Project will use state of the art deep learning methods to securely open up SERL data to many more researchers
A new UCL project entitled ‘AI Generated Synthetic Smart Meter Data’ began on October 1st, 2025, which will explore using artificial intelligence tools to increase the value of the Smart Energy Research Lab Observatory (SERL) dataset.
The SERL dataset contains fine-grained gas and electricity smart meter data for 13,000 GB households that has been gathered since 2019. This smart meter data is complemented by responses from contextual surveys with rich detail about occupants and buildings, as well as Energy Performance Certificate, weather, and other relevant data including energy tariffs.
The next step for SERL
Whilst ground-breaking for UK energy research in its depth and up-to-datedness, SERL fine-grained data is only accessible to accredited to UK academic researchers who have gone through a set of rigorous approval processes. Though entirely necessary to ensure that the personal data of project participants is protected and used only for the correct purposes, it can result in a barrier for researchers to access the data. The SERL data resources have the potential to be highly instrumental in many projects that are looking to address the key energy challenges facing the UK and the world, such as energy security, affordability, and sustainability.
The new AI-EDOL project – funded by the UKRI Engineering and Physical Sciences Research Council and running for six months from October 2025 to March 2026 – aims to address this very issue.
Transforming the data
By using state of the art machine learning methods to create a ‘synthetic’ version of the SERL dataset, the value of the data can be unleashed and made more easily accessible to many more researchers without compromising on the data security and privacy that is the central tenet of SERL’s work.
Project leader Dr Eoghan McKenna said “Our project will train advanced AI models called Generative Pretrained Transformers (GPTs) to learn the complex patterns in real energy data and generate completely synthetic datasets that look like real household energy data but contain no actual personal information. This synthetic data will have the statistical properties researchers need while being completely privacy-preserving and will greatly increase the numbers of researchers that can access this ground-breaking data without compromising participant privacy.”
The project aligns with the EPSRC’s AI for Science objectives including: ‘developing AI capabilities across research fields to accelerate scientific discovery; increasing access to well-governed, high-quality datasets for AI; building interdisciplinary collaborations between AI and energy researchers; and embedding AI as a research tool in a fair and inclusive way.’
Synthetic-SERL
Principal Investigator Professor Tadj Oreszczyn explains “Over six months, we will create “Synthetic-SERL”, the first dual-fuel synthetic smart meter dataset with long temporal sequences. This will include half-hourly gas and electricity data for 13,000 virtual households across full calendar years, each with contextual information about building and occupant characteristics and weather. We will rigorously test this synthetic data to ensure it provides genuine research utility while passing strict privacy audits.”
The intention is to publish the new dataset – along with the training code and tools to integrate the data into workflows – under open licences so that is freely available to researchers worldwide.
Opening up access
As well as benefiting academics by giving increased access to this valuable data, the project will allow industrial partners to test business cases and grid plans to facilitate investment in clean energy technology. Government departments will also be able to use the data to evaluate energy saving policies, interventions, and initiatives to make tackle issues such as energy poverty and achieving net zero.
The project builds on the foundations set by SERL as a world-leading energy resources and supercharges it by enabling access to its insights to a far greater number of users, facilitating their work and in the process learn more about how AI can inform our understanding of energy use and in turn address major societal challenges.
Get involved
The project team will be running a workshop and a hackathon across the 6 months the project runs for potential users and interested parties. If you would like to learn more about these activities or about the project in general, please contact [email protected]