Jointed Task of Multi-scale CNN Based Power Transformer Fault Diagnosis with Vibration and Sound Signals
Mingtao Sun, Xiaojing Bai, Wenbiao Zhang, Lingling Ye
Abstract
The operation status of transformer is closely related to the safe and stable operation of power system. In the process of transformer mechanical fault analysis, fault diagnosis based on vibration signal and sound signal is two feasible methods. However, the acquisition cost of vibration signal is high, and the fault samples of sound signal are few. Therefore, it is a feasible method to combine vibration signal with sound signal to carry out transformer fault diagnosis. In this paper, we propose a jointed learning framework with vibration and sound signals for power transformer fault diagnosis, which integrates multi-branch input, multi-scale residual learning and joint learning technology to identify the mechanical fault types of transformers. The experimental results show that the proposed method has a good discrimination effect on fault characteristics, which can be trained through fault samples of vibration signals, and can achieve high-precision and robust identification for vibration signals or sound signals.