Posted by: @robl

↑

A few ideas to consider:

The ashp is outside - no doubt it's measuring produced power at the unit, so will include any pipe losses that the oil boiler didn't have.

Does the ashp controller "know" it has glycol in it? - it makes a noticeable change to calculated heatflow.

How does the ashp actually calculate heatflow?  I know it seems obvious that it will be based on temperature measurements and flowrate, but are we really sure it doesn't just base it on electricity used?  I'm betting there's no accuracy spec on the heat produced.  Colour me cynical.

I bet the ashp has a immersion in it.  Any chance it's stuck on?  

For all the above, I recommend an independent, mid certified, heat meter.  I would put it where the pipes come into the house.

I had understood the figures provided by @cathoderay to be measured electricity consumption (8 ish MWh/year) which he had multiplied by an  assumed COP of 3 to get 24MWh/year.  I didn't see any actual heat measurements.  Maybe I misread.  

If this is the case then pipe losses come into play together with any inaccuracy in the electricity consumption measurements which, if its a decent meter, should be small.  Hopefully @cathoderay can clarify.  Also perhaps he can tell us what the quiescent consumption is.

Its beginning to feel to me as if the explanation might be a combination of:

• previous intermittent heating actually resulted in house getting quite cold at night and when not heated during the day, so constant heating is resulting in significantly higher losses than previously
• poor modulation depth of Midea combined with likely oversizing leading to excessive short cycling
• pipe losses as you say
• immersion stuck on (surely not(?!) - is it supplied from the same power source?
• and some other things yet to be determined included some that have been speculated about, perhaps each contributing 10-20%.

we already know for a fact that either the first or the second is the case, we cant tell which and either would be a major contributor.

We need the extra data requested above to do some more diagnosis.  Hopefully when @cathoderay is back online he can provide some more info.

@cathoderay

I think that we went through many of these aspects quite some time ago, but it does no harm to have a recap.

Your calculations based around the oil boiler contain a number of assumptions, which may or may not be accurate. You state that the oil consumption was between 1000 and 1200 litres, which value did you use in your calculations? The oil energy input is not actual, but derived. What is the degree of error to be expected from the derived calculations?

You assess that the indoor temperature was approximately the same for oil boiler against heat pump, even though the oil boiler could be off for extended periods. A 1C difference in indoor temperature could equate to a 10% difference in energy demand. You also estimate that the oil boiler was 80% efficient, which again could be a source of error in the calculations.

There appears to be no allowance made for producing hot water using the immersion heater.

On the heat pump side you estimate a COP of 3, which I feel may be on the generous side, since the heat pump is operating for longer periods during colder weather conditions, when thermal energy demand is higher and heat pump efficiency is lower. This is further compounded by the fact that you have a plate heat exchanger (PHE) and secondary water pump, that could be increasing the required LWT by up to 5C and in the process reducing the efficiency of the heat pump even further. Other forum members have already proven that a PHE can significantly lower overall efficiency.

I doubt that the data provided by the Midea controller is particularly accurate, for both temperature measurements and particularly flow rate. If the Delta T between LWT and RWT is showing a value of 5C, when the true value is 4C, any calculated thermal energy output would be 25% higher than the true value.

I don't think that your 'paradox' is due to any one thing, but possibly a combination of quite a number.

Posted by: @derek-m

↑

I don't think that your 'paradox' is due to any one thing, but possibly a combination of quite a number.

Completely agree here.  However, reducing it to the basics which I think we know irrespective of assumptions about COP or the accuracy of data from the Midea:

• The annual oil consumption was between 1000 and 1200l, equating, at the mid point to about 11.4MWh at 100% efficiency, less if the oil boiler is less than 100%
• The annual electricity consumption for the ASHP is apparently about 8MWh, measured by a dedicated meter.

That's absolutely dreadful, however its spun, so something, or some things (as you suggest) are wrong.

Furthermore we know for certain that either the HP, which does not modulate down well, is well oversized or the average house temperature when it was heated by oil was significantly lower (not just a degree, several degrees) than it is when heated by ASHP, possibly because of intermittent heating, or a combo of both

I would venture to speculate that one of these is a major part of the cause and then the rest is made up of a combination of the factors you list plus the factors @robi has listed (possibly plus some others yet to be discovered.

