@derek-m - sorry to hear about the flu - never any fun, and the post viral fatigue doesn't help either.

Controlling the heat pump: all very interesting stuff, more comments in a moment but first to answer your questions:

Posted by: @derek-m

↑

1) Do you wish to keep the indoor temperature constant, or be able to make changes within a specified range?

2) What degree of accuracy would you like to achieve?

3) Would you be happier with a more rapid response (but less efficient), or a slower response with better efficiency?

(1) by and large, constant. There may conceivably be times when I want it to run hotter or cooler, and if there are, I can make any necessary changes - I think it will probably just mean changing the desired temperature in the code.

(2) one or two degrees - I'm being realistic here, doors open, setbacks get programmed in etc. I am also not sure how accurate the temperature sensors and thermometers are, apart from knowing they don't always agree! The room sensor claims +/- 0.5 degrees accuracy, but it is not clear whether that applies at room temp ranges. The real setting is the one where I am comfortable, and then the desired temp is whatever the sensor reads at that point. 

(3) this is at the heart of the matter: the whole idea is to get a more rapid response/recovery after a setback. Initially I favour leaning towards rapid but less efficient, and then see how that less efficiency turns out in practice, and perhaps tweak the speed of recovery a bit if it looks like a rapid recovery costs a lot.

Posted by: @derek-m

↑

Using the Midea data that you supplied some time ago, it would appear that a 1C change in LWT will cause a 6% to 9% change in thermal energy output, which could vary the indoor temperature by between 0.45C to 0.9C. Since the effect appears to be greater at warmer outside temperatures, I would suggest only changing the cold end of the WC curve, the 55C @ -4C end. It may be necessary to adjust the warm end setting to allow adequate control during warmer periods.

This is very useful, and I agree best to start by changing the cold end of the curve. In the trial setback last December, and on other unintentional (eg power cut) setbacks, the IAT dropped by about 0.5 degrees or a bit less per hour (depending on OAT), as can be seen in these two somewhat inelegant charts:

This is the December 2022 trial 6 hour (2200 to 0400) overnight setback, when the OAT (outdoor air temp) fell from around 5 degrees to about 1-2 degrees. The IAT (indoor air temp) falls by about 3 degrees, and then takes ages to recover - by 2100 it is still not quite back where it was. It is this slow recovery that is the trigger for wanting to try a recovery boost.

This is the April Fool's Day 2023 power cut, which lasted for about 5 hours in the morning (the period where the ambient temp disappears, because that now comes from the heat pump - and in passing that is a topic for another day - is the ambient taken from a heat pump useful, given the heat pump itself affects the ambient temp around it?). Because the OAT is a bit higher than it was in December, the fall in IAT is a bit slower, and the recovery a bit faster, but it still takes many hours.  

All of which is to say my building appears to have a relatively high thermal mass, and the IAT changes slowly. The falls only amount to a few degrees, but they also take a lot of effort (energy) to recover from.

I think I get the idea behind Proportional Action (Gain) Control - the corrective action is proportional to the distance the actual IAT is from the desired IAT, but I can't help wondering whether it, and especially the Integral Action (Reset) element, over-complicates things given the reality is we can only make integer changes to the WCC (weather compensation curve) settings. I am not sure what we gain by knowing we should have a lower end setting of 55.34862 when in practice it can only be 55 or 56. In a way, this is yet another modelling problem: the 'model' tells us we need such and such a value, ignoring the fact that in the real world, there is no such value.

The take away from all this seems to me to be that I probably only need small changes to the WCC. These should have relatively large effects both on the change in IAT and efficiency (COP), such that it seems sensible to start with a 1 degree change and go from there, ie empirically determine what works (and doesn't). In code, which runs say every 15 minutes:

# the desired IAT
dIAT = 19

# the standard WCC left hand (lower OAT) LWT setting
sLWT = 55

# compare actual IAT (aIAT) to desired IAT and if differenrce is greater than 2, then raise current LWT (cLWT) by 1 degree
if dIAT - aIAT > 2 then cLWT = sLWT + 1 else cLWT = sLWT   

# send the current LWT to the wired controller (perhaps add code to only do this when cLWT has changed)
instrument.write_register(265, cLWT, 0)
             

As I said earlier, this will no doubt need tweaking and amendments eg if the difference is > 4, then maybe increase LWT by 2, but the general principle is there in 4 lines of code. I don't think this can get into an out of control runaway state, because it is in effect anchored to the sLWT.

