Research on Fire Risk Assessment for Distribution Lines Corridors: A Case Study in Sichuan Province, China

In mountainous areas of Sichuan Province in China southwestern, because distribution lines corridors often traverse jungles and mountainous situations which has complex geographic environment, accidents of forest fires caused by distribution lines may occur. Currently, most of the forest fire prediction studies mainly focus on area and rank assessment, and there are fewer prediction studies for forest fire spreading rate. In this paper, for an area near a 10kV line corridor in Xichang City, Sichuan Province, China, multi-source data such as meteorology, topography, combustibles, etc. are used as influencing factors. Several forest fires spreading speed prediction models are establish using Wang Zhengfei model, random forest regression algorithm (RFR), long-short-term memory network (LSTM), convolutional neural network algorithm (CNN), and extremum learning machine (ELM), respectively. Through the experimental analysis and comparison, it is obtained that the prediction model based on LSTM has the best performance in terms of decidable coefficient (R ² ) and mean absolute error (MAE) evaluation indexes, and the prediction accuracy can reach 92%, which can be effectively used for the prediction of fire risk of power lines. The research results can provide support for line corridor fire prevention to a certain extent.