Wind power has the characteristics of randomness and volatility, so conventional power system analysis, dispatch and control meet new challenges. This paper studies the topic of power system optimal dispatch including wind farms in detail, establishes a dispatch model based on multi-objective particle swarm algorithm. This model considers the running risk caused by cost of power generation, emissions of polluting gases, and short-term fluctuations of wind farm output power, adds reserve positive and negative rotation capacity into constraints, thus reduces the impact of power fluctuations and prediction error on the system. The searching unit ability of multi-target particle swarm algorithm is modified by mixing genetic operators in algorithm design. The capability of global optimization of the model is sharply improved. The simulation result of ten-generator system shows that the proposed method is correct and effective, and it can not only reduce the infeasible solutions in the optimization process, but also resolve the conflict between the various optimization targets, ensuring all the objective function as optimal as possible.