In order to make wind farms dispatchable and smooth its output power，an optimal control scheme of battery energy storage system（BESS）based on state forecast is proposed.This method introduces a predictive control model based on the short-term dispatch of wind farms combined with BESS.The model forecasts the ultimate battery state according to the wind power forecast results.In consideration of forecast errors，fluctuation limits and operating constraints，the output power of BESS is adjusted in real time to reduce the peaks of power fluctuation under extreme conditions.In order to improve the forecast accuracy，a neural network combination forecasting model based on the chaos theory and the general linear method is developed.Case study results show that the proposed control scheme enables the grid power of the wind power and energy storage system to track the dispatch instructions and reduces the peaks of power fluctuation while enabling the wind farm dispatchable.