In the planetary gear transmission system, each gear-teeth cannot be arbitrary positive integers, which are subjected to drive ratio, matching and coaxial requirements. The module of gear only can take certain discrete values for the limit of national standards. It is difficult to obtain the valid solution with the classical restrictions random (CRR) optimization method based on mathematics plan theory. A discrete variables intelligent optimization method with simulation annealing algorithm was presented. The mathematical model of optimization design for two-stage 2Z-X epicyclic gearing was established and an example was presented subsequently. Compared with the CRR method, this method can obtain the global optimal solution more easily and quickly and can be used in engineering design without any post-treatment.