When the ultra-short term power error reported by wind farms to the dispatching center is relatively serious, a huge obstacle is brought to large-scale grid connection of wind power and the competitiveness of wind power is seriously affected. A wind storage combined output model is proposed that uses the energy storage system to track the wind power prediction curve. Firstly, the ultra-short term prediction of wind power is carried out through long and short term neural network. Furthermore, the economic impact of the construction cost of the whole life cycle of the energy storage system and the penalty cost of the error of the forecast curve reported by the wind farm after the introduction of the energy storage system is considered through the combined output of wind and energy storage. Finally, the wind storage joint dispatching plan is determined. Based on the measured data of a wind farm in Jilin Province, this paper compares the economic cost and wind power utilization of wind storage farms under different tracking modes. The simulation results show that the power generation cost of the wind storage joint generation model proposed in this paper is 0.2316 yuan/kWh, which is 22.67% lower than that of the non-storage mode. At the same time, the wind power utilization rate is increased by 17.48%. The root mean square error is reduced by 0.07 and the mean absolute error is reduced by 0.08. The results show that the strategy proposed can effectively reduce the wind power grid-connected power error and improve the wind power utilization rate while ensuring the economy of the wind storage power station.