The large scale of power grid users and the multidimensional superposition of marketing business make power marketing data reports interdependent and have a large quantity and scale. At the same time, there are complex cascading relationships, which leads to extremely high maintenance and analysis complexity of marketing data and strict requirements for data quality. In response to this issue, the article proposes an intelligent analysis method for power marketing report metadata based on graph link model. Using graph theory models to represent the correlation between metadata in power marketing reports and construct a connection topology between multiple sources of metadata. Sort out the association rules between marketing business systems and develop mapping standards between data tables based on data dictionaries, expert knowledge, and other factors. Using nodes and edges to represent metadata entities and the connection relationships between entities, a metadata association topology is formed and a topology graph is constructed based on it. After reduction and standardization, an association matrix is formed. Using complex network community search algorithms to perform topology search, a metadata graph link model suitable for fast retrieval and analysis applications is constructed to achieve adaptive growth of metadata topology tree. The objective function for measuring the complexity of the design diagram structure is to describe the structural complexity of the metadata topology graph using the evaluation value of the average node connectivity, in order to find the optimal complexity metadata connectivity graph. Finally, the applicability of the model was validated through scenarios such as metadata retrieval, batch updating, and correction of marketing reports. Compared with traditional relational models for data retrieval, updating, and correction, the metadata representation method based on graph link model reconstructs the topology of the connection relationships between metadata, and uses directional search instead of recursive search to improve the efficiency of metadata analysis, providing a method reference for power sales report data analysis.