Accurate wind speed prediction can promote large-scale wind power integration and ensure the safe and stable operation of a power system. There is a problem in that traditional point prediction methods find it difficult to represent the probability credibility of prediction results. This paper proposes a hybrid interval prediction model based on fuzzy information granulation, an improved long short-term memory network and an autoregressive integrated moving average model. First, the original wind speed data is decomposed by a complete set empirical mode decomposition model of adaptive noise, and the new sequence is reconstructed according to fuzzy entropy. Then the fuzzy information of each sequence is granulated to obtain the maximum, minimum and average values. Finally, the improved long short-term memory network model is used to predict the high-frequency series, and the autoregressive integrated moving average model is used to predict the low-frequency series and the remainder, and then the obtained upper and lower bounds are summed to obtain the final wind speed interval. Example analysis shows that the wind speed prediction interval obtained by this model can accurately cover the measured wind speed and provide more valuable decision-making information for power system dispatching. This work is supported by the National Natural Science Foundation of China (No. 71774054).