Automating the Classification of Field Leakage Current Waveforms

Abstract

Leakage current monitoring is widely employed to investigate the performance of high voltage insulators and the development of surface activity. Field measurements offer an exact view of experienced activity and insulators’ performance, which are strongly correlated to local conditions. The required long term monitoring however, results to the accumulation of vast amounts of data. Therefore, an identification system for the classification of field leakage current waveforms rises as a necessity. In this paper, a number of 500 leakage current waveforms recorded on a composite post insulator installed at a 150 kV High Voltage Substation suffering from intense marine pollution, are investigated. The insulator was monitored for a period of 13 months. An identification system is designed based on the considered data employing Fourier analysis, wavelet multiresolution analysis and a neural network. Results show the large impact of noise in field measurements and the effectiveness of the discussed system on the considered data set.