@cathoderay Many a sleepless night could be devoted to attempting to analyse this conununundrumdrum I think! 😳 Toodles.

Posted by: @cathoderay

↑

In contrast, my suggested approach uses an equilibrium state, the IAT just needs to stay pretty much the same, and in that state the heat loss must be the same as the energy delivered, give or take. I don't even think the whole house needs to be at the same IAT, all that matters is that everything is stable. This is easy enough to achieve with a heat pump, given the way we run them, and a fossil fuel system could I think be adjusted to do something similar, at something similar enough, but that is not much use for historical data that may well have come from timed heating. The problem here is the same as in your proposal, we are now dealing with a dynamic state, one in which the heat supply and loss are in a state of flux. The key question here it seems to me is whether averages can be relied upon. Is 10kW for 12 hours followed by 12 hours off the same as 5kW for 24 hours? That bit should be true, both are 120kWh, but the problem is the heat loss will vary, as the IAT changes, in the timed scenario. The question then becomes, can we use the concept of an average heat loss? Is the heat loss from 12 hours at 20 degrees IAT followed by 12 hours at 16 degrees IAT is the same as the loss from 24 hours at 18 degrees IAT? I know in practice it doesn't happen like that, with step changes, this is just a hypothetical question to consider the principle. In practice you would calculate the actual average IAT over the period.

 

An equilibrium state is pretty much spot on, because it means that the fabric of the house will neither absorb nor emit energy, which means you are measuring heat loss not changes in the state of the fabric.

Average loss is just fine provided we do a reasonable measurement of heat into the house (as opposed to heat into the heating system).  However with a varying OAT you aren't ever in precise equilibrium so, to avoid distortions due to energy absorbed or emitted by the fabric, you need to measure over a sufficiently long period of time so that the heat capacity of the house doesn't matter because the amount of energy involved is insignificant compared to the the heat loss.  For most houses that will be several days (ie significantly longer than it takes for the house to cool to ambient when the heating is switched off.  

Plotting degree days vs energy consumed is pretty much this measurement, over a long period of time, albeit with the uncertainty of the efficiency of the heating system.  If you put a heat monitor onto the output of the boiler and plotted degree days vs energy this uncertainty could be eliminated.

Posted by: @cathoderay

↑

@majordennisbloodnok - as you well know, the general view here is the only dumb question is the unasked one.

Yes, indeed, and in this case I'm following my own advice. Sometimes it even works....

Posted by: @cathoderay

↑

...

The method you suggest is certainly ingenious, but to my mind it over-complicates things. The other thing that occurs to me is that it is a measurement of a dynamic state, a house that is cooling down, and that further complicates things, because the rate of cooling depends very much on the temperature difference between the building and outside.

...

You may be right, but in essence all I'm suggesting is working out how quickly a known temperature difference erodes so I'm not entirely sure that's all that complicated. What I'm not sure about, of course, is the algorithms necessary to turn that knowledge into actual quantities of energy.

Posted by: @cathoderay

↑

...

In contrast, my suggested approach uses an equilibrium state, the IAT just needs to stay pretty much the same, and in that state the heat loss must be the same as the energy delivered, give or take. I don't even think the whole house needs to be at the same IAT, all that matters is that everything is stable. This is easy enough to achieve with a heat pump, given the way we run them, and a fossil fuel system could I think be adjusted to do something similar, at something similar enough, but that is not much use for historical data that may well have come from timed heating. The problem here is the same as in your proposal, we are now dealing with a dynamic state, one in which the heat supply and loss are in a state of flux. The key question here it seems to me is whether averages can be relied upon. Is 10kW for 12 hours followed by 12 hours off the same as 5kW for 24 hours? That bit should be true, both are 120kWh, but the problem is the heat loss will vary, as the IAT changes, in the timed scenario. The question then becomes, can we use the concept of an average heat loss? Is the heat loss from 12 hours at 20 degrees IAT followed by 12 hours at 16 degrees IAT is the same as the loss from 24 hours at 18 degrees IAT? I know in practice it doesn't happen like that, with step changes, this is just a hypothetical question to consider the principle. In practice you would calculate the actual average IAT over the period. 

In fact, I think the bigger issue with your approach - not insurmountable, but necessary nonetheless to state - is that whilst your equilibrium state can easily measure energy in (in terms of power necessary to run a heat pump or, in another life, oil/gas needed to run a boiler) it doesn't necessarily measure the actual amount of heat energy delivered into the property. On the other hand, if one uses flow rates and water deltas then presumably that could be done and I can't think of too many people on the forum likely to be better set than you to do exactly that. Nonetheless, if the purpose of a heat loss calculation is to answer questions BEFORE installing a heat pump, that would make that approach rather less practical for the average punter.

@majordennisbloodnok @cathoderay 

As I say above the major issue with any short term measurement is the heat capacity of the house which will, in most cases, introduce a large error/uncertainty which will be there even if one could precisely measure the energy put in.  Since its impossible to measure the temperature at multiple points in the fabric, the only way to deal with this is time - either so you ensure that the fabric returns to the thermal state it started in, or so that any changes in the thermal state (and thus retained energy) represent an amount of energy which is insignificant compared to the energy lost.  The first of these is only possible if the OAT remains pretty constant, which is a rarity, so we need in practice to rely on the second. 

