In view of the characteristics of wind turbines SCADA monitoring data, such as massive, high-dimensional and complex, a fault diagnosis method of wind turbines based on manifold semisupervised K-means clustering algorithm was proposed. Wind turbines SCADA data were analyzed, the state parameters of wind turbines were extracted to form the characteristic data set, and the traditional K-means clustering algorithm was optimized. The manifold distance was used as the similarity measure, and the semisupervised K-means clustering analysis was carried out for SCADA data. The experimental results show that the improved algorithm is more effective than the traditional K-means clustering algorithm in identifying the state of wind turbine