RL2Grid
收藏资源简介:
RL2Grid是由麻省理工学院等机构合作开发的一个强化学习基准,旨在加速电网控制领域的进步,并推动强化学习方法的成熟。该数据集基于法国RTE公司开发的电网模拟框架Grid2Op构建,提供了标准化的任务、状态和动作空间以及奖励结构,以便对强化学习方法进行系统评估和比较。RL2Grid的任务涵盖了处理电网中的组合性大量可能动作的复杂电网操作。数据集还包括了由运营商专家提供的实际控制启发式方法和安全约束,以确保RL2Grid符合电网操作要求。
RL2Grid is a reinforcement learning benchmark co-developed by the Massachusetts Institute of Technology (MIT) and other collaborating institutions, aimed at accelerating advancements in power grid control and promoting the maturation of reinforcement learning methods. This dataset is constructed based on the Grid2Op power grid simulation framework developed by the French energy company RTE, and provides standardized task, state, action spaces as well as reward structures to enable systematic evaluation and comparison of reinforcement learning approaches. The tasks covered by RL2Grid involve complex power grid operations that handle the large combinatorial set of possible actions within power grids. Additionally, the dataset includes practical control heuristics and safety constraints provided by grid operation experts, ensuring that RL2Grid complies with real-world power grid operation requirements.

- 1RL2Grid: Benchmarking Reinforcement Learning in Power Grid Operations麻省理工学院 · 2025年



