Spatial Distribution Prediction of Wildfire Risk in Transmission Line Corridors Based on Dynamic Bayesian Networks

The wildfire disasters near the transmission line corridor may lead to a large area of power outage. Accurate prediction of potential wildfire disasters can provide auxiliary guidance for power emergency repair. This paper proposes a prediction model of wildfire distribution in transmission corridor based on dynamic Bayesian network (DBN). Firstly, a total of 17 wildfire factors in four categories of meteorology, human, vegetation and geography were selected. The random forest algorithm is used to select the optimal factor set. The DBN model of different time scales is established by eliminating the coupling relationship between factors through static Bayesian network. With the decrease of time scale, the effect of DBN model is gradually improved, and the accuracy of DBN model with 3 days as time scale is 86.39%. Finally, the model was used to predict the occurrence of wildfires during the Spring Festival in Zhaoqing City, Guangdong Province in 2023, and a spatio-temporal distribution map of wildfires with a resolution of 1km× 1km was drawn. The validity of the distribution map is verified by real fire points.