Many parameters and complex boundaries are involved in the spatial arrangement of an underground powerhouse in a hydropower station, necessitating referencing many relevant cases and specifications. However, in practical applications, retrieving such cases or specifications is difficult, and there is a lack of knowledge regarding cascading logic among design parameters. To address this issue, this study proposes a novel methodology for constructing a targeted knowledge graph, in this case, a knowledge graph for building information modeling (BIM) for underground powerhouses in hydropower plants. Subsequently, based on this knowledge graph, this study develops a question-and-answer (Q&A) system to facilitate subsequent applications. First, the ontology skeleton of the spatial arrangement design of an underground powerhouse in a hydropower station, that represents the knowledge organization structure of the knowledge graph, is constructed by carefully analyzing the requirements for intelligent modeling of underground powerhouses. A large volume of unstructured data is identified based on the optical character recognition (OCR) technology; the collected data are divided into words to extract correlation knowledge using THU Lexical Analyzer for Chinese (THULAC). Subsequently, the knowledge triad of the spatial arrangement of the underground powerhouse is extracted based on ChatGPT and stored in Neo4j, a knowledge base, to build a knowledge graph. Finally, the knowledge graph is employed to realize the query of knowledge and parameter recommendation to assist the digital intelligent design of the spatial arrangement of an underground powerhouse in a pumped storage hydropower station.