In this study, we propose a short-term wind power prediction method combining variational mode decomposition (VMD) and hybrid kernel extreme learning machine (HKELM) in order to improve the prediction accuracy of wind power. Sparrow search algorithm (SSA) is used to optimize the key parameters in VMD and HKELM such that it addresses the difficulty in tuning parameters and achieves adaptive selection of the parameters of the two algorithms. Firstly, the SSA is designed to optimize the fitness function of the VMD based on three signal decomposition indexes. The optimized VMD is then used to decompose the wind power into a set of smooth subcomponents. Then, a hybrid kernel function is formed by combining the radial basis kernel function and polynomial kernel function. An HKELM prediction model with learning and generalization capabilities is built for each component combined with meteorological features, and SSA is used to optimize the model parameters to fully utilize the model performance. Finally, the final prediction results are obtained by superimposing the predicted values of each component. Our simulation experiments used actual data from a wind farm in Inner Mongolia, China, and the results show that the proposed method has higher prediction accuracy than other prediction methods.