This is a renewed attempt to solve the setback riddle - do setbacks save money without compromising comfort? So far this apparently simple question has eluded a confident answer, largely because the empirical results and theoretical predictions do not agree. This means either one of them, or possibly both, are wrong. The initial posts here have been moved from another thread as they no longer related to that thread's title.
Midea 14kW (for now...) ASHP heating both building and DHW
Apropos setback running, I have just done this plot for the 2023-24 heating season, plotting mean daily ambient temperature against daily energy in, for (a) days with a 6 hour overnight setback and (b) days with no setback:
On the face of it, there do seem to be two separate populations...but it's complicated, eg on the setback days, the mean indoor temperature may well have been less, so I would expect to use less energy. The problem, as ever, is how to compare like with like. Once I have more data, I will start a new thread.
Midea 14kW (for now...) ASHP heating both building and DHW
Apropos setback running, I have just done this plot for the 2023-24 heating season, plotting mean daily ambient temperature against daily energy in, for (a) days with a 6 hour overnight setback and (b) days with no setback:
On the face of it, there do seem to be two separate populations...but it's complicated, eg on the setback days, the mean indoor temperature may well have been less, so I would expect to use less energy. The problem, as ever, is how to compare like with like. Once I have more data, I will start a new thread.
Thank you for sharing that data. The graph looks fairly conclusive to me, and my own rough data looks similar (I don't have detailed monitoring, just my own daily observations recorded in a spreadsheet).
I turn my heating off for 5-6h overnight. The 1kW per hour I save more than offsets the extra energy we use reheating the house in the morning. Like you say, the reason less energy is used overall is because the average house temperature is lower due to the setback period. The monetary saving is further enhanced by the fact that for the first couple hours during the reheat, we are on half price electricity. So we are (a) using less electricity overall, and (b) shifting a chunk of what we are using from standard rate to cheap rate. There may even be arguments to be made around efficiency - we are not running our heat pump at a time of day/night when it is least efficient (coldest), and have maybe also avoided a few efficiency-killing defrost cycles.
If we were on a flat rate tariff, I think the monetary savings may be far lower, and easier to justify constant running for the extra comfort. However, for us, the bedroom gets too warm if we leave the heat pump on all night, even on minimum flow temps, and I don't see the point of turning down the TRV in the bedroom and continuing to heat the rest of the house at extra cost when we don't need it, so for us an overnight setback/switch off is also the preferred option for comfort.
@old_scientist - thanks for your comments. @jamespa will be along soon to comment if I don't add something, and even if I do he may still add something. The fundamental problem is the empirical evidence is at odds with the theoretical predictions, which ultimately means one, or possibly, both of them are wrong.
A fundamental part of all this is the comfort level. Sure I can save a lot of money if I do the ultimate setback, and turn my heat pump off forever. But I won't be comfortable. I can also achieve savings if I turn it off for a week while I am on holiday (and my comfort is determined elsewhere). Where it gets much more complicated is with frequent eg daily setbacks. We can all agree there are saving during the setback, we can see them. It is what happens during the recovery that makes all the difference, both to energy use and comfort. To raise the room temp back to where it should be does need some extra energy over and above what would have been used in the recovery hours had there not been a setback, but how much extra energy?
Here is my data from the last 48 hours (I restarted the setback running on Monday night) which usefully has a mild night and a colder night:
During the first setback, I saved ~6kWh, and there is minimal increased usage in the recovery period, perhaps a kWh or two, and some of that is explained by the slight drop in OAT. The IAT also recovered well, back at over 19 degrees (my preferred comfort level) by 0700. Last night on the other hand was notably colder, and the energy use during the recovery period is clearly higher - but again, how much of that is down to the lower OAT, and the malign effects of the defrost cycles? The IAT was also a bit slower to recover, but it was still over 19 degrees by 0700. One of the things I am currently interested in is whether opening up all the valves on my system means it can recover faster, because it can pump heat in faster.
It is teasing out the different effects all these many variables which change constantly that make answering the question (do setbacks save money without compromising comfort?) both challenging and fascinating all at the same time. I think the next step is to to calculate (it is not currently calculated automatically) 24 hour mean IAT values, and see how they vary with and without setbacks. It might also be worth looking at the daily IAT at 0700 (which is already available) as a proxy for has the recovery been adequate.
Midea 14kW (for now...) ASHP heating both building and DHW
@cathoderay I think it's safe to say you have far more data, and of a far higher quality than I. What you have measured agrees with my casual observations. My heat pump ticks over at around 1kW. If I switch it off for 6 hours, I save around 6kW. When I turn it back on, it may use 2kW in the first hour, and maybe 1.5kW in the next hour, before settling back to 1kW once the house is back up to temperature (it definitely does not use 2kW for 6 hours, so I have definitely used less energy overall). I estimate I've saved maybe 4kWh at standard rate, and the 'extra' 2kWh that I've used to reheat has been shifted into the cheap off peak slot at half the price (happy days).
Comfort is subjective. For sure the temperature in the house is lower than if the heating was left on, but in our case leaving the heating on would make the bedrooms too hot and therefore uncomfortable for sleeping. Others may feel differently and like their bedroom warmer in which case they may prefer not to do an overnight setback.
@old_scientist - thanks for your comments. @jamespa will be along soon to comment if I don't add something, and even if I do he may still add something. The fundamental problem is the empirical evidence is at odds with the theoretical predictions, which ultimately means one, or possibly, both of them are wrong.
I think @cathoderay is correct and it goes even further (unfortunately). Robust empirical evidence (from which we might draw general as opposed to specific conclusions) is very difficult to collect indeed because we don't have access to a house in a laboratory, in the real world the conditions keep changing thus disturbing the experiment, and the effects we are looking for are either quite small or swamped by effects we dont know about/understand.
