Abstract: A wind speed prediction model is proposed in this paper to deal with nonlinearity of wind speed series.The proposed model is based on improved complementary ensemble empirical mode decomposition（CEEMD） and extreme learning machine（ELM）.Firstly,in the proposed model,the wind speed series are decomposed using improved CEEMD,and some new time series are obtained utilizing phase space reconstruction,lowering the inconsistency of wind speed series.After that,the processed wind speed series are predicted using the optimal input parameters,found using improved cuckoo search（CS） algorithm,of the ELM model.At the end of this paper,simulations are conducted and the prediction model is proved to be reasonable by comparing the relative error of different models before and after the prediction process.