A wavelet transform technique for de-noising partial discharge signals

Partial discharge (PD) diagnosis is essential to identify the nature of insulation defects causing discharge. The problem of PD signal recognition has been approached in a number of ways. Most of the approaches are based on laboratory experiments or on signals acquired during off-line tests of industrial apparatus. On-line testing is vastly preferable as the equipment can remain in service, and the operators can monitor the insulation condition continuously. Interferences from noise sources have been a persistent problem, which have increased with the advent of solid-state power switching electronics. Use of wavelet transform technique offers many advantages over conventional digital filters and is ideally suited to process non- stationary signals (transients) often encountered in high voltage testing and measurements. In this paper, an empirical wavelet- based method is proposed to recover PD pulses mixed with excessive noise/interference. A critical assessment of the proposed method is carried out by processing simulated PD signals along with noise signals using MAT LAB software.