Posted by: @josephiah

↑

will see what happens with some different settings on different days...

This is absolutely the way to go, and the only way to get real world answers. 

Your sudden jump at 0700 (which I think may refer to 0600 to 0700, depends when you setback ends) is interesting in that it is such a big jump, around three times (or 300% of) the steady state level, and it lasts for four hours, in other words, it is huge recovery boost. I wonder what causes that? Some logic in the controller? I very much doubt the weather compensation curve would produce that pattern, if you are running weather compensation, If it is timed fixed LWTs, then the early morning ones may be higher than they need to be. You might be able to get a clearer picture of what is going on if you can overlay the outside and indoor temps on the energy use chart, or plot them above the energy chart, as I have taken to doing. Even if produced in totally different ways, you could arrange the plots to a convenient layout on any image editor.

In contrast, my recovery boosts are almost invisible by comparison, usually only an hour or two, and the maximum increase above the baseline level is around 50% (cf your 300%), this boost being the result of my default weather compensation curve as adapted by my auto-adaption script.

As a caution to the unwary, here is my last 48 hours data. Apparently the system went ballistic at 1800 yesterday, and if the chart is to be believed, I should be well on the way to the moon by now. But I'm still here:

I think the clue is in the actual value for the Set LWT, it is 255, a number not without significance in the computing world, being the top of the 0 to 255 range, that being the range of decimal numbers that can be represented by 8 bits (as in for example the 8 bits per channel in RGB colours). Here is the chart with the Set LWT removed: 

Last night's setback and recovery has followed the relatively mild OAT temp pattern, that is, minimal boost because the WCC has set the LWT lowish because of the mild OAT, and my auto-adaption script hasn't taken this into account, and has only added a small boost, and as a result, the recovery has been slow (room temp 17.7 at just before 0900, still a little way to go after six hours of recovery). I'm going to tweak the code to deal with this, slightly bigger boosts when the OAT is mild (the counter-intuitive thing I mentioned previously).

More pressing in the meantime is the problem with the DHW, failing to heat when it should. I have now confirmed that it is the auto-adaption script that causes this, by writing to the modbus registers. Although this is being done automatically by a computer script, it seems the wired controller sees it as being exactly as if I had manually gone into the FOR SERVICEMAN menu, set the WCC end points, and exited the FOR SERVICEMAN menu; and, for some bizarre reason, doing that turns the system off (the Midea technical manual confirms this observed behaviour is indeed the default behaviour). I have been caught out by this before, doing a manual WCC change in the early days, and then forgetting to turn the system back on again, with the result that the DHW failed to reheat. I need to add some code that turns the DHW back on again after a register write, but that part of the modbus register uses single bits (1 or 0, DHW on or off) and it looks like I am going to have to do some bitwise maths to get things working, and bitwise maths is not for the faint-hearted...       

Posted by: @cathoderay

↑

in other words, it is huge recovery boost.

Although, roughly speaking, the overshoot hill (above the average line) fits the size of the of the preceding valley (exactly, in fact, over the course of the whole 24hr period 😉 ). So it doesn't seem inappropriate.

But what I am doing now is reducing the delta between the daytime and offset temps; expecting to see it tick over a little higher overnight, then have less of a recovery hill (i.e. smooth out the valley/hill).

Yes, I am looking at temperature, but, despite the system measuring the OAT, it doesn't log it in an accessible way (grr), so unless/until I get myself a separate temp logger I'm having to go off the nearest weather station (which is a good few miles away and lower altitude).

Our system is using both weather comp and room comp, and so far its behaviour has looked pretty sensible to me, modulating temperatures nicely. The occasional software gremlins aside (which aren't their fault), I'd say our installers have firmly bucked the stereotypical trend...

A quick look at the frost outside and the consumption figures morning suggests we are into our first spell of defrost-cycle territory 🫣

Posted by: @josephiah

↑

Although, roughly speaking, the overshoot hill (above the average line) fits the size of the of the preceding valley (exactly, in fact, over the course of the whole 24hr period ). So it doesn't seem inappropriate.

But we aren't trying to back fill the preceding valley, so to speak. Consider setbacks of varying lengths, let's say four hours and eight hours (set by changing start time of setback, same end time), and in both cases the nadir IAT at the end of the setback is the same. If we had to back fill the preceding valley with the recovery boost, then the eight hour setback would need twice the extra energy needed by the four hour setback, and I don't think that is the case. Instead, the primary determinant of the extra energy needed is the current IAT at the end of the setback, or more accurately the difference between the current IAT and the desired IAT. If they are the same, no boost is need, if they are far apart, a big boost is needed etc.

