Posted by: @cathoderay

↑

Posted by: @cathoderay

↑

There isn't really enough data here, meaning I am being somewhat speculative, but I think there are one or two periods where comparisons can be tentatively made...

 

 

I think it might take a bit of thought about how to analyse data to minimise the amount needed.  Part of the problem is that the signal is quite small relative to the noise.  For example if setback averages 2C for 9hr per day, at a typical IAT-OAT delta of 10C, then the daily saving in energy lost from the house is 9/24 x 2/10 ie 7.5%.  Unless cop variations work in your favour this, means that the maximum saving in energy in is also 7.5%.  That's not a lot to dig out of the quite considerable noise.

There, as you may be aware, signal processing techniques to extract signal from noise, but I can't immediately think of one easily applied to heat pumps.

In the absence of any high probability that we will get a run of days when the OAT profile is identical, enabling a true control experiment to be run, I've been giving some thought to possible experimental techniques to extract the signal from the noise.  Two come to mind, but both require quite a lot of measurement:

'Synchronous detection' (This technique is frequently used to extract signal out of noise in many walks of life - its not problem specific.  There are almost certainly people who know how to do this both well and efficiently, its 30 years since I last used the technique):

establish a regular pattern of days when you set back and days when you dont (perhaps 2 days on and 2 days off, or, in order to fit it into a weekly schedule, 2 days on, 2 days off, 2 days on, 1 day off).  Im suggesting 2 days on/off to ensure that recovery truly happens, it might need 3.

Then run for several weeks, measuring the total energy consumed in each 24 hr period (or perhaps each rolling 24hr period with measurements at hourly points)

Then analyse the data looking for signals with a periodicity which matches the pattern of setback/not set back (it may be necessary to employ some signal processing techniques on the data to find the component of the noisy signal with that varies with the period of interest

If you collect and analyse enough data this should give a tolerably reliable indication of the actual saving even with variations in OAT profile.

There are several problems with this technique not least

1. that you will need to keep it going for at least a couple of months (I guess) to get enough data

2. Im not entirely sure how to deal with the likely lag between stimulus and response (see also note 1)

3. related to the - above extracting the synchronous signal is not so easy so far as I am aware, but someone who knows about signal processing can very likely correct me (see also note 1) 

'Degree days' (obviously this is problem specific):

run the heat pump in setback mode for a substantial period of time (perhaps 30 days or more)

run the heatpump without setback for a similar time

as above measure the total energy consumed in each 24 hr period (or perhaps each rolling 24hr period with measurements at hourly points)

Get degree-days for your location from degree-days.net

Make two plots of energy vs degree days, one with setback, the other without (possibly time shifting the boundaries for max correlation)

Compare the slopes

the main problem with this technique is that, certainly for my house, the degree day vs consumption plot, whilst reasonably well correlated, is still a bit of a scattergram, so it may require several months of data to get a proper result.(see also note 2)

Lab measurements

The alternative to either of these is of course a lab where the heat pump is tested in controlled conditions.  I don't have one of these but there are some around.  Does anyone have any contact with someone in possession of a heat pump lab that might be persuaded to devote some time to the problem, which is of considerable significance?

Notes

(1) thinking about extracting the synchronous signal  - it may be as simple as multiplying the energy consumption from alternate periods by 1 and -1 and then summing (the result being the signal).  Dealing with the lag might be as simple as doing that but time shifting the period boundaries (in the data analysis) until the signal is maximised).  Someone who knows about signal analysis please advise!

(2) in principle degree days could be applied as a correction to individual measurements to enable them to be robustly compared.  However if, like me, your degree day vs consumption plot is not particularly well correlated, then the uncertainty of the correction might be not much better than the uncertainty arising out of the variable conditions.

Posted by: @derek-m

↑

I am not certain why you denigrate anyone who is trying to answer the question that you initially posed.

For the umpteenth time, and for the avoidance of all doubt whatsoever and forever, I am not denigrating you. What I am doing is vigorously questioning the modelling approach, which from experience I have found to be far from perfect on many occasions, especially when black boxes are involved.

Posted by: @derek-m

↑

A value of 250kWh was chosen, as stated previously, to represent a medium to high thermal mass building, though as you stated previously there does not appear to be a standard method to quantify a building's thermal mass.

I already know this, what I am asking is why 250kWh? Moreover, why do the units change (W on the spreadsheet, kWh here, with neither being the standard units seen elsewhere), and why is it thermal mass here, while on the spreadsheet it is thermal capacity? If there is no standard way to quantify a building's thermal mass, should we even be using it? Or if we do, then shouldn't we explain how we attempted to quantify it?