We need to find out from @cathoderay some more data to allow us to determine the extent of the undersizing and/or deviation of the average temp under the previous oil regime from the current.  With a bit of luck that will get us quite a lot closer and we can then assess several 10% effects, either that or the mystery will continue.

Thank you all very much for your comments which I will reply to shortly. In the mean time, here is some data from the recent power cut, which by coincidence happened on April Fool's day (roughly 0900 to 1300), but don't be fooled, this is real data. This shows among other things the rate of cooling of the house when the power and so heating is off, and also the rate of recovery when the power/heating come back on. All the data in the chart apart from room temp comes from the Midea modbus minute data (that is, the data logged every minute from the Midea wired controller over modbus) and so it just stops during the power cut. The room temp data comes from battery powered room temp data logger in the kitchen. This of course caries on logging during the power cut, but it only collects hourly data, which is why there are points rather than a line. The design room temp for the kitchen is 18 degrees:

I think the main observation to be made is that the rate of cooling and recovery during and after the power cut is I suggest slow, consistent with the high thermal mass of an old stone wall cottage with a solid concrete ground floor, though it should be noted that the outside ambient temp wasn't that cold, at around 10 to 11 degrees. The peak in amps/LWT/RWT immediately after the power cut is a coincidence, it is the DHW heating coming on, as you will recall, it is timed to come on between 1300 and 1400. This means the heating will have remained off during that period, so the heating off period was actually a little longer than the power cut.

Other things of possible interest: (1) cycling is clearly visible, in the variations in amps in and LWT/RWT (2) the flow rate remains remarkably constant, meaning the modulation of output is done by varying the LWT/RWT delta t (note how it is lower during the warmer part of the day on 1st April, and then increases overnight as the outdoor ambient temp falls), plus in this example, when it can't get the delta t low enough, by cycling and (3) this is a 'dumb' weather comp curve in operation: it only knows the outside ambient temp, not that at 1300 the room temp was below where it should be, and so the output could do with a boost to achieve a catch up. One day my python/minimalmodbus scripts might be able to make the weather comp curve into a smart or adaptive curve: if desired room temp minus actual room temp > 2, then boost LWT = current LWT plus 1 etc. You get the idea.

There is another period of potential interest back in early December 2022 when I manually put in an overnight setback, and the house cooled and then had to recover. I wasn't collecting modbus data then, but I should be able to dig out outdoor ambient temp (from the Met office data) and room temp (from the data logger). As I recall, at the time, I wasn't impressed with the heat pump's ability to recover from the setback, and didn't apply the setback again after that one night trial.

There may be a delay should I feel duty bound to keep an eye on developments in Westminster during the day... 

Here's a rather ugly chart from the data logger's software showing the kitchen room temp during the overnight setback to 16 degrees on the room stat 2200 to 0400 on the night of 5-6th Dec 2022, with an overlay showing the setback period. It's a gee-whizz graph (I can't zero the minimum on the Y axis) but what it actually shows is that it takes ~6 hours to drop ~3 degrees (it goes from 19 to just below 16 degrees, don't know whether the room stat fired as it should have done or whether the setback period ended) ie 0.5 degrees/hour but equally the recovery is slow, it only reaches the design temp, 18 degrees,  again around noon on the 6th, a whole 8 hours after the setback ended. The outside ambient temp fell from 5.6 to 2 degrees during the setback, and hovered between 1.3 and 4.7 during the recovery period.

At the time, this slow recovery persuaded me overnight setbacks are not feasible using a 'dumb' weather curve, the recovery period is too slow, because the 'dumb' curve doesn't know that the room temp is less than it should be. It needs some form of adaptive curve, that factors in the room temp is below par and a boost is needed, along the lines described earlier.      

