GridGuard
Physics-informed LLM optimization for safer distribution-grid operation.
GridGuard uses an LLM planning layer but checks candidate actions using power-system constraints before execution.
GridGuard explores how large language models can support distribution-grid operation when combined with physics-based screening and optimization. The goal is not to let an LLM directly control the grid, but to use it as a reasoning and planning layer whose actions are checked against power-system constraints.
The project studies restoration, voltage support, cyber-screening, and operator decision support using feeder simulation and constraint-aware validation.
Keywords: physics-informed LLMs, distribution systems, active distribution networks, safe optimization, operator decision support.