If your data is on and presented by an app, they can do weird things, as they are the manufacturer marking their own homework, plenty of scope for the heat pump equivalent of dieselgate . I always try to verify app data (did so with the Midea app, and as a result, ditched it because it didn't make any sense). If you have a dedicated kWh meter for the heat pump, taking hourly readings at relevant times will give a clear indication of what you are really using, if not, your main kWh meter usage will be mostly the heat pump when it is on, assuming you don't have other power hungry devices, and it should at least give you an idea of whether the Vaillant figures are credible.

You can get cheap data loggers that will record hourly temp data that you can download and plot to your hearts content (google rc-4 data logger). I used one of these for my IAT before I had the modbus sensor, not as elegant as using modbus but it does get you the data. You might also be able to put a second one in a protected location outside, to get the OAT.   

I too am awaiting the arrival of the season of defrost cycles with interest. Overall energy use will certainly skyrocket, but it will be the extra energy needed during the recovery boost that I will focus on.      

Posted by: @cathoderay

↑

But we aren't trying to back fill the preceding valley, so to speak.

Agreed - my analysis is over-simplistic so far - we should expect to reduce consumption purely because the average temperature differential to outside is lower by using a setback. But then we should also expect to lose some of that saving due to reduced efficiency of the heat pump during setback recovery. The question is: which of those wins.

What I was trying to get across (poorly) is that the average across the 24hr period didn't look too different from the kind of 'level pegging' rates in the non-recovery periods. If there had been a wild difference one way or the other here, that might have told us something interesting; as it is, the changes are too subtle for my crude datasets at present, and I'll need to improve those to draw any finer conclusions.

And then of course, we have comfort considerations, which aren't always aligned with efficiency considerations. On the whole though, I'm pretty happy with our current setup, and I suspect I'm getting into marginal gains territory. All to be revised in due course once we get our PV/battery/ToU tariff...

Posted by: @josephiah

↑

we should expect to reduce consumption purely because the average temperature differential to outside is lower by using a setback. But then we should also expect to lose some of that saving due to reduced efficiency of the heat pump during setback recovery. The question is: which of those wins.

"Which of those wins" - that is exactly the question that I am trying here to answer empirically. It is fair to say that a number of opinions have been expressed in a thousand or more posts here on this forum and elsewhere on which one wins, and they most emphatically do not come to the same conclusions. My own reading of the theory (or rather the laws of thermodynamics) is that, all other things like OATs remaining constant, then a setback must reduce overall energy consumption, because, to use words rather than formulae, the summed delta t between the IAT and OAT over a 24 hour period will be less. Put very simply, at a given OAT, you use more energy to keep the house at an IAT of 20 than you do if you keep it at 20 for 12 hours and 15 for the other twelve hours.

The complication is the extra energy needed, over and above what you would have used anyway, to raise the IAT up from 15 to 20 degrees. We can see this 'extra energy' concept when, for example, we boil and then simmer a pan of water on the stove: the period of raising the temp to boiling point needs a higher setting on the stove than it needs to maintain the simmer once it has got there. Now it is probably very simple to calculate that extra energy to get to the boil for one litre of water in a pan on the stove, and then compare that to the amount needed to keep it on the simmer, but to do the same thing for a house with a probably unique set of characteristics that sits in a constantly changing environment is all but impossible, hence my suggestion that they only way we can get a reasonable estimate of that extra energy is empirically, by observing how much energy is actually used.

BTB, I think perceived comfort and warmth has a lot to do with how we are warmed, with radiant warmth always making us feel warmer than convected/conducted warmth. A given radiator can warm you in two ways: it can warm the room air by conduction, and that warmer air then moves by convection to warm you by conduction, because you are now in warmer air, or it can (also), if it is warm enough, and you are near enough, warm you by radiation, and that always feels warmer/more comfortable. It chiefly how we feel warm basking in the sun, its why we all love log fires (red faces and cold ankles notwithstanding), and it is how those far infrared heaters claim to work, they warm you, not the room. The relevance of all this to heat pumps is that for much of the time the rads are not hot enough to give a useful radiant effect, but during a boost in colder OATs they do start to be hot enough to radiate discernable heat, eg when I walk past my kitchen radiator at such a time, I can feel the warmth, and it feels good!          

I have now determined that the 1800 run of the auto-adapt script yesterday somehow turned off the WCC, and put the heat pump into fixed flow temp mode at 40 degrees LWT, as can be seen in the chart below, where the actual LWT zig zags between 35 and 45 degrees to average 40 degrees, with nary a variation let alone a boost in sight, while the Set LWT is at 255. I have now turned WCC control back on and as can be seen on the extreme right, the Set LWT is now back in sane territory:

I have also learnt that the modbus register that holds the Set LWT actually holds the 'hydraulic module curve T1S calculated value 1', ie the calculated LWT based on the WCC. When WC is off, this value apparently isn't calculated, and drifts off to some arbitrary and irrelevant value.  