The existing monitoring setup already records hourly energy in and out, and will reveal in due course the actual energy use for both continuous (24/7) and intermittent (setback and recovery using the above code for the recovery boost) running.    

I've already thought of one possible tweak that incorporates your proportional gain principle:

# set the gain factor
gF = 0.5

# get the difference between the desired IAT and actual IAT
deltaIAT = dIAT - aIAT

# determine the new LWT and round it to a whole number
nLWT = round(sLWT + (deltaIAT * gF))
 

Thus if the deltaIAT is 2.1 degrees, and the standard LWT is 55 degrees, the new LWT will be set to 56 degrees.  But even this may be over-complicating things. No doubt other ideas will emerge in due course...

Temperature monitoring in action: the last 24 hours, upper green plot is the IAT, collected from the recently installed MD02 sensor, lower blue plot is the OAT (ambient) from the heat pump. With the heat pump off, as it is at the moment, apart from timed DHW heating, it should reflect real OAT, but when the heat pump runs, it affects the OAT in its vicinity, and I think this can be seen in the Ambient (green) plot, at around 1400 yesterday, when the heat pump came on to heat the DHW. I think it shows it lowered the OAT - the DHW timer is set to come on at 1300 - but there is then a curious uptick around 1400, by which time the heat pump would have been off for about 30 mins.

But what about when the heat pump is running most or all of the time? The surrounding air is perceptibly cooler, and so the whole heating system is monitored and controlled (because the WCC uses the OAT as the input) by a OAT that is not a reflection of the more general OAT. Does this matter? Perhaps not. If the WCC is set up in such a way that factors in the measured OAT is lower than the more general OAT, then maybe the system accommodates the discrepancy. 

The other thing the chart shows is the effect of the thermal mass of the building. While the OAT dropped around 8 degrees, the IAT dropped by less than one degree:

BTB yes that chart does come from Home Assassin and the ambient (byte_09) data is from Midea_ac_lan but that is just because it is there for historical reasons. The MD02 temp comes from the modbus collected data stored in a csv file (in HA jargon it is a file sensor) and the ambient, collected over modbus, is also in that csv data file.      

@cathoderay

I hate to burst your bubble, but what you propose will not be possible to achieve, since it would defy the basic Laws of Physics and Thermodynamics.

From the modelling (sorry simulations) that I have carried out, the test data showed that at a constant OAT of 10C, to maintain an IAT of 20C would require an electrical energy input of approximately 1.62kW, giving a thermal energy output of 6.76kW at a COP of 4.17. The calculated LWT is 42.3C.

Stopping the heat pump for 1 hour would save 1.62kWh of electrical energy, but the building would lose 6.76kWh of thermal energy, and hence cause the indoor temperature to fall to 19.73C.

If the heat pump is restarted and the LWT setting raised to 46.5C, the electrical energy input will increase to 2.2kW, giving a thermal energy output of 8.8kW at a COP of 4. Under these operating conditions the thermal energy supplied by the heat pump will exceed the building heat loss by approximately 2.2kW, so it will take approximately 3 hours for the heat pump to replace the lost energy and bring the IAT back to 20C. Total electrical energy consumption is 130W higher throughout the 24 hour period.

During a 2 hour shutdown, the building loses approximately 13.3kWh of thermal energy with the IAT reducing to 19.47C, and takes 6 hours of higher output heat pump operation to restore the iAT to 20C. Total electrical energy consumption is 200W higher throughout the 24 hour period.

During a 3 hour shutdown, the building loses approximately 19.7kWh of thermal energy with the IAT reducing to 19.21C, and takes 9 hours of higher output heat pump operation to restore the iAT to 20C. Total electrical energy consumption is 250W higher throughout the 24 hour period.

