Automatic Matching of Virtual Circuits Based on Multi-Dimensional Typical Template Library and Case-Based Reasoning

To address the issue of ROC value degradation caused by complex semantic descriptions of virtual terminals and heterogeneous device models in the configuration of virtual circuits in smart substations, an automatic matching method for virtual circuits integrating a multi-dimensional typical template library and case-based reasoning is proposed. By parsing historical SCD files, the association relationships between virtual terminals of different IEDs are extracted, and a structured typical template library is constructed based on multi-dimensional features such as voltage level, device type, and bay unit. A Large Language Model (LLM) is introduced for deep semantic understanding and vector representation of virtual terminal texts, enabling preliminary screening based on semantic similarity. On this basis, Case-Based Reasoning (CBR) technology is integrated to retrieve similar historical cases for logical mapping and matching decision optimization. Furthermore, a multi-modal semantic fusion strategy is introduced, incorporating graphical topological features into the similarity calculation to further enhance matching accuracy. Experimental results demonstrate that the proposed method achieves a matching accuracy of 0.95 and an Area Under the ROC Curve (AUC) of 0.97, effectively enabling automatic matching of virtual circuits in smart substations and providing technical support for the intelligent configuration and reliable operation of secondary systems.