Posted by: @jamespa

↑

According to the databook the minimum output of the Midea unit at LWT55 OAT 7C is 9kW

I agree that is what the Engineering databook shows (page 45, 9050W to confirm we're on the same page) but I am not sure what the max/norm/min rows really mean. Max probably means what it says, but I think it can go a lot less than 'min'. Here's the reported (current total minus previous total) and  calculated (from the midea modbus LWT/RWT delta t and flow rate) output during a period with variations in ambient temp (3-4th April 2023). Somewhat reassuringly, the Midea reported output (which is always an integer, because it only reports integer values) and the calculated value (to 2 decimal places) are in agreement. These are hourly values, ie for the last hour before the data point:

As can be seen, when it is milder, the output drops down to 4 kWh (middle of the day on the 4th, the spike at 1300 is the timed DHW heating), 4/14 = 29% of nominal output. I seem to recall that @derek-m normally suggests that most heat pumps can modulate down to about 25% of their nominal output (@derek-m, please correct me if I am wrong).

What we are up against here is obfuscated manufacturer's data - they don't define what max/norm/min mean (as far as I can see, it's the 'Capacity Level' whatever that means). What is clear is that real world output is far less than this 'min' figure. 

Posted by: @jamespa

↑

Any idea what the cycle time/frequency is?

See the April Fool's Day chart, where the cycling can be seen. About 6 times in 3 hours so twice an hour or every 30 mins give or take. But note that is during a relatively mild spell (~10-11 degrees). If we take a period that has more marked highs and lows we get this (apologies for the clutter):

On the left hand side, with cold overnight ambient temps, there is no cycling (the spikes are defrosts, note how the LWT dips below the RWT), on the right hand side, during the warmer daytime period, we get true cycling (LWT stays above RWT).  

Posted by: @robl

↑

The ashp is outside - no doubt it's measuring produced power at the unit, so will include any pipe losses that the oil boiler didn't have.

Does the ashp controller "know" it has glycol in it? - it makes a noticeable change to calculated heatflow.

How does the ashp actually calculate heatflow?  I know it seems obvious that it will be based on temperature measurements and flowrate, but are we really sure it doesn't just base it on electricity used?  I'm betting there's no accuracy spec on the heat produced.  Colour me cynical.

I bet the ashp has a immersion in it.  Any chance it's stuck on?  

For all the above, I recommend an independent, mid certified, heat meter.  I would put it where the pipes come into the house.

Thanks for your pertinent questions. Answers, as best I can:

It is outside, but the external pipe runs are short, perhaps 6ft flow and 6ft return and lagged so minimal losses.

No body has any idea how Midea calculate the energy in and out, or what assumptions are made, including whether the system has glycol in it (it does). There is some reassurance (colour me cynical as well!) in that the Midea reported energy in/out, which is based on total lifetime energy in and out as current value minus previous value, agrees with my calculated value using LWT/RWT delta t and flow rate, with an included adjustment for glycol (fluid specific heat 4.05 instead of 4.2 for water alone), see previous chart showing this correlation. Rather more reassuringly the Midea reported energy in data agrees fairly closely with my external dedicate kWH in meter, on average (it varies over time) about 3% difference (Midea 3% less), which isn't too bad. But the energy out is still all Midea collected data, and we don't know how or where the data is collected, or how accurate it is, or what calculations or 'corrections' are done.

There is no immersion heater in the heat pump, or if there is, it is not connected.

I have considered a full independent heat meter installation, but at the moment am not inclined to fit one, on the grounds of cost and disruption, allied to the fact I do have a way of verifying the Midea data, through a basic inline analogue flow meter on the primary circuit (tolerable agreement with the Midea reported flow rate) and I can stick thermometers on the flow and return pipes. These however are currently rather suspect in principle (IR thermometer plus portable thermocouple on a multimeter) but in practice the readings are not absurd.

There is also a question of whether a full heat meter set up would actually be more useful on the secondary circuit (I have a plate heat exchanger) as that is what actually goes round the house, but that then means I can't measure the DHW heating as it is on the primary circuit.

Posted by: @jamespa

↑

I had understood the figures provided by @cathoderay to be measured electricity consumption (8 ish MWh/year) which he had multiplied by an  assumed COP of 3 to get 24MWh/year.  I didn't see any actual heat measurements.  Maybe I misread.  

If this is the case then pipe losses come into play together with any inaccuracy in the electricity consumption measurements which, if its a decent meter, should be small.  Hopefully @cathoderay can clarify.  Also perhaps he can tell us what the quiescent consumption is.

