Phase determination of partial discharge source in three-phase transmission lines using discrete wavelet transform and probabilistic neural networks

Abstract

This paper proposes an approach to determining the phase where partial discharges (PDs) occur in three-phase transmission lines with discrete wavelet transform (DWT) and probabilistic neural networks (PNNs) using the high frequency current transformer (HFCT). For accurately determine the PD source, the peak absolute value and average power of PD signal are used in the proposed method. The accurate ratios of the PD occurrence prediction using conventional observations of PD signals via oscilloscopes are improved by the proposed method. The standard PD pulse calibrator is installed and PD signals are individually injected into the power cable in each phase of three phase transmission lines. The electricity signal is simultaneously detected on the grounding line by the HFCTs. Furthermore, noises of measured signals are filtered by DWT for which a suitable mother function should be chosen. According to Kirchhoff’s circuit law (KCL) and the concept of power delivery, the peak absolute value and average power of PD signal are applied as input of the PNN. Finally, experimental results validate that the proposed approach can precisely determine the phase location of PD sources in the field.