Another possibly (or possibly not) interesting insight. I have now changed (in the data file) the Set LWT during the non WC period to what it actually was, 40 degrees, rather than the thar' she blows 255 degrees. The 'interesting insight' is that even with no WCC change in the Set LWT, and no auto-adaptive boost, ie the Set LWT is a constant 40 degrees before during and after the setback, the first hour of the recovery does use more energy, in effect providing  a measure of the extra energy needed to get the heat pump started from cold. It is not very much at all. The red bar, energy out, is noticeably taller, but the energy in (green bar) is only marginally increased, at 1.19 kWh in the first hour of recovery, compared to around 1.1 kWh in the hours preceding the setback, and around 1.12 kWh in hours 2 3 and 4 of the recovery: 

Maybe cycling isn't as ruinous as we generally suppose.    

@cathoderay

I fully agree that the heat loss will reduce as the IAT reduces at a constant OAT.

If the LWT is now returned to the value it was prior to setback being initiated, the IAT will start to increase because the heat pump is providing more thermal energy than the present heat loss. The increase in IAT will initially be more rapid, but this rate will slow as the IAT increases and with it the heat loss, so for the system to return to the original IAT may take quite a number of hours if not longer. This method will provide the maximum energy saving.

To recover the IAT more rapidly obviously requires setting the LWT to a value higher than its original setting. More thermal energy will therefore be put into the home in the same time period, but this reduces the heat pump efficiency, so obviously uses more electrical energy.

Modeling clearly shows that a lengthy setback (several hours) with a very lengthy recovery (20 hours or more) can provide a reduction in energy consumption approaching 20%, but achieving a faster recovery by raising the LWT, quickly lowers the energy reduction down to single figures. For true energy saving measurement the IAT would need to be returned back to its original value pre-setback value, since recovering the last 0.2C or even 0.1C will further lower the energy saving that is being accomplished.

Posted by: @cathoderay

↑

The complication is the extra energy needed...

Agree with just about everything you've written there, but I think you may be missing a point (although you may well have covered this elsewhere) about the reduced efficiency of the heat pump during the recovery period.  This is why a smaller setback may work out better in terms of efficiency than a larger one, despite the larger average differential between IAT and OAT.  Finding that crossover point (which will be different with every house/install, and no doubt varies with weather/temp conditions!) is the interesting part.

Also agree on the "how we perceive different types of heating" part.  I'm edging towards our optimum efficiency/comfort setup being probably a very small setback (say 1°C) - enough to turn the system off for a decent stretch overnight and also provide that artificial trigger first thing to address that "warm radiator" comfort effect/preference in the mornings.

Posted by: @josephiah

↑

I think you may be missing a point (although you may well have covered this elsewhere) about the reduced efficiency of the heat pump during the recovery period.

I have in fact covered this elsewhere, things do get rather scattered all over the place, and @derek-m also covers it above. It is one of the factors that makes modelling, as opposed to measuring, the extra energy use so very complicated. @derek-m will correct me if I am wrong, but to model it, you have to use the manufacturer's efficiency data for the the different permutations and combinations, tricky enough of itself, and furthermore you don't actually know whether the manufacturer's data - no doubt obtained under ideal laboratory conditions - actually applies in the real word. For example, perhaps they used water in the system and you have some glycol. And so on, and so forth.

This is why I plug the use of observed/empirical data. In practice, there is no whatiffery about an observed value, and the only realistic source of error is measurement error (which is not to be dismissed, which is why I verify measurement sources against each other, see below). That said, one complication remains: since we are interested in determining the extra energy used in a recovery boost over and above that which would have been used anyway, we need to estimate what that would have been used anyway figure is. This is not a hopeless task, we can use, for example, observations from periods when setback plus recovery boost was not in use, and - good grief, what has become of me! - use some lightweight whatiffery (what if the setback and recovery boost hadn't been applied?) to get a reasonable estimate of that would have been used anyway figure.

I am currently verifying my calculated energy in figures against the external heat pump kWh meter, and it looks as though my calculated figures are a bit lower than they should be, very possibly because they are derived only from the energy calculation parameters (amps x volts) and may not include standing energy use, circulating pumps (x2) use etc. It all depends on where Midea takes the amps and volts readings: at the compressor, or where the mains supply cable enters the heat pump enclosure. I suppose there are also legitimate questions as to what should and shouldn't count as part of the energy in - my preference would be for all energy used by the heat pump to be included, whatever it is used for.

Early indications are that my calculated daily energy in values are consistently close to 80% of the external heat pump meter values. Establishing consistency is important: it it persists, then I can add an empirically determined correction factor to my calculated figure, and be reasonably sure that corrected figure is close to the real one. If that won't wash, then I do still have the option of fitting a modbus enabled external kWh meter, but that means getting and fitting the meter, which I would prefer to avoid if at all possible. I don't mind working on a 13 amp plug, but working near the main distribution board gets a bit hairy for me. It's another of those 'what could possibly go wrong?' situations.       

