In recent years, the implementation of wind power integration policy and the increase of wind power penetration have played a positive role in energy conservation and emission reduction. However, due to the obvious uncertainty and correlation of wind power output, the economic dispatch of power system is also facing great challenges. In this paper, a novel pair copula method is applied to formulate the dependence of multiple wind farms. A large number of stochastic scenarios, in which the complicated dependence of multiple wind farms are considered, are generated to represent the uncertainties of wind power based on quasi-Monte Carlo (QMC) simulations. To solving the stochastic economic dispatch problem with wind power output, a risk constrained mean-variance (MV) model considering risk constraints is constructed. The MV model considers economic cost and economic risk under the uncertainties of wind power simultaneously, among which economic risk is calculated by means of least variance of fuel cost. In order to adapt to the actual scheduling situation, the probability density function (PDF) obtained for fuel cost is established, and a predefined level of confidence interval is proposed to improve the MV model. For solving the multi-objective stochastic economic dispatch problem, group search optimizer with multiple producers (GSOMP) is employed in this paper. The effectiveness of the proposed pair copula method and the improved MV model are validated via numerical simulations with a modified IEEE 30-bus system.