There has indeed been a lot of discussion about whether setbacks save money and/or compromise comfort, including this mega-thread. The answer is not straight forward, with theory and empirical results seemingly at odds at times.
All can agree that a long setback eg going on holiday for two weeks in winter and leaving the house with just frost protection on eg main thermostat set to 10 degrees (which is still a setback, albeit a big one) will save money, but doesn't compromise comfort, as you are not at home.
The same applies to a one week frost protection setback. The saving is more than large enough to offset any extra energy needed during the initial reheat, and a net saving is made.
The same applies to shorter and smaller setbacks until at some point, things start to get muddy, typically when we get to the most common setback, an over-night one of a few degrees. Rather curiously, no one has managed to work out when the obvious savings become less obvious, but suffice it to say we have got there when we get to over-night setbacks.
On the face of it, a setback should save money, because the heating is off for several hours and the average temperature of the house is lower over the 24 hour period. The problem is the house does use more energy during the recovery period, and the question becomes how much extra energy over and above what it would have used if on continuous running does the house use during this recovery boost? Is it enough to wipe out the known real gains during the setback itself?
I think it is fair to say the theory has failed to answer this one, except in the broadest brush terms, with high level statements about the conservation of energy etc. What happens at the individual house level is the theory fails because there are just too many variables, and the models can't cope with that level of complexity. Remember the real question we need to answer is how much extra energy did the house use during the reheat, and is this more or less than that saved during the setback?
The alternative to modelling the answer (an approach I have dubbed whatiffery, in fact whatiffery is any model that asks what if questions eg weather forecasts: what if the Atlantic low tracks east rather than north east?) is to measure the data, ie get an empirical answer. If you have a reasonably comprehensive verified monitoring system as I do, then I can plot the key parameter, hourly energy consumed, over time and see what actually happens. Here is a typical 7 day period from spring this year with over-night setbacks triggered by taking the main room stat down to 16 degrees for six hours between 2100 and 0300. For the rest of the time the heating is always on in weather compensation mode, crucially with a DIY script that boosts output when the actual room temp is below the desired room temp. This script is what achieves the recovery, without it, the house takes too long to recover, and comfort is compromised:
This is my standard monitoring chart which includes a lot of variables. The ones we are interested in are the outside air temp, the OAT, which shows a fairly steady pattern over the period (which helps, less variation), with an average around 10 degrees, the indoor air temp, the IAT, which can be seen to vary as expected, and, in the green bars in the lower chart, the energy consumed in the previous hour. I can confirm that during this week, the house was comfortable to live in.
On most mornings, there is an increase in energy use immediately after the setback. I can easily determine how much energy I actually used during the period by summing the hourly energy use. Now comes the tricky bit: is that extra energy more or less than that saved during the setback? Here I have to do some whatiffery of my own: what if the setback hadn't happened? How much energy would I have used?
As it happens, my energy use in standard weather compensation mode is, unsurprisingly, very closely tied to the OAT. I can use that to determine what the energy use would have been with no setback, ie running continuously in weather compensation mode, by doing a regression of the hourly energy in on the OAT for the hours when the heating was on, and then use the regression equation to predict what the hourly energy use would have been for each and every hour during the whole period. As a sanity check, I have plotted the actual measured energy use with the setback against the predicted (by the regression equation) energy use without the setback:
In a way, this chart really is the defining chart of this argument (it should really be a bar chart, but the result is too messy, so I used lines). It shows the setbacks and subsequent recovery peaks in the blue line, and the predicted hourly use without the setback in the orange line. We can see the predictions make sense: when away from the setback periods, the orange predicted line is very close to the blue actual line, with a bit of smoothing, which is to be expected because we used a single regression based equation throughout. The occasional middle of the day drops in the actual blue line use, if you are wondering, are the hours when the DHW heating was on, so less space heating that hour.
And now for the answer to the question 'did the setback save money?'. The answer is that in my house, and in this week, yes it did, quite a bit in fact. The actual measured use (sum of the hourly values) for the week was 164kWh, while the predicted use (again, sum of the hourly values), had I not had a setback, came out at 227kWh. The setback saved me 63kWh that week, which is a saving of 28% for that week.
