Abstract:Aiming at the problem that the wind power of wind farms is strong and the prediction accuracy of medium and long-term wind power is not high, this paper proposes a method for predicting mid- and long-term wind power of wind farms based on high-level meteorological data. Firstly, by regularizing and normalizing high-level meteorological data, we can find and improve meteorological factors that are strongly related to wind power. Secondly, we combine the atmospheric motion equation with the descending gradient equation to establish an evolutionary physical model of high-level meteorological data. Class and mining algorithms classify multi-dimensional massive upper-level atmospheric data, and train and correct the derived upper-level atmospheric data model based on data. Finally, build high-level atmospheric motion data and wind farm based on model and big data machine learning methods. The law between historical data is based on the combination of statistical analysis and physical model to predict the long-term wind power of wind farms. The mid- and long-term wind power of a wind farm is predicted by combining wind resource data from a certain area in southwestern China. It is proved that the proposed method can effectively improve the accuracy of mid- and long-term wind power prediction of wind farms.