Authored by Andrew Schein, Director of Trials & Analysis, and Louise Bernard, Senior Data Scientist

We’ve just published a revised version of our heat pump working paper, and it includes some additions we want to share in blogpost form. As a reminder, our paper looked at heat pumps and time-of-use pricing, investigating how they actually affect energy demand, using real customer data, at scale. We used real consumption data from Octopus Energy customers to identify the causal impact of air-source heat pump installations, and the additional effect of adopting Cosy, a time-of-use tariff tailored to heat pump users.

As a reminder, we found:

• Heat pumps reduced households’ total energy use by 40%, and carbon emissions by 36% (increasing to 68% emissions savings over the assumed 20-year lifetime of the heat pump as the electricity grid decarbonises over time).
• Time-of-use pricing for heat pumps was very effective at shifting demand, halving consumption during the evening peak and reducing annual consumer bills by 18% compared to standard tariffs.
• The current UK heat pump subsidy, the Boiler Upgrade Scheme, provides positive welfare benefits relative to costs, with £1.27 of societal benefits for every £1 spent.

Thanks to some great feedback on the first version, we’ve put electricity and gas impacts of heat pumps on one chart, to make the fuel substitution easier to understand intuitively and visually, and shown how these impacts change depending on outdoor temperature.

How heat pumps affect both gas and electricity use, by temperature

From this, we can estimate an “inferred COP” (Coefficient of Performance). It’s not exactly a technical COP. You’re seeing real-world gas reductions and electricity increases, with all the behavioural shifts and seasonal changes baked in. But we believe that’s a feature of this analysis: it reflects actual changes in demand when households move from gas boilers to heat pumps.

How the fuel substitution ratio varies by temperature

On average, we see an inferred COP of around 3 to 3.4, with lower values at colder temperatures and higher ones when it’s milder.

That’s broadly in line with engineering models, but with a few real-world twists. In cold conditions, our inferred COP is lower than model estimates, perhaps due to defrost cycles or users turning up the heating more. As it gets milder, the inferred COP actually goes above model projections, which could reflect weather compensation settings or people just using less heat.

Please refer to the full working paper for more information about how the ratio was created.

How welfare impacts change with temperature

We’ve also used this temperature-specific data to estimate how the impact of the UK Government’s flagship heat pump subsidy, the Boiler Upgrade Scheme (BUS), changes with the weather. We use the marginal value of public funds (MVPF) as a way to measure how much social benefit you get per pound of government spending.

Even though the COP falls as it gets colder, the MVPF actually improves. That’s because absolute carbon savings are higher on cold days, so the same £7,500 subsidy (via the BUS) delivers more bang for its buck. This might support the case for higher heat pump subsidies in colder parts of the UK, like Scotland.

There’s also an important result here for households: above ~10ºC, a heat pump saves money compared to a gas boiler. Below that, it’s more expensive, but that’s also when the social benefits (in the form of carbon reductions) are highest. The private-versus-social trade-off flips as the mercury falls.

Time-of-use tariff impacts by temperature

Now, onto Cosy, the heat-pump specific time-of-use tariff.

In the original paper, we showed that Cosy shifts consumption from peak to off-peak hours, and this effect is bigger when it’s cold. Readers of our working paper wanted to understand what happened on extremely cold days with sub-zero temperatures.

We only had 38 days where at least some of the regions in Great Britain had average temperatures below zero (out of 915 days in our analysis). When we grouped temperature into ranges, we found that the “sub-zero” category had the biggest shift in consumption.

In summary

We believe there’s an interesting symmetry in these findings:

• In cold weather, the CO2e reduction from heat pumps increases.
• Cosy, in turn, shifts more electric load away from the peak.

These findings might sound intuitive in hindsight, but they challenge some widely held views that heat pumps and heat flexibility become less useful in cold snaps. Our findings suggest the opposite.