Diagnosis of AD and DSV Winding Faults Based on FRA Method and Random Forest Algorithm

• Published in: 2023 IEEE 4th International Conference on Electrical Materials and Power Equipment (ICEMPE) pp. 1 - 4
• Main Authors: Wang, Song, Qiu, Shengxuan, Xie, Fei, Yang, Siying, Yu, Kerui, Li, Tong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07.05.2023
• Subjects: axial displacement (AD); Classification algorithms; disc space variation (DSV); Fault diagnosis; Feature extraction; Frequency response; frequency response analysis method; Maintenance engineering; power transformer; random forest algorithm; Training; Windings
• DOI: 10.1109/ICEMPE57831.2023.10139451

Frequency response analysis (FRA) method is one of the most important diagnostic methods for detecting the faults of transformer winding. To accurately diagnose axial displacement (AD) and disc space variation (DSV) faults of transformer winding, in this paper, a diagnosis method which combines the FRA method with the random forest algorithm (RF) is proposed to diagnose the fault types of winding. First, an experimental winding model with AD and DSV faults is set. Then, a plenty of FRA measurements are conducted to acquire the FRA dataset of the winding. Furthermore, the variation features of the FRA traces are extracted by using some typical statistical indices. Finally, the trained RF model is used to classify AD and DSV winding faults. The average accuracy rates of this RF model for classification of AD and DSV faults are obtained by random 1000 tests. The results show that the proposed method is feasible, and it has very high accuracy for diagnosing and identifying the AD and DSV winding faults.