经典遗传算法利用单一种群对种群个体进行交叉、变异和选择操作，在进化过程中的超级个体易产生过早收敛现象，粗粒度并行遗传算法利用多个子种群进行进化计算，各子群体分别独立进行遗传操作，相互交换最优个体后继续进化。该文证明了该算法的搜索过程是一个有限时齐遍历马尔柯夫链，给出粗粒度并行遗传算法全局最优收敛性证明。对于旅行商问题TSP利用粗粒度并行遗传算法进行了求解，以解决经典遗传算法的收敛到局部最优值问题。仿真结果表明，算法的收敛性能优于经典遗传算法。-Classic genetic algorithm using a single population of individuals in a population cross, mutation and selection operation, the super individuals in the evolutionary process is easy to produce premature convergence phenomenon, coarse-grained parallel genetic algorithm using multiple sub-populations of evolutionary computation, various sub-groups, respectively, independent The genetic manipulation, the exchange of best individual continue to evolve. This paper shows that the search process of the algorithm is a finite homogeneous traverse the Markov chain, given the coarse-grained parallel genetic algorithm global optimal convergence proof. For the traveling salesman problem TSP coarse-grained parallel genetic algorithm to solve to solve the classic genetic algorithm converges to a local optimum value. The simulation results show that the convergence of the algorithm is superior to the classical genetic algorithm.