There is strong randomness in wind power, and its output is not controllable. Large scale wind power grid-connection to dispatch system brings new uncertainty. This paper constructs the wind-thermal random optimal model based on improved PSO algorithm and achieves the coordination of power cost and pollution emissions of thermal power units. Aimed at drawback of basic PSO algorithm, the paper uses chaos optimization search process to speed up the search and convergence speed. This improved algorithm could improve the coordination between global and local searching ability and apt to search out global optimum quickly. Finally, multi-objective optimal scheduling is illustrated by using the thermal-10 node. The results show that the improved PSO arithmetic solution set uniform distribution, predict accuracy and qualified rate is out of the solution of the highest degree of coordination, get each time specific unit output of the plan, according to the calculation results make corresponding power generation plan, to access a fairly large wind power dispatch system has the actual reference value, so the proposed method is feasible and effective.