I read somewhere that the people who do this measurement take 3 weeks, which seems reasonable.  If the house takes ~ one day to cool to OAT once the heating is switched off, then 3 weeks= 21 days = about 5% max error due to the uncertainty in amount of retained heat.

Posted by: @majordennisbloodnok

↑

it doesn't necessarily measure the actual amount of heat energy delivered into the property. On the other hand, if one uses flow rates and water deltas then presumably that could be done

That is exactly what I have been doing, since early last year (late March)! On my standard chart, which is based on minute by minute data, you can see the RWT and LWT, and I can calculate the delta t, flow rate is also recorded, but not plotted, the chart is already quite busy enough. Both delta t and flow rate come from the Midea sensors over modbus, but the have been semi-verified, see posts passim. The specific heat of the circulating fluid is a constant, so long as the fluid stays constant. That all comes together to get the red bars on the lower half of my charts, the energy delivered over the last hour. The scatter plots I posted recently plot this energy delivered against the OAT. Link to post with recent chart.

Posted by: @cathoderay

↑

Posted by: @majordennisbloodnok

↑

it doesn't necessarily measure the actual amount of heat energy delivered into the property. On the other hand, if one uses flow rates and water deltas then presumably that could be done

That is exactly what I have been doing, since early last year (late March)! On my standard chart, which is based on minute by minute data, you can see the RWT and LWT, and I can calculate the delta t, flow rate is also recorded, but not plotted, the chart is already quite busy enough. Both delta t and flow rate come from the Midea sensors over modbus, but the have been semi-verified, see posts passim. The specific heat of the circulating fluid is a constant, so long as the fluid stays constant. That all comes together to get the red bars on the lower half of my charts, the energy delivered over the last hour. The scatter plots I posted recently plot this energy delivered against the OAT. Link to post with recent chart.

 

Which is pretty good I would say

The only check I would want to do is plot daily energy against degree days.  There is a risk that plotting instantaneous vs OAT depresses the low end, because its so rare that low temperatures occur for more than a few hours overnight and thus retained heat 'smudges' the energy input.  Plotting vs degree days eliminates this risk.  Im not saying that the risk is manifested, but as its easy to get degree days its a check well worth doing.

And of course I can do this, measure energy out, but only because I have the data I do. Since most people don't have access to this data for their heating system, they can't do this. Given we are trying to develop a universally applicable system, I have been barking up the wrong tree. Using energy in, which most people do have access to, via daily readings of their meter, and then adjusting that by efficiency to get energy out is an option, but one fraught with unknowns, not least how do you know the efficiency of your fossil fuel boiler? Back to the drawing board... 

@jamespa - we cross posted. I think there is a fundamental flaw in my suggested approach as described above, most people don't have access to the data they need. I am nonetheless working on plotting daily energy delivered against average daily OAT and will do the same with degree days if I can find a suitable data set. Or maybe I can calculate my own degree days, from the data I have?

Posted by: @cathoderay

↑

...will do the same with degree days if I can find a suitable data set. Or maybe I can calculate my own degree days, from the data I have?

try https://www.degreedays.net/

Posted by: @cathoderay

↑

Using energy in, which most people do have access to, via daily readings of their meter, and then adjusting that by efficiency to get energy out is an option, but one fraught with unknowns, not least how do you know the efficiency of your fossil fuel boiler?

If you have a smart meter you have, or at worse your energy supplier has, half hourly gas meter readings.  That's pretty close to energy in (cooking is negligible)

I grant that the efficiency of the gas boiler is an unknown, but one can probably bracket it.  If its actually condensing (no plumes of vapour) then its likely 100-110% efficient.  If its a condensing boiler thats not condensing then maybe 90-100% efficient.  If its older (to be not a condensing boiler it will need to be 20 years old) then rather less and rather less certain (but there are likely stats out there).

For an older boiler if you were to assume 90% you are overestimating by maybe as much as 20%.  Otherwise you are likely within 10% I would guess, but needs some validation to be certain).  

Good enough for a sanity check on other data certainly.  

@jamespa - thanks. I was reading something on that website as you posted on how degree days are calculated. It seems the preferred method integrates the hourly differences between the hourly mean temp and the degree day baseline (basically an 'area between the lines' calculation) but there is also a simpler method that just uses the daily mean temp. As I have both (hourly means and daily means, and hourly actuals for that matter) I should be able to cook up something. One question though, isn't the degree day value just the OAT shifted by degree day baseline value, such that a linear regression will have the same multiplier, only the intercept will change? Or have I missed something?

Posted by: @cathoderay

↑

One question though, isn't the degree day value just the OAT shifted by degree day baseline value, such that a linear regression will have the same multiplier, only the intercept will change? Or have I missed something?

It is this averaged over 24hrs.  

The reason I think that the daily averaging is useful is that it distinguishes days when its cold all day from isolated hourly periods when its cold.  If its -2 for a couple of hours but otherwise +4, I wouldn't expect the energy in during the cold periods to be the as low as it is on a day when its -2 all day, unless weather compensation is in place.  WC is rarely applied to boilers in the UK and since this is a discussion about heat pump sizing the algorithm has to work with a boiler.

So, reviewing all that has been discussed above, what do we think should be the most basic steps forward?

Not looking to revolutionise MCS in one swoop, but optimising all the info that is available and accessible - incl. A blower test for example

I'm thinking of suggestions that could be reasonably implemented within the skillset of a typical design and installation company, and would undoubtedly improve the system design and product spec'n mistakes that are still seeding doubt in the industry.