Separation of multiple partial discharge sources in power transformer

Abstract

Partial discharge (PD) in power transformer degrades the dielectric insulation and results in insulation failure and breakdown after a long period. In fact, a partial discharge detector can gather signals from two or more PD sources which increase the difficulty on pattern recognition and insulation state assessment. How to separate multiple PD signals is meaningful to subsequent data processing. In this paper, a PD separation algorithm based on cumulative energy (CE) function is introduced. The CE functions in time domain (TCE) and frequency domain (FCE) are calculated from PD waveforms and their FFT spectra, respectively. By using an oblique line to cross the CE curves, width features are extracted from the intersection points between them. Through the mathematical morphology gradient (MMG) operation, sharpness features are extracted to characterize the rise steepness of CE. A clustering algorithm is adopted to discover different clusters in feature space and separate PD signals. After that, a transformer experiment platform is introduced and some different defects are placed at different places in this transformer to produce PD signals. The pulse current method is applied to gather their PD waveforms and to record phase resolved partial discharge (PRPD) which can be used to analyse discharge characteristics of different defects. The separation algorithm is examined with mixed PD current pulses acquired from experiments. The results prove the feasibility of separation algorithm with different PD defects.