Integrated energy and thermal management for battery energy storage systems in range-extended electric vehicles based on generative adversarial imitation learning

Battery energy storage control and thermal regulation play a critical role in improving the energy efficiency and battery lifetime of range-extended electric vehicles (REEVs). However, the strong nonlinear coupling among battery power demand, thermal behavior, and range extender operation introduces a high-dimensional storage control problem, making it difficult to simultaneously achieve global optimality and real-time control. To address this challenge, a global optimization-guided generative adversarial imitation learning (GAIL)-based integrated energy and thermal management strategy is proposed to optimize battery storage operation and thermal regulation in REEVs. First, the coupling mechanism between energy management and thermal management is analyzed, and a multi-objective optimization framework is established to jointly minimize fuel consumption and battery aging cost. Offline dynamic programming (DP) is employed to generate globally optimal state–action trajectories, which serve as expert dataset for policy learning. Based on these expert data, a GAIL framework is developed to learn coordinated control policies for range extender power allocation and battery cooling demand through adversarial imitation learning. Simulation results demonstrate that the proposed strategy achieves 97.82% of DP performance in weighted total cost while maintaining real-time feasibility. Compared with rule-based and reinforcement learning-based strategies, fuel consumption is reduced by 3.31%–10.65%, and battery aging cost decreases by 1.39%–1.67%. Furthermore, evaluation on an unseen real-world driving cycle provides preliminary evidence of the proposed strategy's adaptability and robustness.