GA-BP Neural Network-based Fault Diagnosis GIS Equipment with SF6/N2 Gas Mixture
In order to diagnose latent faults in SF6/N2 gas mixture GIS equipment,an experimental platform was con-structed to explore the relationship between the content of SO2 gas detected by ultraviolet fluorescence method and the dis-charge quantity.BP and GA-BP neural networks were established to predict the discharge amount under four typical defect types,and the increase in SO2 concentration was added as one of the input parameters.The predictive results were com-pared and analyzed to select the optimal diagnostic algorithm.The results show that there is a positive correlation between the discharge quantity and the concentration of SO2 for each type of discharge defect.The prediction accuracy and fitting degree of the GA-BP neural network model are 0.86 and 0.9386 respectively,with an average relative error of 1.23％.The proposed method has a significant advantage with respect to evaluation results and can provide fundamental data for estab-lishing a diagnostic algorithm for latent faults in SF6/N2 gas mixture GIS equipment.
SF6/N2 gas mixturesulfur dioxide concentrationGA-BP neural network modellatent fault
万方数据