So the conclusion would be that under constant OAT conditions, using an overnight setback would actually be detrimental, and use slightly more electrical energy.

I am now exploring ways in which it may be possible to achieve a reduction in total energy consumption, details of which I will post in the near future.

Please feel free to question my findings, which I feel some forum members will wish to do.

@derek-m - don't worry, there isn't a bubble to be burst, as I don't know the answer, which if I did would be the bubble. I am still at the I don't know stage, don't know whether continuous running or set back and recovery boost uses more energy. As a compliment to your simulations (sorry modelling), I plan to collect some empirical data, and see what that shows.

The core question is whether the extra energy during a recovery (over and above what would otherwise have been used) is more or less than the energy saved during the setback. On the face of it, having a setback lowers the average IAT over a 24 hour period, and that would seem to suggest that overall, less energy must be required to maintain a lower average IAT. But it is not that simple, because of the recovery boost, which uses more energy per se, and on top of that produces heat at a lower efficiency. Those two things are what makes it hard to quantify.

An unadjusted WCC takes no account of the IAT, it only uses the OAT, and so I am pretty sure that if there was no recovery boost, there must be an overall energy saving. In fact, I can demonstrate this. Unfortunately, I don't have all the necessary data from the December 2022 setback, but I do have data for the April Fool's Day 2023 power cut, which simulated (sorry, modelled) a setback without a recovery boost. Despite much the same OAT on the three days in the chart below, there is visibly less energy consumed/produced on the 1st April:

Setback + no recovery boost = energy saved. The actual numbers for energy consumed for the three days shown as 30 26 and 37 kWh respectively, meaning I saved about 5-10 kWh during a 5 hour or so setback. These are not trivial numbers given the current price of electricity.

The 64 million dollar question remains: at what level of post recovery boost, as needed to get a reasonably quick post setback recovery, does the extra energy used override and cancel out the energy saved during a setback and recovery without a boost? The problem with a simulated answer is it may well fail to include all the necessary variables, the formulas might be wrong, who knows? That's why we collect real world data, which is what I propose to do.    

Yet another reasonably thorough google search has reminded me that the only established fact about continuous vs intermittent plus boost heating is that no one really knows the answer, let alone can prove it. They are a lot of confident assertions - there's even one above, in which a mythical bubble is burst - but the problem is these confident assertions go both ways, and by and large, for the same building, they can't both be right. The 'for the same building' qualification is important, because heat loss is crucial to all this. Imagine a hypothetical building with zero heat loss, and then another of identical size sitting in the same ambient air with no walls, say an open barn in a field, and consider the heat loss, and therefore demand, for each. Heat loss matters, hugely.

What I think we do know is that intermittent heating without a recovery boost must always, in like for like conditions, be cheaper to run than always on heating. The logic for this statement, beyond the obvious saving while the heating is off, is that the average IAT over 24 hours with intermittent heating will be lower than with continuous heating, meaning the energy needed to maintain that lower average IAT must be less than that needed to maintain a higher IAT. I think we can say amen to that, though it should be noted that there is a school of thought that has it that condensation builds up in the walls during cooler periods, and that drying that condensation out somehow uses more energy. But that condensation in the walls is the Higgs boson of the heating world, I am not sure anyone has ever seen it, let alone quantified it. That said, I should add, and this shows just how complicated all this is, that I have noticed that running a heat pump continuously does make me perceive the house is drier, and more comfortable to be in, than it was with the intermittent oil fired heating.

The uncertainty is how much extra energy is needed, over and above that used in continuous running, to achieve a reasonably rapid recovery from a cooling period. This can be boiled down, so to speak, to what on the face of it seems like a simple question: how much energy is needed to raise the IAT by one degree over one hour? Given that figure, we can, given the temperature rise and time frame desired, then calculate the extra energy needed as the figure we get less the amount of energy needed to keep the building at a steady IAT, which is normally assumed to have a linear relationship with the OAT (for example, if my house has a heat loss of 10kW at 18 IAT / -2 OAT and 0kW at 18 IAT / 18 OAT then the steady state heat loss at 8 OAT will be 5kW etc etc).     

