@jamespa Ah, understood. Thank you for the explanation!

Another thought has occurred to me. Maybe the monolithic house, once it is up to temperature, is such a huge heat (energy) store that you can turn off the relatively low level heat (energy) input for periods (and thereby save energy) with only a minimal effect on the total energy stored. It's a matter of proportionality, the energy in the system is so massive compared to the energy in/out that changes in the energy in make relatively little difference. An analogy: imagine a huge American mega-truck cruising along a long straight and level interstate at 70mph. Every now and then the truck driver takes the engine out of gear, but the truck hardly slows down at all, and the extra throttle needed when the driver does put it back in gear to get back to 70mph is minimal. 

But then again, shouldn't the conservation of energy still apply, the extra throttle need to regain 70mph be the same as the energy saved while the engine was out of gear, give or take? 

Posted by: @cathoderay

↑

Another thought has occurred to me. Maybe the monolithic house, once it is up to temperature, is such a huge heat (energy) store that you can turn off the relatively low level heat (energy) input for periods (and thereby save energy) with only a minimal effect on the total energy stored. It's a matter of proportionality, the energy in the system is so massive compared to the energy in/out that changes in the energy in make relatively little difference. An analogy: imagine a huge American mega-truck cruising along a long straight and level interstate at 70mph. Every now and then the truck driver takes the engine out of gear, but the truck hardly slows down at all, and the extra throttle needed when the driver does put it back in gear to get back to 70mph is minimal. 

But then again, shouldn't the conservation of energy still apply, the extra throttle need to regain 70mph be the same as the energy saved while the engine was out of gear, give or take? 

Yes and Yes.  Conservation of energy must eventually apply, but may take a long while for equilibrium to be re-attained, and my suspiscion is that in some cases it never is.  My house takes a couple of days really to stabilise so a daily setback actually (I suspect) puts it into a different 'place', with different thermal gradients through the fabric, than the situation where it is fully stable.  one implication of this is that I would need to sum area under curve for a long time (24hrs+) to capture whats happening.

I dont (yet) understand the full implications of this but suspect that considering the house as a reserve of air (which is what you measure and experience) surrounded by a monolithic mass would be a sufficiently good approximation to reveal what is happening.

I will look at your latest data a bit later today.

I wonder if we should move this discussion to a thread of its own?  Im not sure if we have the capability to do that but I think @mars does. 

Posted by: @cathoderay

↑

Can you please clarify what this graph is, it appears to have both setback and no setback on the same days?

Also when you did the AUC calculation resulting in 6%, 9%, which curve was that under, relative to what other curve?

I don't think I am clear what you have calculated, apologies for being slow.

Posted by: @jamespa

↑

one implication of this is that I would need to sum area under curve for a long time (24hrs+) to capture whats happening.

The AUC calculations I did were for the whole week, with hourly intervals.

I think the idea that daily setbacks put the whole in a 'different place' is definitely worth exploring. With steady no setback running with a constant IAT, the principle of conservation of energy is easy to apply, energy in equals energy out. But when i switch to daily setbacks, perhaps some other dynamic comes into play, beyond the visible changes in IAT and energy use.

I've given another round of thought to the actual with setback vs predicted without setback chart (which has within it the predicted 28% saving) and it still seems coherent. The actual with setback energy use is reasonably well verified by the independent kWh meter. The predicted energy use without the setback matches the setback running pretty well when the system approaches equilibrium in the latter part of the day, and an eyeball assessment of the predictions for the setback and recovery period interval look very plausible, if anything a little on the low side. Try in your minds eye joining up the orange (no setback prediction) line from the start to the end of each 2100-0300 period, and I think you will find it puts the line above the orange line. This is partly explained (see earlier post) by the increase in OAT caused by the heat pump turning off, but even so, the orange line does look very credible, at least to me: 

All of which is to say I can't find a glaring of for that matter not so glaring error in what that chart tells us. 

I will see if I can find two minimum of seven day actual periods with very similar OATs and IATs, one with daily setbacks, the other without. If I can do that (not a given, nature is fickle), then we can remove the predicted (whatiffery) data, and compare actual use to actual use.    

