These are useful observations (same for comment from @Blue). I wonder whether there would be a way to parse these systems into groups. Probably beyond the scope of current HeatPumpMon but might be interesting to energy companies.
That’s exactly what I do - I run it for DHW at night when the tariff is low (but ambient low) and at 1pm when hopefully the ambient is hih and I might also be generatin some solar. I also run it just before any room heating is likely to kick in so that it’s making use of a warm unit.
@TrystanLea Can you tell me if the Profile Explorer App on emoncms uses just the previous year’s data or does it use all available data? (Similar question for this aggregated data too!)
Another useful feature might be to generate the profile for the coldest day, though that might be just particularly useful for me as I’m trying to understand worst-case grid demand profile a the moment!
Rachel
1 Like
Hello @Rachel it is just the last 12 months. Can you select the coldest day from the aggregated data?
E.g is it the 18th of January 2024?
https://emoncms.org/heatpumpmonitororg/graph/476418
Is it a bit strange that demand peaked overnight on that day I wonder?
If we could do aggregation on the fly, based on filtered systems that would be interesting… would be good to understand if there are particular systems driving this shape…
Yes that’s a way to find it thanks; is there a way to know the number of heat pumps in the sample at that time? (I assume the normalised demand is the total devided by the number of active systems?)
I think the reason night-time looks exagerated is the non-zero y-axis. It doesn’t look that disimilar to the normalised monthly average if you reset the axis. This is one of the artefacts I’m interested in - it feels to me that given the way we operate heat pumps, with a set-back overnight, but often more continuous daytime load, we may move, as a country, to morning peaking during winters rathert than the normal evening peaking!
Yes @Rachel, you should be able to get to that from the feed view>
Emoncms - feed view
1 Like
We had a bit of fun at the recent HeatGeek Hackathon trying to look at some of the topics we have also explored here in terms of grid implications. I thought I would share some of the data and graphs produced looking at one small element of this; the relationship between the aggregated COP of groups of heat pumps and outside temperature.
COP of top 20, top 50 and all systems January 2024 - Note 15-19th Jan cold snap
zooming in on the cold snap, low of -4C on the morning of January 18th 2024:
John Ewbank put together this nice graph to visualise COP vs outside temperature for these different groups of heat pumps.
Or visualised separately:
Top 20 systems
Top 50 systems
All systems
Producing the above turned into a good opportunity to improve the HeatpumpMonitor.org API (Il do a separate post on that) and also to improve the aggregation scripts. The data is all available here https://emoncms.org/heatpumpmonitororg/feed/view, zoom over to January 2024 to see the data (includes aggregated electric input, heat output, outside temperature and COP).
6 Likes
That’s all really interesting stuff; especially noting that the spread drops as the temperature falls. Also would be interesting to compare the electricity input at time of system peak demand for those systems operating flexibily and those targetting long and low and max CoP.
1 Like
I updated the aggregated electric demand graph recently following a discussion on the topic, it’s quite an interesting chart, in yellow aggregated heat pump demand in watts (left axis), in blue number of systems in the aggregation. Emoncms - graph 476418
If we divide the aggregated demand by the number of systems we get the average demand per system: Emoncms - graph 476418
We hit an all time peak of 1.031 MW (572 systems) on the 21st of November 2025:
On a per system basis the highest demand so far was 7:30am Saturday January 11th 2025: 1.836 kW per system.
If we simply multiply 1.836 kW by 30 million households to project a future peak grid demand this would translate to 55 GW of electric demand. The 24h average for the 11th of January was 1.59 kW which would translate to 48 GW. Which gives us a best case idea for peak capacity savings if the demand profile was flattened on the highest demand day - we could potentially avoid the need for ~7 GW of peak generation (~13%).
Peak demand coincided unfortunately with a period of low wind power output. I’ve scaled the wind data here so that on an annual basis there is enough wind power to cover 120% of the heat pump demand, there’s a lot of oversupply at other times. Emoncms - graph 47641897699 (scale wind by 0.0620).
zooming out to a year view:
I think the challenge of meeting these peaks of heat pump demand is a fascinating problem, it’s all very theoretical at this point given that heat pumps are only 1-2% of heating demand, a lot could change in the decades that it will take to get anywhere near having to grapple with this in reality. E.g break-throughts in storage costs, availability of long duration energy storage using e-methane or e-methanol, cheap load following fission or fusion.. If we had to solve this problem with only the technology that we have today we’d manage a little bit of peak shaving with batteries but have to use peaking gas engines to cover the bulk of the gap between renewable supply and heat pump demand - or use *very well controlled* hybrid heat pumps.
