The transition to net-zero carbon requires the integration of millions of grid-edge devices into the Great Britain power system over the next decade. Examples of these devices include electric vehicles (EVs), heat pumps (HPs), and distributed batteries. When aggregated together, flexibility equivalent to multiple large power plants can be obtained from these devices with only small adjustments in usage.
The major challenge is that computational complexity and communication requirements make traditional centralised dispatch infeasible for grid-edge devices. This has motivated research into AI-based strategies which leverage multi-agent reinforcement learning for scalability and adaptivity. However, despite successes in other domains, and significant work by the power systems research community, the lack of safety guarantees has prevented industrial adoption of these AI-based approaches by system operators and flexibility aggregators. SAGEflex is part of the Advanced Research + Invention Agency (ARIA) programme on Safegaurded AI, which involves close collaboration between experts in mathematics, machine learning and application domains. Through SAGEflex, our ambition is to give power system operators confidence in adopting safeguarded AI for grid-edge devices, unlocking large amounts of clean low-cost flexibility for Great Britain’s power grid. This will accelerate decarbonisation and help lower customer energy bills. SAGEflex is a collaboration between the Power Systems Architecture Lab (PSAL), led by Professor Thomas Morstyn, and the Foerster Lab for AI Research (FLAIR), led by Professor Jakob Foerster.