Posted by: @jamespa

↑

Can you please clarify what this graph is, it appears to have both setback and no setback on the same days?

Also when you did the AUC calculation resulting in 6%, 9%, which curve was that under, relative to what other curve?

I don't think I am clear what you have calculated, apologies for being slow.

Sorry, I hope it is all there, but I didn't explain it well.

Starting with your IAT/OAT delta t values (19/10 to 16/10 for 6 hours etc), instead of just using broad averages, I calculated the delta t for each hour. Because the IAT was always more than the OAT. all the delta t values are all positive. I then plotted these values, with the Y axis baseline at zero. In the chart, the blue line is the delta t values from the actual daily setback regime, the orange line is the predicted delta t with an IAT of 19 had I not had a setback on the day in question (hence both actual setback and predicted no setback for each day). The AUC calculations are for the AUCs you see on the chart, between the blue line and the baseline for the actual setback, between the orange line and the baseline for the no setback predictions.        

Posted by: @cathoderay

↑

by the increase in OAT caused by the heat pump turning off

This confuses me.  Do I correctly recall that your heat pump (and more importantly OAT sensor) is in a bit of a well so the OAT sensor doesn't measure the OAT the house sees, it measures the (different) OAT that the heat pump sees (you might call it the AIT - air intake temperature), or was that someone else?

Also you say that the 'predicted deltaT' is based on IAT=19.  Was that the IAT when the energy vs OAT curve which you use to predict energy was measured?  My vague recollection is that it was more like 19.7, but I could well be wrong (but it matters).

Posted by: @jamespa

↑

This confuses me.  Do I correctly recall that your heat pump (and more importantly OAT sensor) is in a bit of a well so the OAT sensor doesn't measure the OAT the house sees, it measures the (different) OAT that the heat pump sees, or was that someone else?

It was probably me, but I think it also applies to many if not most heat pumps, because the immediate environment, specifically the air temperature, is affected by the cooler exhaust air, and the OAT is usually (but not always) in the heat pump, typically measuring the temperature of the intake air. I don't buy the 'well' explanation, if anything my heat pump is slightly raised above the immediate ground. The chill happens because the heat pump sucks in air from its' immediate neighbourhood, and inless you duct the exhaust air away, it ends up in the immediate vicinity. Having a wall behind the heat pump, as is often the case, increases the effect, the air going in at the back has to come from the sides and to some extent the front.

This chilling effect is real, and is clearly visible on OAT charts, but the amount is not that great, usually one or two degrees of chilling, see many past charts.

Posted by: @jamespa

↑

Also your post crossed with one in which I ask for clarification on a couple of points, most importantly the details of what curves precisely (between which the area was measured) resulted in an AUC calculation of 6%, 9%.  

Hopefully now clarified. I used the zero delta t as the baseline (bottom of the area for which the AUC is calculated) on the grounds it is the equilibrium state, when the IAT and OAT are the same, and there is, on paper at least, no energy entering or leaving the house from/to the environment.

Your post changed while I was relying, it now has:

Posted by: @jamespa

↑

Also you say that the 'predicted deltaT' is based on IAT=19.  Was that the IAT when the energy vs OAT curve which you use to predict energy was measured?

Not quite, the IAT=19 comes from the desired IAT (which is 19) and the fact that in no setback running the IAT is normally about 18, though it does vary a degree or two either way from time to time, due to solar gain or wind chill for example. In other words 19 degrees is the IAT I would expect in no setback running, and I used that to determine the no setback delta t (= 19 minus OAT).

Posted by: @cathoderay

↑

This chilling effect is real, and is clearly visible on OAT charts, but the amount is not that great, usually one or two degrees of chilling, see many past charts.

 

 

It looked bigger on the charts you posted recently, perhaps more like 2-3 C

The difference does make a fair difference to some of the calculations particularly at moderate OATs, and also to the total heat loss calculation we were discussing a few weeks ago.  Really this means that 'OAT' is in fact 'AIT' (air intake temperature).  It would be interesting to do some experiments cycling your heat pump so that it gets more 'recovery' time, but one thing at a time.

