Prediction of Wind Turbine Blade Icing Based on LSTM-SVM

Abstract

As one of the important parts of wind turbines, icing of wind turbine blade will lead to loss of power generation, and even cause blade to break and injure workers in severe cases. Therefore, the prediction of blade icing has always been the focus of the wind power industry. To solve the problem, this paper proposes a wind turbine blade icing prediction model based on LSTM-SVM. Firstly, we analyzed the time series data of a wind farm and the mechanism of blade icing to obtain the features sensitive to wind turbine blade icing. Then, the LSTM is used to predict these feature parameters for future moments. Finally, an SVM model was employed to diagnose the icing status of wind turbine blades based on the predicted feature parameters, thus achieving the prediction of blade icing. The model was verified by experimental data and was able to accurately predict whether the wind turbine blade would icing or not. Compared with traditional algorithms, it can predict the icing status of wind turbine blade in advance and has more accurate prediction results, making it more suitable for practical applications of blade icing prediction.