Real-time Wildfire Risk Assessment Model for Transmission Corridors Based on Feature Engineering and Boosting Integrated Learning Model

Wildfire disasters may cause trippings of the overhead transmission lines and threaten the stable operation of the power grid. In this paper, a real-time bushfire risk assessment model for the transmission corridors based on the feature engineering and the Boosting algorithm is proposed. First of all, the original data of the 20 characteristics in four categories, including the human activity behavior, the geographic factors, the real-time meteorological situations and the regional historical bushfire disasters, which have affected the occurrence of bushfires in the transmission corridors, are collected, extracted and cleaned. Then the quadratic polynomial is used to derive the features, generating up to 236 features. Considering the time complexity of the model calculation, the Wrapper method combined with the five-fold cross-validation method is used to iteratively obtain 100 features with the highest importance to construct a feature subset as the model input to build a real-time bushfire risk assessment model of the transmission corridors based on the Boosting algorithm. In order to minimize the log-loss function as the optimization objective, the Bayesian optimization algorithm is used to search for the parameter space and obtain the optimal model parameters. Finally, the accuracy, precision and recall of the model are verified on the test set. The model performs well with the accuracy of 96.4%, and recall of 88.1%, which is able to effectively evaluate the real-time risks of the transmission corridors in real time. © 2023 Power System Technology Press. All rights reserved.