Combined Prediction Method of Transmission Line Icing Based on ARIMA-CSSVR

Abstract

Real-time prediction of icing on power transmission lines is of great significance for grid disaster warning. Under the same terrain conditions, micro-meteorological such as air humidity, temperature, and wind speed are the main reasons that affect the icing of power lines. The construction and prediction accuracy of the existing mechanism-based and statistical model are difficult to meet the requirements of practical applications, while related intelligent computing models ignore the effect of time accumulation. This paper proposes a combined prediction model based on ARIMA-CSSVR. ARIMA predicts the linear growth of icing on power transmission lines. SVR based on CS optimized is used to fit the nonlinear errors contained in the ARIMA predicted time series. Then combined the two results as the final prediction. A numerical example is used to compare it with other machine learning algorithms, and the prediction advantage of this method is verified.