Gas analysis technique for gas insulated switchgear condition monitoring and diagnosis

Abstract

Sulphur hexafluoride gas insulated switchgear (GIS) is widely used in electrical power supply system and therefore needs regular preventive maintenance. Usual diagnosis methods used are based on acoustic, optical, electrical and ultra high frequency techniques. A new method with great potential is using gas by-products analysis. Previous gas by-products research is confined to a plane-plane electrode instead of typical coaxial GIS configuration, a limited number of defect types and the by-products analysis using gas chromatography. In this thesis, partial discharge experiments using a purposely designed coaxial GIS chamber were carried out to expand the diagnosis database for a new set of simulated defects represented by three categories, namely sole defect, hybrid defect, and material dependent defect. A total of eight defects namely, free conducting particle, electrode to dielectric void, electrode protrusion, fixed particle aluminium on spacer, fixed copper particle on spacer, electrode protrusion-fixed copper particle hybrid, electrode protrusion-free copper particle hybrid, and electrode to dielectric void-free copper particle hybrid were simulated. In each experiment lasting up to 50 hours, continually applied voltage at 0.2 MPa pressure, samples of gas by-products were taken at 10 hour intervals for an off-line Fourier transform infrared spectrometer gas analyses. A total of 12 gas by-products due to partial discharge activity in all defects were detected. Arranged according to significance, these are hexafluoroethane, sulphur dioxide, sulfuryl fluoride, octafluoropropane, silicon tetrafluoride, thionyl fluoride, carbon monoxide, disulfur decafluoride, hydrogen fluoride, tetrafluoromethane, carbonyl sulphide and tetrafluoride. Arranged according to significance, the most harmful gases are produced by the defects such as electrode protrusion-fixed copper particle hybrid, fixed copper particle, electrode protrusion-free copper particle hybrid and electrode protrusion. The type, number, concentration and chemical stability of by-product gases are found to be closely correlated to the type of defect. Further analyses using pattern recognition with eight algorithms based on the presence and concentration of the gas by-products were carried out. The random forest algorithm successfully recognises a given defect with an accuracy of 87.5%. The performance of the random forest algorithm is 1.5 times better than the next best algorithm. This research illustrates the feasibility and applicability of an effective GIS diagnostic using gas by-products analyses, in particular, using the random forest pattern recognition.