Study on Icing Prediction for High-Speed Railway Catenary Oriented to Numerical Model and Deep Learning

The pantograph-catenary system, serving as the 'throat' for the energy supply of high-speed railways, inevitably experience current degradation caused by ice accumulation on the catenary when high-speed trains pass through areas with high relative humidity, low temperature, and high wind speeds. To explore the impact of the complex service environment on the growth process of catenary icing, this paper, based on the study of the mechanism of catenary icing, proposes an ice prediction method for catenary icing oriented to numerical model and deep learning. Firstly, based on the calculation of key parameters such as capture rate, freezing coefficient, and collision coefficient, a numerical model for catenary icing under time-varying meteorological conditions is developed. Secondly, study the impact of four factors, including wind speed, temperature, liquid water content (LWC), and median volume diameter (MVD), on the evolution of catenary icing. Finally, by integrating the convolutional neural network-long short-term memory (CNN-LSTM), a prediction model for catenary icing is established. The study addressed challenges such as unclear mechanisms of catenary icing, difficulties in detecting icing states, and insufficient sample data. It laid a theoretical foundation for icing forecasting and warning, selection of de-icing strategies, and research on the dancing of catenary icing.