Note the italics in the previous paragraph, because that is where the arguments start. This is a classic n=1 study, in this case in a small old leaky listed building in southern England with a relatively high thermal mass for its size during a period of middle of the road OATs. To what extent, if any, can these findings be generalised to other buildings and other settings? Over to you...
re-read this a few times and have to say a superb post 👍
All can agree that a long setback eg going on holiday for two weeks in winter and leaving the house with just frost protection on eg main thermostat set to 10 degrees (which is still a setback, albeit a big one) will save money, but doesn't compromise comfort, as you are not at home.
The same applies to a one week frost protection setback. The saving is more than large enough to offset any extra energy needed during the initial reheat, and a net saving is made.
The same applies to shorter and smaller setbacks until at some point, things start to get muddy, typically when we get to the most common setback, an over-night one of a few degrees. Rather curiously, no one has managed to work out when the obvious savings become less obvious, but suffice it to say we have got there when we get to over-night setbacks.
On the face of it, a setback should save money, because the heating is off for several hours and the average temperature of the house is lower over the 24 hour period. The problem is the house does use more energy during the recovery period, and the question becomes how much extra energy over and above what it would have used if on continuous running does the house use during this recovery boost? Is it enough to wipe out the known real gains during the setback itself?
I think it is fair to say the theory has failed to answer this one, except in the broadest brush terms, with high level statements about the conservation of energy etc. What happens at the individual house level is the theory fails because there are just too many variables, and the models can't cope with that level of complexity. Remember the real question we need to answer is how much extra energy did the house use during the reheat, and is this more or less than that saved during the setback?
All agreed, with the possible comment that theoretical models are (currently) invariably based on well adjusted/designed heat pump systems, and we know from other evidence that a fair proportion of heat pump systems are not well adjusted/designed.
Understanding poorly adjusted/designed heat pump systems theoretically is going to be difficult for some while. However there is currently no reason to suppose that they will act consistently, so you cant reliably extrapolate from one poorly adjusted/designed system to another or indeed from a theory based on a well adjusted/designed heat pump system to an arbitrary poorly adjusted/designed one.
To what extent, if any, can these findings be generalised to other buildings and other settings? Over to you...
So far as I can see they cant.
In broad terms (so far as I understand your post and diagrams) you are seeing a 28% reduction in daily energy use for a roughly 50% reduction in IAT-OAT (~19-10 vs 16-10) for 25% (6hrs) of the day, The latter translates to a reduction in daily heat loss from the house of 50%*25% =12.5%.
So far as I can see it simply cannot be the case, unless conservation of energy does not apply to your part of the world which neither you nor I believe, that your reduction in energy is wholly due to the reduced loss from the house resulting from lower average temperature. Other factors must be at work, which will vary significantly from system to system. These may include (but are not limited to:
standing energy consumption due to pumps etc
some factor associated with your 'predicted' energy use (which you describe as what-iffery)
some non-linearity/inefficiency of your heating system eg oversizing
I am not for a minute disputing your experimental results, they are doubtless true. However the 'simple' explanation (which would be the basis on which the result could be simply extrapolated elsewhere) prima facie defies the laws of thermodynamics and therefore is, with a very high degree of certainty, incomplete.
It would be good to get to the bottom of this but, until we do, using this (or any other unexplained) result to draw more general conclusions would be a mistake.
So, in saying 'over to you' must, I would add a warning (for the benefit of those who may not read the whole history and/or be familiar with scientific methods, not 'aimed' at @cathoseray ) that there is no evidence, theoretical or experimental, to support drawing general conclusions from one experimental result and to do so would thus be somewhat foolhardy.
Without a doubt we need more data and a better understanding. Until we have both then individual experimental results apply to that case only, as you indeed say.
More data please from anyone who takes comprehensive measurements!
This post was modified 4 months ago 3 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.
We are, like Sisyphus, condemned to roll this ball forever! I will get back to you tomorrow.
I don't think so and hope not. I grant that heating, while fundamentally simple, is actually rather complex. However there is, in principle, a vast amount of data to which we don't have access but heat pump manufacturers and others do. From this it should be possible to extract more useful experimental evidence.
Furthermore, models can be improved by taking into account some more of the real world factors, such as pumps that run continuously, the contribution to heating from people, animals, and electrical equipment, and a more sophisticated model of a house than the 'monolithic heatstore' model that spreadsheet jockeys are more or less constrained to use.
