Multi-Output Physical-Informed Neural Networks Based Impedance Identification of Buck Converter in More Electric Aircraft Power Systems

With the increasing complexity of more electric aircraft (MEA) power systems, particularly due to the rising nonlinear constant power loads (CPLs), the stability analysis of MEA power systems is challenging. The data-driven models represented by neural networks, with their strong ability to characterize nonlinear properties, are superior in the field of power system component modeling. This paper focuses on the buck converter in the DC microgrid of a multi-electric aircraft as the research object, and proposes an Impedance measurement methods based on Multi-Output Physical Information Neural Networks (MO-PINN), which can realize the measurement of impedance with minimal reliance on experimental data. Simulation verifies the feasibility of the method, which introduces a new idea and approach for analyzing the stability of MEA dynamical systems.