Abstract: A large number of historical data will be generated during the operation of wind farms, which plays a very important role in improving the operation quality of wind farms. However, abnormal data in the data collected by the monitoring system of wind farms account for a high proportion, which has a serious impact on wind power prediction, unit state monitoring and other work. In order to accurately identify outliers in running large data ofwind farm, this pa-per puts forward anidentification algorithm of abnormal data for wind farms with conditional probability distribution characteristics. Abnormal data identification model is established, and wind farm measured data and synthetic data are chosen as the research object, using the identification model to deal with two kinds of data, abnormal data identification results are obtained. The results show that the proposed algorithm can effectively identify all kinds of abnormal data and solve the problem of abnormal power point identification.