The distribution network structure is complex, the equipment is numerous, and the frequency of pattern and mode changes is high. These characteristics lead to certain difficulties in power distribution automation operation and maintenance graph management. This paper adopts the mathematical statistics method of machine learning to analyze the multi-version hierarchical subscription mechanism of the distribution network graph. We conduct a breadth search on the distribution network graph to realize the automatic topology of the network. This paper implements a dynamic display system of distribution network monitoring information. The research results show that the graph-digital-analog integrated system has practical significance for data integration, application integration, and interoperability between systems.