A huge amount of historical data has been generated during the operation of wind farms,and the improvement of data quality is the prerequisite work for achieving high-efficient and intelligent maintenance of wind farms. Therefore,the distribution characteristics and formation mechanism of wind power data in wind farms are analyzed,and a variance change rate criterion and quartile combined method to identify abnormal wind power data is proposed. Firstly,the original wind power curve is preprocessed by physical rules,and the obviously abnormal data is eliminated. Then,the abnormal power data points of the accumulation type of the wind power curve are identified and cleaned by the wind power variance change rate criterion method,and the threshold value of the criterion is automatically obtained through the box plot. After that,the quartile method is used to identify and clean the discrete abnormal data points. Finally,the feasibility of the proposed algorithm is verified by an example. The results show that the proposed algorithm has the advantages of easy implementation,high efficiency,and strong universality. The anomaly recognition performance of the proposed method is superior to the local outlier factor (LOF) or Thompson tau-quartile algorithms,and the value of its time consumption is 9.6 s or 0.49 s lower than that of the LOF or Thompson tau-quartile algorithm,respectively. The universality of the proposed algorithm has been verified at 5 wind farms in different locations.