ABSTRACTTo address the shortcomings of conventional dissolved gas in oil analysis technology in transformer fault diagnosis, the GRU‐dual transformer and BiLSTM fusion method for transformer fault diagnosis is proposed. Firstly, the time series waveforms are independently transformed into two different image representations by using the GASF and RP techniques. Then, the images of these two types are sent into the transformer encoder, respectively, to extract complementary features. Finally, BiLSTM is introduced in front of the output additional layer to enhance the method's error detection proficiency by capturing the bidirectional time sequence dependence and realizing the fine optimization of model parameters to enhance diagnostic precision. The experimental results show that the proposed method significantly enhances the performance of transformer fault diagnosis.