Rock creep model generally contains several parameters. To obtain the global optimal solution of the parameters, Microevolution Algorithm (MA) was employed for the inversion of non-stationary parameters. In this paper, parameters of the creep model are directly inversed with the least square error between measured creep values and calculated creep values as the optimization criterion function. The computation results demonstrate that microevolution algorithm can maximize the use of all test data and avoid the difficulty of selecting initial parameters in traditional optimization algorithm. Moreover, microevolution algorithm is simple and effective, and offers higher accuracy than Chaos Particle Swarm Optimization (CPSO). In this sense, it can be applied to the inversion of parameters in other creep models and has high application value for engineering.