A Novel Deep Learning Model for Transmission Line Icing Thickness Prediction

Abstract

Monitoring and prediction icing thickness of overhead transmission lines are important factor threatening for the safe operation of power grid. A novel transmission line icing thickness prediction method is proposed by employing deep learning model. For the proposed method, a novel deep learning model based on parallel coordinates has been used for converting the matrix consisting of multi-dimensional icing meteorological monitoring data to simplified parallel coordinates plot. The icing thickness values has been classified into several levels and convolutional neural networks model was constructed to analyze the relationship between converted simplified parallel coordinates plots and measured icing thickness. The experiment results in Hunan demonstrate the effectiveness of proposed method and the icing thickness prediction accuracy exceeded 92%. The icing thickness values can be accuracy predicted in all cases, which makes it essential in the decision-making of de-icing when the icing thickness is increasing rapidly or the transmission line is heavily iced and monitoring image is not available.