To identify the key nodes of public opinion information dissemination based on the centrality of user groups, and to control the public opinion information dissemination in social networks, a cooperative game assignment method based on the centrality of user groups is proposed. The identified key nodes are marked as information separators based on the Shapley value, and then the SA-SVCGID algorithm is proposed. Finally, the numerical simulation and algorithm comparison are carried out based on the infectious disease model. The results show that the proposed algorithm is better than other public opinion control algorithms in homogeneous, heterogeneous and real social networks. When the number of control nodes increases, the proportion of infected nodes and immune nodes will decrease and increase rapidly, respectively.