Since the underlying thermodynamics is very simple indeed, it should be possible to explain experimental results with such a slightly more sophisticated model, and develop the slightly more sophisticated model based on the learnings from a set of experimental results. The two go hand in hand and that's how science progresses as I am sure you would agree.
Once we have done this we will be in a position to make more general statements.
This is, unfortunately, probably a more complex analysis using data to which we don't have ready access than can be readily done on this forum (although it may turn out that a simple variation to the monolithic model is in fact sufficient to explain a usefully large class of experimental results). However it is nothing like as difficult as, for example, predicting the weather across the UK, which is now possible with a pretty high degree of accuracy (provided you interpret the results sensibly).
Until we have an analysis that does this, and experimental results to underpin it (or vice versa) then the best advice has to be to try it for your system, but be sceptical of any apparent results and test that they do survive in a variety of relevant conditions. In the end if you can find a lower cost way to operate your system in a way that you find comfortable then that's all that matters, but be wary of making any assumptions that you cannot personally test or extrapolating it to any situation other than the specific one tested.
This post was modified 4 months ago 2 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.
Understanding poorly adjusted/designed heat pump systems theoretically is going to be difficult for some while.
This is an important point, the effects of poor (and for that matter mediocre and good) design and adjustment are effectively unknown unknowns. While me may (think we) know they are there, we don't know what they are, or how they actually affect things, and this is why models struggle. You have to identify and quantify the effects to include them in a model that works from the bottom up, but we don't know their identity or quantity, and so can't (yet) include them. The empirical method does the opposite. It works from the top down, making a basic assumption, which is not unreasonable, that the measured values you get from the data already have all the unknown unknowns baked in.
you cant reliably extrapolate from one poorly adjusted/designed system to another or indeed from a theory based on a well adjusted/designed heat pump system to an arbitrary poorly adjusted/designed one
Agree, apples and oranges, but that applies mostly to the particular findings, the details of that case, the particular regression equation that applies to that particular building, with that particular installation, operated in that particular way. But the pattern, rather than the detail, may be generalisable. If we run the heating with a setback, the energy use during the setback will drop to zero, or near zero if ancillaries are running, and the IAT will fall. If we build in some sort of post setback recovery boost, the energy used during recovery will be more that it would have been without the setback and boost. What we then need to do is quantify those changes, fully accepting that the detail will vary between installations. The results I posted in my recent post are my results for my installation for one week with middling OATs. Had I done the same thing for a week with OATs around zero, I suspect the details would have been different, but the pattern similar, but I don't know. Maybe the defrost cycles will screw things up? Perhaps I can find a week where the OAT was low and stable.
In broad terms (so far as I understand your post and diagrams) you are seeing a 28% reduction in daily energy use for a roughly 50% reduction in IAT-OAT (~19-10 vs 16-10) for 25% (6hrs) of the day, The latter translates to a reduction in daily heat loss from the house of 50%*25% =12.5%.
I agree, on the face of it this is a problem, something somewhere is wrong. The somewhere lies within one or more of only three things, firstly my data (observations with a setback), secondly my whatiffery (called whatiffery on purpose, predictions for no setback), and thirdly your derivation of 12.5%, with "a roughly 50% reduction in IAT-OAT (~19-10 vs 16-10)" being the first thing I would look at.
Going through these three things in turn, I am reasonably confident my observed data is good enough, because I have the independent heat pump only kWh meter. This needs manual reading, because it is not hooked into the modbus system, and I don't have hourly data, but I do have weekly data, but it is Monday to Monday (25 Mar 2024 to 1 Apr 2024), not Tuesday to Tuesday (the charts above), and the number is 181kWh, compared to the modbus derived calculated value of 164kWh. However the kWh meter value also includes DHW heating, which the calculated value excludes, and there may also be some background ancillary use. The total space and DHW heating calculated value for the Monday to Monday kWh meter reading period is 170kWh, a 6% under-estimation of the 181kWh actual use, which could be ancillary use or measurement error or both. Whatever the explanation, the values are I suggest near enough to be credible.
