Short-term photovoltaic power forecasting is of great significance for maintaining the security and stability of the power grid and coordinating the utilization of resources. In this paper, a short-term photovoltaic power generation prediction method based on Kmeans algorithm and Support Vector Machine (SVM) is proposed. According to short-term photovoltaic power generation characteristics and seasonal characteristics, the training set of the prediction model is organized. The Kmeans algorithm is used to cluster the training set. Each class of data obtained by clustering is used to train a SVM. The SVM of the same type is used as the forecast sample for power generation prediction. Experiments show that the prediction accuracy of the proposed model is better than that of the traditional BP model and SVM model, so it has a good engineering application value. This work is supported by Basic Public Benefit Research Program of Zhejiang Province (No. LGF18F020017).