An Expert Trajectory Guided Online Energy Management Strategy for Aviation Fuel Cell Hybrid Power System

To meet the higher power output and real-time control requirements of aviation fuel cell hybrid power system, an online energy management strategy (EMS) for multienergy sources becomes essential. A reinforcement learning (RL) based online EMS is proposed in the article, which integrates expert trajectory guidance with autonomous exploration mechanism. During the offline phase, the strategy mimics power allocation actions based on trajectories generated by the dimension-reduction dynamic programming (DRDP) strategy. Subsequently, the trained agent is deployed online for performance validation. Hardware-in-the-loop (HIL) test results demonstrate that compared with demo-guided TD3 and demo-guided SAC, the proposed strategy can effectively reduce hydrogen consumption rate by 1.34%–2.45%, suppress the state of charge (SOC) fluctuations, and enhance operational efficiency by 2.43%–5.56%. While inheriting global optimization capabilities, the proposed strategy significantly enhances rapid response and adaptive capacity under complex operational conditions, which significantly improves the economic performance and operational stability of aviation fuel cell hybrid power systems.