A hybrid energy storage consisting of battery and super-capacitor is adopted to smooth wind power fluctuation and improve grid-connected operation capability of wind power. According to the historical data of wind farm output, and considering the grid-connected power volatility standards, a moving average method is used to decompose the original wind power into grid-connected power and energy storage power. Then the energy storage power is decomposed by wavelet packet decomposition into low frequency components and high frequency components, which are respectively allocated to battery and super-capacitor. By analyzing battery life quantitative model, a capacity optimal allocation model using minimum energy storage annual cost as objective is established. Results show that the proposed method is effective and economically practical. It provides a theoretical basis for energy storage planning in wind farm.