Abstract: Taking a foreign object short-circuit fault analysis report on the 1000 kV Huaiwu II line as an example, this paper proposes a method for evaluating the probability of electrical tower component faults based on the construction of a domain knowledge graph. This method integrates large-scale language models to amalgamate expert experiences, utilizes the TextRank algorithm to process fault-related data, and establishes a fault inspection knowledge graph based on the Neo4j graph database. Building upon this, it models the distribution of faults across multiple components in multiple towers, facilitating the rapid and accurate identification and localization of faulty components across multi-level towers. This method provides data support for efficient and rational route planning for drone inspections and underscores its considerable practical value.