AI-Assisted Energy Management System for an Ultra-Long-Endurance Solar-Powered High-Altitude UAV

Ultra-long-endurance High-Altitude Long-Endurance (HALE) unmanned aerial vehicles (UAVs) operating in the lower stratosphere are fundamentally constrained by energy availability rather than aerodynamic performance. Sustained multi-week flight requires strict energy-neutral or energy-positive operation over repeated diurnal cycles, while simultaneously preserving the health of onboard energy storage systems. This paper presents a physically realizable, certifiable, systems modelling and supervisory control Energy Management System (EMS) for a solar-powered HALE UAV equipped with wing-integrated photovoltaic arrays, lithium-ion batteries, and a hydrogen fuel cell. The proposed EMS does not directly control the aircraft; instead, it provides energy-optimal setpoints and constraints to an existing lower-level flight control system. Artificial intelligence is employed strictly as a predictive decisionsupport layer to improve disturbance forecasting and optimization robustness, without introducing black-box control or online learning. The EMS formulation integrates energy balance equations, altitude-based energy buffering, and storage degradation awareness within a constrained Model Predictive Control (MPC) framework. Offline-trained learning models enhance solar irradiance and disturbance prediction while respecting hard safety constraints. Simulation-based analysis demonstrates that the proposed architecture enables repeatable energy-neutral 24-hour operation under environmental uncertainty, making it suitable for ultra-long-endurance HALE missions.