I’ve created various little modelling tools over the years to explore this if anyone is interested in playing about with these:
-
UK Grid simulator with option to add heat pump demand based on HeatpumpMonitor.org aggregated heat pump demand: UK Grid Simulator you can get to 90% supply demand matching with 30% oversupply and 10 kWh of battery per household. You need a big long duration energy store to get to 100%.
Here’s the same Jan 11th peak in this tool/model:
- Individual household storage simulator: Storage simulator
- ZeroCarbonBritain hourly energy model, most advanced model, default scenario is the ZeroCarbonBritain scenario from CAT with very ambitious insulation/retrofit levels to reduce space heating demand, there is also a baseline scenario that assumes we just fit heat pumps and dont insulate to compare against ZeroCarbonBritain Hourly Model
2 Likes
A few things it would be good to improve with the aggregation:
-
Dividing the total electric demand by the number of systems is not necessarily representative of the average heat pump electricity consumption if the distribution of the capacities of heat pumps in the sample is not representative.
-
It would be good to have aggregated heat output and average outside temperature. With aggregated heat output we could calculate the average COP for all systems over time..
I’ve created a provisional heat pump demand aggregation page on HeatpumpMonitor.org here:
https://heatpumpmonitor.org/aggregation
This shows the normalised electric demand per household scaled up by a chosen number of households, default 30 million (roughly expected number of UK households in 2030).
It’s also possible to overlay wind power output, also scaled to provide 120% of heat pump electricity demand on an annual basis.
Interestingly the current cold snap is also unusually matched by a period of high wind power output!
Perhaps air change rates and heat loss in homes will be even greater this week for many?
New maximum demand peak this morning at 6:30am, 1.2MW from 596 heat pumps!
potentially about 61 GW with 30m heat pumps:
1 Like
Latest version of aggregation tool is much improved:
https://heatpumpmonitor.org/aggregation
- Aggregated heat output and COP
- Aggregated average outside temperature
- Improved scaling of heat and electric output to reflect average UK household ~9100 kWh of space and water heating demand per year.
The improved scaling drops the maximum heat pump electric demand for 30 million households from 55-61 GW above to 40GW - quite a difference!
Plotting outside temperature highlights how the morning peak in electric demand is clearly exacerbated by the colder outside temperatures that coincide with coming off set back:
Here’s an interesting pattern:
I assume: first half on the left, warm, windy, cloudy weather, performance is good with warm outside temperatures, demand is fairly stable night to day time. Then in the second half perhaps sunny & colder weather? high solar gains in the day, lower early morning COP & before sun comes out, then warm day time + solar gains = lower demand and higher cop so even lower electric = high amplitude fluctuations..
1 Like
This is awesome
1 Like
Any thoughts on what the default number of households should be? E.g it’s aggregating ~600 systems and then normalising that to produce an “average” system. The default view could either match the latest value for the total number of systems on HeatpumpMonitor or it could scale to 1 household, 300,000 (roughly total number of MCS installs), or 30 million ( 100% uptake of air source heat pumps in a theoretical future)..
Scaled to a single system gives a peak electric consumption of ~1.4 kW and heat output of 3.5 kW. This “feels” like a really small amount of electric.. e.g at the coldest point it’s less than a single supplemental electric heater:
Scaled to 300,000 installs, roughly number of MCS installs, peak electric is now 409 MW, I dont really have an intuitive reference point for this number, other than that current grid electric demand peaks around 50,000-55,000 MW and so heat pumps installed to date, barely make up 1% of peak electric demand in 2025/2026. Main take away here is that it’s really not something to worry about just yet!
Then at 30 million, a theoretical 100% uptake of heat pumps in a future grid (no hybrids, mostly air source, no other sources of heat, high performance SPF 3.8): 40GW (roughly 55GW at SPF 2.8). This is a much more thought provoking number especially when considered alongside periods where wind output is low - as it highlights the need in such scenarios for very large capacities of backup generation or long duration energy storage.
1 Like
300k is good.
A single one is too abstract, 30m is far too speculative. Maybe just have them as easy toggles?
I think papers tend to use percentage penetration rates @ 20%, 50%, 100% of households.
1 Like