Posted by: @derek-m

↑

@cathoderay

Could you please provide all the relevant data and information in a timely manner, so that a correct peer review of your suggestions can be made.

@derek-m - you have made three requests for data all within three hours, one at 8 pm, another just one hour later at 9:01 pm and then this rather petulant one two hours later at 10:54 pm, all on a Sunday evening. I have no idea what the urgency is, or why you feel the need to make repeated requests, or why it can't wait until the next day. Earlier you asked a rather pointless and I dare say is provocative question about why I changed the start dates. Does it even matter, as long as the periods cover 24 hours and the periods match across the data, which they do? What's so pressing about this 'correct review' of my 'suggestions' that it needs doing in a 'timely manner' at 11pm on a Sunday evening? 

We're clearly heading for another row, with you as usual taking the provocative lead, and me reacting with increasing irritation. This time I am not going to play ball. Here's the data used to generate the charts:

Posted by: @jamespa

↑

Without detailed specs not really possible to tell.

Agreed, I went through the Midea manuals and could not find anything. But I am not sure it matters that much. We have the compressor kWh in which you wanted, and then the remainder, the difference between the compressor kWh and the metered kWh. I have a 'good enough' (for me at least) way of estimating the total use from the compressor use. From a pragmatic, rather than analytical, point of view, what matters is the total energy use by the heat pump, as that is what is actually used, and actually paid for.

I keep returning to the when does a no-brainer (long ie days long setbacks obviously save energy) become a brain teaser (shorter setbacks are more complicated) question, and how and why does that happen. I'm sure the answer has something to teach us.

@cathoderay

Dates and times are important otherwise we may be comparing apples with oranges.

Could you please remind me how you arrived at the 1.18 Correction factor.

Posted by: @derek-m

↑

Dates and times are important otherwise we may be comparing apples with oranges.

Could you please remind me how you arrived at the 1.18 Correction factor.

The dates and times were in the post with the charts. 

The 1.18 factor was a guaverage, a portmanteau word. As I have pointed out before, I have limited widely separated in time records for the external kWh meter, because the only way of reading it at present is by taking a manual reading. I normally read it at 0700 on Monday mornings. The guaverage comes from the week ended 27/11/2023 07:00, when the metered value was 225.7, and the calculated value was 191.503, giving a correction factor of 1.1786 which I rounded to 1.18. As I showed in one of the charts, it seems to work well enough, even when applied to a different, though similar in weather, period.

I may do some daily meter readings during the current cold weather, and then test how well the correction factor works or doesn't work at sustained lower OATs. The heat pump's performance at the moment is pretty awful, frequent defrosts, very poor COP and hourly energy in values high enough to make my ears pop. I suspect, but won't know until I do the assessment, that the defrosts will alter the correction factor, possibly by quite a lot.

We shall just have to wait and see.

@cathoderay

I am sorry to hear that your heat pump is struggling, though the humidity is allegedly lower today which may help.

Posted by: @derek-m

↑

I am sorry to hear that your heat pump is struggling, though the humidity is allegedly lower today which may help.

Thanks for the kind thoughts. At least it is sunny as well as less humid, and the cold won't last forever. 

@cathoderay

I did some maths over my lunchbreak and I think I've got an explanation for why there is a polynomial in the relationship. I think it is indeed the cold well; the temperature measured by the heat pump is lower than the temperature that the building loses heat to. I'll do my best to type the maths out here, but sadly there's no LaTeX on this forum so it might be hard to follow!

Heat Loss ~ IAT - OAT*, where OAT* is the ACTUAL temperature that the building loses heat to.

