Integrated Design of 12  <scp>kV</scp> Intelligent Ring Main Unit With Multi‐Model Fusion and Fault Diagnosis Method

ABSTRACT The existing fault diagnosis methods for 12 kV ring main units face several limitations, including slow perception speed, low cross‐modal data fusion accuracy, and ineffective avoidance of signal interference. The accuracy of fault diagnosis is low, and the isolation speed is slow. This article proposes an integrated design method for the 12 kV intelligent ring main unit and constructs a fault diagnosis and isolation model. Firstly, an integrated hardware platform is constructed that significantly improves the perception speed and fusion potential. Then, an innovative fusion diagnostic model, named GAT‐Transformer‐DRSN, is proposed. The proposed model consists of three core components: a GAT‐based topology fault locator, which analyzes fault propagation paths; a spatiotemporal Transformer‐based cross‐modal evolution analyzer, which improves fusion accuracy; and a DRSN‐based anti‐interference feature extractor, which suppresses the impact of signal interference. Finally, a self‐optimizing closed loop is formed based on weighted fusion output diagnosis results and isolation instructions. The experimental validation utilizes fault data from the American Electric Power Research Institute (EPRI) and actual operational data from a provincial distribution network in China. The proposed GAT‐Transformer‐DRSN achieved the highest performance across all datasets, with fault diagnosis metrics including a Macro‐F1 score of 0.9531, recognition accuracies of 96.28% and 96.45%, and false alarm rates of 0.75% and 0.72%, outperforming all other comparative methods.