In the future net-zero energy system, household demand for electricity will significantly increase because of the electrification of heat and transport. Alongside this, there will be more complexity in the energy system overall, with more diverse and variable sources of renewable generation.

Increased uptake of low-carbon heating systems, such as heat pumps will not only place greater demand on the electricity grid but also change when peaks in energy demand occur because heat pumps and gas boilers generally have a different half-hourly consumption profile. However, this increase in demand could be offset by the potential of green heat technologies to use energy differently or at different times. 

A better understanding of energy consumption is crucial for developing strategies to reduce carbon emissions and manage a net-zero energy system. Additionally, identifying consumption trends can also help promote energy-conscious behaviours and help households save money by shifting their energy usage.

Smart meters have the potential to not only provide information to individual householders about their energy consumption but also to identify patterns of usage across the entire energy system.

We have been analysing smart meter data to uncover information about energy consumption habits, and how household appliances, physical property characteristics and demographic factors influence energy usage. Our ultimate goal is to create a set of energy-use profiles representative of Great Britain’s population. This project update presents the prototype resulting from our first phase of work.

In the first phase of our project, we have developed a prototype set of profiles based on half-hourly electricity smart meter data that group households together based on how they use energy. The goal of this phase was to understand whether smart meter data could be used to identify different patterns of energy usage and characteristics of homes and households associated with each usage pattern. 

The second phase of the work will focus on improving the existing profiles in response to stakeholder feedback and potential use-cases we identify. The final set of profiles will include gas consumption, alongside electricity, data. This will allow us to identify meaningful energy-use profiles that are representative of the population of Great Britain.

The analyses presented were conducted using Smart Energy Research Lab (SERL) observatory data. Households in the analysis are pseudo-anonymised to protect their identities, and all published results are based on 10 or more households.

The underlying data behind our analyses is half-hourly electricity consumption data from smart meters. We started by pre-specifying behaviours, lifestyles and property characteristics that we theorised might contribute to differences in energy consumption patterns. Then we mapped these to features we could extract from the SERL dataset. These features should capture the main drivers of energy usage, while being meaningful to users and futureproof (ie, not tied to a specific time in the past). For this reason, we focused our analysis on periods after the Covid-19 lockdowns, as consumption in this period is likely non-typical. 

As an example, households with home-working patterns might have a higher electricity consumption from 9:00am to 5:00pm on weekdays compared to those who work onsite.

You can find the behaviours, lifestyles and property characteristics as well as respective features created so far in Table 1 of the technical appendix.

We input these features into a clustering model, grouping households that consume energy in similar ways together and ensuring each group has a distinct energy usage profile. 

We created two different prototypes: one with four energy-use profiles and another with nine energy-use profiles. It is important to note that there isn’t a ground truth on how many energy-use profiles there are in Great Britain, but profiles should:

• be distinct from each other, and households in the same profile should be as similar as possible in their energy-use patterns
• be useful, eg, knowing that 60% of the population follows a certain pattern of electricity usage might be true, but not very useful for the purposes described above
• reflect different behaviours, demographics & heating systems, eg, we expect bigger families to consume more energy than smaller families, and consumption patterns for households using heat pumps versus gas boilers to be different.

We should note that only smart meter consumption data is being used to create the profiles. No information about households and their respective homes is used to group households – only to contextualise the profiles after they’ve been identified.

After identifying the energy-use profiles, we bring in contextual data about the households and the properties they live in to identify what households in the same profile have in common and how profiles differ from each other. For instance, the number of people living in a household and the appliances they own. 

Contextual information is drawn from Energy Performance Certificates and surveys to households run by SERL. You can find the contextual information we’re currently including in Table 2 of the technical appendix.

For this analysis, we focused on the period between September 2021 and August 2022. Results are based on a sample of 10,211 households, designed to represent Great Britain across regions and index of multiple deprivation (IMD) quintiles. Some of the key learnings are below:

• Households are not evenly distributed across profiles, reflecting that different proportions of the population have different energy consumption behaviours. 
• As expected, households in deprived areas consume less electricity on average than those in least deprived areas. 
• A small group of households (5% of households) uses disproportionately more electricity (five times more per household) than households in the most commonly observed profile (60% of households). 
• Our ‘overnight electricity consumers’ account for 1-3% of households, and a higher than average proportion of them live in the least deprived areas, in homes with solar panels and own electric vehicles.

Explore the dashboard below with our main findings so far:

The main caveat of the current prototypes is that they only use electricity consumption data. We should note that contextual information is not complete for every household in the dataset, and that we can only release results that include 10 or more households for statistical disclosure purposes. As a result, the fewer households within a given profile the less information we are able to publish.

Our first phase of work confirmed that we can use smart meter data to develop a set of ‘energy profiles’ corresponding to different patterns of energy usage. We learned that these profiles cover different proportions of the population that consume electricity in varying amounts and at different times.

In our next phase of work, we will improve the existing profiles by:

• including gas consumption data to cluster households
• Improve and expand our smart meter features, for example by accounting for inter-day consumption variability
• add new contextual features, including the presence of heat pumps and other renewable technologies, and the tenure of households.

Alongside this, we’re engaging with external stakeholders to understand how they think these energy profiles could be used in the energy sector.

At the end of the next phase of the project, we will have a good understanding of how households in Great Britain consume energy and the factors influencing consumption, such as the type of heating system installed in the property. This could help energy suppliers or system planners make decisions about new products and services, or about how a net-zero energy system will need to evolve to serve homes in the future. 

If you have any thoughts or suggestions about any of the above please fill in our feedback form. To learn more about the methodology and assumptions behind our analysis, read our technical appendix.

Acknowledgements

The authors would like to acknowledge the suggestions provided by the whole sustainable future team at Nesta during this first phase of the project. Particular thanks go to Elizabeth Gallagher and Daniel Lewis for advising on data science, Max Wollard for reviewing this project update, Michael Fell for energy-related domain knowledge support and Elin Price for leading on communications, and Chris Williamson, Nesta alumnus, for setting up this project. Thank you to Martin Pullinger and James O’Toole from the team at University College London for support with SERL and UK Data Service-related queries.