The next thing to consider is my predicted values, which I deliberately called whatiffery to highlight the fact they are potentially, but not necessarily, suspect. They just need closer scrutiny. And in fact we can test them, and the best way to test them is to plot the predicted values against the actual values during periods of steady running, ie no setback or recovery boost in operation. I have already done this, but here is the plot again for ease of reference:
The fact is, the predictions in the steady running periods are, give or take, very close to the actual values. It is therefore not unreasonable to suppose the predictions for the setback and recovery periods aren't far out; and furthermore, visually they also make sense. There is however a quirk I can't explain, the first predicted value for no setback in the first hour of a setback is always a bit lower than its neighbours. This may be down to the fact the measured outside air temperature (OAT) at the heat pump increases at the start of a setback, because the heat pump has stopped running. To investigate this, I added the OAT to the plot (green line). The result is not pretty, and I had to do some axis jiggery pokery to get the OAT where I wanted it, above the energy plots, but there does appear to be a rise in heat pump OAT at the start of most setbacks, see red arrows. In a way, this is a sort of unknown unknown negative (OAT up, kWh down) feedback loop, that only comes to light when we look at the actual data, becomes apparent:
In summary, there is no particular reason to suppose that either the actual measured values during setback running, or the predicted values for no setback running, are wildly out. If we take an extreme case (crude sensitivity analysis), for example the actual value is 10% below what it should be, and the predicted value is 10% above what it should be, then we get something like 180kWh actual with setback, 204kWh predicted with no setback, which comes out at a 12% saving. But what evidence is there that the actual measurements were 10% under, and the predictions 10% over?
In passing, I should perhaps mention that I should perhaps be using areas under the curves, rather than sums for the energy values but (a) calculating areas under the curves is complex, (b) I think despite this I did do it once, and the results were very close and (c) we can do eyeball areas under the curves, and the area under the predicted curve does seem larger than that under the actual curve, at least to this eyeball.
That leaves the third thing to look at, the appliance of science, specifically the "roughly 50% reduction in IAT-OAT (~19-10 vs 16-10) for 25% (6hrs) of the day" statement that gives rise to the 12.5% figure. The problem here is that correcting the roughly 50% to a more accurate 9 vs 6 delta t gives a 33% reduction, which in turn gives a 24 reduction of only 8%, which is even more at odds with the actual vs predicted difference. Something somewhere is wrong - but where is it? Perhaps the error is using averages that mask more extreme events? I don't know the amswer, any thoughts welcome.
So, in saying 'over to you' must, I would add a warning (for the benefit of those who may not read the whole history and/or be familiar with scientific methods, not 'aimed' at @cathoseray ) that there is no evidence, theoretical or experimental, to support drawing general conclusions from one experimental result and to do so would thus be somewhat foolhardy.
Exactly, that's why I said 'over to you', to encourage debate, because something somewhere is amiss, and like you I want to get to the bottom of it, and at the end of the day, again like you, I think we need more data.
However there is, in principle, a vast amount of data to which we don't have access but heat pump manufacturers and others do. From this it should be possible to extract more useful experimental evidence.
I agree, the data is no doubt there, but I fear it will stay there, as manufacturers are very likely to guard their (which in fact is also our data) data very closely, unless and until it paints them in a favourable light. I'm not sure they have any interesting in determining whether setbacks save money without compromising comfort. What is needed is an independent/academic study, but academia ain't what it used to be these days, and there is also the efficacy/effectiveness problem, the more tightly you control the experiment, the more un-(real)worldly the result.
Since the underlying thermodynamics is very simple indeed, it should be possible to explain experimental results with such a slightly more sophisticated model, and develop the slightly more sophisticated model based on the learnings from a set of experimental results. The two go hand in hand and that's how science progresses as I am sure you would agree.
Again, I agree. Hypothesis, experiment, re-work hypothesis to take into account experimental results. My own view is that while the theoretical thermodynamics may be simple, the real world systems are extremely complex, and that is why the hypotheses/re-worked hypotheses struggle. We do know about solar gain, but imagine a world where we didn't. Maybe there are other effects going on that we haven't yet spotted - the unknown unknowns - that materially effect what happens in the real world.
However it is nothing like as difficult as, for example, predicting the weather across the UK, which is now possible with a pretty high degree of accuracy (provided you interpret the results sensibly).
Ho hum, I think you already know my views on the accuracy or otherwise of weather forecasts. The mention of needing to 'interpret the results sensibly' is perhaps telling. I'm still collecting the daily 0600 Inshore Weather Forecasts, but am still stuck on how to compare them to the actual weather (wind speed and direction). The problem is the two (forecast and actual) are in fundamentally different formats, the former being free text, the latter a table of rows and columns. How do I compare, for example, a forecast of 'North 2 to 4, becoming variable 3 or less, then west or southwest 3 to 5 later' for a region to 24 hourly rows of actual wind speed (knots) and direction (degrees) data from a single location in that region. One day, when I want to have a laugh, I might try an AI chatbot, and see what it comes up with...
