Classification of partial discharge sources by SF6 decomposition by-products

Abstract

This paper studies the decomposition characteristics of Sulfur hexafluoride (SF6) under partial discharge (PD) of three common types of defects in gas-insulated switchgear, and classifies the type of defect based on the byproducts. To extract the effect of each defect on insulation gas, the decomposition characteristics under partial discharge of two protrusion, free conducting particles and fixed conducting particles defect configurations were studied. The generated products and their concentrations were used as feature parameters in the support vector machine (SVM) algorithm in Python environment to predict the type of defect causing the PD. The artificial defects were designed and placed in a coaxial gas chamber simulating a GIS configuration. During the stress cycle of up to 126 hours, the applied voltage was regularly increased in steps and hence higher PD levels were measured. Gas samples were collected every 18 hours and were analyzed using Fourier transform infrared (FTIR) spectroscopy gas analyzer. Results have shown that the decomposition of SF6 under the three defects are not the same and the SVM was successfully able to identify the type of the defect based on the supplied gas concentrations.