Digital twin-driven fault diagnosis of power substations by multi-modal fusion learning

Nature Communications, Published online: 20 May 2026; doi:10.1038/s41467-026-73483-5 The study builds a digital twin of a power substation and combines topology, alarms, waveforms, and measurements using attention-based graph models to diagnose fault location, fault type, and protection failures with robust performance under noise and missing data.