Centre for Net Zero recently presented our generative AI work at the world's leading machine learning conference, the 12th International Conference on Learning Representations (ICLR 2024). AI researchers from around the world gathered in Vienna to hear from leading academics on everything from next-gen AI agents, to data-intensive and optimised ML techniques.

Our working paper was accepted by Climate Change AI’s (CCAI) Workshop, “Tackling Climate Change with Machine Learning”. In the paper, we present Faraday, our AI model trained on hundreds of millions of smart meter data. It outputs household-level synthetic load profiles conditioned on information about the property, such as type/size and low carbon technology ownership. Faraday outputs were compared against actual substation readings to show that the model can be used for real-world applications by anyone interested in modelling energy grids of the future.

Unlocking smart meter data is essential to successful transitions to energy systems that require real time optimisation of supply and demand. Find out more here about the open data community that we've launched with The Linux Foundation, OpenSynth, which is designed to democratise global access to synthetic smart meter data.