Wavelet Transform and Approximate Entropy Based Identification of Faults in Power Swings

Based on wavelet transform and approximate entropy, a novel algorithm to detect and identify faults during power swings is proposed in this paper. Although currents will change similarly during power swings with or without faults, there is a difference between them which determines their different frequency distributions. Wavelet transform, which has a capability of time-frequency localization, can be employed to decompose the signals into several frequency-bands in order to make it possible to extract feature details. Approximate entropy, which has been used successfully in mechanical diagnosis, is applied to represent the signal feature intuitively so as to form the criterions for fault detection and identification. Numerical simulations with PS CAD and practical analysis allow us to conclude that this novel algorithm can not only detect faults but also identify various faults during power swings quickly and accurately. Moreover, a short sequence of data is needed by this algorithm, which will make it ideal to be employed into a protection relay for real-time applications and would put on a better performance than traditional blocking algorithms.