Most existing decision-making trial and evaluation laboratory (DEMATEL) methods are only suitable for small and simple systems, and do not consider whether expert consensus has been reached. In this study, a group hierarchical DEMATEL method is designed for reaching consensus in complex systems. The hierarchical DEMATEL method is extended from individual decisions to group decisions and its decision-making process is analyzed. Then, the limitations of expert knowledge and experience are considered, and the basic probability distribution (BPA) function is employed to obtain judgment opinions on each level of the system. The initial direct-relation (IDR) matrix construction method is designed and a consensus-reaching method is constructed from three levels of the hierarchical system. For reaching consensus, the process of feedback adjustment is established with the help of an automatic opinion correction mechanism. The steps of the group hierarchical DEMATEL method for reaching consensus are summarized by following the principle of the classical DEMATEL method. Finally, an illustrative example is provided to demonstrate the superiority of the proposed method and algorithms, indicating that it can improve decision-making efficiency for complex systems.