Abstract: In response to the pain point of traditional substation operation ticket preparation and verification heavily relying on manual experience, an intelligent ticket generation and verification algorithm that integrates semantic understanding and knowledge graph is proposed. In the ticket generation stage, BERT and attention mechanisms are used to parse the semantics of scheduling commands, generate structured operation tasks, and embed device topology information using graph convolutional networks to achieve intelligent retrieval and parameter replacement of classic tickets. At the same time, batch ticket generation and graphic assisted supplementation are supported. In the verification phase, a multi-layer assembly line that combines rigidity and flexibility is constructed: the rigid layer performs "five defenses" and secondary protection logic verification based on the topological adjacency matrix and equipment state equation, while the flexible layer relies on the Guangming Power Big Model to perform a five dimensional rating of the ticket"s compliance. The experimental results show that the complete model has an accuracy rate of 95.9% in ticket generation, a comprehensive verification recall rate of 93.4%, and a compressed invoicing and review time of 1.9 minutes and 4.2 minutes, respectively, significantly improving the intelligence level and safety control capability of substation operation ticket generation.