Research on Icing Status Forecasting of Wind Turbine Blades Based on Machine Learning

Abstract

In order to respond to the problem of wind turbine blade icing disaster in a timely manner, preventive measures have been taken to prevent possible icing disaster and ensure the safe and efficient operation of wind plants. Based on monitoring data from wind plants supervisory control and data acquisition (SCADA), this paper proposes a wind turbine blade based on bidirectional long short-term memory (Bi-LSTM) and support vector machine (SVM) model to forecast the icing state. Firstly, the principal component analysis (PCA) is used to reduce the dimensionality of the monitoring data of the wind turbine blade icing state. The data features after screening are preprocessed. Secondly, the Bi-LSTM and SVM model are trained based on historical data, and the training results show that the model has good accuracy. Finally, the real data is input into the trained Bi-LSTM prediction model for data feature prediction, and then the prediction output result is input into the SVM model to determine whether the wind turbine blades will have an icing disaster. The analysis of examples shows that the proposed method can accurately predict the icing status of wind turbine blades.