For the purpose to reduce the cost on energy storage for smoothing wind power fluctuation, this paper proposes a new method which selects a hybrid energy storage with certain confidence level to smooth wind power fluctuations. Firstly, comparing the smoothing effect of mixed energy storage for the fluctuation of typical daily wind power output with different time constants, and the reference power of energy storage is obtained. Then, the empirical mode decomposition (EMD) is utilized to decompose the energy storage reference power into a series of intrinsic mode function (IMF). According to the least alias in the instantaneous frequency-time curve, the energy storage reference power is divided into a power type high-frequency energy storage configuration and an energy-type low-frequency energy storage configuration. A reasonable energy storage type is then chosen based on energy storage cost on this basis. Finally, the actual operation condition of a wind farm in Hunan is simulated. It shown that the capacity and power required for energy storage to smooth wind power fluctuations is significantly reduced after considering the confidence level.