Adaptive Optics(AO) based on Stochastic Parallel Gradient Descent(SPGD) corrects the wave-front aberration by directly optimizing the performance evaluation function of the system. It may realize real-time correction and has great application potential. The convergence rate of SPGD algorithm is closely related with the number of the control units, the selection of the performance evaluation function, the value of gain coefficient and disturbance amplitude. Through the simulation method, the influence of the number of control units on the convergence rate was studied when SPGD AO system was used for static wave-front aberration correction. A conclusion was reached that there was a linear relationship between the number of control units and the required number of iterative for convergence. SPGD AO system used for dynamic wave-front aberration correction under different iterative rate and wind speed was simulated, and suggestions for the selection of number of control units were provided.