No, you didn't misread, that is correct. Interestingly, the Midea kWh out is probably not dissimilar. The current lifetime energy out figure is 35,600 kWh, covering Feb 2022 to date. However, most of that usage will have been over the 2022/23 heating season, as the system wasn't fully running in Feb 2022 (commissioning problems) meaning the active heating months have been Mar to Apr 2022 and Oct 2022 to Apr 2023 (inclusive; summer DHW use is relatively very small, cam probably be ignored). That means a very crude average of around 4000 kWh per heating month, or 28,000 kWh over a 7 month heating season. The May to May 8823 energy in (from the dedicated external kWh meter) x 3 (COP) = 26,500 kWh out isn'y a million miles from 28,000 kWh. Note that these figures are based on totally independent data, Midea reported data vs external kWh meter, apart from the COP, which is based on a sensible general estimate influenced by the COP based on the reported Midea data (3 is about normal, and as it happens the SCOP based on reported Midea data is about 3). This suggests the house is swallowing up huge amounts of energy, and to that extent the paradox remains.      

Not sure what "quiescent" consumption - too long and too Latin a word for me! - but if it means background use with the heat pump off, it is about 40 kWh per week in winter (measured as the difference between the dedicated heat pump kWh meter and the whole house kWh meter). 

Posted by: @jamespa

↑

Its beginning to feel to me as if the explanation might be a combination of:

• previous intermittent heating actually resulted in house getting quite cold at night and when not heated during the day, so constant heating is resulting in significantly higher losses than previously
• poor modulation depth of Midea combined with likely oversizing leading to excessive short cycling
• pipe losses as you say
• immersion stuck on (surely not(?!) - is it supplied from the same power source?
• and some other things yet to be determined included some that have been speculated about, perhaps each contributing 10-20%.

we already know for a fact that either the first or the second is the case, we cant tell which and either would be a major contributor.

We need the extra data requested above to do some more diagnosis. 

I think the first one may well be significant, but is it significant enough, if it is to carry the burden of explaining the paradox? Let us imagine that for example using oil I heated the house for 50% of the time, vs 100% of the time with the heat pump. That could reasonably explain twice the use, but not the observed three times or more use. Here for ease of reference is the original 'paradox chart':

Not sure I agree about the poor Midea modulation, see previous comments, nor the over-sizing. I am reasonably confident the 12.3kW total heat loss is about right, because I have checked and checked and checked it 1000 times over (I had the usual installer games of pulling heat losses out of hats etc, but as it happens I did the original room by room losses for the oil CH when I installed it, so I  had a fairly good idea what the right figure is). There is also some observational confirmation that it is about 12kW in that when it gets down to sustained sub zero temps (as happened in the Dec 2022 cold snap), the heat pump can't quite cope, room temps fall by a degree or two, when its warmer it manages fine. But this is of course complicated by the fact that at -2 outside my so called 14kW unit is struggling to put out 11.3kW but this still suggests the heat loss is in the say 11.5 to 12.3 kWh range (my logic is that the point where the indoor temp starts to fall is the point at which the max heat pump output is just failing to match the heat loss, if that happens at an output of 12kW then the heat loss is 12kW). 

I've deliberately stymied the immersion heater. The heat pump doesn't have one (or if it does, it is not connected) and the so-called cylinder Legionella cycle one is definitely not in use (I don't believe in scalding myself for in practice a non-existent risk). I only keep the immersion wired in in case the heat pump goes on the blink, meaning I can maintain a supply of hot water.

@cathoderay 

Lots of data here.

Firstly I would start by reframing your question from

'why does my house take more energy to heat using ASHP than using oil' to

'why am I apparently not achieving the SCOP I expect.' (note the insertion of 'apparently' here to leave open the possibility that there are some other things going on.) 

I don't believe your house has gone from 11MWh/year to 23MWh/year and I doubt you do either, hence the reframing.

Now lets look at what we know about the house:

• under the previous, intermittent heating, regime you burned about 11.4MWh/year  (based on oil consumed) (1)
• if this were 24 hour heating, it would indicate a peak demand of about 6kW at -2C, whereas your system is designed for 12 and has a 14kW unit 
• under the new, essentially constant heating, regime you consume about 8.8MWh/year of electricity (based on a meter specific to the ASHP(?)(2)
• your house temperature falls, if unheated, by about 0.08-0.09 degrees per hour per degree temp difference between OAT and IDT (indoor temp). (3)

NOTES

(1), (2) I checked degree days for 2022-23 and 2021-22 and they are about the same, so those two years are comparable.  The previous year 2020-21 was about 10% 'colder'.

