I think people, including me, do this simulation exercise as a bit of a sense check and to ascertain how low temperature heating affects their comfort.  Obviously the latter can only be judged in the colder months, unless the boiler is running WC, which few do.  What I discovered is that running low (ISH) and slow made the house more comfortable for no measurable extra gas consumption, largely because temperature was more stable rather than being affected by thermostat hysterisis.

Regarding consumption figures, I agree that similar heating profile is needed to carry across heat loss ( although one could calculate a correction factor) but can't see why WC matters in this case.  The calculation is about energy to the house not energy to the boiler/hp, WC affects the latter but, assuming a similar heating profile, not the latter.  

If we conclude that consumption is not a way to sense check the gigo spreadsheet, I think it's fair to say we currently do not have a way to determine heat loss with any reasonable certainty/accuracy for a typical retrofit with a patchwork of construction standards and unknown air change value.  Since this is the majority of the housing stock we need to tackle, this is a major problem to say the least.

Posted by: @johnmo

↑

looks like a good chance of sh*te in - sh*te out, if you are not careful.

I do agree, it is a minefield, however, consider the logic of my approach which only needs a steady IAT (if that isn't the case, all bets are off) and a way of determining energy delivered to the building:

1. Given a steady IAT, then the energy delivered to the building must equal the heat loss of the building. That's why the IAT is steady.

2. If we can measure or derive the energy delivered to the building (whatever the heat source, and however it is modulated, be it by weather compensation, on-off cycling or a combination of both) at various OATs (the main determinant of actual day to day heat loss) at that steady IAT, then we can plot ideally hourly average OATs against energy in for that hour. The more data points the better (we want a regression to the mean effect). 

3. We can use a visual check (and R squared) to assess how well they are correlated. If they are all over the place, bin the data, they are of no use.

4. If they are well correlated (as they were in my plot posted earlier), then we can plot a regression line and determine what the heat supplied to, and therefore lost from, the building is at various OATs, including the design OAT, whatever that is (depends on location, and possibly other factors).

I do appreciate the fact I used energy in and then multiplied it by efficiency to get energy out (delivered) might have caused confusion, but I was only trying to demonstrate the principle of the method. Re-doing the plot with the energy values adjusted to energy out rather than energy in didn't appeal, but could be done, and would give the same result.

The whole thing is based on the simple premise that when the IAT is stable, then the energy delivered to the building must equal the energy lost, ie the heat loss. If you can determine the energy delivered at a particular OAT, then that is the heat loss at that OAT. Get as many data points as possible, plot them, and if they stack up, then you have a useful estimate of building heat loss at various OATs, including the design OAT.

The point of posting my plot earlier was to show that the heat loss determined by the method described above, 12kW, is actually very close to the calculated heat loss, 12.3kW, ie the method can work, but as I keep on saying, it needs a steady IAT. In a way, you could say that controls for the 1001 other variables that could otherwise confound the situation.

There is some other semi-independent evidence that my heat loss at design temp (-1 to -2 depending on what source you use in my case) is 12 (and a bit) kW. At these OATs my IAT starts to fall a bit, because at these OATs the heat pump, according to Midea's engineering data, is only producing 11 (and a bit) kW. As this is below the actual heat loss at these low OATs, the IAT falls.

I think I can confidently say that if you can achieve the steady IAT requirement, it doesn't matter how you achieve it, just that you do achieve it, and you can get a good enough estimate of the energy delivered to (not consumed by) that steady state IAT building at various OATs, then you have a very good chance of getting a good enough estimate of heat loss for that building at the design OAT.

I appreciate my demonstration of the method is a classic n=1 demonstration. I don't suppose anyone else has enough data to replicate my findings? All you need is (preferably hourly) data for (a) OAT and (b) energy delivered, for a house that has a steady IAT. The heat source doesn't matter, nor does the control logic, all you need is OAT, energy delivered, and a steady IAT. If you have that sort of data, then almost certainly you will know how to plot them...and then all you need to do is compare the empirically determined heat loss at design OAT with the calculated heat loss at the same OAT.          

Posted by: @jamespa

↑

If we conclude that consumption is not a way to sense check the gigo spreadsheet, I think it's fair to say we currently do not have a way to determine heat loss with any reasonable certainty/accuracy for a typical retrofit with a patchwork of construction standards and unknown air change value.  Since this is the majority of the housing stock we need to tackle, this is a major problem to say the least.

