K-means Clustering Study on the Evolution Rule of Ice Thickness for Segmental Prediction of Icing Process

Ice cover prediction is of great significance for judging the future ice cover situation and arranging ice melting work in advance. In this paper, 11 ice cover characteristic parameters describing the ice thickness-time curve of ice cover events are proposed, as well as the methods of normalization and hierarchical K-Means clustering. The 282 ice-covering events were clustered three times, and four types of ice-cover thickness change patterns were obtained, namely, spike type, saturated type, multi-peak type and artificial ice melting type, and their regular characteristics were summarized and analyzed. Based on the clustering results of the type of ice cover thickness change rule, the distribution law of the relevant parameters such as the voltage level where the line is located, the year of ice cover, the number of days of ice cover and the maximum ice cover thickness were statistically analyzed, which laid the foundation for the fine prediction of ice cover on transmission lines.