Forecasting the fault and danger in power system and limiting their incidence have been the emphasis in modern power system study. The abundant real-time data gathering in power system contain the complexity and uncertainty of the system, so mining and fusing a series of universal applicable quantities from these data to detect system fault and its stability is essential. Combining wavelet analysis, the modern time-frequency analysis technique, with entropy theory by exploiting their virtue will resolve the problem. In this paper, the feasibility of the wavelet entropy application in power system fault detection and distinguish is analyzed, the wavelet entropy concept based on wavelet analysis is defined, the computation methods of the two kinds of wavelet entropies are established, and the application of the wavelet entropy in transmission line fault detection is simulated and analyzed.