A fault location method for secondary virtual circuits in smart substations based on the integration of GAE and GCN

The secondary virtual circuits in smart substations enable digital and networked information exchange between devices. However, their extensive logical connections, complex structure, and implicit interdependencies pose significant challenges for fault localization. To address this issue, this paper proposes a fault localization method for secondary virtual circuits that integrates a Graph Autoencoder (GAE) and a Graph Convolutional Network (GCN). First, by analyzing fault events in secondary equipment, multi-dimensional key features-including device self-test status, alarm information, communication message status, and electrical quantity sampling values-are extracted to construct a unified feature representation. Using the logical topology of virtual circuits as the initial graph structure, the GAE is employed to model complex relationships between fault events and automatically learn latent node associations, thereby constructing an enhanced graph that better reflects fault propagation characteristics and effectively captures implicit topological dependencies of faults. Subsequently, a fault localization model incorporating an attention mechanism is developed based on the GCN, which aggregates neighbor node information through multi-layer convolution and adaptively highlights critical nodes and features to suppress redundant interference. Meanwhile, the Slime Mould Algorithm (SMA) is introduced to optimize the model’s hyperparameters, addressing the limitation of traditional fixed hyperparameters in adapting to complex fault scenarios. Finally, case study results demonstrate that the proposed method can effectively improve the fault localization accuracy of secondary virtual circuits in smart substations while maintaining good localization performance under noise interference and reduced-training-sample conditions.