A hybrid algorithm based on s transform and affinity propagation clustering for separation of two simultaneously artificial partial discharge sources

Abstract

This paper presents a hybrid algorithm for separation of two simultaneous partial discharge (PD) sources of oil-paper insulation based on S transform (ST) and affinity propagation clustering (APC). Similarities between PD pulses are acquired by comparisons of the corresponding ST-amplitude (STA) matrices, which are input of APC to realize the PD pulses separation and obtain two sub-groups of PD pulses having similar time-frequency characteristics. A classification-based model for separation results validation are developed using a support vector machine with particle swarm optimization (PSO-SVM) classifier and 27 phase-resolved partial discharge (PRPD) statistical features. Artificial defect models are made to simulate two PD sources simultaneously active. Several PD data of different two simultaneous PD sources are acquired in laboratory and adopted for algorithm testing. It is shown that ST computes very fast and is suitable for online PD applications. The separation results of PD data produced in laboratory are verified by the developed validation model, which demonstrate that ST combined with APC can effectively eliminate pulse-shaped noises (PSN) and separate pulses of two simultaneous PD sources. Comparisons with typical separation methods from the state of the art provide better separation performance of the proposed ST combined with APC algorithm for two simultaneous PD sources. The obtained results in this work provide a solid basis for the data mining technique that can be used to facilitate PD diagnosis of transformers.