This paper uses CVaR as the risk measure and applies EVT to model the tails of the return series so as to estimate risk of assets more accurately. This paper also applies pair Copula to capture the inter-dependence structure between assets and constructs pair Copula-GARCH-EVT model; then, we combine it with Monte Carlo Simulation and Mean-CVaR model to optimize portfolio. Finally, an empirical study of four Indexes from Chinese Stock Market is performed and the result suggests that pair Copula can better characterize the inter-dependence structure between assets and the performance of pair Copula-GARCH-EVT-CVaR model is better than that of multivariate t Copula-GARCH-EVT-CVaR model in portfolio optimization.