This paper focuses on addressing mobile communication fault repair within an electric-power Communication network in State Grid Zigong; that is, a long-standing problem of delay in transferring faults due to the traditional paper-tickets system. A project-based repair system, which integrates topological Reasoning and online Workflow Control technologies to build a mobile application of the iStateGrid platform. A Mobile Optical Cable Graph Neural Network (MOC-GNN) was proposed to represent cable segments, optical Distribution frames, Devices, Service Paths, Repair Tickets, Roles, Telemetry windows as type Nodes, Model serviceTraversal, physical adjacency, Ticket Transfer, Authorisation, Alarm Propagation, Image Evidence, Historical Co-failure as typed Edges. Using a project-aligned digital-twin dataset with 18,240 fault events, 61,800 normal windows, 124 access rings, and 4,812 service paths, the model achieves 93.8% top-1 localization accuracy, 98.4% top-3 accuracy, 92.4% macro-F1, 78 m median error, and 226 m 95th-percentile error, outperforming rule-based, CNN, XGBoost, GAT, R-GCN, and HGT baselines. Approximately a 64% reduction in the average ticket-leave waiting time after adopting the new work-flow plan. The ablation results show that service-path relationships primarily impact the accuracy of localisation; The other features inRBAC, OCR, NLG, signatureandRedis-state contribute to improved compliance,ticket confirmation, report receipt and mobility-related aspects. The results support deployment of the micro-application as a topology-intelligent repair system rather than a simple electronic ticket tool.