But I digress. What we need more is more heat pump data, running with and without setbacks.
Midea 14kW (for now...) ASHP heating both building and DHW
The empirical method does the opposite. It works from the top down, making a basic assumption, which is not unreasonable, that the measured values you get from the data already have all the unknown unknowns baked in.
True (which is one reason I advocate it for pump sizing), but it doesn't support extrapolation to other scenarios (ie houses) unless you have enough examples to answer statistically.
People are asking 'should I' and the empirical method will not answer that question unless we have lots of examples, more than we are likely to get. So we need a combo of empirical and theory.
But the pattern, rather than the detail, may be generalisable. If we run the heating with a setback, the energy use during the setback will drop to zero, or near zero if ancillaries are running, and the IAT will fall. If we build in some sort of post setback recovery boost, the energy used during recovery will be more that it would have been without the setback and boost. What we then need to do is quantify those changes, fully accepting that the detail will vary between installations
All agreed. I think that a good starting point is that, if you dont increase the FT and it recovers 'in time' then its relatively unlikely that you will use more energy than you would without setback, and relatively likely that you will use less.
agree, on the face of it this is a problem, something somewhere is wrong...
I think its more likely that its a combination and also that other things are at play. I don't doubt the basic thermodynamics, so that I wouldn't question (apart from the numbers).
The simple model of a house as a monolithic entity with heat capacity is clearly very simple. The air in the house will behave differently to the solid fabric, and thinking about this should give us some pointers. I listed in my previous posts some further factors and I would add that I have a suspicion that setback is often accompanied by an acceptance of a lower temperature for the non-setback time, perhaps by as much as half a degree, which when the OAT is 10C is significant. This means that the pseudo-control (the bit you describe as what-iffery) may be inaccurate. Add all these together and you could quite plausibly explain the discrepancy. The question is, which one(s) are dominant (if any) and how will they vary from house to house. Unfortunately we will probably be unable to run a real control other than in a lab.
Exactly, that's why I said 'over to you', to encourage debate, because something somewhere is amiss, and like you I want to get to the bottom of it, and at the end of the day, again like you, I think we need more data.
Agreed we should encourage debate, and preferably some actual facts. However (IMHO) we have to think about the (majority?) who dip in and out and aren't interested in/don't follow the detail. We also have to remember that people often believe what they want to believe.
I wouldn't want someone 'dipping' in to go away with the impression that setback will save 28%, or indeed any other number above that which is explicable by the current theory. It may, but it may not, and equally it may save a lot less (or, in come circumstances, cost). We simply don't know and until we have experiment and theory which are consistent wont know.
What is needed is an independent/academic study, but academia ain't what it used to be these days, and there is also the efficacy/effectiveness problem, the more tightly you control the experiment, the more un-(real)worldly the result.
Completely agree. There are some academic studies of heating systems from time to time. If I were in government I would be sponsoring a whole lot more because its clear (a) that its an important topic and (b) that there is a load of both missing information and misinformation out there.
The mention of needing to 'interpret the results [of weather forecasts] sensibly' is perhaps telling.
I guess I had something particular in mind. I spend quite a lot of time outdoors and am therefore interested in the predicted rain and temperature. Temperature predictions are generally good, but when the forecast tells me that there is (eg) a 30% probability of rain its basically useless. However if you look at the weather map you can get more useful information because it tells you whether the rain clouds are scattered but small (in which case you almost certainly will get rain but only for a short time) or large but only just graze your location (in which case the uncertainty that the 30% is describing is whether it hits you or not). These tend to cause me to make different decisions about what i do!
My real concern, which I am trying to protect against with cautions about the experimental results, is that someone who isn't following this in detail will jump to the conclusion that setback is a very good thing without having first set up their heat pump to run at its optimum when on 24*7. This could easily lock them into a position where their flow temperature is set higher than it needs to be in order to effect a rapid recovery (like the gas boiler they are used to), but adds 10-30% net to their cost, and reduces their comfort in addition. I wouldn't want that to happen as a result of our debates!