The problem is there is a notable lack of a method to calculate this extra heat needed, and, of course, 'every building is different', and not only that, the weather changes. That is why I am keen to get an empirical answer - what really happens in the real world. Does @derek-m's method factor in the heating Higgs boson (the invisible spook condensation in the walls element, if it is indeed a real effect)? I suspect not, unless it is already baked into the data he used to do his calculations. This will be best done once the heating season is in full swing, but in the meantime, I do have the April Fool's Day power cut data, and I think it shows some interesting things. Here are two plots, the upper various parameters as shown in the legend, and the second the kWh in and out. The former is modbus based, ie values read directly from the wired controller over the wired connection, the second is, for historical reasons, based on midea_ac_LAN, a python module that snoops on the Midea app data, which may or may not have been pre-cooked by Midea. The former shows minute data, the second hourly data.

The power cut, on the morning of April Fool's Day, can be clearly seen. The two big spikes in the LWT/RWT are DHW cycles. Other things to note are (a) the Midea steady state flow rate, meaning all the variation in output is done by changing the LWT/RWT delta t, which tends to be on the low side (it hardly ever reaches 5 degrees) and (b) by and large, the LWT goes up and down as it should in response to the ambient, ie the weather compensation is working, this being particularly clear from 1800 onward on the 2nd April; likewise amps in go up and down as expected, but there is also very evident cycling, perhaps because the ambient wasn't that cold.

In passing, the midea_ac_LAN data does have some odd spots, eg managing somehow to produce energy out with no energy in, and perhaps some times merging two hours, as in the early hours of 2nd April, but I suspect this mostly down to using integer values for kWh, plus or minus a bit of Midea pre-cooking the data. Now that I have amps in and volts in, I can actually calculate my own kWh in. In due course, I may yet add that modbus enabled replacement for my current dedicated heat pump kWh meter.

Now let us consider the recovery period, which happened without any boost, as the heat pump had no way of knowing there had been a power cut, and as a result the IAT was below what it should be, 19 degrees. What we see is the amps in almost certainly dropped a small amount for a while, as the ambient was a little higher, and, from the lower chart, the heat pump was using 1kWh per hour for most of the eight hour or so recovery period, the 2kWh hour at the start of the recovery period is almost certainly partly the DHW cycle, which usually takes about 30 mins. During that recovery, the IAT rose from 17.8 to 19.6 degrees, ie it overshot a bit (solar gain? I seem to recall it was sunny that day, and it was April) but in general, on average (though it is clearly not linear), 1 kWh per hour of energy in raises the IAT by around 0.25 degrees an hour in the conditions prevailing at the time.

What I don't yet know is what the relationship is between that energy input and subsequent temperature rise. If it is a simple linear relationship, ie it would take 4 kWh for an hour to get the IAT to rise by 1 degree, then to get the recovery to happen in say two rather than eight hours, I might need to use 2 x 4 = 8 kWh, but as I would have used 2 x 1 kWh anyway, the extra energy used is 6 kWh, perhaps a bit more than the energy saved during the power cut.

But this is pure speculation, I just don't know how much extra energy is needed, and without some very robust and satisfactory explanations of how any 'models' work, I am disinclined to trust anything they suggest. That's why I am going to get an empirical answer once the heating is back on and the weather a bit colder.            

Posted by: @derek-m

↑

I am now exploring ways in which it may be possible to achieve a reduction in total energy consumption

a flock of sheep stacked around the house? any related clothing? wool knitted underwear is bliss

Posted by: @fazel

↑

Posted by: @derek-m

↑

I am now exploring ways in which it may be possible to achieve a reduction in total energy consumption

a flock of sheep stacked around the house? any related clothing? wool knitted underwear is bliss

 

Don't you find it rather itchy? 🙄 

@cathoderay

You are indeed correct that the subject is rather complex, and I agree with much of what you have stated, along with the need to test some of the theories on a functioning system.

As you are aware I have been carrying out some 'whatiffery' which has helped identify what I feel are some of the more important factors when dealing with the subject of overnight temperature setback.

