Entropy-augmented deep reinforcement learning with adaptive exploration for integrated energy and motor thermal management in hybrid electric vehicles

The driving motor of hybrid electric vehicles (HEVs) exhibits synchronous accumulation of energy and heat under time-varying driving scenarios. The independent processing of energy management and thermal management in the vehicle control unit overlooks the inherent energy-heat coupling relationship, thereby further limiting both fuel economy and thermal safety performance. A deep reinforcement learning (DRL) based integrated energy and motor thermal management strategy is proposed to address the multi-objective optimization between the dual subsystems. To address the sparse-reward issue caused by motor overheating and excessive battery discharge, an adaptive entropy regularization method based on the gradient of the value function is proposed to achieve a self-adaptive balance between exploration and exploitation. Comparative experiments show that the proposed method achieves an approximate global optimum, improving temperature-control performance, while the economic performance gap is only 8.10%, 2.24%, and 5.43% across various standard driving conditions. Test results under real-world prolonged high-load operating conditions further demonstrate the adaptability and robustness of the proposed method.