Abstract:Today''s power plants face many challenges, including a wide variety of power equipment, a large number of equipment, many types of faults, complex data coupling relationships, and massive fault information data. Knowledge graph can integrate various information, visualize and present, and support intelligent application, which helps people better acquire, manage and apply knowledge, thus improving efficiency and creating value. The use of knowledge graph to analyze power plant failure data helps to study the power plant equipment failure situation in depth. In the process of constructing knowledge graph, relation extraction is one of the key steps, and its accuracy directly affects the quality of the final knowledge graph construction. In this paper, we propose a relationship extraction tool for the construction of knowledge graph of key power plant equipment failures, which can visually represent the massive and heterogeneous data in the failure information as well as the related fault processing, and at the same time support the user to interactively participate in the process of relationship extraction, and optimize the relationship extraction model through iterative training. In the experimental testing phase, the validation using real power plant equipment fault data demonstrates the effectiveness of the tool in significantly improving the accuracy of relationship extraction. As a result, the quality of the constructed knowledge graph is improved, which provides important support for power plant managers to better operate, maintain and manage power generation equipment, and provides a strong support for controlling power plant related data as well as promoting the construction of a complete power plant.