Nuclear power plants generate a vast amount of heterogeneous documents throughout their construction and operation phases. Traditional manual management methods often fail to efficiently retrieve embedded knowledge, resulting in information silos and low management efficiency. To address this issue, this paper explores the use of artificial intelligence (AI) techniques for digital processing of nuclear power plant documents and subsequently constructing a knowledge graph specific to the nuclear power domain to facilitate knowledge management and intelligent applications. Firstly, AI methodologies for document classification, optical character recognition (OCR), and natural language processing (NLP) are introduced, establishing a structured document-processing workflow. Subsequently, techniques for constructing knowledge graphs are presented, including domain-specific knowledge extraction, ontology modeling, and storage, integrating extensive nuclear-related knowledge into interconnected networks. Through case studies, the paper demonstrates typical application scenarios of the constructed knowledge graphs in nuclear power plant operation and maintenance, such as intelligent question-answering retrieval, fault diagnosis assistance, and compliance verification of safety regulations. Results indicate that AI-based document digitalization significantly enhances document management efficiency and reduces human error. Integrating a knowledge graph effectively consolidates heterogeneous nuclear knowledge, strengthening retrieval and reasoning capabilities. In practical applications, the proposed methods enhance knowledge utilization efficiency and intelligent decision-making capabilities in nuclear enterprises. Finally, the paper summarizes the research outcomes, identifies existing limitations, and outlines future development directions, providing valuable insights for the digital transformation and intelligent operation and maintenance within the nuclear industry.