๐ฏ 39+ environments over 7 base grids, including:
๐น Topology optimization actions
๐น Redispatch & curtailment operations
๐น Idle & recovery heuristics
๐น CMDP-based constraints (overload, islanding, load shedding)
๐ฏ 39+ environments over 7 base grids, including:
๐น Topology optimization actions
๐น Redispatch & curtailment operations
๐น Idle & recovery heuristics
๐น CMDP-based constraints (overload, islanding, load shedding)
๐ฆ RL2Grid is a benchmark for training & evaluating RL agents on a multitude of realistic grid control tasks, modeling:
โ
AC power flow
โ
Multi-step physics simulation
โ
Operator heuristics
โ
Real-time contingencies
โ
Safety constraints
Why power grids?
๐ Complex, dynamic, hierarchical systems
๐ Traditional solvers hit scalability walls
๐ช๏ธ Control challenges driven by VRE, demand-side volatility
RL is promising for grid control, but it must be grounded and advanced realistic benchmarks!
๐จ Excited to share RL2Grid! A standardized suite of RL tasks for realistic power grid operations โก Built with TSOs, RL2Grid aims to bring RL closer to real-world critical infrastructure.
๐ Preprint: arxiv.org/abs/2503.23101
๐ป Code: github.com/emarche/RL2G...
Details below ๐