Digital-twin-based visual monitoring and automated maintenance decision support for substation equipment health

Abstract

The digitalization of modern substations has transformed automation-control infrastructures into cyber-physical environments that continuously generate multi-rate and heterogeneous monitoring data. Conventional platforms, however, usually present isolated alarms and cannot maintain an interpretable link among field assets, virtual models, degradation evidence, and maintenance actions. This paper develops a digital-twin-based visual monitoring and automated maintenance decision-support system for substation equipment. The system integrates equipment geometry and topology, synchronized sensor streams, robust residual modeling, an explicit health-index function, risk scoring, and three-dimensional operation-and-maintenance visualization. All model functions, feature weights, correction gains, thresholds, and calibration rules are specified. A reproducible 110-kV prototype case contains 8,640 samples per continuous channel and 240 labeled event windows. On a held-out set of 48 windows, the proposed method achieved 93.8% accuracy, 95.8% recall, and a 93.9% F1 score; its median event-detection delay was 3.5 min, compared with 12.0 min for threshold-only monitoring. The results demonstrate improved multi-source interpretation and maintenance traceability, while field validation over multiple seasons remains necessary. By converting synchronized health evidence into prioritized maintenance recommendations and traceable alarm explanations, the proposed system links diagnostic analytics with decision-support and supervisory automation-control workflows, supporting earlier intervention and more consistent operator responses without bypassing certified protection functions.