Wildfire prediction for transmission lines provides references for optimizing dispatching resources and pre-deploying firefighting equipment. To address the issues of modeling challenges in wildfire prediction for transmission lines

we put forward a novel approach for wildfire prediction for transmission lines based on residual-temporal convolutional network with multi-source data integration. A multi-source dataset was constructed by using resampling techniques and integrating six categories and fourteen influencing factors

including meteorological elements

vegetation conditions

drought index

human activities

topographical features and transmission lines. Through temporal sequence matching and spatial alignment

a spatio-temporal five-dimensional tensor structure was designed for multi-source data input. Multi-layer residual units designed with two-dimensional group convolution

together with dilated causal convolution

were used to capture spatio-temporal correlation among variables and to fit high-dimensional nonlinear relationships between input features and output labels. Finally

California in the United States was selected as the study area for case analysis. Moreover

the method comparison is conducted and wildfire risk level distribution maps for transmission lines are visualized

validating the effectiveness of the proposed method for wildfire prediction for transmission lines and providing guidance for relevant authorities in conducting wildfire prevention and control efforts for power grids.