@cathoderay

Please find attached the results for the 24 hour period from 21:00 on 12th Dec. 2023 to 21:00 on 13th Dec. 2023.

The first table uses LWT's averaged from the whole of the minute data during any 1 hour period, the second table averages the LWT only when the compressor was producing thermal energy. It would be interesting to know which method may give the more aligned result.

The time shown is at the end of each 1 hour period.

The OAT value used has been averaged over each 1 hour period from the raw data, no adjustment has been made for deviation from the true ambient temperature.

If anyone has any questions then please ask.

@derek-m - thanks. I've got various other things to be getting on with today, but later on I will plot the two predictions against the observed values and post the result.

The period in this spreadsheet is as you note in the tables a no setback period. Are you also able to do a prediction covering one or more of the setback periods (16/17 Dec onward) (a) with a 6 hour 2100 to 0300 setback and (b) without a setback?  

Posted by: @cathoderay

↑

@derek-m - thanks. I've got various other things to be getting on with today, but later on I will plot the two predictions against the observed values and post the result.

The period in this spreadsheet is as you note in the tables a no setback period. Are you also able to do a prediction covering one or more of the setback periods (16/17 Dec onward) (a) with a 6 hour 2100 to 0300 setback and (b) without a setback?  

Yes. No Problem.

Here are charts of @derek-m's model predictions vs my observed (based on amps x volts) values for energy in (PI in @derek-m's spreadsheet), first scatter plots and then bar charts. They are hourly energy in values over a 24 hour period, 2100 12 Dec to 2100 13 Dec 2023, no setback and a moderate relatively stable OAT (range 7.1 - 8.9). The text box on the left hand side of the scatter plot gives the 24 hour totals, 'Midea' is the value based on the Midea total (cumulative) values (current value - value one hour ago). I have checked for errors typos etc but E&OE as ever, I am only human.  

Edit: bar charts temporarily removed because there was an error, correct version will be posted shortly

Edit: corrected bar charts now added

@cathoderay

Please find attached the previous results with those from 13th to 16th Dec. added.

Please also note that as yet the spreadsheet makes no correction for DHW heating, since, as yet, I am undecided as to how much DHW heating may contribute to overall building heating.

@cathoderay

Thank you, it will be interesting to see how other days and weather conditions compare.

@derek-m - sorry, I did make an error in one of the bar charts (second lower chart had same data as first), I have now corrected that. I have also done the same thing, this time using my combined OAT vs energy in regression line equation (the basis I used to calculate predicted values in my earlier observed vs predicted charts):

I will shortly post all the charts in one image, to make direct comparison easier. 

Here are all the charts in one image, to (I hope - you need a large monitor) make direct comparison easier: 

This is a comparison, albeit only for one rather unexceptional but not atypical day, of predictions based on observed data (using my regression line) and a model worked up from scratch. I think the comparison is interesting: despite having the worst R squared value, my regression line equation makes the best 24 hour value prediction. If we take the 'true' value to be 35.32 kWh (with the Midea 36 being a rounding error), then my regression over-estimates by 2.6 kWh, while your closest model prediction under-estimates by 3.05 kWh. @jamespa will no doubt correct me, but what I think is going on here is a sort of regression to the mean: the hourly values are a bit out, but over a 24 hour period the 'outness' cancels itself out, if that makes sense. I'm also if I admit it a bit suspicious of the third R squared value...it looks a bit low to me. All thoughts and comments welcome. Please bear in mind the goal is to be able to predict accurately what the 24 hour energy in would have been, had a setback not been applied, which can then be compared to the actual use with a setback applied.    

Posted by: @derek-m

↑

Please find attached the previous results with those from 13th to 16th Dec. added.

Please also note that as yet the spreadsheet makes no correction for DHW heating, since, as yet, I am undecided as to how much DHW heating may contribute to overall building heating.

Thanks, I will do what I can as soon as I can. The really interesting 24 periods will be the setback periods, but the non-setback ones are also very useful. 

@cathoderay

Please find attached the results file with the remaining data added up to the 18th December, which is all the data provided.

Could you now post the un-redacted raw data so that meaningful comparison can be made.

Posted by: @derek-m

↑

Could you now post the un-redacted raw data so that meaningful comparison can be made.

Will do, but juggling many things at the moment for all the usual reasons, so may be a delay. But will do as soon as I can.

PS what is unmeaningful about the charts I have already posted?

Posted by: @cathoderay

↑

Posted by: @derek-m

↑

Could you now post the un-redacted raw data so that meaningful comparison can be made.

Will do, but juggling many things at the moment for all the usual reasons, so may be a delay. But will do as soon as I can.

PS what is unmeaningful about the charts I have already posted?

Nothing.

I am used to looking at the big picture, identifying probable problem areas, and then looking in detail at these areas.