Abstract: In order to improve the prediction accuracy of ultra-short-term wind power,a hybrid intelligent wind power prediction algorithm including multi-decomposition strategy and error correction is proposed in this paper. The wind power is preprocessed by the multi-decomposition strategy through the improved empirical mode decomposition. and the sequence with large prediction error is decomposed further by wavelet transform,which reduces the non-stationarity of the wind power and greatly reduces the prediction error of the high-frequency sub-sequence. Then,the adaptive neural fuzzy inference system based on particle swarm optimization is used to correct the error,which reduces the maximum error caused by the single prediction algorithm and improves the application range of the algorithm. Finally,the feasibility and accuracy of the proposed algorithm based on the analysis of the measured data of a wind farm in Liaoning are verified.