Energy management strategies for new energy HEVs based on reinforcement learning

Introduction In order to solve the problems of low accuracy and multi-objective optimization imbalance in traditional hybrid electric vehicle energy management strategies under dynamic conditions. Methods A study was conducted to design an improved reinforcement learning energy management strategy based on dual delay deep deterministic strategy gradient (TD3), aiming to improve fuel economy, extend battery life, and enhance strategy robustness. Firstly, a multi energy system dynamics model was constructed, which includes an engine, power battery, and electric motor. Secondly, in order to solve the problems of slow convergence and easy getting stuck in local optima in traditional reinforcement learning for multi-objective optimization, adaptive reward functions and priority experience replay mechanisms are introduced. Results The results indicate that the initial value of the state of charge for all three strategies is 0.5, and the research strategy maintains it at 0.5. Discussion ITD3 can more accurately control the state of charge, making it close to the initial value and reducing excessive energy consumption; Overall, compared with traditional strategies, this research strategy exhibits better battery state of charge retention ability under two typical operating conditions. This strategy can achieve precise energy management, effectively reduce costs, improve energy utilization efficiency, support environmental sustainability, and provide better solutions for energy management of new energy hybrid vehicles.