Design of a hybrid thermal management system for lithium-ion batteries based on reinforcement learning

This study proposes a reinforcement-learning-based intelligent hybrid thermal management system that achieves efficient thermal regulation for prismatic battery modules. This research employs computational fluid dynamics simulations and experimental validation to develop an autonomous decision-making actor-critic-proportional-integral-derivative control strategy. The cooling structure incorporated droplet-shaped spoilers in a rectangular pipe and S-shaped spoilers in a hexagonal pipe. By integrating the control algorithm with a hybrid cooling structure design, the system dynamically optimizes the proportional-integral-derivative parameters in real-time to minimize the response time. The battery maximum temperature is maintained between 28.12 °C and 28.91 °C with a temperature difference under 2.03 °C. Compared with fixed flow rate control, the system reduces energy consumption by 19% to 20%, significantly improving both control precision and energy efficiency performance.