A new short-term wind power forecast combination algorithm, which is based on particle swarm optimization and support vector machine (SVM), combined with error correction algorithm, has been proposed. Firstly, an analysis and cleaning have been made of the original data; then an optimization can be achieved of the parameters of support vector machine by particle swarm optimization algorithm, followed by a prediction of the wind power. The empirical modal algorithm is used to filter the primary prediction to achieve the effect of noise reduction, thus working out the primary prediction error. Finally, the error correction algorithm is used to correct the one-time prediction error, thus obtaining the final prediction value. The simulation and test results show that the combined algorithm can improve the prediction accuracy better than the traditional single algorithm.