Effective data cleaning methods can improve the quality of wind turbine measurement data. The quality of wind turbine data plays a very important role in wind resource assessment, wind power accurate prediction, and performance diagnosis of wind turbines. There are many uncertainties in the data collection and monitoring systems of different wind turbines for fault diagnosis, which result in uneven quality of wind measurement data for wind turbines. This paper proposes a new method for identifying the probability interval of wind power. This method uses the characteristic changes between wind speed and power to clean and correct the effective data of wind turbine measurement data. It can effectively improve the utilisation rate of wind turbine data. This paper selects wind turbine data from a wind farm in the northern part of Ulanqab, Inner Mongolia Autonomous Region from 2020 to 2022. By sequentially subjecting the data to rationality and validity tests, wind power interval checks, and finally, cleaning and correcting abnormal data, which are carried out by utilising the correlation of the turbine. The final results indicate that: by using the wind power interval method, it is difficult to distinguish abnormal wind speeds. This method can improve data quality and enhance the accuracy of wind speed and power. According to statistics, the data integrity has been significantly improved from 68.7%-92.5% to 90.1%-92.7%. Above all, the data integrity has been significantly improved. This method achieves mutual calibration between wind speed and power through the wind power probability interval recognition method. It provides fundamental support data for predicting and regulating the power generation of wind farms. It provides guidance and a basis for more refined meteorological service products for power and other related sectors.