In order to predict wind speed accurately, this paper combines CART, CEEMDAN, echo state network and error correction strategy to propose a short-term wind speed prediction method with a multi-processing strategy. CART is applied to reconstruct the original dataset to get the training data. CEEMDAN is employed to extract the feature information. Then, ESN is used to model the wind speed based on the features. Finally, the model is modified by an error correction strategy. The proposed method can predict wind speed accurately, guide the production of a wind farm and improve the automation level of production. This work is supported by National Key Research and Development Program of China (No. 2018YFB1500803).