GenAI modelling

Faraday is a pioneering model that outputs synthetic smart meter data conditioned on specific household characteristics


Modelling the impact of low carbon technology adoption on consumer energy consumption is challenging without access to smart meter data. Centre for Net Zero’s unique access to Octopus Energy’s data allows us to train a generative model conditioned on information that is vital to prepare intelligent grid systems for the future, when there will be a high penetration of low carbon technologies in homes.

Faraday is a model based on Conditional Variational Auto-encoder (VAE) and Gaussian Mixture Model (GMM). Using over 300 million smart meter readings, from the UK’s largest electricity provider, it generates daily load “profiles” consisting of half-hourly kWh consumption for a given set of user-specified inputs (such as low carbon technology, property type, season), outputting synthetic data that can be safely be shared with third parties.

Use cases

Researchers, policy makers and DNOs are exploring multiple downstream applications of the tool, simulating how households consume energy and modelling specific scenarios. Users include the Department for Energy Security & Net Zero, Ofgem, ARUP and the universities of Oxford, Cambridge, Manchester, Birmingham and more. Access to the tool is free; get in touch to find out more.

Open data community

Our Faraday model is the foundation for OpenSynth, an open data community that we created in partnership with the Linux Foundation. OpenSynth will democratise access to synthetic energy demand data, enabling widespread, global access to smart meter datasets.

Our papers


Trained on +300 million smart meter readings


User testers, from government departments and regulators to industry and grid operators.