If we're getting R-squareds of 0.9 with just the OAT vs energy consumed, then we're probably not going to be able to discern a 5-10% difference in energy consumption that hypothetically might result from a setback. You need more modelling and more data.
I'm not sure I agree about the modelling, though I do 100% agree about more data. The R square values just tell us there is good correlation, which can also be seen visually, along with the fact it is better at higher OATs than lower OATs (defrost cycles, probably). But the methodology I suggest here is in principle very simple (and definitely not of itself a model, though it does rely on a regression that some might call a model), that just compares observed energy in with setback to expected energy in without setback. Within the measurement limits, we know observed energy in during the setback, all we need to do is estimate the expected energy in value, were the setback was not in place, which we do with the regression equation (and here the high R squared value is an asset), and then compare that to the observed value. Any difference in the two represents energy saved (or additional cost, if it goes the other way).
Midea 14kW (for now...) ASHP heating both building and DHW
@cathoderay@derek-m Health warning: This post doesn't contain any conclusions, just some more observations to trigger thought, pursuant to the overriding issue that the apparent saving cannot, so far as I can see, be explained by the saving in heat lost from the house.
Further to my comments above here is an interesting plot of the hourly data during the 'control' period (I haven't excluded the DHW periods so the correlation isn't quite as good)
Note that
correlation of energy energy in with OAT is MUCH better than the correlation of heating energy out
the latter appears to have an intercept of 2kWh at 20C whereas you would expect it to be 0 or thereabouts at 20C. There is a lot of uncertainty on this number but its definitely looking like a positive intercept. Its a similar intercept if you plot heating energy out vs (IAT-OAT).
So Im thinking two things:
(1) is perhaps fairly easily explained by WC. If the only control loop is the WC control loop then the system is only marginally responsive to IAT and thus the correlation with IAT will be better than the correlation with OAT assuming that the emitters are oversized (so that they just emit everything thrown at them). Oversized emitters are probably not uncommon, and certainly a case worth considering (@derek-m isn't there an emitter capacity parameter in your model).
(2) suggests the possibility of a vampire load. If there is a vampire load then this could be part or most of the apparent saving through setback, particularly in view of (1)
If the vague speculations in (1) and (2) are correct, then the implication of this and the data may be that 'batch heating' is cost effective in the particular circumstances we have here.
OK, I think I may be getting some ideas, its by no means fully there but its going in the right direction, I think.
I think that all the modelling (that Derek-M principally has done) and at least much of the thinking that I have done to date tacitly assumes (for the purpose of comparing setback behaviour with 24*7 behaviour and ensuring that the simulation/thinking ignores cases where comfort is compromised) that the heating system is somehow controlled, whether by the emitter capacity combined with WC, or by some other feedback loop, to achieve the desired IAT. In this case then heating energy to house will be on average proportional to IAT-OAT, and energy lost from the fabric during setback must be replaced. The latter will be manifested by increased energy consumption after the setback period, possibly for quite a long time, either because its forced by some control loop (such as @cathoderay's compensation routine) or because with IAT lower following setback the emitters emit more at a fixed flow temperature.
However what if the tacit assumption is incorrect and, in fact, the way the system works with simply
a fixed WC curve
a fixed flow rate
(Both of which we know to be the case in the @cathoderay system, with the exception of the short forced 'recovery'. Its doubtless also the case in some, possibly many, others.)
and
the emitters are sufficiently large that they just emit all that the heat pump throws at them (either because they are larger than necessary or the WC curve is set a bit low), the system operates differently.
To a good approximation in this case the heat pump outputs an amount of energy which is unresponsive to the IAT and responds only to OAT. The correlation of heating energy in with OAT will be very strong (which we see), the correlation between heating energy and (IAT-OAT) weaker (which again we see). Setback will, in this circumstance, save almost exactly the amount of energy that would have been supplied to the house during the 'off' period. This again is what we see - in the pivot below - the difference between 'actual' and expected' is almost exactly equal to the energy that would have been consumed during setback.
If, in addition, there is a vampire load (a sink for energy which is not delivered to the house), then this will further increase the energy 'saved' during setback without affecting the house. There will always be some vampire load, if only the pump plus any standing consumption which the heat pump itself has when delivering heat. Of course any heat loss from the system outside the insulated envelope is a further vampire load.
So If we work on the principle that the @cathoderay data is from a house where he above applies, what we see in terms of energy consumption is exactly what we get.
This subtly changes the question we need to answer to rationalise the data with the laws of physics. It now becomes, 'why does the house temperature not fall even outside the setback period, as a result of the energy lost from the fabric during setback?' instead of 'why is the consumption during the setback period so much lower?'
