Abstract: Wind speed is nonlinear and non-stationary. In order to improve the prediction accuracy of ultra short term wind speed,an adaptive hybrid prediction model with error correction is proposed. A fully ensemble empirical modal decomposition model with adaptive noise and an improved variational modal decomposition model are used to decompose the sample data series and the future prediction error series respectively,while the fuzzy entropy of each sub-series is calculated to determine the complexity of the series. Further,the improved long and short term network is applied to predict the higher complexity series and the autoregressive integrated moving average model to predict the lower complexity series. Finally,the prediction results and the wind speed error prediction values are summed to obtain the error-corrected ultra-short-term wind speed prediction values. The results show that the correction of forecast errors and double decomposition can effectively improve the performance of point prediction,and the proposed model has excellent prediction accuracy in multiple scenarios compared with the benchmark model.