Enhancing Prony's method for fault location

Abstract

The paper presents a phasor estimator combining Prony's method and discrete analytic signal generation for use in impedance-based fault location. The precision of impedance-based fault location hinges on accurately extracting the fundamental component in the presence of transient components, including inter-harmonics, which compromises the accuracy of the industry-adopted full-cycle Discrete Fourier Transform (DFT) method. The analytic signal gives rise to half the model order, excluding the purely damped exponential, thus producing more accurate results from Prony's process while also contributing to its speedup. The discrete analytic signal is generated so that its length is half of the original signal, meaning its sampling frequency is effectively halved, yielding better numerical conditioning and faster processing in Prony's method. The proposal does not trade accuracy for speed, yet the speed up makes fault location closer to the vision of installing its software module on protection devices. The proposed Prony's method with discrete-time analytic signals is compared with the classical Prony's method applied to discrete-time real-valued signals, the full-cycle DFT with special provisions for eliminating the decaying DC component, and Kung's method. The results obtained by the different approaches show the superiority of Prony's method when combined with analytic signal generation. © 2022 Elsevier Ltd