Demagnetization Modeling and Fault Diagnosis in a 25-kW PMSG Using Finite Element Method and Machine Learning Techniques

In wind power power systems, the reliable functioning of Permanent Magnet Synchronous Generators (PMSG) depends on condition monitoring and fault diagnosis. Using 3D simulation models, we present a diagnostic approach for identifying multiple demagnetization problems in PMSG. Specifically, we use the four states of demagnetization: healthy condition, 50% unipolar magnet breakage, 75% demagnetization, and 100% demagnetization. The major goal is to improve PMSG monitoring capabilities, which will increase operating efficiency, optimize maintenance procedures, and boost wind energy extraction reliability. In order to accomplish this, we created a sophisticated defect diagnosis method that combines the Discrete Wavelet Transform (DWT), Kruskal Wallis, machine learning algorithms, and the Finite Element Method (FEM) to offer insights into the machine's basic principles and physical behavior. The proposed technique was evaluated and validated across four different scenarios of demagnetization faults, evaluating both faulty and healthy PMSG situations utilizing current and flux outputs. The simulation results demonstrate the approach's effectiveness and reliability.