Transformer inter-turn short-circuit fault identification method based on multi-parameter feature fusion and adaptive coupled Swin transformer

Abstract

To address the limitations of conventional inter-turn short-circuit fault identification in transformers, which relies on a single monitoring parameter and fails to fully exploit the complementary characteristics of heterogeneous multi-source information, a multi-parameter fault identification method based on information entropy-adaptive fusion is proposed. An electromagnetic–thermal–circuit multiphysics coupling model of an oil-immersed transformer is established to reveal the response mechanisms of the high-voltage winding current, winding power loss, and tank temperature under different fault severities, and a multi-parameter dataset covering various inter-turn short-circuit levels is constructed. Deep features of each monitoring parameter are extracted using the Swin transformer. Information entropy is introduced to quantify the uncertainty of each classifier output, based on which adaptive fusion weights are determined to achieve collaborative decision-making. Simulation results demonstrate that the proposed method effectively exploits the complementary characteristics of heterogeneous physical parameters across the considered fault severities. The proposed method achieves an average recognition accuracy of 90.1%, outperforming the best single-parameter model and conventional fusion strategies, while reducing misclassification between adjacent fault levels. These results demonstrate the effectiveness of the proposed fusion strategy within the present simulation framework and provide a basis for subsequent experimental and field validation.