Research on Transformer Partial Discharge Feature Extraction Based on Empirical Wavelet Transform and Multiscale Permutation Entropy

Abstract Aiming to extract efficiently the fault features of partial discharge in process diagnosis power transformer, a method combining Empirical Wavelet Transform (EWT) with Multiscale Permutation Entropy (MPE) is advanced transformers discharge. Firstly, four different pulse signals are analyzed by EWT method, and signal decomposed according frequency domain characteristics obtain intrinsic mode function (IMF) signal. Then, calculated multi-scale permutation entropy IMFs complete feature extraction. Finally, semaphore used as eigenvector Support Vector Machine (SVM) for glitch diagnosis, accurate systematization transformer realized. Compared Continuous (CWT), Mode Decomposition (EMD), Ensemble (EEMD) extraction way, it shows that raised EWT-MPE more valid diagnosing analyzing faults, accuracy classification 96.43%.