(3) Calculated by taking the slope at the beginning of the fall off both on April 1 and during the set back, which are similar once you compensate for the difference in OAT (outdoor air temp),  The setback plot seems to show a currently unexplained falling off as IDT gets close to 16C, suggesting that there might have been some heating going on as the target temperature is approached.

So far we have made no assumptions about COP or the accuracy of any MIDEA data

You don't say how long the heating was off during the day or during the night when it was heated by oil, but if the OAT were say 4-7C, not atypical during the heating season, the house temperature could easily have fallen by between 5 and 10C during the 'off' period.  Since you were asleep or (presumably) out, you wouldn't notice this.  Without a lot more thought I cant work out how to calculate accurately how much this reduces the heat loss over the heating period, but it certainly feels like it might reduce the time-averaged house temperature by 3-5C.  This information from the degree-day site suggests that this is in the right ballpark, but I would stress it may be out by a factor of 2.  In other words you were doing quite a good job of minimising your costs given your use pattern.  Incidentally condensing oil boilers are about 90% efficient, so you may be underestimating the delivered heat.

Without even more thought I cant calculate accurately how much this changes the energy demand to keep the house heated, it 'feels' like it might be perhaps 20-25% (particularly since the heating is off during the night when its typically coldest) but interestingly, if you follow the instructions on the degree day site for dealing with intermittent heating, it suggests a figure more like 30% (which may be because of the night time effect).

If we now correct 11.4MWh/year by say 25% we get an estimate of what the (oil) consumption would have been had the heating been on 24*7 - 15MWh approx.  Incidentally this would correspond to a peak demand at -2C of 8kW or thereabouts, much more believable than 6kW, given your description of the house and the calculations.

I think that's about all we know about the house at present. Note that everything beyond the four bullets is rough and ready ballpark figures only.  However I think that this does tell us that you were doing quite a good job of minimising your costs given your use pattern.

----------

So now lest see what the graphs tell us about your Midea (assuming we believe them, which we probably don't entirely so this is as much a sanity check as anything else).  I took some measurement from the April 1 graph, in particular the stable period between 14:00 and 18:00 as follows:

OAT =10, IDT=19, flow rate = 1.4cu m/hr = 23l/min, Mark/Space ratio of input current = 4:1; input current when on =5A, LWT = 40/45, DeltaT=4 (approx)

From which we can deduce

electrical power input when on =  1.15kW

average power = 0.92kW

power output when on = 6.4kW (assuming water not glycol)

COP = 5.5

@LWT = 40 the spec sheet says min o/p=6.6 with a COP of 4.92 and an input power of 1.3kW, so all within the same ballpark.  All this tells us of course is that Midea are consistent, not that they are right, because all the input data is from Midea! 

One other sanity check would be useful which is to compare the rate of change of flow temp during the heating period, with the expected rate of change (given the input power and the system volume).  

One other thing about your Midea is that I cant find a SCOP at 55, only a COP at various AW combinations.  The one they quote A7W55 is only 2.95, which suggests to me you should be expecting perhaps 2.5 @ 58 which I believe is what you are running at

------------

So where does this get us so far?  I would say that

• almost for certain what we know about the house tells us that your intermittent heating regime was pretty efficient, and that 24*7 without set back is probably materially increasing your annual consumption.    
• compensating for this alone gets us to an apparent SCOP of perhaps (very roughly) 1.5-2, compared to 2.5-3 expected, so there is still a way to go.
• the data output from the Midea on April 1 is broadly consistent with their databook
• you might be able to improve on your annual consumption (but not your COP) if you implement setback (has your usage pattern changed so you need the house heated during the day, wheras previously you did not?

Unless we can find a way to get a more accurate handle on the contribution of the intermittent heating, there will remain a lot of uncertainty in this part, but I think its still clear that other factors are involved many of which have been referred to by others above.

