Small-Sample Fault Diagnosis Method for High-Voltage Circuit Breakers via Data Augmentation and Deep Learning
Qiuyu Yang, Zixuan Wang, Jiangjun Ruan, Zhijian Zhuang
Abstract
Deep learning algorithms have been successfully applied in the field of fault diagnosis. However, the deep learning-based diagnostic approaches require rich diverse samples to support model training, which is undoubtedly a challenging task for high-voltage circuit breakers (HVCBs) because their fault data are difficult to obtain. In this article, an intelligent diagnosis method based on data augmentation and deep learning is proposed for HVCB under small-sample conditions. We use the existing small-sample data of HVCB to generate time-frequency domain fault samples of different types and levels based on the improved deep convolutional generative adversarial network (DCGAN) method. In this process, we use the structural similarity principle and perceptual hash algorithm (PHA) to guarantee the quality of the generated samples. Finally, by using the abundant generated fault samples, the intelligent diagnosis model is constructed with the modified AlexNet and tested with a real industrial HVCB. It is demonstrated that the proposed method can effectively address the problem of small-sample fault diagnosis of HVCB.