In the operation and planning of power systems，it is important to model the dependence of random wind energy in order to get accurate results.Different from the widely used linear correlation analysis，the Copula function is introduced for modeling the dependence structure of wind speed and wind power among different wind farms，and the joint distribution functions of wind speed and wind power are established respectively.By studying the actual wind speed data from four wind farms in Jiuquan of Gansu province，it is found that wind speed correlation between each two wind farms shows unsymmetrical thick tail characteristic.As this phenomenon cannot be described by a single kind of Copula function，a mixed Copula function，i.e.convex combination of some single Copula functions，is constructed based on selecting the candidate single Copula functions according to goodness of fit test.By studying the actual wind power data from two wind farms in Fujian province，it is found that Gumbel-Copula function is suitable for modeling the dependence structure between output power of two wind farms.A four-dimensional dependence structure model for wind speed is finally presented.