That said, I am not sure you do use it! Looking at rows 21 to 24 in the play area, they appear to keep and hourly profit and loss / balance sheet for energy, with row 24 being the current 'balance in the bank', but I don't think any of these cells (the ones in row 24) are referenced by any other cells, suggesting the thermal mass/capacity is not incorporated into the model (it doesn't affect anything), though it is reported as part of the output. In other words, it is a dependent variable, and not an independent (predictor) variable.   

Posted by: @derek-m

↑

2b) It is indeed true that heat pumps do vary their LWT in the real World, possibly due to variations in OAT, or maybe because the heat pump is cycling, though on some charts I have seen recently there have been periods when the LWT has been fairly constant. How many decimal places would you prefer.

You know perfectly well the point I am making about decimal places, spurious accuracy, no heat pump is ever going to maintain a LWT for a 24 hour period to an accuracy of two decimal places. I'm sure there may have been charts showing periods of 'fairly constant' LWTs (to two decimals places?) but by far and away the dominant pattern for LWTs is they vary all the time. A model that doesn't have that dominant pattern is rather like a model of the sea in which the water never goes up and down...

I do applaud your efforts, I really do, as I do your willingness to put them in the public domain, but surely at least some of the reason for doing that is to have others study, review and critique your work? Surely we should see this ultimately as collaborative work, in which we attempt to get answers to the practical questions we have - like does a setback save energy without compromising comfort? You favour modelling, I favour direct observation, but that's fine, and we both benefit because each can be compared to the other. Just as I question you modelling methods, you and others can question my observational practices, as @jamespa has recently done.   

Posted by: @jamespa

↑

In the absence of any high probability that we will get a run of days when the OAT profile is identical, enabling a true control experiment to be run, I've been giving some thought to possible experimental techniques to extract the signal from the noise.  Two come to mind, but both require quite a lot of measurement:

Interesting, I will give it some thought. Python is good at automated data collection, and even a goon like me can write the code, given @derek-m's helpful assistance eg in working out how to read binary data.

In the meantime, I was going to ask you another question, if I may, as another way of looking at the question of whether setbacks save energy. It has been touched upon before, but this is if you like a succinct form of the question. Imagine we have three identical buildings with identical heat pumps all in a field together, but not so close that they can affect each other. By some freak of nature, the weather and OAT stay constant all the time, the OAT being 5 degrees. In the first building, the heat pump maintains the IAT at a steady average of 20 degrees over a day. In the second building, the heat pump maintains a steady average IAT of 10 degrees over a day. Now, I think we would all agree that the second building will use less energy that the first building. But what about the third building, which maintains an average IAT of 15 degrees, with that average coming about because it has a 12 hour period of 10 degrees, followed by a 12 hour period of 20 degrees. How much energy will it use, compared to the other two?

The general question is whether a building with a lower overall average IAT must, all other things being the same, always use less energy than a building with a higher overall average IAT? And then, if the answer is yes, does it not follow that the same building, when running with a setback as compared to no setback, must use less energy, because its overall average IAT is lower?        

@jamespa

Sorry, Freudian slip, it should have been labelled Wh.

The data, including Energy In Reduction and Energy Out Reduction, was just extracted from the modeling tool. I did that rather than post the modeling tool, which is not too simple to operate.

I provided the two tables to highlight the fact that one of the main factors that affects the level of energy reduction that may be achievable, is dependent upon how quickly IAT is to be recovered once the setback period has ended. A faster recovery, requires a higher LWT, which in turn may reduce the efficiency, and in so doing use more electrical energy.

I pointed out quite some time ago that reducing IAT by 1C would provide a reduction in energy consumption in the order of 10%. In my case reducing IAT down to 20C would be approaching divorce territory, let alone going to 19C or below.

For reference, I'm using Excel 2010.

The major problem I experienced with the latest modeling tool was how to extract the relevant COP values for each 1 hour period for the relevant LWT, and changing operating conditions. The COP and the maximum energy output values are therefore extracted from the manufacturers data contained within the tables in the Detailed Data sheet.

The underlying philosophy is that the Energy Supply value in cells D21 to AA21, is a measure of how much thermal energy would be emitted by the heat emitters at the relevant LWT contained in cells D18 to AA18.

The COP at this LWT is derived from the manufacturers tables, via the WC Table.

The electrical energy input (PI) in cells D19 to AA19 is obtained by dividing the Energy Supply value by the COP value.

Extracting thermal energy output from the manufacturers data does not work, because at an OAT of say 10C and a LWT of 40C, the thermal energy output from the heat pump could be anywhere between 4960W and 16520W when running constantly. The output could be even lower if cycling. The actual energy output is dependent upon the loading, which is how much thermal energy is being extracted by the heat emitters. This of course effects COP and PI.

Please feel free to take a copy of the spreadsheet and modify it to your hearts content.

@jamespa

I think that you will find that degree days don't work well with heat pumps, because the COP varies quite considerably with OAT.

