A wavelet transform approach to adaptive extraction of partial discharge pulses from interferences

This paper is concerned with an application of discrete wavelet transform (DWT) to adaptive extraction of partial discharge signals (PDs) from unknown and variable noise. The proposed method can be applied to both on-line energized network (energized with power frequency voltage) and some offline energized network (energized with approximate power frequency voltage). This paper describes an adaptive extraction system (AES) which performs partition of original discrete sequences to more than one time-frames, implements DWT of each time-frame and identifies noise from PDs. AES tunes sets of thresholds according to obtained noise characteristics, and then it modifies DWT coefficients on different scales of DWT to separate noise from PDs. The AES approach produces a significantly improved signal-to-noise ratio (SNR) and almost undistorted partial discharge pulse waveshape at the output of AES. Simulation and experiment studies associated with this AES approach are presented in this paper.