Icing Thickness Prediction Model of Transmission Line Based on Linear Interpolation Method and Support Vector Machine

In recent years, transmission line icing disasters occur frequently, that bring great threat to the safe and stable operation of power system. Therefore, an effective icing prediction model is very important. However, the current forecasting model mechanism is complex and faces many difficulties. This paper selects temperature, relative humidity and wind speed as input meteorological parameters and icing thickness as output based on grey correlation degree method to predict icing thickness at the initial stage of ice growth. The sample data were expanded by the interpolation method, that were used as training sets for the BP neural network and SVM regression methods respectively. An improved icing thickness prediction model was obtained. The prediction results show that compared with the traditional BP neural network and SVM regression methods, the proposed method can effectively predict the icing thickness, and the average relative prediction error was 5.742%. The method described in this paper is expected to be applied in line operation and maintenance.