A Wavelet-based Tool for On-line Partial Discharge Diagnostics

Partial discharge (PD) measurements provide useful information on the condition of electrical insulation in high and medium voltage components. However, UHF on-line PD measurements are often affected by external noise and disturbances of various nature which can hide the discharge signal and strongly affect the sensitivity and reliability of the measurement. In such a harsh environment, the extraction of partial discharge becomes a fundamental topic. The application of Wavelet Packet Transform (WPT) to extract partial discharge signals buried in excessive noise is described in this paper. The denoising technique, applied to numerical signals, is based on an energy criterion in conjunction with hard-thresholding