I suspect this is a combination of factors
1. It does fall, albeit only by ~0.3C on average which is not alone enough to explain the behaviour
2. There is some evidence of a significant vampire load in addition to the pump. The vampire load does not contribute to the warming of the house
3. The setback experiment was done at a fairly mild OAT, around 10C (possibly higher as it is said that the OAT sensor in the system in question reports lower values than actual). At mild temperatures the heating from electrical equipment and occupants becomes significant, so house temperatures don't fall as much as would otherwise be expected if heating is switched off.
4. Some other mitigating factors that I haven't yet thought about (for example secondary heating, although we don't know if there is any)
It would be good to get some info from @cathoderay on possible vampire loads and secondary heating sources. This might help us understand whether together they could account for the observations, or whether there are other factors still to be discovered.
As I say this doesnt reconcile the figures yet, but it does perhaps subtly modify the way to think about it.
This post was modified 1 year ago 4 times by JamesPa
4kW peak of solar PV since 2011; EV and a 1930s house which has been partially renovated to improve its efficiency. 7kW Vaillant heat pump.
@jamespa - all very interesting, with much to think about. The sun is shinning and I have other things to do today, but I will think about things as and when I can, and reply later.
Midea 14kW (for now...) ASHP heating both building and DHW
CathodeRay stated that the total capacity of his heat emitters is 23kW @ DT50, which he did state was limited by room and location, and the calculated heat loss is 12.4kW, though having run some data through my spreadsheet it would appear to be on the high side.
At the moment I have only had a chance to put one set of data into the spreadsheet, the 24 hour period from 9pm on the 5th November to 9pm on the 6th November.
I would prefer to check more dates and OAT's before posting any findings.
CathodeRay stated that the total capacity of his heat emitters is 23kW @ DT50, which he did state was limited by room and location, and the calculated heat loss is 12.4kW, though having run some data through my spreadsheet it would appear to be on the high side.
At the moment I have only had a chance to put one set of data into the spreadsheet, the 24 hour period from 9pm on the 5th November to 9pm on the 6th November.
I would prefer to check more dates and OAT's before posting any findings.
We are all doing this part time!
Scatter plots of IAT vs heating energy out and IAT - OAT vs heating energy out suggest 8.5kW at -2/20 and with a vampire load (apparent heating requirement at IAT=OAT or OAT=20) of nearly 2kW, a figure which you might expect to be either small or even negative due to other heating sources. There are some things in the data which are challenging to explain!
4kW peak of solar PV since 2011; EV and a 1930s house which has been partially renovated to improve its efficiency. 7kW Vaillant heat pump.
With reference to the now christened Vampire load, on my Viessmann i have one meter monitoring all the energy to the entire heating system, the other is part of the heat pump setup these give a consistent but different reading. Very old school here so every morning about 7am i take a manual reading of both. If the heat pump is on 24 hours the difference is 2.6KWh or 0.11 per hour, when i was experimenting with 2x 6 hour setbacks in spring it was 1.4KWh, so i am confidant the 2 meters are reading the same, and my pump uses 0.11KWh for each hour its running, that isn't used for energy production. So in a 8 hour setback i save 0.88KWh.
Initially i used the room stats as a on off to introduce setback, however the Viessmann service engineer told me that wasn't the best way to run their system. It seems my Viessmann is like a sprinter in the set position waiting for the gun to go, it keeps its loop of water, at the WC curve required temperature expecting to "go" at any minute, when this loop cools it reheats it thus using energy. The best way for me to introduce setback was to use the Viessmann controller, which stops the heating but still allows it to defrost if required. I realize the heating produced wasn't all lost, as it was mostly in the fabric of the building,however the hourly stop start wasn't the most efficient.
CathodeRay stated that the total capacity of his heat emitters is 23kW @ DT50, which he did state was limited by room and location, and the calculated heat loss is 12.4kW, though having run some data through my spreadsheet it would appear to be on the high side.
At the moment I have only had a chance to put one set of data into the spreadsheet, the 24 hour period from 9pm on the 5th November to 9pm on the 6th November.
I would prefer to check more dates and OAT's before posting any findings.
We are all doing this part time!
Scatter plots of IAT vs heating energy out and IAT - OAT vs heating energy out suggest 8.5kW at -2/20 and with a vampire load (apparent heating requirement at IAT=OAT or OAT=20) of nearly 2kW, a figure which you might expect to be either small or even negative due to other heating sources. There are some things in the data which are challenging to explain!
A statistical approach is not really used in the field of Instrumentation and Control Systems, since the primary purpose is to avoid deviation, not have a standard one. As an Engineer I am more interested in cause and effect, since this is more often than not used in identifying the root cause of any problem scenario.
