Abstract: Under the background of the“dual carbon”strategy and the high proportion of renewable energy grid-connected policy,in order to accurately quantify the consumption cost of new energy such as wind power and the risk loss caused by its randomness to support power dispatching decisions,the CVaR（conditional value-at-risk）is used to model the conditional risk value of abandoned wind and abandoned load caused by the randomness of wind power,and the Copula function is used to calculate the continuous Markov chain wind speed model to predict wind power output,and then a short-term multi-objective optimal scheduling model of hydro-thermal-wind system（MOS-HTW）with minimum of uncertain risk loss,power generation cost and pollution emission is established. Meanwhile,strength Pareto evolutionary algorithm 2（SPEA2） is improved to solve the model efficiently in three aspects:variable external population size,enhanced local search ability and elite population elimination rule based on K-nearest neighbor distance（KND）. The simulation results show that CVaR can well model the uncertain risk of wind power,and find a better Pareto optimal solution set by improving SPEA2.