Posted by: @cathoderay

↑

In the meantime, I was going to ask you another question, if I may, as another way of looking at the question of whether setbacks save energy. It has been touched upon before, but this is if you like a succinct form of the question. Imagine we have three identical buildings with identical heat pumps all in a field together, but not so close that they can affect each other. By some freak of nature, the weather and OAT stay constant all the time, the OAT being 5 degrees. In the first building, the heat pump maintains the IAT at a steady average of 20 degrees over a day. In the second building, the heat pump maintains a steady average IAT of 10 degrees over a day. Now, I think we would all agree that the second building will use less energy that the first building.

I cant comment on everyone, but I agree.

Posted by: @cathoderay

↑

But what about the third building, which maintains an average IAT of 15 degrees, with that average coming about because it has a 12 hour period of 10 degrees, followed by a 12 hour period of 20 degrees. How much energy will it use, compared to the other two?

The general question is whether a building with a lower overall average IAT must, all other things being the same, always use less energy than a building with a higher overall average IAT? And then, if the answer is yes, does it not follow that the same building, when running with a setback as compared to no setback, must use less energy, because its overall average IAT is lower?        

Maybe.  The problem is that the heat pump has to work harder to heat the building from 10 to 20 than it would keeping it at an average of 15. So whilst its undoubtedly true that the loss from the building is very similar in both scenarios, that does not necessarily mean that the energy input to the heat pump required to make up that loss is.

This in fact is the whole crux of this issue.  We assume (it may not be true) that fossil fuel boilers have a linear relation between heat in and heat out which is largely independent of flow temperature, providing that the FT is low enough that the boiler is condensing.  Heat pumps definitely don't behave in either of these ways and in particular are significantly more efficient at lower flow temperatures.  That's why spatial zoning is not always a good thing (heat geek have a very plausible worked example which proves this) and why temporal zoning (ie setback) may also not always be a good thing.  

Its pretty clear that the answer depends on several factors, including building 'thermal mass'.  Contrary to what @derek-m has suggested (I think) several posts ago, my current view is that setback is more likely to be successful with buildings that have a low thermal mass, and less likely to be successful in buildings with a high thermal mass.  My rationale for this is as follows:

• Consider first the limit condition of a building with zero thermal mass.  In this case the building will respond instantly to changes in energy supplied, the heat pump does not have any 'making up' to do at the end of setback, and thus its obvious that setback must save energy in as well as energy out.
• Now consider in contrast a building with a very high thermal mass (and/or extremely good insulation), so high that it remains warm for days after the heating is turned off.  In this case the building acts simply an integrator of the total energy supplied to the building, and anyway wont respond in time to setback to have any material effect.  So its equally obvious that the most efficient way to run it is to deliver that energy to the building in the most efficient way the heat pump can manage ie by leaving it on 24*7 at the lowest FT capable of delivering the required amount of energy and in this case turning the heat pump off overnight will result in a higher overall energy consumption, because the only way to achieve the same IAT is to raise the FT.

Posted by: @derek-m

↑

Sorry, Freudian slip, it should have been labelled Wh.

The data, including Energy In Reduction and Energy Out Reduction, was just extracted from the modeling tool. I did that rather than post the modeling tool, which is not too simple to operate.

Thanks

So in the first example, if I understand it correctly,

the total energy in is the sum of the 'energy supply' line /4.5 = 22055Wh, and in the second the sum of the 'energy supply' line /4.35 = 25889Wh.

the reduction in energy in the first case is 4819Wh, 22%, however of this 6% must be attributed to the fact that the 'standing' IAT has reduced from 20 to 19.45, so the act of setback is apparently saving 16% at the expense of quite an extended recovery time

the reduction in energy in in the second case is 991Wh ie 4%, but the recovery time is much better.

Have I got these correct?

Posted by: @derek-m

↑

xtracting thermal energy output from the manufacturers data does not work, because at an OAT of say 10C and a LWT of 40C, the thermal energy output from the heat pump could be anywhere between 4960W and 16520W when running constantly. The output could be even lower if cycling. The actual energy output is dependent upon the loading, which is how much thermal energy is being extracted by the heat emitters. This of course effects COP and PI.

Indeed, although from the manufacturers tables (with interpolation) is where, ultimately, COP and Max/Min o/p must come, given the load that the emitters impose.. 

I can more or less see a neat-ish way to do this in a fairly transparent custom function if you ignore the variation of cop with load (which may be a sufficiently good approximation and is anyway data that is not available for many heat pumps) but I cant currently work out a neat way to do it (other than creating intermediate tables, as you have) if you wish to take account of the variation of COP with load.  

I can only comment from a seat on the pants experience of how my system operates but it falls closely within what I perceive to be the test building  makeup, ie 14 Kw Ecodan to large radiators in a leaky timber frame house in exposed conditions, Auto Adapt in use since commissioning 4 years ago.  This would appear to meet @JamesPa's zero, or near zero thermal mass building. 

