Genetic algorithm (GA), as a kind of important intelligence computing tool, is a wide research content in the academic circle and the application domain now. In this paper, for the mutation operation of GA, by combining with the essential feature, we establish a genetic algorithm based on schema mutation (denoted by SM-GA, for short). Further, we discuss the global convergence of CM-GA by using the Markov chain theory, and analyze the performance of SM-GA through an example. All the results indicate that, SM-GA is higher than the ordinary binary code genetic algorithm (denoted by B2GA, for short) in convergence precision. There was no significant difference between SM-GA and B2GA in convergence time. SM-GA overcomes the problem that B2GA can not converge strongly to some extent.