关键词: 全电飞机, 模型预测控制, 机载电力系统, 电作动, 机器学习, Saber仿真

Abstract: With the advancement of all-electric aircraft technologies, impulsive, nonlinear, and high-power loads typified by electro-hydrostatic actuators and electromechanical actuators are being increasingly deployed, posing substantial challenges to power balance in onboard electrical power systems of all-electric aircraft. The prevailing source-load decoupled control paradigm lacks an effective coordination mechanism between the generation side and the load side, leaving the power source unable to proactively predict and accommodate uncertain, abrupt load demands. Consequently, power imbalance may induce grid instability and degrade power quality. To address these issues, a source–load coordinated model predictive control (MPC) strategy for the all-electric aircraft onboard electrical power system is proposed. An MPC-based optimal controller is developed based on dynamic models of a three-stage generator and an energy storage battery system, together with a machine-learning-based load forecasting model for electromechanical actuators. Based on the predicted load demand, control commands including the excitation voltage of the three-stage generator and the duty ratio of the bidirectional DC/DC converter in the battery energy storage system are updated online, there-by endowing the onboard power system with proactive prediction and adaptation capabilities against sudden and uncertain load-side demand variations and enabling coordinated source–load optimal control. A Saber-based software-in-the-loop embedded simulation platform is established to validate the proposed strategy. Simulation results demonstrate that the proposed control strategy significantly suppresses bus voltage fluctuations under abrupt and uncertain load conditions, shortens the post-disturbance bus-voltage recovery time, and effectively improves the dynamic stability and response speed of the DC bus voltage.

Key words: all-electric aircraft, model predictive control, onboard electrical power system, EHA/EMA, machine learning, saber simulation