In the case of the data from @cathoderay the matter is further complicated by the fact that the 'OAT' sensor is in fact an air intake sensor which we know is itself affected by whether the heat pump is running or not. Thus an 'OAT' of 0C when the heat pump is switched on is not, in reality, the same as an 'OAT' of 0C when its switched off. Since the experimental method used depends on the two being equivalent, the conclusions cannot straightforwardly be relied upon.
We also have to recognise the possibility of confirmation bias and be careful (or specific) about what IAT conditions we are comparing, as @cathoderay alludes to. The latter point, put another way, means we need to define fairly carefully what the 'control' is otherwise it becomes purely subjective based on one person's sense of what they feel.
Finally ToU tarrifs can (and almost certainly do) completely swamp any savings in energy, so the question can rapidly become academic.
By the time we consider all of this drawing robust general conclusions about comparative costs becomes close to impossible.
Nevertheless I have been giving particular thought to the 'control' in a situation where we are not considering ToU tarrifs, if only so we can understand the theory better (and also draw more robust general conclusions from specific experiments on which we can overlay ToU tarrifs).
I think the two states to compare are
1. A heat pump optimally set up running 24x7 and maintaining a constant IAT with either a constant OAT or one which varies in a defined fashion
with
2. A heat pump as above which is operated with a defined set back (or is switched off for a defined number of hours) and which maintains the IAT to within a defined tolerance of the constant IAT in case 1, for a defined number of hours in each period of 24 hours, and where at the end of each period of 24 hours the IAT is the same as it was at the start of the day.
The final condition is (2) is necessary to ensure that the comparison is a true one and not one which, if continued for any period of time, would drift to an alternate equilibrium IAT.
Once we understand how these two compare I think it might be possible to infer at least the general behaviour of deviations from these ideals in at least some circumstances.
I have made a start on a model encompassing this concept. I was going to wait to publish it until I had written a description, but that has got delayed, so I am posting it anyway. The basic premise is that a house can be represented by a (single) IAT and heat capacity and that, min order to compensate for setback in case (2), it is necessary to modify the flow temperature so that the system remains in equilibrium and meets the conditions set out above. The physics is actually very simple, but unfortunately the differential equations for both cooling and reheating cannot be solved analytically and any attempt I have so far found at approximation (in order that they can be solved analytically) makes requires making an assumption which isnt true. So instead they are 'solved' numerically (using Eulers method.). The way I have done it does introduce granularity in selecting the flow temperature, which can lead to a sprinkling of apparently anomalous results, I am thinking about how this could be refined.
The conclusion so far is that there are circumstances where a setback, even allowing for diurnal OAT variation, can reduce total consumption, but they generally occur in fairly lossy houses and at milder OATs, or alternatively in very low loss houses where the fixed consumption of the water pump etc, and the contribution to the heating from occupants and waste heat from electrical items, becomes significant. However savings (in consumption) are small. Likewise penalties from deviating from the 'optimum' are small. The upshot of these latter observations is that ToU tarrifs will almost certainly dominate and setback for comfort reasons is unlikely to harm. It should be noted that in the current model setback means 'switch off'. I think it would be possible to extend the method to model 'turn down', which I may at some point do if I have a few weeks when I am otherwise bored (unlikely).
I dont, at this stage, want to get into an extended debate about experiment vs theory because I have yet to see the results from any experiment in which the conditions are adequately controlled or monitored for the results to be particularly meaningful other than in the specific case, and the model is not yet properly testes. However I would be interested in comments on the model 9including errors in the logic) if anyone can be bothered, which I wouldn't expect anyone would!. Once I have dealt with the granularity issue I hope to republish (hopefully also with a description) and try to draw some more robust conclusions. However I fear they may amount to 'the physics show that, in theory, setback makes only a marginal difference either way so if you do setback at all do it for reasons of comfort or to take advantage of a ToU tarrif, or alternatively because in your specific situation (for reasons which may be specific to your situation) you have satisfied yourself it makes financial sense.' In a sense that would be a defeat both for the theoretical and experimental approach, but it seems ever more likely.
what would the plot look like if you used (Inside temp - Outside temp) as the x-axis?
My guess is the scatter will increase, because the the heat pump doesn't directly know what the IAT is. I say 'doesn't directly know' because unusually for a Midea unit, it does get some input on the IAT, from my auto-adapt script, which tweaks the weather curve and so LWT according to the IAT.
I can do the plot, the way the data is logged means I can work out the DT at hourly intervals and the average them over 24 hours. I will now do the plot and post the result once it is done.
Midea 14kW (for now...) ASHP heating both building and DHW
Blimey, that wasn't easy! I hoped I could do it using a pivot chart, but failed because I have been using noon to noon 24 hour periods, and days in a pivot chart are always midnight to midnight (unless someone knows a way to change that). Here's the result (all rows, not subdivided into setback/no setback):
Much as I expected, the two visibly separate samples have disappeared, I suspect because the introduction of the IAT component has increased the scatter.
Midea 14kW (for now...) ASHP heating both building and DHW
Much as I expected, the two visibly separate samples have disappeared, I suspect because the introduction of the IAT component has increased the scatter.
Potentially a significant finding though since the expectation from the physics is that energy use is proportional to the difference. However are there still two populations hidden in the noise?
We really do need to find a way to calibrate out the OAT/AIT discrepancy so like with like can be compared.
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.
However are there still two populations hidden in the noise?
Yes and no. There are definitely setback data points and no setback data points, but I think any statistical difference between them has disappeared because of the added noise introduced by adding the IAT component. Here's the plot with the setback/no setback points identified. Haven't run any stats on them because my MK 1 Eyeball tells me they are now from the same population:
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.
