The Dataset for a Machine Learning Forecast Model of an Overhead Power Lines Icing

The problem of overhead power line phase conductors and ground wire icing is currently very strong. While the climate is changing, the line icing forecast become less accurate. The periods of strong icing changes and even the regions that were not considered as dangerous for lines icing are facing a “new” problem for grids. The ice is dangerous for an overhead power line and is a consequence of the stability of a power system. There are several measures to prevent and control ice damage to a powerline conductor. But the first step of using these measures and instruments is icing forecast in time and location. The article presents the approach for using a machine learning algorithm and dataset collection to accurately forecast ice on power line phase conductors and ground wires. The aim of research is to collect the dataset and research accuracy in forecasting the ice formation without special sensors. The results demonstrate a satisfactory level of performance the accuracy of approximately 81% on the test set that are satisfying and may be testing on a real grid.