Abstract: An amount of unstructured and semi-structured data including specification and inventory, they are generated in the design and construction of power grid infrastructure engineering can be used as an important source of basic data from power grid equipment and assets, their data values have not been fully explored. On the basis of data preprocessing, this paper applies the dependency model and Bidirectional long short-term memory to natural language processing, and constructs the infrastructure engineering knowledge map based on Neo4j. The map transforms the natural languages in different files into nodes and relationships in the knowledge base and realizes intelligent retrieval function. Finally, an example is given to verify that the proposed knowledge map can store valuable information in engineering data hierarchically and provide a new structured data source for transportation, inspection and scheduling departments.