In this paper，combined with extreme value theory ( EVT) ，four categories of time-varying Copula models were constructed，the dynamic extreme value contingency coefficients between stock indexes were fitted and risk measurement was conducted on each portfolio． By means of backtesting analysis method，a comparative study on the precision of the risk measurement of different time-varying Copula-EVT-ES models was made．The empirical results show that in the condition of extreme market volatility，the time-varying Copula-ES model combined with EVT can effectively measure the extreme value risk at the tails of portfolios，and the risk models can get more precise measurements for short positions than long positions． The time-varying t Copula-EVT-ES model which is conducive to depicting the dependency of the fat tail extreme values of variables has a better performance in portfolio risk measurement; in high risks，time-varying SJC Copula-EVT-ES model is also worthy of special attention，especially for dual portfolios．