The next place I would go (based largely on the suggestions of others above) is:

• look at plots when OAT is around 0, 5 and 15 (or the max OAT at which you heat) just to see what these tell us (if anything).  Also when its (supposedly) not heating at all (is there a quiescent demand, ie is the Midea drawing current when its not heating?)?
• check your electrical power meter, compare it with what the Midea is saying.  Do some 'sanity checks'
• measure DT across the HEX and between HEX and Pump to see if losses are contributing significantly
• confirm, for the avoidance of doubt, whether water heating is included in either oil or ASHP figures and any relevant characteristics of same.

I'm not sure this makes any conclusion/measurements more comfortable, and it may well say something about the economics of heat pumps in a situation where the use pattern is such that intermittent heating is acceptable.  However until we are a bit closer to understanding and have assessed the effect of appropriate set backs I wouldn't want to jump to that conclusion.  This is, however, potentially an interesting case study.

@cathoderay 

Lots of data here.

Firstly I would start by reframing your question from

'why does my house take more energy to heat using ASHP than using oil' to

'why am I apparently not achieving the SCOP I expect.' (note the insertion of 'apparently' here to leave open the possibility that there are some other things going on.) 

I don't believe your house has gone from 11MWh/year to 23MWh/year and I doubt you do either, hence the reframing.

Now lets look at what we know about the house:

• under the previous, intermittent heating, regime you burned about 11.4MWh/year  (based on oil consumed) (1)
• if this were 24 hour heating, it would indicate a peak demand of about 6kW at -2C, whereas your system is designed for 12 and has a 14kW unit 
• under the new, essentially constant heating, regime you consume about 8.8MWh/year of electricity (based on a meter specific to the ASHP(?)(2)
• your house temperature falls, if unheated, by about 0.08-0.09 degrees per hour per degree temp difference between OAT and IDT (indoor temp). (3)

NOTES

(1), (2) I checked degree days for 2022-23 and 2021-22 and they are about the same, so those two years are comparable.  The previous year 2020-21 was about 10% 'colder'.

(3) Calculated by taking the slope at the beginning of the fall off both on April 1 and during the set back, which are similar once you compensate for the difference in OAT (outdoor air temp),  The setback plot seems to show a currently unexplained falling off as IDT gets close to 16C, suggesting that there might have been some heating going on as the target temperature is approached.

So far we have made no assumptions about COP or the accuracy of any MIDEA data

You don't say how long the heating was off during the day or during the night when it was heated by oil, but if the OAT were say 4-7C, not atypical during the heating season, the house temperature could easily have fallen by between 5 and 10C during the 'off' period.  Since you were asleep or (presumably) out, you wouldn't notice this.  Without a lot more thought I cant work out how to calculate accurately how much this reduces the heat loss over the heating period, but it certainly feels like it might reduce the time-averaged house temperature by 3-5C.  This information from the degree-day site suggests that this is in the right ballpark, but I would stress it may be out by a factor of 2.  In other words you were doing quite a good job of minimising your costs given your use pattern.  Incidentally condensing oil boilers are about 90% efficient, so you may be underestimating the delivered heat.

Without even more thought I cant calculate accurately how much this changes the energy demand to keep the house heated, it 'feels' like it might be perhaps 20-25% (particularly since the heating is off during the night when its typically coldest) but interestingly, if you follow the instructions on the degree day site for dealing with intermittent heating, it suggests a figure more like 30% (which may be because of the night time effect).

If we now correct 11.4MWh/year by say 25% we get an estimate of what the (oil) consumption would have been had the heating been on 24*7 - 15MWh approx.  Incidentally this would correspond to a peak demand at -2C of 8kW or thereabouts, much more believable than 6kW, given your description of the house and the calculations.

I think that's about all we know about the house at present. Note that everything beyond the four bullets is rough and ready ballpark figures only.  However I think that this does tell us that you were doing quite a good job of minimising your costs given your use pattern and that this is a material contributor, but certainly doesn't explain everything.

