Detection for Axial Displacement Fault of Winding in Autotransformer Based on Dynamic Frequency Division of FRA and HHO-RF

To address the issue that traditional Frequency Response Analysis (FRA) methods are primarily used for conventional transformers and cannot accurately identify common axial displacement faults in autotransformer windings, a fault diagnosis method based on dynamic frequency division of FRA curves and Harris Hawks Optimization-Random Forest (HHO-RF) is proposed. Firstly, based on the existing experimental conditions in the laboratory, an axial displacement fault simulation platform for autotransformers was constructed to obtain frequency response data of the windings under different axial displacement faults. Secondly, the points where the extreme values of the amplitude-frequency characteristic curve coincide with the zero-crossing points of the phase-frequency characteristic curve are taken as preliminary frequency division points. The frequency response data is then dynamically divided into multiple frequency bands, and polar plots for each frequency band are drawn. Next, four gray-level co-occurrence matrix (GLCM) texture features and four Tamura texture features related to the polar plots are extracted. Finally, the state of the autotransformer windings is analyzed through the variation patterns of the graphs and features, and the HHO-RF optimization algorithm is used to classify and diagnose the faults. The results indicate that after frequency division, the generated polar plots have fewer overlapping points, making it more conducive to extracting image features. When faults of different degrees occur, the variation trends in the polar plots become more pronounced. The fault diagnosis accuracy achieved through the parameter optimization algorithm exceeds 90%, which is approximately 10% higher than that of the standard RF algorithm.