Background:: With the depletion of fossil energy and the increasingly serious environmental pollution, the task of developing renewable energy is imminent. As a green and pollutionfree renewable energy, the penetration of wind energy in the power grid continues to rise. Objective:: In order to reduce the volatility and randomness of wind power series and increase the accuracy of wind power prediction, a wind power combination model based on the ICEEMDAN (improved adaptive noise full set empirical mode decomposition) method is proposed. objective: In order to reduce the volatility and randomness of wind power series and increase the accuracy of wind power prediction, a wind power combination model based on ICEEMDAN (improved adaptive noise full set empirical mode decomposition) method is proposed Methodology:: First, the complex original wind power data have been decomposed into several relatively simple subsequences using the ICEEMDAN method. Aiming at the different lengths of coarse grain time series and data loss in traditional multi-scale entropy, a fine composite multiscale dispersion entropy is proposed to calculate the entropy value of each decomposition component, and divide the high- and low-frequency modal components to predict the modal components of different frequencies; secondly, differential moving autoregressive model (ARIMA) and shortterm memory neural network (LSTM) are used to establish the prediction models of high- and low-frequency components, respectively. Results:: Finally, the prediction results of each component have been superimposed and reconstructed to obtain the final prediction results. The effectiveness of the combined model is verified by the actual operation data of a European wind farm. result: The effectiveness of the combined model is verified by the actual operation data of a European wind farm. The results show that compared with the other four single and combined forecasting models, the combined model in this paper has higher forecasting accuracy. Conclusion:: As the effectiveness of the combined model is verified by the actual operation data of a European wind farm, the results have shown that compared to the other four single and combined forecasting models, the combined model in this paper has higher forecasting accuracy. Therefore, the model proposed in this article can be used for predicting wind power with significant fluctuations, which will help to provide support for optimized scheduling and energy storage configuration of wind farms, thereby reducing costs and increasing income for the power grid and wind farms. conclusion: ICEEMDAN decomposition algorithm can effectively reduce the influence of white noise on mode decomposition, and the prediction accuracy of wind power can be better improved by putting different frequency power components into different prediction models