Distributed energy storage only works at scale if operators can estimate battery health and detect faults across thousands of assets they do not own. Pooling raw telemetry is often impossible for privacy and commercial reasons, and naive federated learning is exposed to poisoning and inference attacks.
This project, funded under the Trailblazer for Recycling and Clean Energy (TRaCE) Lab to Market scheme, builds a federated battery management system in which:
- battery health models are trained across sites without centralising data;
- the training process is hardened against poisoning, backdoor and gradient-inversion attacks;
- the software interfaces of the battery management system are secured and monitored in operation.
Wei Song leads the project as Lead Chief Investigator. It anchors RISE’s critical infrastructure direction and provides a real-world testbed for the R, S and E layers.