Electronic transformer (ET) is wildly used in digital smart substation. However, the actual running time of ETs is generally short, and most of them have higher failure rate than that of traditional transformers. Since the accuracy and stability of ETs data collection affect the safe operation of smart grid, it is necessary to study the fault diagnosis method of ETs. According to the characteristics of ETs, this paper proposes an on-line fault diagnosis method of ET based on wavelet transform theory. After noise reduction the output signal waveform of the transformer, the high frequency decomposition coefficient is obtained by wavelet transform, so as to diagnose and locate the three kinds of faults: drift deviation fault, variable ratio deviation fault, and fixed deviation fault.