Analysis of the Applicability of Predicting Transmission Lines Icing by Atmospheric Reanalysis Data Based on the BP Neural Network Algorithm

The phenomenon of ice accumulation on power transmission lines leading to flashover trips, mechanical damage to conductors, hardware, and towers is a major factor affecting the safe operation of China’s power grid during winter. Accurate prediction of ice accumulation on power transmission lines is a crucial foundation for the power grid to carry out anti-icing and disaster mitigation work. In the field of power transmission line maintenance, numerous scholars have proposed a multitude of ice accumulation prediction models based on divergent principles. However, these models vary in structure and have varying degrees of limitations in applicability. To identify a more broadly applicable ice accumulation prediction model, this study employed the BP neural network algorithm to assess the correlation between numerous meteorological factors in atmospheric reanalysis data and ice accumulation on transmission lines. The analysis identified 31 strongly correlated meteorological data points, 9 weakly correlated data points, and 16 uncorrelated data points. All meteorological data were meticulously categorized into four distinct groups: strongly correlated, weakly correlated, correlated, and uncorrelated. These data were then employed as inputs to predict icing conditions. The prediction accuracy of models utilizing disparate datasets was evaluated.