Prediction of Wildfire-induced Trips of Overhead Transmission Line based on data mining

Once a wildfire occurs near the transmission line, the continuous burning of the wildfire would trip the transmission line with a low success rate of reclosing. In this paper, data mining tools including Support Vector Machine(SVM), Back Propagation Neural Network(BPNN) and Random Forest(RF) were used to predict the events of wildfire-induced trips. As the number of wildfire-induced trips are much less than that of wildfires, the SMOTE, Random undersampling and EasyEnsemble were used to balance the data. The results showed that EasyEnsemble performs best among the three data-balance approaches and the model established by EasyEnsemble combined with RF has the best generalization with an AUC of 0.97. And the accuracy, recall and precision are all 91.7%.