The basis of traveling wave-based fault location methods is to extract the arrival times of transient signals in the power network. In this paper, a new method for extracting the traveling wave arrival times is presented. For this purpose, the aerial modes of three-phase voltage signals are extracted and two sequential sliding windows of unequal length move along with this signal. By fitting a line to the samples inside each of the two windows and calculating the angle between them, the traveling wave arrival times can be determined with high accuracy. Because this takes place in the time domain, there is no need to switch between time and frequency domains similar to those in Fourier transforms. On the other hand, fitting curves reduce the negative effects of noise and sampling frequency changes. EMTP-ATP is applied to perform the transient simulations and the results are then analysed in MATLAB to conduct the sensitivity analysis against the measurement noises, the sampling frequency, and the fault parameters. The proposed technique is compared to the common techniques such as discrete wavelet transforms and Hilbert Huang transforms. The results demonstrate that the proposed technique has acceptable performance and covers the drawbacks of common methods.
INTRODUCTIONThe motivation for addressing the issue of traveling wave arrival time (TWAT) detection, the work done in this field, and what is discussed in this article can be summarized as follows.
MotivationAccurate fault location in transmission lines reduces power system recovery time, associated costs, and financial losses, especially in deregulated environments. Fault location in the power grid is done in three ways; impedance-based methods [1, 2], traveling wave-based fault location (TWFL) [3-5], and artificial intelligence-based methods [6, 7]. The impedancebased method use power frequency components of voltages and currents while the traveling wave-based methods adopt high-frequency transient components generated by the fault or switching operations. Artificial intelligence/machine learning (AI/ML) methods are generally based on soft computing.In recent decades, with the advancement of signal processing and the possibility of signal sampling at high frequencies,This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.