Thermal Mass.

One of the factors that I have been exploring is the effect of the thermal mass or thermal capacity of the building, and how this may influence both the rate of heat loss and the IAT. I incorporated thermal mass (TM) into the calculations as the quantity of thermal energy needed to raise the indoor temperature from 0C to 20C, and then tried different values of TM in the simulations with the following results.

A setback lasting several hours in a building with a low TM could possibly provide a reduction in energy consumption of over 10%, totaling 5kWh or more, but as with most things in life this comes at a price, in the form of a large reduction in IAT by the end of the setback period.

A building with a large TM does not suffer such a large reduction in IAT, but conversely does not achieve the same degree of energy reduction.

Heat Pump Size (Capacity).

One of the often discussed problems is correct sizing of a heat pump.

If a 12kW heat pump is operating in the upper end of its range at say 10kW, and is then stopped for a 1 hours period, the loss of energy from the building will be in the region of 10kWh. If when restarted the heat pump can achieve its full output of 12kW, the excess of supply over demand would be in the region of 2kWh, so would take approaching 5 hours running at full output to replace the lost 10kWh. When running in the region of full output heat pumps are often less efficient and therefore require more energy in.

Variations in OAT.

This may seem obvious at first glance, but there are even more aspects to consider.

During milder weather conditions the amount of heat energy lost will be lower, so hence the reduction of energy consumption will also be lower in kWh terms. Several days ago in took approximately 9 hours for the IAT to fall by 1C in our home, this evening the same 1C reduction has taken only 3 hours. This brings in the question of how best to control any setback period, should it be for a timed period, in which case the IAT may fall lower than desired and take much longer to recover, or should it be controlled on temperature, where the heat pump restarts once the IAT has fallen to a designated temperature?

During colder weather conditions it is probable that the IAT may not have returned to an acceptable level when the home occupants arise from their slumbers.

Simulations have shown that for each 1 hour of setback, it can take in the region of at least 2 hours (even with a high level of boost) to recover the original IAT. In some cases the system actually consumed more energy than it would have done without a setback. So it is once more a balancing act, to setback and recover more slowly and save more energy, or setback and try to recover rapidly, but in the process use the same or maybe more energy?

In an Ideal World.

OAT on most days does not remain constant, it is often lowest just around dawn and highest mid afternoon.

For some homeowners it may therefore be possible to take advantage of these temperature differences. Run the heat pump to recover or even boost the IAT during the warmer day time period, and setback or at least reduce the IAT during the nighttime period.

This of course is easier said than done, since after an overnight setback the IAT will probably not have recovered by breakfast time. If the home is unoccupied for much of the day, an alternative could be to initiate a setback at say 8am, when the OAT is still quite cool, then start the recovery and boost after lunch when the OAT will likely be higher. Simulations show a 7% to over 10% reduction in energy consumption could probably be achieved.

For those wishing to go the whole hog and have the finances and room to accommodate it, a thermal heat store could be used to accumulate thermal energy during the milder period of the day for use during the colder period.

Posted by: @fazel

↑

wool knitted underwear is bliss

Wool is wonderful, just be sure not to put those bloomers on inside out and back to front!

@derek-m - thank you for an extremely useful review of the many factors in involved. I agree the importance of thermal mass is often overlooked. It's effects can be seen in the charts I posted a couple of days ago, OAT fell by 8 degrees, IAT by less than 1 degree. I agree with your thinking that a low thermal mass is perhaps somewhat counter-intuitively likely to work better with a setback. Although the building cools more rapidly, it also reheats more rapidly, in other words it is more responsive. One sees the same effect when comparing cooking in a light aluminium frying pan against cooking in a heavy cast iron one.

