Partial discharge signals separation using cumulative energy function and mathematical morphology gradient

Partial discharge (PD) measurement and interpretation have become a powerful tool for condition monitoring and failure risk assessment of high voltage power equipment insulation. The occurrence of multiple discharge sources affects interpretation accuracy. This paper presents a PD signal separation algorithm using cumulative energy (CE) function parameters clustering technique. The waveform of PD signals are acquired by digital detection instruments with high sampling rate. Cumulative energy functions in time domain (TCE) and frequency domain (FCE) are calculated from PD waveforms and their FFT spectrums, respectively. Mathematical morphology gradient (MMG) operation is applied to the TCE and FCE to describe their variation characteristics. The feature parameters including width, sharpness and gravity are extracted from CEs and MMGs in both time and frequency domain, and compose a six-dimension feature space. The improved density-based spatial clustering of applications with noise (IDBSCAN) clustering algorithm is adopted to discover clusters in the feature space. The proposed separation algorithm is examined with mixed current impulse signals acquired from PD experiments on artificial multi-defect models and an on-site transformer. The separation results indicate that the proposed algorithm is effective for separating mixed PD signals initiated from multiple sources.