Fault diagnosis and trend forecast of transformer based on acoustic recognition

After discussed the difficulty of monitoring transformer fault online, this paper proposes a fault diagnosis scheme based on acoustic wave analysis. The time and frequency domain signal processing is utilized to analyze acoustic signal so that equipments´ running status can be identified and the trend of development be predicted. Firstly, a new scheme of transformer fault diagnosis and trend forecast is designed. Transformer acoustic signal acquisition, noise elimination method in diagnosis system is represented. Secondly, the new algorithm of the signal strangeness detection and trend-based forecasting based on wavelet analysis is put forward. Lastly, the wavelet packet algorithm is utilized to analyze acoustic signal and extract frequency & time-domain features which relative with its development trend, and then the predicting consequence can be use to assess the fault type. The actual measurement test demonstrated that the serial wavelet packet transformation (WPT) algorithms have high availability and feasibility to diagnosis and forecast the transformer faults accuracy.