提出一种求解动态优化问题的多群体单变量边缘分布算法(MUMDA).首先,利用多个概率模型(对应多个群体)将搜索空间分成几个部分,通过对不同区域的搜索或探索将好解进行迁移,扩大搜索空间,增加种群多样性,跟踪最优解的变化;然后,利用对UMDA收敛性的证明分析了所提出算法的有效性;最后,对两个动态优化问题进行仿真计算,并与传统UMDA和基于随机迁移的UMDA(iUMDA)进行了比较,结果表明,MUMDA能快速适应环境的变化,跟踪最优解. A multi-group univariate edge distribution algorithm (MUMDA) is proposed to solve the dynamic optimization problem.Firstly, the search space is divided into several parts by using multiple probability models (corresponding to multiple groups), and the search or exploration of different regions will be good Then we use the proof of the convergence of UMDA to analyze the validity of the proposed algorithm. Finally, we simulate the two dynamic optimization problems, Compared with traditional UMDA and randomized UMDA (iUMDA), the results show that MUMDA can quickly adapt to the changes of the environment and track the optimal solution.