A multi-objective DRL energy management system for fuel-cell/battery/supercapacitor powered H-VTOL-UAV system to achieve overall efficiency improvement

Due to the unique combination of vertical lift capabilities and efficient cruising performance, Hybrid Vertical TakeOff and Landing Unmanned Aerial Vehicles (H-VTOL-UAV) have seen surging adoption in fields such as logistics, surveillance, and patrol. The continued advancement of these platforms relies heavily on the optimization of energy efficiency. Nevertheless, current research on hybrid power energy management strategy (EMS) is largely restricted to conventional aircraft architectures, resulting in a scarcity of strategies specifically tailored for HVTOL-UAV. To address this deficiency, this study introduces a multi-objective Deep Reinforcement Learning (DRL) framework designed to maximize the system efficiency of fuel cell hybrid power systems. We formulate a comprehensive modeling approach that includes the hybrid powertrain, the UAV power demand profile, and an efficiency-oriented learning algorithm. Comparative experiments against the Equivalent Consumption Minimization Strategy (ECMS) verify that the proposed method significantly enhances overall system efficiency while effectively mitigating fuel cell stress across various flight conditions.