In view of the traditional clustering algorithm’s problem that is hard to select the initial clustering centers

more sensitive to noise

slow convergence rate and easy to fall into the local optimum

a hierarchical clustering algorithm based on density peaks for weather category is proposed in this paper. Firstly

the parameters of density peaks on meteorological data are determined

and meteorological data are divided into different categories using hierarchical clustering algorithm. Then

the support vector machine（SVM） is used to identify the unknown weather types. Finally radial basis function（RBF） neural network is adopted to establish the photovoltaic short-term power prediction model. The simulation results showed that the proposed method can improve the classification accuracy of weather types effectively

accelerate the speed of searching optimums

and improve the robustness of the outliers

which proved that the accuracy is still higher under circumstance of small samples

and still be able to achieve the prediction of power value in weather have great fluctuations.