Energy digital objects come from a wide range of heterogeneous and complex sources, and the differences in data formats and cross correlations between data sources make it difficult to fuse the data. However, cross modal data fusion can improve the management efficiency of energy objects, and optimizing resource allocation can help improve the level of integration and development of energy objects. The key to solving this problem is to use a knowledge graph to represent the correlation between energy digital objects and find a connection medium for cross modal data to achieve the construction of a knowledge graph. Using semantic elements such as entities, attributes, and relationships to describe multi-source heterogeneous data objects, integrating dispersed data into a unified semantic knowledge model to reveal the inherent connections and potential patterns between data. Consider the semantic similarity of data entity connection relationships and integrate object semantic connection topologies across data sources. Convert the connection relationship into object vector groups based on semantic embedding technology and calculate their similarity. Utilize the structural information in the knowledge graph to assist in obtaining object spatial features, and calculate the similarity of connection relationships based on the context and path information between entities. Capture semantic information in graph structures using graph convolutional networks to improve the accuracy of relationship matching.