Let OAT* = OAT . f(OAT), where OAT without the * is the MEASURED temperature, as reported by the heat pump, inside the cold well, and f(OAT) is a correction factor that is a function of the OAT -- the higher the OAT, the bigger the correction factor. f(OAT) is proportional to how hard the heat pump is running, which is proportional to how much heat is being demanded by the house which is proportional to the actual temperature outside the house, OAT*. So we have OAT* ~ OAT . (IAT - OAT*). We can continue to substitute OAT* here ad nauseam, but let's assume that this second OAT* is our measured OAT from the heat pump. Now we have

OAT* ~ OAT.(IAT - OAT) = OAT.IAT - OAT^2

and

Heat Loss ~ IAT - OAT.IAT + OAT^2

or written as a polynomial in OAT:

Heat Loss ~ OAT^2 - IAT.OAT + IAT

Finally we divide through by COP to get the heat pump power draw, which is COP = IAT / (IAT - OAT) and we get:

Power Draw ~ [(IAT - OAT) / IAT] *OAT^2 - (IAT - OAT) * OAT + (IAT - OAT)

In R that would be as follows -- though note that you'll need to convert all your temps to Kelvin, probably:

lm_htg_in_quadratic <- lm(htg_kWh_in ~ I(ambient^2 * IATmOAT/MD02_tmp) + I(IATmOAT*ambient) + IATmOAT,
                          data = midea_data_clean_nosetback)

When I use that formula I get a pretty nice residuals plot, and a 27.5% setback saving. I still find this to be suspiciously high as it's higher than the theoretical maximum you could possibly save from turning something off for a quarter of the day. But the fact that a cold well might result in such a hit to your heat pump performance that the simple act of turning it off for a few hours and letting the air temperature equalise again can get you a 20% energy saving over the long run is something I'd like to dig into further. If you are able to take temperature measurements independently of the heat pump itself, from a different side of the building, away from the cold well, and then do all of this again, that would be a great "little" experiment.

I still haven't figured out why the COP is linear with temperature though, instead of inversely proportional.

In any case, this is enough evidence for me to think about experimenting with my own system, perhaps by shutting it off between 4pm and 7pm when the CO2 emissions of our grid are quite high. Lots to investigate I think!

@scrchngwsl - thanks, though I will need to put my thinking cap on to get my head round that lot! 

Posted by: @scrchngwsl

↑

If you are able to take temperature measurements independently of the heat pump itself, from a different side of the building, away from the cold well

I have thought about how to do this a few times. The problem is finding a modbus enabled sensor that can run off USB voltages and is weatherproof. The indoor sensor I have works fine, and I could get another one - but it isn't waterproof, very far from it, and making it waterproof will inevitably insulate it from the environment, and compromise it's ability to get an accurate OAT reading. So far, I haven't come up with an answer!

@scrchngwsl

Well done. It looks like you are getting there.

If I am understanding your formulas correctly, they should produce somewhat similar results to those from the spreadsheet simulator, though there appears to be some differences, since I could only achieve an energy saving over 20% by failing to fully recover the heat balance over the 24 hour period.

I don't know if you have looked at the spreadsheet, but it may be useful to do a comparison, to see if my philosophies are missing something, or you need to further develop your work, or possibly both.

The spreadsheet tries to simulate both the home and the heating system, so it requires some initial information about the home. The calculated heat loss needs to be entered along with the relevant OAT and IAT. Next is the total capacity of the heat emitters at the standard reference temperature. Followed by a desired heat pump Delta T. The final parameter is an estimation of the thermal capacity of the building, which dictates how long it takes to warm up and how quickly it is likely to cool down.

I have tried to develop a type of universal simulation spreadsheet, that is not limited to one specific home, or one make or model of heat pump, but can be used with any, provided that manufacturer's data tables are available, and the required information about the home is also available.

Obviously all the above initial parameters can be changed by the user, which enables experimentation (or as CR likes to call it, 'whatiffery', which I think is a good description) to be performed, such as what would be the effect of increasing insulation levels to reduce the calculated heat loss, or what would happen if larger heat emitters are installed. It should even be possible to simulate the effect of changing to UFH with a large concrete slab, and thereby increasing the thermal mass of the building.

The main limiting factor will be how accurate is this initial data.