Posted by: @jamespa

↑

Posted by: @jamespa

↑

Also you say that the 'predicted deltaT' is based on IAT=19.  Was that the IAT when the energy vs OAT curve which you use to predict energy was measured?

 

 

Not quite, the IAT=19 comes from the desired IAT (which is 19) and the fact that in no setback running the IAT is normally about 18

Unless the IAT was in fact 19 when you did the energy vs OAT curve, we need to correct for that!

Posted by: @jamespa

↑

It looked bigger on the charts you posted recently, perhaps more like 2-3 C

It does vary. Here is the chart again:

On this chart it is more often 2-3 degrees. You can also the the warming caused by the heat pump turning off, and cooling caused by the heat pump starting.  

Posted by: @jamespa

↑

The difference does make a fair difference to some of the calculations particularly at moderate OATs, and also to the total heat loss calculation we were discussing a few weeks ago.  It would be interesting to do some experiments cycling your heat pump so that it gets more 'recovery' time, but one thing at a time.

Also you say that the 'predicted deltaT' is based on IAT=19.  Was that the IAT when the energy vs OAT curve which you use to predict energy was measured?  My vague recollection is that it was more like 19.7, but I could well be wrong (but it matters).

The whole system is by nature volatile. Here is the IAT for the last month (eek! from HA, though the data is modbus data). It is currently 18.6, and I have been running without a setback for the last few weeks, and to complicate things even more, I have turned the heating off altogether when it has been mild. In the middle of October it was steady at a little over 19 degrees, it is now a little under, with more variation:

The predicted no setback energy use came from the regression equation for the data points where the heating was on during the test period. I did this because, although a smaller data set, it was from the period of interest, meaning, I hoped, I had excluded extraneous variables creeping in and changing the equation. This is the plot with the regression equation: 

The entries in the predicted energy in column are calculated from the OAT as predicted energy in = -0.1604 x OAT + 2.8173. As can be sen from the combines setback/no setback energy in plot, it is, unsurprisingly, very good at predicting actual energy in in the second half of the day, ie post setback and recovery periods. 

Posted by: @jamespa

↑

Unless the IAT was in fact 19 when you did the energy vs OAT curve, we need to correct for that!

The problem is it is volatile! The above regression chart comes from a period when the IAT did vary, because I had a daily setback, but as I said, it only uses point when the heating was on, setback hours are excluded (they would show as a line of dots on the x axis, whatever the OAT, and through the regression equation way off).

More to the point, the OAT/energy in plot doesn't know about the IAT (except to the extent my auto-adapt script boosts energy in, which does happen, but only infrequently during the boost periods). We on the other hand do use IAT, but chiefly to confirm the house is in overall (longish interval) heat balance. The plot and equation just say this is what the energy out will be when the OAT is this.     

Posted by: @cathoderay

↑

The problem is it is volatile! The above regression chart comes from a period when the IAT did vary, because I had a daily setback, but as I said, it only uses point when the heating was on, setback hours are excluded (they would show as a line of dots on the x axis, whatever the OAT, and through the regression equation way off).

AIT (air intake temperature) and energy will correlate well whatever the IAT is not least because the heat pump controller reacts to AIT alone and the rad output wont vary that much if the IAT changes by a degree.

The problem is that, if the IAT was (for example) 18 when the regression was done, but (for example) 19 when the setback was done, and the average AIT was 10 (meaning that the OAT was probably 12), then we are comparing predicted energy when the delta T was 18-12 with actual energy when the deltaT was 19-12 which is a 16% difference on top of the difference due to setback.   Add to this the fact that there is an unaccounted for heat input of say 600W, enough to sustain an IAT OAT difference of nearly 2C with no input from the heat pump, the proportionate effect of a 1C misalignment is even greater.  You don't need many 1C misalignments (if there are in the same direction) to account for a material discrepancy between theory and measurement. 

Currently, I fear, there are too many ~1C uncertainties to make much more progress.  Really we need to measure OAT (not AIT) and IAT during both setback and control and recalibrate before we have much hope of explaining what is happening with any conviction.