Just to clarify: am I right in thinking you are treating the heat in a building a bit like the amount of water in a bath that has a standing leak, and a tap that tops up the level on various schedules, either continuously to intermittently? In the steady state condition (with a low OAT), using litres rather than kWh because it is a bath, the bath leaks 10L an hour, and the tap supplies 10L an hour, and the water levels stays constant. Turn off the tap, and the bath then loses 10L over the next hour, but no water is added, so the water level drops. Next turn on the tap again, at the same rate as before, without any boost, and by and large the water level will very gradually increase, as the leak will be marginally less from the lower head of water, until the water level gets back to what it was before. This is in effect what happens with a setback with no recovery boost. Now comes the hard to get your head round bit, at least for me. I know I need to add 10L to get the water back to its previous level. But this 10L needs to be in addition to the volume need to match the current leak. But I think - I am not sure - your analysis uses a more static assessment, 10 kWh lost in setback, boost output during recovery from 10 to 12 kWh, which gives 2 kWh extra, deficit is 10 kWh so takes 5 hours to replace the deficit. What confuses me is going from 10 to 12 kWh is actually quite a substantial increase, 20%, yet when we talk about boosting the LWT, we tend to talk on 1 degree increments, which is, at around 50 degrees, only a 2% increase.

The Midea engineering data output tables (I'm using the one for my 14kW heat pump) don't really help with this at all. For example, at an OAT of 5 degrees, operating at 'normal' capacity level, the output at a LWT of 50 is given as 10500 kWh, at a COP of 2.92 and a power input of 3595 kWh, while at a LWT of 55, the output remains almost exactly the same, at 10568 kWh, with as expected, a lower COP of 2.81, and a power input 3767 kWh. In other words, according to Midea, dialling in a 5 degree increase in LWT has no effect on energy out, while at the same time it increases energy in! I think the answer to this riddle is that the higher LWT will heat the house more rapidly (greater delta t between rads and rooms), despite the fact the total energy out is pretty much the same, but it will use an extra 172 kWh of energy to do it - but that is only a 4.8% increase, not a 20% increase, to boost the LWT not by 1 degree but a whopping 5 degrees.

@cathoderay

One of the benefits of using simulations is the ability to vary one parameter, whilst keeping others constant, to see the effect of that one variable.

Your analogy of a leaking bathtub is indeed correct. My philosophy was to assume that at a given IAT, the building would store a certain quantity of thermal energy at 20C, with at an IAT of 10C only storing half this amount. I then performed a set of calculations with a TM of 100kWh, 250kWh, 500kWh, 750kWh and 1000kWh at the designated 20C.

By recalculating at 1 hour intervals it is possible to take account of the reduction in heat loss as the DT between IAT and OAT reduces. So if the TM is 100kWh at 20C, and loses 10kWh over a 1 hour period, the resulting TM will now be 90kWh, which would indicate that the IAT will have fallen from 20C to 18C. At an IAT of 18C the rate of heat loss will have reduced to 9kWh, so after a further 1 hour period the TM will now be 81kWh at an IAT of 16.2C.

Carrying out the same calculations with a starting TM of 500kWh at 20C would produce the following:-

Hour.   Heat Loss     TM.         IAT.

   0                        500          20

   1          10           490          19.6

   2          9.8          480.2       19.21

So the actual thermal energy lost from a building with a larger TM will be greater than that from one with a smaller TM, and hence require more thermal energy to be replaced to restore IAT.

As you correctly noted, the Midea unit struggles to produce a higher thermal energy output when operating near maximum operating conditions. My mention that setback can adversely affect even the operation of an Ecodan, which can increase its output near maximum operating conditions, but does so at lower efficiency. Employing a setback which causes the heat pump to move from nominal operating conditions to near maximum, can reduce overall efficiency for no added benefit.

One of the unknown factors is the mode in which the heat pump is being controlled, how the temperature setback is achieved, and how the controller responds to such a setback.

Posted by: @cathoderay

↑

I agree with your thinking that a low thermal mass is perhaps somewhat counter-intuitively likely to work better with a setback

Posted by: @derek-m

↑

One of the benefits of using simulations is the ability to vary one parameter, whilst keeping others constant, to see the effect of that one variable.

Agreed wholeheartedly.  It's often quite difficult in the real world to create conditions that tease out the effects of different factors.  Modelling/simulations allows you to do this. 

Obviously you still need to treat models with care, but that's true also of real world measurements where you haven't got control of key variables, which is the case with central heating systems.