Condition Recognition of High-Voltage Circuit Breakers Using Data Augmentation Vision Transformer

An improved data augmentation vision transformer-based method for the recognition of the High Voltage Circuit Breaker (HVCB) condition is proposed in this paper. The method utilizes the vision transformer approach to differentiate states by enhancing the time-frequency domain features of vibration signals through Continuous Wavelet Transform (CWT). A novel vision transformer network is constructed to enhance local feature extraction capabilities. In the performance verification phase, vibration data from HVCB in four different states were collected for comparison. The results demonstrate that the average condition recognition rate of the proposed method reaches 98.91 %showcasing a significant improvement over conventional vision transformer methods. Moreover, it outperforms current mainstream CNN models, providing evidence of the effectiveness of the vision transformer approach.