As a new kind of intelligence optimization method,genetic algorithm,with the features of simple structure and strong adaptability,achieves great success in many real fields,but still there are some shortcomings such as greater computation complexity and more chance of being trapped into local states.This paper analyzes the deficiency of the existing genetic operation and the essential characteristics of creature evolution to improve evolution efficiency,and proposes a composite genetic algorithm based on multi-reserving strategy(MRS-CGA for short).Moreover,it analyzes the performances of MRS-CGA by the theory of Markov chains and simulation technology.All the results indicate that,MRS-CGA is essentially the extension of ordinary GA,and obviously better than ordinary GA in computation efficiency and convergence performance.