Approximate the optimal energy management strategy for plug-in hybrid electric vehicle from the constant-sum game

Efficient energy management in plug-in hybrid electric vehicles (PHEVs) necessitates addressing the intricacies of bi-level optimization posed by dynamic driving conditions. This paper models the energy management problem as a constant-sum game between driving demands and the energy management system (EMS), reformulating adaptability under varying conditions into an equilibrium computation task. A Policy Migration-Monte Carlo Tree Search (PM-MCTS) framework is proposed, combining policy and value networks trained on Dynamic Programming (DP) results with heuristic search. By leveraging optimal policy transfer and guiding Monte Carlo Tree Search, PM-MCTS achieves computationally efficient and adaptive energy management. Experimental results demonstrate that PM-MCTS improves fuel economy by 22.39 % compared to charge-depleting/charge-sustaining strategies and by 13.71 % over adaptive dynamic programming, while maintaining a deviation of only 6.21–7.13 % from DP-optimal trajectories. These findings establish PM-MCTS as a robust and scalable approach for real-time energy management in PHEVs.