Abstract:Serious icing coating of transmission lines may cause the decline of the mechanical and electrical performance of transmission lines, threatening the safe and stable running of power systems. Line icing prediction technology is one of the difficulties in the field of anti-icing power grid. In this paper, we take advantage of the massive monitoring data of natural icing coating of power transmission lines, and conduct data manipulation such as outlier processing, missing value imputation. To evaluate prediction accuracy, we propose an interval accuracy evaluation method based on icing tension floating interval. This paper conducts the data-driven icing prediction technology of transmission lines based on new deep learning, and blends historical monitoring pull, micrometeorological data and future weather forecast, and builds the pull time series and first-order difference pull time series icing prediction models. These models achieve the accurate pull prediction of the icing monitoring terminal in the next 24 hours, and can predict whether or not the icing coating and the icing coating degree. It is helpful for the decision of anti-icing and melting ice operation, and ensure the stable and safe operation of power system.