GRNN–DP–MPC Co-optimization for predictive energy management in hybrid UAVs

As hybrid unmanned aerial vehicles (UAVs) proliferate in low-altitude applications, effective energy management for multi-source hybrid energy systems has become a core enabling capability. Yet many existing strategies suffer from weak real-time prediction and imperfect energy allocation. To overcome these limitations, we develop a predictive energy management strategy (PEMS) that integrates a generalized regression neural network (GRNN) with dynamic programming (DP). The proposed GRNN–DP predictive optimization model is instantiated on a hybrid UAV power system: the GRNN delivers high-accuracy voltage forecasts, DP provides globally optimal references, and the power split is executed online within a model predictive control (MPC) framework. Acting cooperatively, this GRNN–DP–MPC scheme cuts computational complexity while curbing lithium-battery energy consumption. Simulation and experimental evidence confirms reduced battery usage under practical operating conditions and improved endurance, thereby offering a new pathway for UAV energy management and yielding both theoretical and practical value for long-duration, low-altitude flight.