Abstract: Aiming at the overhead transmission line ice cover prediction problem, this paper proposes a prediction model based on graph convolutional neural network. Firstly, a graph model containing overhead transmission line topology and environmental factors is constructed by integrating the relative air humidity, wind speed, air temperature, and related data such as conductor surface temperature, conductor temperature, ambient humidity, and conductor tension change, defining the nodes as the monitoring points of the line, and the edges represent the spatial relationship between the monitoring points and the environmental impact supervision. Then, a graph convolutional neural network is used to extract features from the graph model, capture the interactions between nodes by passing node information layer by layer, and introduce an attention mechanism to weight the features of different nodes to improve the prediction performance; finally, supervised learning is performed using historical ice cover data to optimize the model parameters and ensure the generalization ability. The experimental results show that the model has high prediction accuracy and robustness under different weather conditions and line environments, providing effective support for the power sector to take timely ice melting measures.

Key words: ice cover prediction, graph convolutional network, attention mechanism, overhead transmission lines, multi-source data fusion