Please don't let me stop others from progressing the problems from different angles.
Have any of the following aspects been accommodated within the statistical approach?
From the raw data I suspect that the OAT reading is negatively affected by the operation of the heat pump, the affect increasing with heat pump loading, meaning that the OAT sensor appears to be reading artificially low by up to several degrees.
The inclusion of a PHE in the system, is artificially increasing the required LWT, which in turn will negatively impact the system efficiency. The temperature drop across the PHE may also vary with heat pump loading, which again may affect both the statistical approach as well as any simulation method.
Because of the above unknowns, I would suggest that checks should be carried out to ascertain how accurate the OAT sensor reading is when compared to the actual OAT, and also test for temperature variations across the PHE during different operating conditions.
CathodeRay stated that the total capacity of his heat emitters is 23kW @ DT50, which he did state was limited by room and location, and the calculated heat loss is 12.4kW, though having run some data through my spreadsheet it would appear to be on the high side.
Not quite right: the available wall space was a significant constraint (no huge rads, allowing a lower LWT), but by using K3s and a high LWT, I could select rads that would fit and more than just matched the predicted heat loss at design (-2 OAT) temps. The relevant figures are 12.7Kw heat loss (based on my calculation, not Freedom's spreadsheet), total nominal output 23.3kW at delta t50, or 13.6Kw at design delta t (average 32). There was a deliberate slight over-sizing of the rads (next size up rather than down when exact room loss fell between two rads sort of thing), the reason my house falls below design IAT in low temps around zero OAT is because it turns out a 14kW Midea heat pump can only produce 11.3kW max in the these conditions, and on average produces less, defrost cycles etc. See posts passim on minute data calculated energy out, and this image: it does occasionally get to about 11, but spends much of its time at considerably lower outputs; note also the cycle in this chart is a defrost, LWT goes lower than RWT making delta LWT/RWT negative and turning the heat pump into an energy black hole where negative energy exists, or, as I prefer to say, a heat thief, because it steals energy from the house:
@jamespa, I'm on the case, just haven't got the answers yet, though not seen any vampires recently.
Midea 14kW (for now...) ASHP heating both building and DHW
CathodeRay stated that the total capacity of his heat emitters is 23kW @ DT50, which he did state was limited by room and location, and the calculated heat loss is 12.4kW, though having run some data through my spreadsheet it would appear to be on the high side.
Not quite right: the available wall space was a significant constraint (no huge rads, allowing a lower LWT), but by using K3s and a high LWT, I could select rads that would fit and more than just matched the predicted heat loss at design (-2 OAT) temps. The relevant figures are 12.7Kw heat loss (based on my calculation, not Freedom's spreadsheet), total nominal output 23.3kW at delta t50, or 13.6Kw at design delta t (average 32). There was a deliberate slight over-sizing of the rads (next size up rather than down when exact room loss fell between two rads sort of thing), the reason my house falls below design IAT in low temps around zero OAT is because it turns out a 14kW Midea heat pump can only produce 11.3kW max in the these conditions, and on average produces less, defrost cycles etc. See posts passim on minute data calculated energy out, and this image: it does occasionally get to about 11, but spends much of its time at considerably lower outputs; note also the cycle in this chart is a defrost, LWT goes lower than RWT making delta LWT/RWT negative and turning the heat pump into an energy black hole where negative energy exists, or, as I prefer to say, a heat thief, because it steals energy from the house:
@jamespa, I'm on the case, just haven't got the answers yet, though not seen any vampires recently.
What you appear to be failing to grasp, is that to get 13.6kW of thermal energy to your radiators, via the PHE, the LWT at the heat pump would probably need to be 60C, with a RWT of 55C. Not very good from the efficiency point of view.
If the heat pump LWT is 55C with a RWT of 50, the temperature at the radiators will probably only be 50C and 45C respectively, which will limit their output to approximately 11kW. That is probably the reason why you cannot get much more than 11kW out of your heat pump.
What you appear to be failing to grasp, is that to get 13.6kW of thermal energy to your radiators, via the PHE, the LWT at the heat pump would probably need to be 60C, with a RWT of 55C. Not very good from the efficiency point of view.
I absolute do get it, and have done so for a time, but only with hindsight. At design/installation time the installer and I hadn't seen the Midea engineering data, and didn't know about the lower outputs. We only had the Freedom spreadsheet, which over-estimated output, which made it look as though it would be OK, albeit very tight with no headroom. There was also the imminent grant deadline, and the fact a 16kW unit would not be available until long after the grant deadline passed, meaning no grant, and that would have meant no heat pump. I also didn't know anything about a PHE being part of the design until I saw it in place. At the time of the installation, Freedom made them mandatory (no PHE, no warranty), and my installer fitted them as standard. I have since learnt what the effects of a PHE (which aren't all bad, but they do interfere with heat transfer).
