This paper presents a risk-constrained coordinative dispatch method for energy storage systems in the large-scale wind-storage systems based on stochastic programming and the risk decision theory.The purpose is to figure out the best application mode and optimal dispatch method for large-scale energy storage systems in power systems with high penetration of wind power.The stochastic unit commitment model is re-formulated as a bi-level optimization model，with a deterministic unit commitment model and a reserve risk decision model as its external level and internal level，respectively.Conditional value-at-risk（CVaR）is adopted to calculate the risk reserves and coordinate the bi-level models.A novel dispatch method used for coordination between wind power，energy storage and conventional thermal units is also proposed.An IEEE30-bus system with two actual wind farms is used for verification purposes.