The Early Warning of Icing Flashover Fault of Transmission Line Based on Support Vector Machine and Partial Mutual Information Method

Icing flashover fault of transmission line is a major cause of power system failure. Current researches mainly focus on the study of the icing flashover characteristics of insulators and the evaluation model of flashover voltage with different influencing factors by artificial icing tests on insulators. However, the evaluation model of flashover voltage, which only considers a factor or a few factors, cannot fully reflect the insulator flashover voltage under the combination of all factors. Moreover, the current characteristics model of icing flashover is difficult to be used to forecast icing flashover fault directly due to the measurement errors. In this paper, support vector machine (SVM) and partial mutual information (PMI) method are adopted to forecast icing flashover fault considering the rapid development of data mining technology. Firstly, since the dimension of the input variable has important influence on the extensive ability of SVM model, PMI method is proposed to select the key factors for input variables. Secondly, the SVM model is trained and tested with the sample data to obtain the prediction model. Finally, the predicted results are compared with those of other models to prove its effectiveness.