The operation and dispatch of smart grid become more difficult with the access of larger scale wind power.Accordingly

we established a hierarchical dispatch model about multi-type flexible loads to solve the problem

in which flexible loads were divided into industrial high energy load

business aggregate load and smart house load

then the three kinds were classified and dispatched according to the capacity. Aiming at three objectives

such as the maximum of absorptive wind generation and flexible load utilization

and minimum of difference of peak-valley

we established a weighted minimum module ideal point function based on the three objectives

and used the modified particle swarm optimization algorithm which contained nonlinear decreasing inertia weight to calculate the function.Finally

we analyzed a distributed network. Simulation results show that grading optimization of multi-type flexible load will increase the utilization of wind generation and flexible load

and reduce difference of peak-valley effectively. Moreover

reasonableness of the proposed model is verified.