Abstract: Because the randomness and instability of photovoltaic power generation will affect the accuracy of power prediction,this paper proposes a short-term photovoltaic power prediction model based on Pearson correlation coefficient（PCC）,K-means algorithm（Kmeans）,variational mode decomposition（VMD）,sparrow search algorithm（SSA）,and kernel based extreme learning machine（KELM）. Firstly,PCC is used to select the main factors as input;K-means algorithm clusters the historical data into sunny,cloudy and rainy days. Secondly,VMD decomposes the original signal to fully extract the input factor information in the set to improve the data quality. SSA optimizes the kernel function parameters and regularization coefficients of KELM model to solve its sensitive problem of parameter selection. Finally,the final prediction result is obtained by superimposing the prediction values of different series. The simulation results show that the PCC-VMD-SSA-KELM model with similar day clustering has small prediction error.