Abstract: manual work，weak generalization ability of static rule-based diagnosis，and low operation and maintenance efficiency caused by data silos throughout the entire life cycle-this study proposes a technical solution for constructing a virtual circuit verification model library and implementing adaptive diagnosis for smart substations. Firstly，a semantic model library of virtual circuits is established based on meta-models and ontology to achieve full-coverage verification at the grammatical，semantic, and logical levels. Secondly, a multi-source data fusion and consistency verification system for the entire life cycle is designed to address the issues of cross-stage data heterogeneity and credibility. Finally，an adaptive fault diagnosis algorithm integrating graph computing and machine learning is developed to overcome the limitation of traditional methods in identifying hidden faults and unknown faults. Experimental results demonstrate that the F1-score of static verification for this solution reaches 100%，which is 66.7% higher than that of traditional tools；the average positioning time for dynamic faults is less than 10 seconds，and its efficiency is 60 times higher than that of traditional methods; the detection rate of unknown faults is 88%，and the accuracy rate is restored to 92% after adaptive learning. This solution can cover the entire life cycle of virtual circuits and provide support for the transformation of operation and maintenance from passive emergency repair to active early warning.