Abnormal Data Processing of Wind Turbine Based on Combined Algorithm and Class Center Imputation

High precision wind power curve is essential to wind power predictions. In order to eliminate a host of abnormal data of a wind turbine, the abnormal data is divided into decentralized data or abandoned data, and the wind speed into low wind speed and high wind speed ranges in this paper. An algorithm based on the combination of the quartile and clustering is proposed to clean the abnormal data of the two types of the wind speed ranges respectively. Moreover, in view of the elimination of a wealth of missing data, the interpolation method based on class center is employed. The numerical results on cleaning and reconstruction of turbine data in a prototype wind farm have positively confirmed the performances of the proposed algorithm.