Posted by: @josephiah

↑

Posted by: @cathoderay

↑

The complication is the extra energy needed...

Agree with just about everything you've written there, but I think you may be missing a point (although you may well have covered this elsewhere) about the reduced efficiency of the heat pump during the recovery period.  This is why a smaller setback may work out better in terms of efficiency than a larger one, despite the larger average differential between IAT and OAT.  Finding that crossover point (which will be different with every house/install, and no doubt varies with weather/temp conditions!) is the interesting part.

Also agree on the "how we perceive different types of heating" part.  I'm edging towards our optimum efficiency/comfort setup being probably a very small setback (say 1°C) - enough to turn the system off for a decent stretch overnight and also provide that artificial trigger first thing to address that "warm radiator" comfort effect/preference in the mornings.

A further factor to consider is how the setback is achieved. If it is by using the WC offset, it may just be reducing the LWT by the offset, and raising it back to the original value when the offset is removed.

If the system is operating in an Ecodan AA type mode, then lowering the setpoint by 1C may cause several degrees change in LWT, with the reverse happening when the setpoint is returned to its original value. This could have a much more marked effect both on the time period required to recover, the quantity of electrical energy used during recovery, and any energy consumption reduction that may have been achieved.

A further major factor is the thermal mass of the home, since this will dictate how low the IAT could potentially fall over a given time period at a particular OAT, and of course how long and how much electrical energy will be required to return the IAT to the desired level.

@derek-m - I agree, see above, the efficiency loss caused by raising the LWT matters, but I still think empirical data is the best way to determine how much it matters.

In the meantime, I may need to ask for your help on doing some bitwise maths. Having determined that setting the WCC endpoints over modbus turns the system, or at least the DHW part of the system, off, I need a way to turn it back on again, and this is done in register 0, which uses bits for the data, specifically bit 2 (of 16 bits in all, numbered 0 to 15 so bit 2 is the third bit) which is o for DHW off, 1 for DHW on. I have a test script that does this, as an initial attempt to read the data before I write to it:

reg0 = instrument.read_register(0)
print(f'Register0={reg0}')
reg0bit0 = (reg0 >> 0) & 0x0
print(f'Register0 bit 0={reg0bit0}') 
reg0bit1 = (reg0 >> 1) & 0x1
print(f'Register0 bit 1={reg0bit1}') 
reg0bit2 = (reg0 >> 2) & 0x2
print(f'Register0 bit 2={reg0bit2}') 
reg0bit3 = (reg0 >> 3) & 0x3
print(f'Register0 bit 3={reg0bit3}')

and the result (when heating is on and DHW off) is 

Register0=2
Register0 bit 0=0
Register0 bit 1=1
Register0 bit 2=0
Register0 bit 3=0

which sort of makes sense (bit 1 is the heating on/off bit, and 0b0010 is decimal 2) but when I manually turn the DHW on (which turns the heating off) I get something like this:

Register0=6
Register0 bit 0=0
Register0 bit 1=1
Register0 bit 2=0
Register0 bit 3=0

which doesn't make any sense to me. Bit 1 should be 0 (off) and bit 2 should be 1 (on), and where on earth has the decimal 6 come from?

I have minimalmodbus running in debug mode and the request/response entries for the above are, respectively:

MinimalModbus debug mode. Will write to instrument (expecting 7 bytes back): 01 03 00 00 00 01 84 0A (8 bytes)                                                                               
MinimalModbus debug mode. Clearing serial buffers for port /dev/ttyUSB0                                                                                                                      
MinimalModbus debug mode. No sleep required before write. Time since previous read: 927230778.75 ms, minimum silent period: 4.01 ms.                                                         
MinimalModbus debug mode. Response from instrument: 01 03 02 00 02 39 85 (7 bytes), roundtrip time: 0.0 ms. Timeout for reading: 50.0 ms.

and 

MinimalModbus debug mode. Will write to instrument (expecting 7 bytes back): 01 03 00 00 00 01 84 0A (8 bytes)
MinimalModbus debug mode. Clearing serial buffers for port /dev/ttyUSB0
MinimalModbus debug mode. No sleep required before write. Time since previous read: 935301814.40 ms, minimum silent period: 4.01 ms.
MinimalModbus debug mode. Response from instrument: 01 03 02 00 06 38 46 (7 bytes), roundtrip time: 0.0 ms. Timeout for reading: 50.0 ms.

I can see the 2 and the 6 in the responses, but that is about it...

Any thoughts on how I might make sense of all this?