Oscillation wave testing for transformer winding fault diagnosis based on a CNN-BiLSTM algorithm

The equivalent circuit parameters of transformer windings are determined by their geometric structure, and variations in thes e parameters are directly associated with winding faults. To achieve high-precision fault diagnosis, this paper establishes an oscillation wave testing platform for transformer windings. This platform simulates four operational conditions, namely healthy state, axial displacement fault, inter-layer short circuit fault, and winding loosening fault, by collecting oscillation wave signals and thereby constructing multidimensional sample sets for reasoning. For transformer winding fault diagnosis, an attention-based CNN-BiLSTM fusion fault diagnosis algorithm is proposed. Evaluation of the model on the test dataset shows that the proposed algorithm can effectively identify winding faults, with superior performance especially in scenarios involving weak faults or complex faults. Transform can accurately identify fault-related features, offering a new approach for the online condition monitoring of traction transformers.