Hindsight is a wonderful thing.
Midea 14kW (for now...) ASHP heating both building and DHW
It would be good to get some info from @cathoderay on possible vampire loads and secondary heating sources. This might help us understand whether together they could account for the observations, or whether there are other factors still to be discovered.
The basic answer to this question (is there a vampire load and/or secondary heating source) is relatively straightforward. As far as I can tell, there is no vampire load, except possibly (a) the PHE, but surely any heat lost from the PHE heats the house, as it is in the airing cupboard, or maybe it becomes a vampire in some other way and (b) the pipework between the heat pump and the heated part of the house. This pipework runs from the back of the heat pump to where it passes through the wall (~1.5m run) into a mostly unheated utility room and then up through an unheated attic to the first floor landing and then the airing cupboard where it connects to the PHE. The total one way run including the ~1.5m outside bit is around 8 - 10m and it is insulated throughout with pipe lagging. It will lose heat that doesn't end up in the house, but surely not a lot.
Having thought about it, and looking again at the energy out correlation plots, it seems to me that the correlation is very poor, and maybe the explanation for the 20 degrees OAT / 2 kWh out 'intercept' is just an artefact of poor data. There are three things that go into determining the energy out, the specific heat of the circulating fluid (a constant), the flow rate (effectively in my Midea system a constant, though it does vary a bit), and the LWT/RWT delta t. This last factor is volatile, and can at times eg during defrost cycles go negative, thereby introducing us to the novel concepts of negative energy and heat thieves - and possibly even vampires? The control period used (28 Nov - 9 Dec) for the energy out correlation plots does include a period when the OAT was low, and there were frequent defrost cycles (and IAT falls a bit due to a combination of heat pump dyscopia (CQC) and the defrost cycles themselves, so for a brief period the house wasn't in energy balance):
Bottom line: is the volatility of the LWT/RWT delta t causing artefacts sufficient to mean the calculated energy has little direct correlation with OAT, and therefore any inferences from the energy out vs OAT plot (such as the 20 degree OAT / 2 kWh out 'intercept') are suspect? Just because a 'dumb' spreadsheet can produce scatter plot with a trend line and equation for that line doesn't guarantee that equation has any real world meaning. We as 'smart' humans have to decide whether that is the case or not, and I for one think the residuals may look a bit dodgy:
Technical note: residuals are the difference between an actual value and its predicted value, and a plot of them gives (a) a visual representation of goodness of fit, in effect a visual version of R squared, and (b) an opportunity to spot bias (systemic as opposed to random error, there are as many potential causes of bias as you care to think of), which is typically apparent as a non-symmetrical distribution of the residuals about zero across the whole range of values.
Midea 14kW (for now...) ASHP heating both building and DHW
Having thought about it, and looking again at the energy out correlation plots, it seems to me that the correlation is very poor, and maybe the explanation for the 20 degrees OAT / 2 kWh out 'intercept' is just an artefact of poor data
I would have reached the same conclusion except that the scattergram of corrected IAT-OAT vs energy out (corrected for the thermal mass of the fabric), which I previously posted, has a similar intercept. This is also the scattergram which shows the best correlation of all of the correlated data points - a fact that's rather encouraging because its the one that has the most solid underpinning in physics. So I am pretty convinced that the intercept is real. As to how it should be interpreted, well that's another matter!
We need to know more about the PHE. Can you tell us anything about temperature differences between the ports? Also, I'm wondering what the quiescent consumption of the heat pump is, and whether its included in the figures that come out of it (something, after all, is accounting for a factor of 1.18).
There has to be something which explains why you can get away with (apparently) dumping 20% less energy into the house for a few days and it only cools on average by a fraction of a degree. So far we don't have a plausible explanation so we need to keep looking. Otherwise any conclusion we might attempt to reach from the experimental results applies only to your specific system in the specific conditions of the experiment, which is not a lot of use to anyone other than you.
... the LWT/RWT delta t. This last factor is volatile, and can at times eg during defrost cycles go negative, thereby introducing us to the novel concepts of negative energy and heat thieves - and possibly even vampires
Heat for defrost does indeed come from the house, which is why I counted both positive and negative values when correlating IAT-OAT vs energy out.
4kW peak of solar PV since 2011; EV and a 1930s house which has been partially renovated to improve its efficiency. 7kW Vaillant heat pump.
Thinking about installing a heat pump but unsure where to start? Already have one but it’s not performing as expected? Or are you locked in a frustrating dispute with an installer or manufacturer? We’re here to help.