The main work is done within the sheet I named 'playroom', which is where the hourly values such as OAT and WC Offset can be entered, to produce the predictive results. The playroom is also where the WC curve settings are entered and can be varied to see the effect. Specifically for CR's system I also added a PHE DT parameter, which can be disabled by entering a value of zero. I have also added a Other Heat Source (OHS), which is to account for human activity within the home such as cooking etc. The WC curve settings are straightforward, with the problem areas being a measurement of the temperature drop across a PHE, if one is installed, and quantifying the effect of human activity, which in milder weather conditions could be providing 10% or more of the required thermal energy to keep the home warm. I have suggested removing the 'human infestation', for scientific purposes of course, but have not had much success in convincing anyone. 😋 

I don't have time to go into great detail, so I will give you a quick overview and you can then ask any questions you wish.

I have tried to simulate, as closely as possible, how I presume a heat pump operating in WC mode would perform. Calculations are made at hourly time intervals, with the ending values of the previous hour being the starting parameters for the present hour.

From the OAT value for each hour and the WC curve settings, the required LWT for each hour is calculated. The WC Offset can then be added to or subtracted from this value.

Using this LWT value and the OAT value, the predicted COP value is extrapolated from the manufacturer's data tables.

The Energy Supply value is where I think that our approaches differ, with this value being calculated from the LWT, any temperature drop across the PHE, the average water temperature at the heat emitters, the average IAT during the hour, with reference to the total capacity of the heat emitters. In basic terms, the maximum thermal energy that the heat emitters can emit under those operating conditions.

The Power Input (PI) is a simple Energy Supply divided by COP calculation, which matches quite closely the PI value extrapolated from the manufacturer's data tables. It would appear that this value is virtually the V x I value without any auxiliary power included.

This discussion has been very useful, because the addition of the OHS parameter now means that I will need to re-assess the Energy Supply equation to account for the reduction of thermal energy required from the heat pump and its effect upon PI.

The Energy Demand is the standard IAT minus OAT times the heat loss per 1C from the calculated heat loss. 

The Energy Balance is the difference between Energy Supply and Energy Demand.

The Energy Balance value is added to, or subtracted from, the Thermal Capacity value, from which is derived the new IAT value.

The process is then repeated for the next hourly period, but because there are circular references, it is necessary to enable iteration, which recalculates the whole worksheet until a stable result is obtained, or the present limit of 100 iterations has been reached.

I hope that the above makes sense.

@cathoderay 

Re the 1.18 correction figure. I have a completely different pump (Viessmann) and have a well insulated property with underfloor throughout. I have a meter that records the heating and hot water, including all circulation pumps, in addition the heat pump has its own meter. All my data is from manually recording the values on a daily basis. For the entire year (2023) my total use is 4033 Kwh, the heat pump meter gives 3410 Kwh, the difference is 1.1826. How can this be a coincidence? if i take a 2 month figure in the heating season, Nov and Dec 23 i get 1128 and 993, a difference of 1.1359, a 5 month figure in summer with no heating is 347 and 276 a difference of 1.26. So with me the hot water gives a bigger difference maybe because the pump has to work hard to get to the higher temperatures, however in summer the heating circulation pumps are all off.

@cathoderay I've had success with this £6 sensor attached to a Raspberry Pi:

Oumefar Digital Temperature Sensor Module Board Kit DS18B20 Electronic Components with Waterproof Stainless Steel Probe for Raspberry Pi https://amzn.eu/d/hzGTum7

The Pi of course has to be indoors (and you need a Pi or similar GPIO device to attach it to), but if you have a garage or shed with power you can put the Pi inside and run the probe outside under the garage door or whatever. That's if you enjoy DIY'ing this stuff and would rather not spend £££ on a waterproof temperature logger.

@derek-m I think I've used your spreadsheet in the (long) past but haven't seen a recent version. I'll go have a look when I get some time this week.

@scrchngwsl

If you wish I will post a copy after I have updated for the OHS capability.