----------

So now lest see what the graphs tell us about your Midea (assuming we believe them, which we probably don't entirely so this is as much a sanity check as anything else).  I took some measurement from the April 1 graph, in particular the stable period between 14:00 and 18:00 as follows:

OAT =10, IDT=19, flow rate = 1.4cu m/hr = 23l/min, Mark/Space ratio of input current = 4:1; input current when on =5A, LWT = 40/45, DeltaT=4 (approx)

From which we can deduce

electrical power input when on =  1.15kW

average power = 0.92kW

power output when on = 6.4kW (assuming water not glycol)

COP = 5.5

@LWT = 40 the spec sheet says min o/p=6.6 with a COP of 4.92 and an input power of 1.3kW, so all within the same ballpark.  All this tells us of course is that Midea are consistent, not that they are right, because all the input data is from Midea!

If we believe 6.4kW at IDT-OAT=9, then that suggests 15.6kW at OAT=-2/IDT=20, which clearly is inconsistent with any of the other figures.  This is not yet explained other than to note that, when I look at my half hourly meeting readings and plot them against OAT or daily against degree days, whilst there is a reasonable correlation the spread is also quite high, perhaps due to 'history'.  The same might be true also in your case.

One other sanity check would be useful which is to compare the rate of change of flow temp during the heating period, with the expected rate of change (given the input power and the system volume).  

One other thing about your Midea is that I cant find a SCOP at 55, only a COP at various AW combinations.  The one they quote A7W55 is only 2.95, which suggests to me you should be expecting perhaps 2.5 @ 58 which I believe is what you are running at

------------

So where does this get us so far?  I would say that

• almost for certain what we know about the house tells us that your intermittent heating regime was pretty efficient, and that 24*7 without set back is probably materially increasing your annual consumption.    
• compensating for this alone gets us to an apparent SCOP of perhaps (very roughly) 1.5-2, compared to 2.5-3 expected, so there is still a way to go.
• the data output from the Midea on April 1 is broadly consistent with their databook
• you might be able to improve on your annual consumption (but not your COP) if you implement setback (has your usage pattern changed so you need the house heated during the day, wheras previously you did not?

Unless we can find a way to get a more accurate handle on the contribution of the intermittent heating, there will remain a lot of uncertainty in this part, but I think its still clear that other factors are involved many of which have been referred to by others above.

The next place I would go (based largely on the suggestions of others above) is:

• look at plots when OAT is around 0, 5 and 15 (or the max OAT at which you heat) just to see what these tell us (if anything).  Also when its (supposedly) not heating at all (is there a quiescent demand, ie is the Midea drawing current when its not heating?)?
• check your electrical power meter, compare it with what the Midea is saying.  Do some 'sanity checks'
• measure DT across the HEX and between HEX and Pump to see if losses are contributing significantly
• confirm, for the avoidance of doubt, whether water heating is included in either oil or ASHP figures and any relevant characteristics of same.

I'm not sure this makes any conclusion/measurements more comfortable, and it may well say something about the economics of heat pumps in a situation where the use pattern is such that intermittent heating is acceptable.  However until we are a bit closer to understanding and have assessed the effect of appropriate set backs I wouldn't want to jump to that conclusion.  This is, however, potentially an interesting case study.

@cathoderay

I just had a look back again and spotted this part of one of the plots you posted, where the current in (and to some extent the LWT) is all over the place.  Not sure if thats significant but its interesting.  As mentioned above some plots at a range of OATs other than 10 would be interesting and might either confirm some things already looking probable, or introduce new ideas.

Posted by: @derek-m

↑

Your calculations based around the oil boiler contain a number of assumptions, which may or may not be accurate. You state that the oil consumption was between 1000 and 1200 litres, which value did you use in your calculations? The oil energy input is not actual, but derived. What is the degree of error to be expected from the derived calculations?

You also estimate that the oil boiler was 80% efficient, which again could be a source of error in the calculations.

I really should have written up my method notes better at the time, but what I think I can say is this. I based the oil calculation on the fill data from the supplier (they meter the amount as they fill the tank, and as this is for billing, I think it should be reasonably accurate) for the last 8 years of oil heating use. However, looking closely at the records (the slips printed out by the tanker) the volumes vary, and I know the starting point (how much oil was left in the tank) varied, and the degree of fill (90%? 95%) also varied. But I also assumed that over time these variations should even out, in effect i know how much oil went into the tank over an 8 year period, and can divide that by 8 to get an estimate of average annual consumption. I suggest that this figure is 'good enough' as it based on oil metered for billing purposes. The average is 930.75L per annum (range 713 to 1100), or an average of 9624 kWh at 10.34 kWh per litre.

