Alternative "weather compensation" control ideas

[ProDave]

The subject of weather compensation for ASHP's somes up from time to time.  The principle seems to be the less cold it is outside, then the lower can be the flow temperature to the under floor heating to meet the needs of the building and so improve the COP of the ASHP.

 

That all sounds fine. BUT lowering the temperature of the water will just result in the ASHP running less hard but probably for a similar period of time.

 

I have been looking at how ours works (without weather compensation) and have thought of an alternative idea.

 

I have the heating timed to come on at 6AM, so by the time we get up at 7:30 there should be some warmth in the floor and the house will be warming up nicely (not that it cools down over night very much, even when very cold outside)  But when it is not very cold outside, by say 8AM the rooms are all up to temperature and it has shut off. 

 

It would seem to me therefore, that rather than traditional weather compensation it would make more sense for the morning start up time to be the thing that varies according to how cold it is outside.  So when we are in a typical (for here) spell of -10 nights then a 6AM start is fine but much above 0 and a 7AM or even later start would make more sense.  It would certainly help self use more of the solar PV generation.

 

So I throw it open to the forum, to find an off the shelf programmable timer for the heating control, that can vary the morning start time according to how cold the outside temperature is.  I strongly suspect such a thing does not exist, but you never know.  It would be quite simple to build such a thing but I have tried to present the "user controls" as a standard boiler programmer,  so looking for ideas that don't make it complex to the average user.

 

Of course the same variable timing could also be applied to the evening turn off time as well,

[Radian]

1 hour ago, ProDave said:

So I throw it open to the forum, to find an off the shelf programmable timer for the heating control, that can vary the morning start time according to how cold the outside temperature is.  I strongly suspect such a thing does not exist, but you never know. 

 

I think you may be right. I can't find anything of this description. But does it actually make sense?

 

1 hour ago, ProDave said:

  It would be quite simple to build such a thing but I have tried to present the "user controls" as a standard boiler programmer,  so looking for ideas that don't make it complex to the average user.

 

Why would it need to present any 'user controls' i.e. you just black-box it?

 

Your house might only cool a tiny bit but it's still measurable and is representative of the actual situation rather than any assumptions. If you can identify the precise indoor temperature set-point for night set back such that it will reach your daytime target in the chosen period following your morning start time, would this not have the desired effect? I might be missing something here but this scheme works well enough for me. My only problem is significant early morning solar gain which, by the time it arrives (if at all), the heat has already been put into the floor and the temperature overshoots as a consequence. I'm trying to set up a ML program to use a solar forecast to modify my set point to solve this but it's proving quite tricky.

 

Most off-the-shelf thermostats have lousy hysteresis and poor resolution but there are sensors with improved granularity. I implemented my control system with sensors having a resolution of 0.1°C and was immediately struck by how sensitive I was to the effects of setting a temperature to some fraction of a degree. They say "save money by turning your heating down by a degree" but this is way too coarse. The funny thing is I implemented an Alexa smart home skill to set the temperature and found its spoken response rounds any given fractions. So if I say "Alexa set the heating to 20.5" it says "OK, autoset to 20". But what comes in from the Lambda function is indeed 20.5 so thankfully my code can act on it.

[Marvin]

Hi @ProDave

 

I have been considering your thoughts and would like to add a bit of mine:

 

2 hours ago, ProDave said:

That all sounds fine. BUT lowering the temperature of the water will just result in the ASHP running less hard but probably for a similar period of time.

So when warmer in the morning, with the temperature difference being less the COP increases?

 

2 hours ago, ProDave said:

I have the heating timed to come on at 6AM, so by the time we get up at 7:30 there should be some warmth in the floor and the house will be warming up nicely (not that it cools down over night very much, even when very cold outside)  But when it is not very cold outside, by say 8AM the rooms are all up to temperature and it has shut off. 

Does this mean that you are using a higher temp than necessary as your floor heats up so quickly and thereby reducing the COP? Also I noticed that often just before the dawn is the coldest time in the night so starting the ASHP then surely generally means that it has to work hardest and therefore reduces the COP?

 

2 hours ago, ProDave said:

 

It would seem to me therefore, that rather than traditional weather compensation it would make more sense for the morning start up time to be the thing that varies according to how cold it is outside.  So when we are in a typical (for here) spell of -10 nights then a 6AM start is fine but much above 0 and a 7AM or even later start would make more sense. 

Surely this would result in a delayed start in warmer weather, but using the same temperature output, thereby shortening the cycle but still reducing the COP?

 

2 hours ago, ProDave said:

It would certainly help self use more of the solar PV generation.

 

Surely this would depend on the individual PV design. In winter we don't have any real PV until 9.30am but in summer I think it will be at about 7.30am when we will need less heating.  

[ProDave]

That is food for thought.

 

I was probably looking at this the wrong way, trying to think of a way to delay the morning start when it is warm.

