High Impedance Fault Detection in Power Systems: A Systematic Review Using PRISMA Guidelines

Abstract

One of the biggest challenges in modern day electrical power distribution systems is high impedance faults (HIFs). They are usually caused by a conductor touching a high resistance area like vegetation, tree branches or ground. The nonlinearity of a HIF, its random and asymmetric nature and its comparatively small magnitudes make it difficult to detect with conventional protection devices, and it is often difficult to tell if a HIF is occurring. A HIF can be serious safety hazard if not detected early on, such as posing an electric shock hazard or fire hazard, especially in distribution networks. A structured review of literature on the detection methods for HIF is presented in this paper, concentrating on the development of the methods from the classical signal processing to the modern intelligent and data driven methods. The studies that have been reviewed are classified according to their method, such as signal-based techniques, harmonic analysis, wavelet transform methods and machine learning based methods. Comparative statistics are then performed to reveal dominant trends in the literature and relative adoption of categories of techniques in recent literature. The results showed that the trend is definitely shifting from deterministic systems to hybrid and intelligent systems especially those that combine wavelet-based feature extraction with machine learning based classifiers. The review also points out that in recent years, data-driven techniques have become more and more popular for their better adaptability and fault detection accuracy under various fault conditions. In addition, the study highlights the need for adaptive and resilient detection systems that can maintain a reliable operation and public safety in modern distribution networks. Last but not least, this work brings a systematic classification of the available HIF methods, highlights the needs for the research and development of more reliable and real-time detection strategies and outlines future research directions.