Retrieval of construction standards for pumped storage power station engineering based on knowledge graph enhanced retrieval generation framework

In response to the problems of term heterogeneity, semantic dispersion, and complex structure in the retrieval of construction standards for pumped storage power stations, this paper constructs a knowledge graph-based Retrieval Augmented Generation (RAG) technology framework. This framework integrates entity relationship extraction from engineering corpus, process knowledge modeling, and cross domain graph construction to achieve semantic organization and precise recall of standard content. Introduce vector indexing and semantic extension mechanism in the retrieval stage, and use GPT model for structured content reorganization in the generation stage. The experiment conducted multidimensional performance evaluation on industry standard datasets, and the results showed that this method outperforms traditional methods in terms of recall rate, structural retention, and user satisfaction, demonstrating strong practicality and engineering promotion value.