Abstract: The monitoring data of modern power system have strong intrinsic relationship and its abnormal and fault signal is often hidden in big data sets, so the traditional Fourier transform analysis methods don't have the ability of time domain localization analysis. A new method based on time-frequency characteristics of wavelet analysis was proposed for singularity detection of transient signals in power system, which through decomposing the multi-scale one-dimensional transient signal wavelet, extracted high frequency and low frequency coefficient, and got fault time information while the signal de-noising; then based on the singularity detection of modulus maxima, the fault location information was obtained, so as to realize the temporal-spatial detection and diagnosis of abnormal signal of power system. By means of the simulation analysis in IEEE 39 bus system, the results show that the proposed method realizes the singular signal time-frequency characteristics of sorting, and initially satisfies the requirement of fault spatial-temporal positioning.

Key words: fault detection, wavelet analysis, singularity detection, modulus maxima, high and low frequency coefficients