Last year we installed a heat pump at home and we now have almost an entire heating season of data for it. Here I present some results. Our overall mean COP (efficiency) is a little better than we were told to expect. During the heating season we have made a number of changes: increasing the night time setback temperature and reducing the configured flow temperature, with impacts on heat demand, electricity consumption or both. It is tempting to focus on efficiency alone but, being practical, it is energy use that we want to minimise. For example you can expect increasing setback temperature to increase efficiency but also heat demand - which wins?
No logging facility on the heat pump.
Our monitoring setup is minimal:
- For the heat pump we have a heat meter and an electricity meter and I take daily readings. The electricity meter measures consumption including both the heat pump and the circulation pump.
- We needed the meters for the RHI anyway, as we have a hybrid system. However, apart from the Daikin Altherma I do not know of any heat pump controllers that give you access to this information. I feel strongly that all installations should have some means of measuring heat supplied and electricity use so it is possible to check on efficiency.
- Hourly temperature data from a local weather station
- Half hourly electricity consumption from my smart meter. (I access this through Carbon Co-op, of which I am a member.)
Overall COP results - 3.1
Our RHI certificate told us the expected COP is 2.96 with a flow temperature of 55°C. Considering only the data up to when I reduced the flow temperature, the actual COP was 3.06. so a little better. Based on the actual weather I estimate the expected COP as about 3.1 so the actual performance is slightly under rather than over but still close - and the LG specs exclude the circulation pump so this is very reasonable.
COP by temperature is pretty much a straight line.
To give you a feel for the performance generally here are some charts for the whole period. Each dot represents a day and I have coloured the dots according to the amount of heat delivered that day. COP is higher on warm days, as you would expect for an air source heat pump. This is pretty much a straight line and it very rarely drops below 2.5.
 |
| COP by mean temperature - each dot represents a day. The dots are coloured according to heat demand - red means a lot of heat was supplied that day. The mean temperature is weighted by electricity demand so that temperatures have more influence when the heat pump is working hard. |
Electricity by temperature is slightly curved
This second chart shows electricity demand by mean temperature. This is slightly curved, with higher electricity use on colder days when the efficiency is low. The heat demand (not shown) is pretty linear with the temperature during the period.
Impact of increasing setback temperature - more heat demand
Initially I set the thermostat to 17°C overnight and when we did not want heat. Then in January I decided to try increasing the setback to 19°C. This was prompted by my analysis in 'Will heating your house constantly use more energy' With a low setback you get a huge demand for heat when the heating comes on in the early morning at which time the temperature outside (and hence efficiency) is often at the minimum for the whole day. A higher setback temperature should bring increased efficiency, though this also depends on the weather.
In our case, however, due to our hybrid setup we have to turn off the heat pump at some point in the night to heat the hot water. Then there is always high demand when it comes on again. As far as I can tell, increasing the setback temperature increased overall heat demand by 10% but the impact on COP was negligible - overall a 10% increase in electricity consumption. Not good.
Reducing flow temperature reversed the increase in electricity - but we are more comfortable
About a month later, I changed the weather compensation curve to reduce the flow temperature by 3°C. Instead of ranging 55-34°C it now ranges 52-31°C This has improved the COP by about 3% (from 3.05 to 3.14 at 5.5°C) and it has also decreased overall heat demand somewhat (which suggests it has also reduced mean room temperatures). The net effect in terms of electricity use takes us pretty much back to where we started - except that my beloved assures me that he is more comfortable now and particularly appreciates the warmer temperatures overnight.
This chart shows the overall effect on electricity consumption of both changes. In cold weather the higher COP from reduced flow temperature dominates while in warmer conditions the higher heat demand dominates. This chart is generated from a statistical model I derived from the actual data (see below for details of how I did this, if you are interested.)
 |
| Impact of the increase in setback temperature and decrease in flow temperature, under a range of external temperature conditions. |
Conclusions
The LG Therma V is not the most efficient heat pump available but it is doing a reasonable job for us. The overall performance is close to what we would expect from its specifications.
My attempts to improve performance have not had any overall impact on electricity use but have made us more comfortable. My beloved has health issues and likes to be warm. He particularly likes the high setback temperature for when he needs to get up in the night. Since we spend most of the day at home as well it makes sense for us to run the system with a steady target temperature. Another household which needs only intermittent heat might save (some) energy by running with a higher flow temperature and shorter heating times.
The hybrid installation we have here is not the best for the heat pump as we have to turn it off to heat the hot water. We still use gas for that (supplementing a solar hot water panel) and also for cooking. In a few years I expect we will get rid of gas altogether.
The rest of this post is for people who are interested in my analysis methods.
Our heat pump supplies space heating only which simplifies the analysis quite a bit but it was still fairly complex. I have used a statistics package (R) that I use a lot in my work.
The effect of weighting the temperatures
How I estimated the overall expected COP
To estimate the expected COP based on the weather, I used another layer of weighting on the temperatures but this has the same function - I weighted the daily mean temperatures by the heat demand that day, so days of high heat demand have more weight. By this reckoning the mean temperature was 5.6. The heat pump is configured with weather compensation and at this temperature (under the initial configuration) I would expect a flow temperature of 48.4. (I believe the weather compensation 'curve' is actually a straight line within a specified range.) Reading off the chart I drew based on the heat pump COP specs (below) I would expect about 3.1 at this outside temperature and flow temperature - quite close to the actual value of 3.06. It would have been nicer to have high resolution data for the flow temperatures and actual COP but this is the best I can do with what I have.
 |
| Chart drawn from the COP spec data for my heat pump which is an LG Therma V. The COP values include the effect of defrost heating. |
Measuring the impact of specific changes
I used a linear regression model for both heat demand and COP, with 'dummy' variables for the changes made during the heating season. The models for heat and COP are roughly linear as you can see in the charts above, so linear regression is fine. This is not so for electricity but it can be calculated from the results for heat and COP. For the regression models I threw everything in to start with and then simplified the model by removing parameters that turned out to be unimportant. The final models for heat and COP look like this.
heat = k0 + k1*mean.temp + k2*diff.temp + k3*mean.temp*highsetback + k4*mean.temp*lowerflow
COP = k0 + k1*wmean.temp + k2*lowerflow
where
- kn are the regression model coefficients.
- mean.temp and wmean.temp are the simple mean and weighted mean temperature for the day
- diff.temp is the difference between simple mean temperature and weighted mean temperature.
- highsetback and lowerflow represent those changes: 0 for before the change and 1 after.
The R-squared for these models were both almost 0.9 which shows they are a fair match to the data. Diff.temp is an interesting parameter as it reflects both weather and day length. It is mostly between 0 and 1 in winter but reaches up to 3°C in March when the days are longer and there is solar gain on sunny days but clear nights are still cold. Diff.temp had a larger impact on heat demand than the flow temperature but less than the change in setback temperature.
Having determined the model coefficients, I applied these to an artificial dataset with a sequence of temperatures and varying other parameters to ascertain the expected heat demand, COP and hence electricity demand. I used this to generate the line graph above comparing before and after the changes.
No comments:
Post a Comment
Comments on this blog are moderated. Your comment will not appear until it has been reviewed.