Abstract: The low-voltage distribution network in mountainous areas of northern Guangdong is characterized by scattered distribution of stations, long power supply radii, distributed photovoltaic penetration rate exceeding 70%, and drastic tidal fluctuation of source and load. The traditional operation and maintenance mode suffers from distorted topological ledgers, insufficient on-site perception, slow fault disposal, extensive line loss management and heavy workload for grassroots teams. A low-voltage lens system is developed by Shaoguan Power Supply Bureau, constructing a four-layer closed-loop digital twin architecture and integrating eight self-developed algorithm clusters. Based on time-series measurement data of existing smart electricity meters, precise phase sequence identification is realized through a dual-engine topology recognition method and multi-dimensional similarity clustering. A hierarchical phase-separated current accumulation model combined with a dynamic current-carrying capacity threshold judgment model is established to achieve graded early warning of line overload. Verified by field engineering tests, the proposed integrated technology requires no additional on-site monitoring hardware and supports full-process digital management including topology correction, phase sequence recognition, fault location, overload diagnosis and hierarchical line loss tracing. The topology recognition accuracy reaches no less than 96%, and the phase sequence discrimination accuracy under complex mountainous working conditions is over 99%. The average power restoration time after faults is reduced by 59.7%, providing a low-cost and replicable digital operation and maintenance paradigm for mountainous rural distribution networks in China.