The house is maintained most of the day at 19C. I calculate (seat of the pants) that the lowest OA temperatures are overnight so the HP is less efficient then and so "setback" to 15C at night and at 0600 increase the set temp to 18C and at 1000 to 19C. The house rarely falls to less than 15C overnight but if it does the HP brings the temp back to 16C and returns to setback, or one one noted occasion when the OAT plummeted in the early hours and the IAT was still above 15C at 15.5C. Go figure.

The morning recovery period only sees the LWT runs at 35C or occasionally 36C/37C and only significantly rises to heat the DHW, which is set to heat on demand and never seems to operate at night anyway, even after bedtime showers.   This does not appear to be the models experience but that could easily be me mangling the configuration or misunderstanding what I am looking at.

Notwithstaning the noted reservations some have, good work guys, thank you.

@cathoderay

There is no 'black box', there are no hidden cells, everything is open to view. If you don't understanding what you are seeing then please ask.

I would have thought by now that you would have realised that 250kWh is the same as 250000Wh. Please let me know which you prefer.

I treat a building's thermal mass as if it were a hot water cylinder. If 200 litres of water is heated from 20C to 50C it will take approximately 7kWh of energy. If this water now cools to 40C, it will have lost about 2.3kWh of energy and to reheat it back to 50C would need 2.3kWh of energy to be added.

If a building has a thermal mass that contains a thermal capacity of 250kWh of energy at an IAT of 20C, with reference to 0C, then if the IAT is reduced to 19C, the thermal energy within the building will now be 237.5kWh, having lost 12.5kWh of energy to the real World. If you are not happy with Thermal Mass or Thermal Capacity then please provide a suitable alternative. If you can provide an alternative method of quantifying a building's thermal capacity then please do so.

The values from row 24 are actually used within the table, seek and ye shall find.

My own home heating controller does actually display IAT to 2 decimal places, but you are correct that it is not the norm. Data is show to 2 decimal places to illustrate the slight changes that can, occur and how the rate of change varies as the system comes towards balance. If you don't like data to 2 decimal places then I am quite happy for you to modify your copy of the modeling tool, if you need guidance then please ask.

Posted by: @jamespa

↑

Posted by: @derek-m

↑

Sorry, Freudian slip, it should have been labelled Wh.

The data, including Energy In Reduction and Energy Out Reduction, was just extracted from the modeling tool. I did that rather than post the modeling tool, which is not too simple to operate.

Thanks

 

So in the first example, if I understand it correctly,

the total energy in is the sum of the 'energy supply' line /4.5 = 22055Wh, and in the second the sum of the 'energy supply' line /4.35 = 25889Wh.

the reduction in energy in the first case is 4819Wh, 22%, however of this 6% must be attributed to the fact that the 'standing' IAT has reduced from 20 to 19.45, so the act of setback is apparently saving 16% at the expense of quite an extended recovery time

the reduction in energy in in the second case is 991Wh ie 4%, but the recovery time is much better.

Have I got these correct?

 

 

 

When using the modeling tool it is first necessary to calculate the reference conditions, which in this case was 24 hour running at an OAT of 10C, giving the initial conditions where Energy In (PI) would be 1120-1121W. Over 24 hours this totals 26893W.

Employing a 6 hour setback reduces PI to 22074, which when subtracted from 26893 gives an Energy In Reduction of 4819. The same method was used to produce the Energy Out Reduction value.

Table 1 was primarily to illustrate that without recovery boost it may not be possible to fully recover IAT. My calculation of energy saving would be given by 4819/26893 = 17.92%.

Table 2 was to illustrate that with recovery boost the possible energy reduction would not be as great. In this case the calculated energy saving would be 991/26893 = 3.68%.

I hope this answers your questions.

Posted by: @jamespa

↑

Posted by: @derek-m

↑

xtracting thermal energy output from the manufacturers data does not work, because at an OAT of say 10C and a LWT of 40C, the thermal energy output from the heat pump could be anywhere between 4960W and 16520W when running constantly. The output could be even lower if cycling. The actual energy output is dependent upon the loading, which is how much thermal energy is being extracted by the heat emitters. This of course effects COP and PI.

Indeed, although from the manufacturers tables (with interpolation) is where, ultimately, COP and Max/Min o/p must come, given the load that the emitters impose.. 

I can more or less see a neat-ish way to do this in a fairly transparent custom function if you ignore the variation of cop with load (which may be a sufficiently good approximation and is anyway data that is not available for many heat pumps) but I cant currently work out a neat way to do it (other than creating intermediate tables, as you have) if you wish to take account of the variation of COP with load.  

 

I think that you will find that the modeling tool already takes account of the variation of COP with load, also with OAT and LWT, by extrapolating the correct values from the manufacturers data.

I can explain if you wish.