 or so. (Or even make this adjustable ?)
Last night there was a very anomalous response from the system to the two bedroom radiators which had to cope with a large plummet in outside temperatures with a set point of 16C...
Main Bedroom.png
From the image it can hopefully be seen (this forum scales down image quality unfortunately) that the set point was 16C all night, however the temperature of the room plummeted below 16C from around 1am and got all the way down to 12C by a little after 4am before any heating was observed.
Note however the HR92 started calling for heat around 1am as expected and ramped up that request progressively to 100% before 2am, so the HR92 was doing everything right, as observed by my independent monitoring system.
Similarly our Son's room did the same thing:
Joshua.png
It also went well below the set point and started calling for heat and was calling for 100% before 2am but as before no heating until around 4am...
Here is the boiler heat demands coming out of the controller:
Boiler Relay.png
To note - I have a three relay system (boiler relay, CH relay and HW relay) and it can clearly be seen that the CH relay demand started at around 1am and worked its way up to 100% by 2am - as would be expected from the heat demands from the zone.
However the boiler relay itself only rose to a maximum of 4% (implying a load scaling factor of 25 which is rediculous) until it suddenly went to 100% at 4am, as a result of the "optimal start" for the living room causing the living room to call for 100% heat.
In past observations I've found load scaling only scales the boiler relay it does not scale the CH relay - so the difference between the two is the load scaling.
It's been said before by a Resideo employee on this forum that if load scaling causes a room to fail to heat there is a timeout where it will eventually increase the heat demand - I've seen this in action before (on previous firmware versions) where usually after half an hour to an hour it will suddenly bump up the heat demand.
However in this firmware version this does not appear to have happened, and the jump to 100% at 4am is the result of another zone coming on.
From this it appears to me that load scaling is fairly well broken in this firmware release and I'm reluctantly going to turn it off.
Great idea, buggy implementation. Two problems that I can see:
1) As with previous firmware versions the scaling factor (derived heuristically somehow) can get ridiculously high - 25x here. There should be a sensible cap on the scaling figure of say 4x, and the method that is used to calculate the scaling factor needs some work as well. Any heat demand of less than 10% (with a relay) will not even fire the boiler at all due to the minimum on time threshold, so a scaling factor of 10 or more basically means a zone cannot call for heat, which is clearly broken.
2) The previous "failsafe" where it would after a period of time start increasing the heat demand beyond the scaled value seems to be gone. Not only did this cause cold rooms in this situation it could even be dangerous in frost protection situations if a zone that needs frost protection gets a ridiculously large scaling factor and is unable to cause the boiler to fire.
I see one other post about load scaling last month but will start a new thread here to hopefully draw more attention to the problem I saw overnight.
I've run with load scaling set to "partial" for years and while it can occasionally scale a zone too much I've found the benefits generally outweigh this occasional glitch.
It's not really documented exactly how load scaling works but the basic idea seems to be to observe how much heat demand a room really needs over time and then apply an invisible scaling factor to the heat demand for the zone before combining it with heat demands from other zones.
For example if it has decided that a load scaling factor of 4 is appropriate for my bathroom, the radiator calling for 100% heat would be scaled down to 25%.
In addition to this, load scaling performs a kind of "ramping" when there is a sudden large change in heat demand - for example a zone being scheduled from off to on. Without load scaling this causes the boiler demand to instantly go to 100% and can result in temperature overshoots in all over zones that are heating.
Load scaling largely avoids this by not going directly to 100%, but scaling down the heat demand request and then gradually creeping it up over around 30 minutes, which gives other zones time to react and avoid large overshoots - this is the feature of load scaling I like and why I use it.
However the heuristics which calculate the load scaling factor for a zone have always been a bit flakey IMHO and can go way too far. I've seen scaling factors of 10 or more applied at certain times and IMHO it should really be clipped at a factor of around 4 or so. (Or even make this adjustable ?)
Last night there was a very anomalous response from the system to the two bedroom radiators which had to cope with a large plummet in outside temperatures with a set point of 16C...
Main Bedroom.png
From the image it can hopefully be seen (this forum scales down image quality unfortunately) that the set point was 16C all night, however the temperature of the room plummeted below 16C from around 1am and got all the way down to 12C by a little after 4am before any heating was observed.
Note however the HR92 started calling for heat around 1am as expected and ramped up that request progressively to 100% before 2am, so the HR92 was doing everything right, as observed by my independent monitoring system.
Similarly our Son's room did the same thing:
Joshua.png
It also went well below the set point and started calling for heat and was calling for 100% before 2am but as before no heating until around 4am...
Here is the boiler heat demands coming out of the controller:
Boiler Relay.png
To note - I have a three relay system (boiler relay, CH relay and HW relay) and it can clearly be seen that the CH relay demand started at around 1am and worked its way up to 100% by 2am - as would be expected from the heat demands from the zone.
However the boiler relay itself only rose to a maximum of 4% (implying a load scaling factor of 25 which is rediculous) until it suddenly went to 100% at 4am, as a result of the "optimal start" for the living room causing the living room to call for 100% heat.
In past observations I've found load scaling only scales the boiler relay it does not scale the CH relay - so the difference between the two is the load scaling.
It's been said before by a Resideo employee on this forum that if load scaling causes a room to fail to heat there is a timeout where it will eventually increase the heat demand - I've seen this in action before (on previous firmware versions) where usually after half an hour to an hour it will suddenly bump up the heat demand.
However in this firmware version this does not appear to have happened, and the jump to 100% at 4am is the result of another zone coming on.
From this it appears to me that load scaling is fairly well broken in this firmware release and I'm reluctantly going to turn it off.
Great idea, buggy implementation. Two problems that I can see:
1) As with previous firmware versions the scaling factor (derived heuristically somehow) can get ridiculously high - 25x here. There should be a sensible cap on the scaling figure of say 4x, and the method that is used to calculate the scaling factor needs some work as well. Any heat demand of less than 10% (with a relay) will not even fire the boiler at all due to the minimum on time threshold, so a scaling factor of 10 or more basically means a zone cannot call for heat, which is clearly broken.
2) The previous "failsafe" where it would after a period of time start increasing the heat demand beyond the scaled value seems to be gone. Not only did this cause cold rooms in this situation it could even be dangerous in frost protection situations if a zone that needs frost protection gets a ridiculously large scaling factor and is unable to cause the boiler to fire.
). There's a single Evotouch controller (acting as the sensor for one zone), 11x HR92s in 9x zones and 1x BDR91 relay box connected to a Worcester Bosch combi boiler with no opentherm.
Comment