I think this is perhaps unduly pessimistic. We have my n=1 experiment described above which showed a consumption based approach can work, as long as you have the necessary data. My empirically derived figure is the same as the calculated figure, give or take, and there is further evidence this is the right figure. The beauty of this method is you don't need to faff about with known and unknown unknowns, instead it just relies on one basic assumption: given a steady IAT, then the heat loss must equal the heat delivered. If you know the latter, you know the heat loss.

I think it might also be useful to determine what counts as 'reasonable certainty/accuracy' for heat loss. Given that no heat loss determination method is ever going to provide an exact answer, what counts as 'good enough'? Within 10% of the actual figure?     

Or pick up a heat meter, we plumbed it in while we had the gas boiler, used it to optimise settings of the boiler and UFH. Luckily for me, I got one new for peanuts from another forum.

Then when it it got cold, I started adjusting the heating system flow and temp up to the calculated heat loss. We had a few days at -9 so had plenty of time to stabilise. Ended pretty close to calculation (new build). But wife not really happy with messing with the heating when it was so cold.

For what it's worth, I have just compared the independent kWh meter energy in readings with the calculated total energy in values (source (2) in the post above, and the data used to plot the green bars and calculate the COP) for two four week periods, one in January of this year (heavy use), and one in April of this year (lighter use, less than half the January use). The results are:

January: kWh meter use = 1451.8, calculated use 1493.34, calculated use over-estimates metered use by 2.86%

April: kWh meter use = 597.8, calculated use 546.50, calculated use under-estimates metered use by 8.58%

Adding the two periods together gives kWh meter use = 2049.6, calculated use 2039.84, calculated use under-estimates metered use by 0.48% 

Posted by: @johnmo

↑

Or pick up a heat meter

Yes, that should give you a gold standard answer, but it is expensive, invasive and needs some technical expertise. Probably not Joe Bloggs' first choice. I don't suppose you (@johnmo) actually have the data needed to test my method for your house? Steady IAT data, OAT and energy delivered, perhaps coming from your heat meter, then plot and do the regression?

@cathoderay 

Since you are convinced (and have sone evidence) that your house heat loss is 12kW, let's take that as read and see where it takes us.  In this case your heat pump performance is pretty as one would expect, and the mystery is how you managed to keep warm with oil or electric using only 10MWh/annum.

Unfortunately you don't recall that time so well, but let's look for clues.

Can you tell us where you live (with whatever accuracy you feel comfortable ), what the design temp is corresponding to 12kW (you have mentioned both -4 and -1) and what the occupancy pattern of the house was when you were heating by electric or oil (I'm presuming you never used the two together).  Also did you have any supplementary heating and when heating by electric, can you confirm that the heating was separately metered?

Posted by: @cathoderay

↑

I think this is perhaps unduly pessimistic. We have my n=1 experiment described above which showed a consumption based approach can work, as long as you have the necessary data. My empirically derived figure is the same as the calculated figure, give or take, and there is further evidence this is the right figure. The beauty of this method is you don't need to faff about with known and unknown unknowns, instead it just relies on one basic assumption: given a steady IAT, then the heat loss must equal the heat delivered. If you know the latter, you know the heat loss.

I agree with the first sentiment, I was being provocative.

Practically speaking one can download smart meter readings and degree days.  These should also correlate tolerably well, maybe with a temporal offset to account for the 'memory' that the house has in its fabric.  Mine do and I can't see why that can't be used to get a good handle on loss.

Posted by: @cathoderay

↑

I think it might also be useful to determine what counts as 'reasonable certainty/accuracy' for heat loss. Given that no heat loss determination method is ever going to provide an exact answer, what counts as 'good enough'? Within 10% of the actual figure?     

Obviously there is no exact answer as it depends on what design choices are triggered as the figure changes.  Based on the design choices I have to make, 10-15% would be good, any worse forces me into choices which are in some way sub optimal.

@jamespa - I've been giving it some thought. As I said it was a long time ago especially the oil period. I probably put the basic timer settings in decades ago. I still have all my marbles, but remembering that sort of thing from that long ago is a long shot by any measure.

