Data preprocessing of wind turbine based on least squares support vector machine and neighbor model

Abstract

Wind turbine data preprocessing is a key step in wind turbine equipment condition assessment, and it helps to improve data quality and data utilization. In this paper, a data preprocessing method has been proposed based on the neighbor model of least squares support vector machine, with the wind speed data as an example. There are strong similarities between the operating conditions of wind turbines with similar wind resources. In this paper, e use the normal data of multiple wind turbine anemometers and the least squares support vector machine (LS-SVM) method to establish the neighbor model between wind speed data of multiple wind turbines. This model reflects the similarity of the wind speed between neighbors. After the model established, the wind speed data containing the outliers will be input to the model. When the wind speed data of one unit is abnormal, the similarity relation between the data and its adjacent units data is destroyed. The prediction residual of the wind speed of this unit will be increased significantly by the neighbor model, indicating that the wind speed data is abnormal data. The method can realize the recognition of wind turbine abnormal data. Based on the actual operation data of a wind farm, the validity of the method is verified.