The effectiveness of different sampling rates in vegetation high-impedance fault classification

Abstract

This paper investigates the alarming fire igniting scenario of High-Impedance Faults (HIF) resulting from the contact of vegetation with the power lines. Our findings are based on a set of experiments performed on a data set of real staged fault-signals sampled by two distinct channels with different band-pass filters. Representations from these two sampling methods are extracted by different signal processing methods and ranked by their discriminative potential. Experimental results obtained by our proposed methodology show the effectiveness of wide band signals sampled at higher frequencies. Their features result in higher separability potential and are more effective at discriminating fault occurrences than ones from the lowfrequency channel. As the approach of employing high-frequency signals in such task may be faced with skepticism, the paper also discusses the possible concerns and feasibility of using higher sampling rate technologies for HIF fault classification.

Vegetation High-Impedance Faults’ High-Frequency Signatures via Sparse Coding

“…Given their sampling rate of around tens of thousand samples per second, which is on the higher end for most works, this becomes especially pertinent to the High-Frequency (HF) spectrum components. Such factor may or not be a concern for the HIFs type target in the cited papers but evidence [3], [4] shows that it has strong relevance to the faults investigated in this paper.…”

“…The voltages, rather than fault currents, were the domain where discriminative analysis took place. Previous works by the authors, based on this data set, proposed an HF voltage-based HIF detection method [3] and a comparative analysis between the information content of the LF and HF channels [4]. The fact that the latter gave evidence to the existence of relevant discriminative information to be existent only in the HF signals is the reason for focusing on this domain.…”

“…All these points are intricate topics that were detailed in previous works. Primarily, the authors proposed a fault detection algorithm based on the high-frequency signals in [3], and then presented a comparative analysis between the fault predictor information of the LF and HF channels in [4]. The latter gave evidence to the existence of relevant information to discriminate between HIFs and normal states to be existent only in the HF signals.…”

“…The present analysis focuses on the staged faults' HF voltage recordings mainly due to their higher information content [4] when compared to LF signals. The lack of specific discussion regarding such signals in the related literature is also a relevant gap to be cited.…”

“…One crucial assumption made by the authors on this and past works [3], [4] is that HIFs should not be addressed as a single unified problem but rather by their distinct types. These disturbances are actually an intricate set of problems depending on many factors such as conducting surface, voltage level, network ground type, and more.…”