1. Best to ignore the standard mains electricity period. It was an infill period when the oil tank was not really usable (minor leaks and I mean minor,  although it did have a small amount of oil that I very occasionally used in very cold weather, but I didn't dare fill it) and getting the heat pump. As many others have found, that was a long and complicated process with both planning and listed building consent needed, cowboy installers, all the usual crap. The house using only standard mains electricity was very cold in winter (enough to make me think I'm doing myself in) because mains electricity is very expensive and I am a mean old bugger so I only used to heat one room, the room I was in. In cold spells, the kitchen would get down to 7 degrees in the morning. It was a highly atypical period, not at all comparable, and not much use for our purposes.

2. I live in a mid-southern county of England, in a small leaky old listed building. Heat loss is always going to be high, unless I do things that I am not going to do. What I have done is pumped up the loft insulation, and I am in the process of adding secondary glazing, which in turn needs window frame works to get a suitable level finish on ye olde frames, which in turns means I can reduce draughts, which in this house are a major cause of heat loss. The heat loss calcs done by my installer and submitted to MCS have a design OAT of -2, in earlier version it was -1.8. The -4 is not a design temp as such, it is my weather curve left hand end point OAT, I wanted the set LWT to carry on going up a bit after the design OAT (otherwise it flat lines).

3. I think my oil CH was probably on for less that 12 hours a day. For almost all of my working life, I worked away from home, ditto my partner, and I would not have had the heating on during the day (frost protection was by a frost stat). I have a very vague recollection the evening on period was 1830 to 2230, 4 hours. The boiler could heat up the house quickly, in 20 mins or so, and of course I was younger, and less likely to feel the cold, so the late start time was OK. I am now only guessing, but I think it is possible the morning on period was also 4 hours or shorter. If so, then the oil heating was only normally on for 8 or less hours, 1/3 of the heat pump time when running without a setback. I do also recall I did occasionally use the 'extra hour' on the time in cold weather, but I doubt that in the grand scheme of things that amounted to very much. I do have an open fire, but that tends only to be used on special occasions, it is not in regular use.

I think the daily hours of use is probably the main explanation. If I only ran the oil CH for 8 hours a day, 1/3 of the heat pump running hours, then that difference seems to me to a plausible explanation for the extra energy delivered by the heat pump. Another possibility is that, although I never felt cold with the oil CH, the house might in fact have been a degree or two less, that the current 19 degrees, but being younger I didn't feel colder. I don't have any IAT records from the oil CH time, just a recollection that I don't remember feeling cold.     

For the OP we found heating the house with the gas boiler with manual WCand our limit of lower water leaving temperature of 60C (since DHW could not be separately set) to be highly useful. The house was warm and because we were also forcing the boiler to condense more economical.

For the ‘conundrum’ side thread since it was not resolved can I add the thoughts that a highly leaky house will behave differently to a more sealed one, and low thermal mass (eg wooden frame)  to a solid walled one. I don’t recall the topic at all unfortunately.

@cathoderay Age and health can be a bit of a buggerance can’t it? In better health and in my youth-and-a-bit, I would be happy to work indoors at 18-19 degrees C (though I should have been working at 20 degrees in precision engineering as best practice required). I am not as mobile these days and 22.5 degrees C. suits me far better. I well remember keeping a paraffin heater running on a low wick 24/7 in the family home in the north easterly facing hallway of our council house, just to stop the living room temperature from dropping 10 degrees each time the hall door was opened! Oh happy days. 😉 Toodles.

Posted by: @cathoderay

↑

but it is expensive, invasive and needs some technical expertise. Probably not Joe Bloggs' first choice. I don't suppose you (@johnmo) actually have the data needed to test my method for your house? Steady IAT data, OAT and energy delivered, perhaps coming from your heat meter, then plot and do the regression?

Lucky me paid about £20 including postage.

I only good data from this month. As that is when I got all the monitoring working together. My issue is solar gain, it screws up a lot of results. I need a day at reasonable stable temp and no sun. Today isn't bad, quite stable no sun. 

All my data is here. There were a couple of cooling days in the data

https://emoncms.org/app/view?name=MyHeatpump&readkey=443f1fd946e1bd45ff4cfebfc25ee64d

I am still fiddling to get the system performance to where I want it