This post was modified 4 months ago 3 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.
I wouldn't want that [blindly adopting a setback] to happen as a result of our debates!
I agree, but I think there are sufficient caveats. In particular, my post that reported my findings ended with a statement that my findings were not definitive and generalisable ('To what extent, if any...'):
"And now for the answer to the question 'did the setback save money?'. The answer is thatin my house, andin this week, yes it did, quite a bit in fact. The actual measured use (sum of the hourly values) for the week was 164kWh, while the predicted use (again, sum of the hourly values), had I not had a setback, came out at 227kWh. The setback saved me 63kWh that week, which is a saving of 28% for that week.
Note the italics in the previous paragraph, because that is where the arguments start. This is a classic n=1 study, in this case in a small old leaky listed building in southern England with a relatively high thermal mass for its size during a period of middle of the road OATs. To what extent, if any, can these findings be generalised to other buildings and other settings? Over to you..."
I am inclined to think the predicted kWh values for no setback probably aren't far out. We have established here and elsewhere that OAT is a very good predictor of both energy in and energy out, particularly with middle range OATs like these. Furthermore, in these middle range OAT no setback conditions, the IAT tends to be pretty stable. However, I think you may be onto something with the IAT behaviour during setback. Although I have been comfortable in the house, there is a lag in recovery, even with the boost, meaning the 50% or 33 x 25% calculation may underestimate the actual deficit. As it is, the calculation only estimates the deficit during the six hour setback. Here is a gee-whizz chart (gee-whizz = Y axis doesn't start at zero, to make effect look more dramatic, I'm doing it here not to be dramatic, but to zoom in on the detail) of the IAT over the setback week:
I've managed to get the X axis gridlines lined up with the setback start and end times (2100 and 0300). I felt warm enough because on most days the actual IAT was within 1 degree of the desired IAT by 0600, and within half a degree by 0900, but there is still a deficit outside the setback period. We need a more sophisticated way of measuring that deficit than the 50% or 30% x 25% method. If it has to be areas under the curve, so be it, but they are not easy to calculate (may have to take the data into python, or even R).
I spend quite a lot of time outdoors and am therefore interested in the predicted rain and temperature. Temperature predictions are generally good, but when the forecast tells me that there is (eg) a 30% probability of rain its basically useless.
I do quite a lot of sailing, which is why I am interested in forecast accuracy, particularly for wind speed and direction, but also for visibility, hence my interest in studying the accuracy of forecasts.
I totally agree the probability based rain forecasts are a spurious nonsense, intended to make the Met Office look clever, but as you say basically useless to the individual looking at the forecast. More generally it is an example of what I call the collapse of the probability function: what may make statistical sense eg it will rain in 50% of places where the forecast is 50% makes no sense for the individual, for whom it is a binary outcome, it either rains or it doesn't. I use 50% because it is a good percentage to ask of people what does a 50% chance or rain actually mean? They soon realise they don't know whether it is going to rain or not!
The collapse of the probability function comes from medicine. Imagine a GP has a waiting room full of ten patients, all of whom happen to have a 10% risk of a fatal heart attack within the next five years. It may well happen to one of them, but which one? For the patients, they can't have a 10% fatal heart attack, its all or nothing, dead or alive. And what do you, the GP, do about the fact you don't know which one it will be?
I agree about the weather charts, the rainfall radar observations (past real world data, not the whatiffery forecast!) in particular is useful for short term DIY forecasting of whether it is going to rain or not.
I am inclined to think the predicted kWh values for no setback probably aren't far out.
Provided the 'control' period from which these were derived is representative of the test period I would agree. If any changes have been made to the system, or the IAT has changed, then less so.
My gut feel is that it does, several hours of a few tenths of a degree difference add up. I cant remember how often your data is read, but if its half an hour or more frequently just taking each period as a rectangle would suffice I think.
There is an additional question which is - does a setback of eg 1C when the OAT is eg 10C actually cause exactly a 10% reduction in energy input to the house from the heating system (once everything is in equilibrium). It certainly causes a 10% reduction in loss from the house, and that must certainly be balanced by energy put into the house. But the energy put into the house is the sum of energy from the heating system and energy from other sources. The other sources are principally humans (and large animals), and electrical equipment. If we take a typical electricity consumption (without heat pump) of ~3.5MWh/annum, electrical equipment amounts to 400W, assuming all electrical energy eventually gets converted to heat. Resting human beings are about 100W each. So we could easily be talking 600W of energy going into the house which has not come from the heating system, 15% of the resting total at OAT 10C. I reckon that needs to be accounted for also.
