Prediction of Conductor Icing Thickness Based on Random Forest and WRF Models

Abstract

In this paper, the ice thickness prediction model of WRF field prediction elements is constructed using Makkonen icing model using comprehensive monitoring data of high-voltage transmission line ice accumulation in five southern provinces from Dec. 13, 2020 to Dec. 19, 2020. For the random forest algorithm, the actual icing thickness derived by conductor tension is inputted, 19 predictor variables are selected, such as tower number, phase, and predicted ice thickness value. The WRF-random forest model for icing prediction is constructed, and the best parameters are found by using Bayesian parameter optimization method, showing a coefficient of determination of 0.968 for the training set and 0.949 for the testing set.