Abstract: To address the challenges posed by the large volume of unstructured data in power supply expansion applications, such as low efficiency, high error rates, and inconsistent standards in manual verification, an intelligent verification system was designed and implemented based on practical business requirements. Integrating OCR, NLP, knowledge graph, and rule engine technologies, the system establishes an automated end-to-end workflow encompassing data collection, parsing and extraction, intelligent verification, and result feedback. This enables precise parsing and compliance verification for 18 types of documents. Practical application shows that the system achieves a field extraction accuracy of 96.8%, improves verification efficiency by more than six times compared to manual methods, and reduces the business error rate from 10.3% to 2.1%. This saves grassroots staff 2～4 hours of verification time daily, effectively mitigating compliance risks and enhancing the "Getting Electricity" service level.