Reinforcement Learning

Our new preprint proposes a hierarchical control framework that combines model predictive control and reinforcement learning to balance second-by-second market decisions with long-term battery health, extending battery lifetime by 84% and increasing profit by 34% compared with MPC baselines.

Our hierarchical MPC–RL approach for multi-timescale battery systems was presented at the 23rd IFAC World Congress in Busan, Korea — separating slow, health-aware scheduling from fast, second-by-second market decisions.