The invention discloses a fan power curve data preprocessing method based on outlier detection. According to the method, based on collected fan data comprising wind speed, active power, environment temperature, a pitch angle, etc. and real-time operating data obtained through a supervisory control (SCADA) system, outlier detection criteria are used to perform abnormal data cleaning, air density obtained through calculation is utilized to correct wind speed data, an outlier detection algorithm is used to detect and eliminate outliers in fan power curve data. According to the data preprocessingmethod, the outlier detection criteria guarantee the effectiveness of fan data information; wind speed correction guarantees the comparability of the fan data; and the outlier detection algorithm guarantees the reliability of final power curve data. The method is based on data driving, does not have a special requirement on data and is high in universality. Compared with the prior art, data credibility is higher, the process is more standard, and a high theoretical property and practicability are achieved.