Signal processing of PD measurements to predict arcing faults in MV switchgears

Abstract

Arcing faults in MV switchgear cause serious hazard to personnel, significant damage to equipment, and often serious process interruptions. Many of the faults develop slowly, e.g. because of insulation degradation or loose connection. An interesting research question is whether these developing faults could be detected before they escalate into devastating high-power faults. Detection of partial discharges (PD) or monitoring of temperature has been suggested in on-line monitoring systems. In this research, a switchgear panel has been subjected to PD in the laboratory and measurements have been captured by different sensors and recorded by high frequency oscilloscope. Generally, the on-line signals are suppressed by high frequency noise, therefore, the de-noising of PD measurement is of paramount importance to get reliable arcing fault prediction results. The discrete wavelet transform (DWT) to de-noise such PD signals has been employed in this paper. Time domain and frequency domain comparisons of original and de-noised PD signal reveal the significance of this technique for arcing fault prediction in medium voltage (MV) switchgears.