I then used the rnorm function in R (the maths program) to distribute that kWh value as weekly values over a 30 week heating season. The rnorm function gives a normal distribution over so many chosen intervals with the option of setting the mean and standard deviation, which I tweaked until I got something that resembled in shape recent (the only period I have weekly data for) weekly use. For some reason I did not document, I also let the total annual usage to creep up to, in the end to 12,534 kWh. There are cryptic references to correction factors, but not what they are correcting for - mea culpa, for which I can only beg forgiveness.

It therefore looks as though I may have overestimated oil usage, but the problem is, this only makes the paradox worse. I may also have under-estimated the oil boiler efficiency - the original sales brochure which I still have for the boiler claims over 90% efficiency, and I used 80%, down-rated to allow for deterioration over time. Let me suggest that in round numbers:

I burnt on average 930L of oil per annum, giving an oil energy in of 9620 kWh per annum.

Increasing the efficiency to 90% gives an oil energy out figure of 8658 kWh per annum.

I used 8823 kWh for the heat pump over the last year, this figure being the most reliable one as it is based on the external dedicated kWh meter.

Estimates of heat pump energy out vary, but they are of the order of 26,000 kWh per annum.

The paradox still remains, and it remains gross, at about 3 (heat pump needs to pump out 3 times more heat to achieve broadly the same comfort level).

Let us say the oil CH was on for 12 of every 24 hours say 0400 to 1000 then 1600 to 2200. Had I had it running 24 hours a day (as the heat pump does) then at worst it would double the energy in/out to 19,240 and 17,316 kWh respectively, but I strongly suspect these are large over-estimates, as the boiler would be unlikely to need to run at the same rate all the time. But even if it did, there is still a big energy out paradox: 17,000kWh for the oil boiler, 26,000kWh for the heat pump. In my mind, that size of difference can't be explained by a few minor wrong assumptions here and there. These figures remove the timing discrepancy, and yet the has to produce over 50% more heat than the oil boiler to achieve the broadly the same 24 hour heating level.

I am not sure whether this is relevant, but it looks (from old paperwork I have) like the oil boiler was set up to deliver 80,000 BTU/hr (23.44kW) which is roughly double the max power of the heat pump. I don't know why it was set to this output, maybe just the old habit of installing and commissioning grossly overtaed boilers, because the estimated heat loss back then (1993) was the same as it is now, around 12kW. It certainly used to cycle a fair amount, whether from it's internal thermostat controlling the LWT, or the room stat responding to the room temp, or both.   

I still have the oil system room stat fixed to wall, set to 18 degrees - the same as the current set temp for the kitchen, confirming I have not radically altered my desired temps. Maybe under the oil system, the room temp did fluctuate, but probably only by a few degrees (see earlier posts about rate of cooling) and in any event I didn't notice because I was either asleep at night or out during the day. As I have said many times, subject to the usual caveats about memory not being reliable, I don't recall the house being significantly cooler with oil. I still wear the same number of layers of clothes in winter as I did in the past.

My calculations don't allow for DHW use, but I only use very little hot water, certainly not enough to explain the paradox.

I'm trying to think of a way I can determine what actual room temps were under oil. I only started using the data logger after I had stopped using oil, but there may have been other reasons to take ad hoc readings. I have always had a thermometer or two in the house, maybe sometime somewhere I kept a record.

Just before posting this, I checked for any new posts, and have just seen your latest posts @jamespa, I really do appreciate all the effort you are putting into this, and do hope I don't appear too think or intransigent to be worthy of your efforts. As you say, and I heartily agree, this can be seen as an interesting case study, and so I dare to hope of some general interest. I will consider what you have said and reply shortly. Some of the above may answer some of your questions.

Very briefly: the 'all over the place' current/LWT bit of the chart shows defrost cycles. At least I am pretty sure that is what they are, because (a) the OAT is cold enough to imply defrosts will happen (b) the LWT goes below the RWT (probably the most reliable diagnostic marker of a defrost cycle) and (c) current  in drops because it is not heating during the defrost. The charts are based on minute data, and so have good temporal resolution.