Integrated Thermal Management Strategy Using High-Value Prioritized Experience Replay Deep Reinforcement Learning for Battery Electric Vehicles

In low-temperature environments, electric vehicles rely on effective thermal management strategies (TMS) to rapidly heat the battery to its optimal operating temperature while maintaining cabin temperature within a comfortable range. Considering the development trend and critical role of TMS, a deep reinforcement learning (DRL)-based TMS is established, integrating a high-value prioritized experience replay mechanism with the soft actor-critic (SAC) algorithm. This approach prioritizes the replay of high-value experiences during training, thereby enhancing the learning efficiency of the agent. Simulation results indicate that, compared to rule-based TMS, the proposed improved SAC (ISAC)-based TMS reduces the energy consumption of the integrated thermal management system by 4.10% and 3.82% under training cycle and test cycle, respectively. Additionally, it ensures the battery reaches the target temperature more efficiently and significantly reduces cabin temperature fluctuations. Sensitivity analysis further confirms the effectiveness and robustness of the proposed strategy. Comparative experiments at initial temperatures of -5°C and 0°C demonstrate that the ISAC-based TMS outperforms the rule-based TMS, which yields 4.41% and 6.49% improvements in energy efficiency, respectively.