In recent years, with the rapid development of smart grid and energy Internet, it is necessary to regulate the power supply in order to ensure the safe and economic operation of the distribution network. Cluster regulation is an important measure to solve the problem of “dimensional disaster” brought by the large number and wide distribution of distributed power supply to regulation and control, and to guarantee the orderly scale and safe and economic access of distributed power supply, therefore, this study constructs an optimal scheduling model of distribution network based on the division of clusters, and constructs a hierarchical control system that uniformly participates in the regulation and control of the distribution network in the form of clusters. The study also constructs a hierarchical control architecture in the form of cluster unified participation in distribution network regulation and control. On this basis, based on the physical reality of integrated grid operation, the economic operation of each link of generation, transmission, distribution and usage is considered in an integrated manner, so as to realize the optimal allocation of global resources of the large grid to establish a cluster-based hierarchical control system. In the optimization of distribution network cluster division, this paper’s method has better cluster division effect compared with both SLM algorithm and traditional k-means algorithm, and the average network loss before and after the optimization of global regulation effect is reduced by 2.39Kw/h, which verifies the application effect of this paper’s method. In this paper, the reactive power adjustment of the layered regulation method for voltage control is 3.33kvar for the system with voltage crossing lower limit scenario and −12.80kvar and −7.73kvar for voltage crossing upper limit scenario respectively. Compared with the traditional local Q(U) control method, the voltage can be adjusted to the qualified range with a faster speed, giving full play to the reactive power regulation capability of the distributed power supply, and realizing the layered control of the distribution network voltage.