Accurate prediction of short-term wind power is very important for reliable operation of power grid. In order to enhance the prediction accuracy, a short-term wind power prediction method based on meteorological scenario correlation was proposed to capture meteorological characteristics fully. Firstly, to address the many discrepancies in climate data, a meteorological feature clustering model under the framework of CKFD-KM is constructed by combining Complete Kernal Fisher Discrimination and K-Means algorithm, and the data value of numerical weather forecast is extracted and aggregated. Secondly, based on different meteorological scenarios, the basic prediction sub-model is selected to build a multi-scenario wind power prediction model.Finally, using the actual data of a place in Shandong province to test, the research proves that the performance of the new model is better than the traditional model.