BP-Neural-Network-Based Aging Degree Estimation of Power Transformer Using Acoustic Signal

In this paper, an aging degree estimation method using acoustic signals is proposed based on a BP neural network. Twenty-eight transformers are taken as research objects. The transformer's internal noise mechanism is analyzed, and the acoustic signals of the high- and low-voltage sidewalls are collected and screened. The BP neural network is used to predict the transformer age in real-time. Comparing the predicted results with their actual operation time provides a sufficient basis for determining the degree of transformer aging and the need for an overhaul. After network training and data testing, the error between the predicted value and the actual value reaches the least. The proposed estimation method can play an innovative role in the process of transformer fault monitoring.