Fault detection of on-load tap changer based on improved transformer

On-load tap changers produce mechanical signals with unstable and nonlinear characteristics, making it difficult to accurately identify faults from vibration signals. To improve the accuracy of fault diagnosis of loaded tap changers, this paper proposes a fault diagnosis method based on variational mode decomposition (VMD) and multidimensional Transformer. Firstly, the dung beetle optimization (DBO) is used to optimize the penalty coefficient and the number of decomposed modes in VMD. Then, the decomposed modal components are input into the multidimensional Transformer encoder for signal feature learning. Finally, the fault of the on-load tap changers are realized by intelligent classification. The experimental results show that the recognition rate of the proposed method for the faults of loose contacts, loose springs, and broken springs of on-load tap changers reaches more than 96%. Compared with the Transformer and BP neural network algorithm, the recognition effect is significantly improved.