To address the energy management challenge for methanol range-extended series hybrid systems in port tugboats, characterized by highly transient and intermittent operations, this study proposes a real-time energy management strategy based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. A high-fidelity forward simulation model was constructed and validated to train the TD3 agent. In simulations of typical port operation cycles, TD3 reduced methanol consumption by approximately 18.5%, 10.2%, and 7.3% compared to rule-based (RB), equivalent consumption minimization strategy (ECMS), and deep deterministic policy gradient (DDPG) approaches, respectively. Emissions such as NOx and carbon dioxide (CO2) were also significantly reduced, while maintaining superior battery state of charge (SOC). Its overall performance approximates global optimal (DP) performance with a gap of less than 2.5%, while retaining real-time online decision-making capability. Hardware-in-the-loop (HIL) testing further demonstrates that TD3 exhibits less than 1.8% performance degradation under actual communication and execution conditions, validating its engineering feasibility and deployment potential. This study provides methodological and experimental foundations for developing high-performance, low-emission, real-time energy management algorithms for port tugboats.
Li et al. (Mon,) studied this question.