It may well turn out that these two, relatively simple, factors together are sufficient to account for the discrepancy. If not then we will need to turn to the more complex matters like the fact that the house fabric isnt monolithic, inefficiencies in the heat pump system etc. Hopefully we can avoid these.
@jamespa How are you guys equating a 1degC drop from 10degC with a “10% drop”?
The Celsius scale extends below 0degC, and 0degC is simply the freezing point of water. Surely 0 Kelvin is zero, and a 1degC difference cannot be equated to a percentage of anything but the entire Kelvin scale?
This post was modified 4 months ago 2 times by ectoplasmosis
@jamespa How are you guys equating a 1degC drop from 10degC with a “10% drop”?
The Celsius scale extends below 0degC, and 0degC is simply the freezing point of water. Surely 0 Kelvin is zero, and a 1degC difference cannot be equated to a percentage of anything but the entire Kelvin scale
The '10%' refers to the change in deltaT between IAT (20C->19C) and OAT (10C).
House loss (which is how the '10%' is being used) is proportional to IAT-OAT, so if the OAT remains at 10C a 1C drop in IAT will reduce the house loss by 10%.
Provided the 'control' period from which these were derived is representative of the test period I would agree.
I deliberately used the regression equation from test week, so they should be representative. The average IAT did change over the period, but only slightly, increasing by about 0.3 degrees. An absolutely steady state running is rare, because there are so many factors at play.
My gut feel is that it does, several hours of a few tenths of a degree difference add up. I cant remember how often your data is read, but if its half an hour or more frequently just taking each period as a rectangle would suffice I think.
I have now managed to do this, it turned out to be easier than I thought, but the results don't get us much further!
This is what I did. I have assumed - please correct me if I have got this wrong - that your reasoning stated simply is the energy use is proportional to the IAT-OAT delta t. For the very simple example where the IAT-OAT delta t with a setback is half what it would be without a setback, then the setback will use half as much energy as no setback running would have used. In this case we are doing an area under the curve (AUC) calculation, but it is so simple we can do it in our heads.
In the real world, the OAT and to a lesser extent the IAT are constantly changing. This means we need to do a formal calculation of the AUC, which can be done using the trapezoid rule. This divides the AUC into a series of narrow width rectangles, and then sums the area for each rectangle. If for example over one hour (the interval used in the data) the delta t went from 10 to 12, then that hour would contribute 11 to AUC ((10+12)/2 x 1), Because the x axis interval is 1, we can drop it, and just uses the average of the start and end delta t.
Note that overall, as mentioned above, there was a very small net gain in IAT during the setback test period of about 0.3 degrees C. Overall, the house wasn't cooling down, it was actually creeping up, but only very very slowly (1.5% increase).
Before calculating the AUC, I plotted the setback/no setback delta t values, using a steady 19 degree IAT for the no setback running, expecting to see a clearly visible difference, but there isn't one:
This I realised is because it is the OAT not the IAT which dominated the delta t. Even with the setback, the IAT is relatively steady compared to the OAT.
I then calculated the AUC, and found the AUC with the setback was 1537, while without the setback it was 1635, giving a setback 'saving' of 6%...
Note that the mean IAT and the mean delta T with and without the setback are IAT 18.4 vs 19.0 degrees (3% 'saving'), delta t 9.2 vs 9.8 (6% 'saving'). In other words, putting all of the above together, it suggests the house was only marginally cooler during the setback period, meaning, by the conservation of energy laws, there should only be marginal savings...
It may well turn out that these two, relatively simple, factors together are sufficient to account for the discrepancy. If not then we will need to turn to the more complex matters like the fact that the house fabric isnt monolithic, inefficiencies in the heat pump system etc. Hopefully we can avoid these.
Given the above AUC and related findings, things now look even muddier. It may be I have made some blunder in the calculations, but I don't think so. The two telling but clashing plots are the one immediately above, which suggests very little difference, very little savings, and the actual with setback vs predicted without setback energy use, re-posted below for ease of reference, which suggests considerable differences, considerable savings:
Midea 14kW (for now...) ASHP heating both building and DHW
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.
