Due to the influence of unplanned shutdown, load limiting operation, extreme weather and other factors in the operation of wind turbines, there are a large number of horizontal and discrete abnormal data in the power curve. However, the existing data cleaning methods are limited by the conditions of the algorithm itself, and the local scattered accumulation data cannot be accurately identified in a single application process, or the characteristics of sample data are limited, so the algorithm cannot be directly applied. Therefore, by combining the advantages of DBSCAN(density-based spatial clustering of applications with noise) clustering algorithm and Laida criterion, a suitable method for cleaning abnormal data of wind turbine power curve is proposed. The example shows that the method can identify the discrete data of the wind turbine under abnormal working conditions with high quality, and has strong generalization ability, and has good application in the power curve data cleaning of wind turbine. It can be used as the basis for data analysis and data mining of wind turbines, and provide effective data guarantee for subsequent efficient use of wind energy, quality and efficiency improvement of wind turbines, strategy optimization and other work.