Fault diagnosis and early warning of relay protection secondary circuits in smart substations based on multi-source data fusion

Abstract

This paper presents a multi-source data-fusion framework for fault diagnosis and early warning of relay protection secondary circuits in smart substations. The proposed method integrates seven data sources—SV statistics, GOOSE statistics, IED self-check records, MMS alarms, physical-link measurements, time-synchronization data, and SCD configuration records—through credibility-weighted Dempster–Shafer evidence fusion to manage uncertain and conflicting information. A graph-based topology model localizes faulty elements via suspicion propagation, while an adaptive health index tracks degradation trends for graded early warning. Experimental validation on a 110 kV IEC 61850 hardware-in-the-loop platform demonstrated 97.1% diagnostic accuracy, 1.8% false-alarm rate, and 95.6% localization accuracy, outperforming single-source analysis by 25.3 percentage points. The framework provided attention-level warnings 53.1 hours before functional failure, enabling preventive maintenance. Results confirm that conflict-aware fusion and topology-based localization substantially improve diagnostic reliability over existing methods.