Abstract: To address the complex demands of smart operation and maintenance as well as station-city integration in railway station districts, this study first innovatively proposes a next-generation smart operation and maintenance system that integrates the domestic DeepSeek large model with digital twin architecture and core technologies. The system architecture comprises three layers: a digital twin foundation layer, a multi-stakeholder business layer, and a large model objective layer, which integrates real-time video feeds and BIM models to establish a dynamic digital foundation. Three key innovations include a self-calibrating travel guidance system based on deep spatiotemporal networks, a space optimization model incorporating multi-objective reinforcement learning, and a swarm intelligence-enabled operation and maintenance system. Applied to the station-city integration project at Wuxi Railway Station, the system achieves dynamic congestion prediction and emergency evacuation simulations, quantitatively coordinates objectives including traffic efficiency, commercial value, and cultural promotion, and enables collaborative equipment control and health management through distributed intelligent agent clusters. This approach effectively facilitates the transformation of transportation hubs into multifunctional urban complexes, providing an intelligent solution for station-city integrated development.