 

Perhaps I should leave the existing controller there (plain ordinary 3 channel boiler time switch) but set the morning start time to a little later than it is, (i.e. to the warm weather start time).  And then add the hidden "black box" controller to turn on the heating earlier (overriding the time switch) IF it is particularly cold outside.  That black box could be very simple indeed.

 

[Marvin]

In tests I have run here we found the ASHP and controls uses less power left on low over night than turning it off and on.

By the time the dawn comes the house is not as cold and so the ASHP seems to have less bad COP work to do.

 

Obviously this is hard to be accurate about because of all the variables, but that is what we do.

 

However, I think this might not work if the building had poor airtightness or poor insulation or no MVHR.

Edited February 19, 2022 by Marvin

[Marvin]

15 minutes ago, ProDave said:

And then add the hidden "black box" controller to turn on the heating earlier (overriding the time switch) IF it is particularly cold outside.  That black box could be very simple indeed.

You could just have a timer and an external thermostat. 

 

The process being:

If the thermostat drops to a given figure, and the timer has already switched on at a given time the ASHP will come on.

 

Done.

[PeterW]

Does the black box though not need to be aware of the internal temperature and the potentially for heat loss the next day though..? That becomes your issue which I think @TerryE has tried to crack using some basic heat Store logic. 
 

I think what you’re describing though is one of the pulse time relays that could have a pair of series wired thermostats (internal and external) that give you the pre-boost to the main programmer. 
 

something to consider is a Shelly that would allow you to do this using scenes - off the shelf component and with a bit logging and monitoring you could validate what the slab, house and external temperature are doing and then do a “fail safe” one hour boost at the last minute to finish off the heating irrespective of external temperature. 

[ProDave]

Yes.  All I think i want to achieve is variable start time in the morning.  My objective I guess is the set time, wants to be the time it reaches the required temperature, not the time it starts to warm up.

 

Internal temperature would be a good measure of that and would be a lot simpler than trying to get a connection to an outside temperature probe.  My little gizmo could just sit on the wall next to the existing programmer, or even hidden behind it inside the wall.

[dpmiller]

It's like the "optimum start" feature on some older digital roomstats- they learn the rate-of-change of the room/house and re-start from the setback temp in time to have regained temperature at a selected time. The lower the internal temp falls overnight, the earlier the heat comes on.

 

but as @Marvin alludes, how much of a point there is to this in a well-insulated property with an ASHP, is questionable

 

[Marvin]

12 minutes ago, ProDave said:

Yes.  All I think i want to achieve is variable start time in the morning.  My objective I guess is the set time, wants to be the time it reaches the required temperature, not the time it starts to warm up.

 

Internal temperature would be a good measure of that and would be a lot simpler than trying to get a connection to an outside temperature probe.  My little gizmo could just sit on the wall next to the existing programmer, or even hidden behind it inside the wall.

A sort of err frost stat ?

Edited February 20, 2022 by Marvin
Add info

[Marvin]

14 minutes ago, ProDave said:

Yes.  All I think i want to achieve is variable start time in the morning.  My objective I guess is the set time, wants to be the time it reaches the required temperature, not the time it starts to warm up.

 

Internal temperature would be a good measure of that and would be a lot simpler than trying to get a connection to an outside temperature probe.  My little gizmo could just sit on the wall next to the existing programmer, or even hidden behind it inside the wall.

One of the challanges is that the inside probe wouldn't take into account the rate of temperature  change and so the heating comes on, but does it need to be alot or a little? Who knows.

 

Surely to keep the inside temperature steady you only have to evaluate the difference between that and the out side temperature at regular intervals and adjust the input accordingly.  

 

Surely a system that was continually monitoring the temperature differences and reacting to this would be best and use the least energy, a sort of weather compensator?

 

And for really clever you could have a system that was linked to the weather forecast and take the next 4 hours expected weather in to consideration.

 

However, the biggest downfall to using a weather compensator would be that it would have to be individually adjusted to all individual factors. Here's some I've thought of:

 

The locations general climate.

Building rotation in relation to solar gain.

Shadowing

Wind exposure 

Altitude

Occupiers habits

Occupiers temperature level expectations.

And so on...

 

 

 

[Radian]

26 minutes ago, ProDave said:

Yes.  All I think i want to achieve is variable start time in the morning.  My objective I guess is the set time, wants to be the time it reaches the required temperature, not the time it starts to warm up.

I think that's what I was suggesting yesterday but I'm not exactly sure what you mean by 'set time'?

[ProDave]

Lots of food for thought.  I won't rush into a solution and will probably be a summer project to test next winter.

 

Simple frost thermostat won't do it, on an extreme cold day that could potentially leave the heating on most of the night.  so would still need to be some combination of temperature probe(s) and some timer logic so all it is doing is varying the turn on time in the morning.

 

I do like the idea of a programmable thermostat with an "optimum start" function.  If anyone knows one still available offering that, let me know.

[JohnMo]

Don't you just need a thermostat with ITLC control?

