GIS Discharge Fault Analysis Based on SF6 Decomposition Product Classification Planning
In order to improve the safety evaluation ability of GIS,based on the robustness of GIS diagnostic model on the basis of machine learning algorithm,empirical probability functions of different fault types processed by Arrhenius chemical reaction model are studied.Six machine learning algorithms are used to establish discharge fault and insulation defect models.The recognition algorithm is trained by learning the aggregate data sets and their ratios of the volume fractions for different gases(SO2,SOF2,SO2F2,CF4,CO2)in the constituent system.The results show that the empirical probability model can effectively identify various GIS insulation defects and their coexistence states.The volume fraction ratio of SO2F2 to SO2 is 4.2,which is a critical point of high energy discharge state and is of great significance for fault warning.In the GIS insulation defect test result graph based on Gaussian distribution,the region y≈0.15 is where multiple discharge faults coexist,which needs to be paid close attention to.
gas insulated switchgearmachine learning algorithmSF6 decomposition productsinsulation defectdischarge fault
万方数据