[PeterW]

@ProDave the Honeywell T4 or T4R both have optimisation algorithms in them and are off the shelf but not that cheap. They are decent units too - an evolution on the older CM range. Also look quite nice !

[SteamyTea]

31 minutes ago, Radian said:

I think that's what I was suggesting yesterday but I'm not exactly sure what you mean by 'set time'?

I have, as many of you know, E7 heating.

By the very nature of this, there is a switch off set time (7 AM UTC).

Because my water and storage heater would come on at 11 PM, then off for an hour, and back on at 1 AM, I put lock out timers on each circuit (my time window has now changed to midnight to 7 AM).  This was to stop things heating up, then cooling down, then heating up again.  I currently have the E7 time window effectively set as an E4 (3 AM to 7 AM).

This gives me plenty of time to heat everything up.

 

I have often thought that a variable time window could be useful, but rather than along the lines that @TerryE uses (signal from MET Office), just looking at the rate of change of the OAT between 8 PM and Midnight could be used to set the start time for the next day's heating window.

I have quickly plotted a few years outside air temperature (OAT) binned it by rate of change in the above 4 hour window and correlated it with the mean OAT.

Needs a bit more thinking about how to use the slope of the time to set the length of the heating window, but I feel there is a reasonable correlation to be useful.

I know from experience that my heating has to come on when the daily mean OAT is below 10°C.  There does seem to be a half decent correlation between the night before temperature change and the day ahead mean temprature.

Below is a quick chart showing July and February slopes.  These are my two extreme months.  Needs a bit more thought, and how to integrate it with the building heating curve may become tricky.  The aim is simple, the heating turns off at the set time with just the right amount of energy having been delivered, and no more.  The objectives are much harder.

Edited February 20, 2022 by SteamyTea

[JohnMo]

Does this do everything you need out the box?

 

A Salus SQ610 has ITLC algorithm for UFH and you can connect external sensors or switches

[ProDave]

15 minutes ago, JohnMo said:

Does this do everything you need out the box?

 

A Salus SQ610 has ITLC algorithm for UFH and you can connect external sensors or switches

That is a possiblility.  Do you know where I can find a simple description of this ITLC algorithm?

[JohnMo]

All I could find was.

 

https://salus-controls.eu/2021/03/07/learn-about-heating-control-algorithms/

 

[TerryE]

5 hours ago, PeterW said:

Does the black box though not need to be aware of the internal temperature and the potentially for heat loss the next day though..? That becomes your issue which I think @TerryE has tried to crack using some basic heat Store logic. 

 

Peter, thanks for the prompt to this interesting discussion.  My only quibble is with the verb "tried" ? I would say that the system has been working really well for over 4 years now.

• IIRC, you have a pretty energy-efficient house with decent internal thermal mass.  This puts you well outside the operating domain of most standard controllers.  A good indication is the 24hr heat-loss without heating (mine is just over 1°C per day).
• I use two straight-line functions to calculate total daily heat requirement:
  • Planned demand in kWh as a function of Δt: (set-point - Met-office predicted temperature)
  • A feedback adjustment as a function of Δt: (yesterday's average room temp - set-point temperature)
• When you schedule the on times isn't that critical. I do mine overnight so I get a ~1°C ripple over the day.  Moving a chunk to an afternoon boost would drop this to nearer ½°C
• I am not sure that you really need some fancy (and expensive) controller.  An RPi zero 2 controlling a relay or two would be perfectly adequate.

Edited February 20, 2022 by TerryE

[SteamyTea]

10 minutes ago, TerryE said:

A good indication is the 24hr heat-loss without heating (mine is just over 1°C per day).

Does that take purely the drop in internal temperature, regardless of external conditions?

[TerryE]

@SteamyTea Nick, that's a typical daily ripple during winter months.  The time constant of heat flow through the walls is ~1 day so I find the Met office forecast is good enough.

Edited February 20, 2022 by TerryE

[SteamyTea]

2 minutes ago, TerryE said:

a typical daily ripple during winter months

Do you every run a stats test to see how what you hope your algorithm is doing, and what is really happening?

Should be fairly easy to compare temperature stability with energy input.

 

[TerryE]

The stats are all aggregated in Home Assistant and Jan and I can look at these through a browser or the HA app. TBH the biggest bumps (about ½°C before the feedback term kicks in) is when we go away for a weekend, say or have guests to stay  (about 3kWh per person plus extra cooking ....) Other than that the ripple is remarkably stable.

Edited February 20, 2022 by TerryE

[SteamyTea]

16 minutes ago, TerryE said:

Other than that the ripple is remarkably stable

Is in my place as well.  Generally less than a °C. And I have basically no control unless I manually twist the input and output knobs on the two storage heaters.

Still be interesting to run a Chi Square on it as that will check the formula/equation/algorithm, rather than the outcome.

Maybe @DamonHD can have some input as his devices are trying to do the same, but via a different pathway (and they may have got a statistician on board now).

Edited February 20, 2022 by SteamyTea