Machine Learning Approaches for Power Transmission Lines Fault Classification: A Systematic Review

Abstract

The era of computational intelligent power systems has resulted in a proliferation of great number of techniques, technologies and products for the power industry. This may be attributed to the unique ability of computationally intelligent systems to demystify the hidden truths in large volumes of power systems data. In this paper, a succinct systematic review of power transmission line fault classification based on ML approaches is presented. The review accounts for the different Machine Learning (ML) approaches in terms of their nature of classification such as the single or multi-class, the nature of the ML such as the use of purely mathematical or AI, and the nature of the hybridization such as the use of dual or multi-hybridizations and for a period of 2006 to 2020. Specifically, the review studies the aspect of power systems classification that deals with the detection and localization of faults on transmission lines. From the reviews conducted, the most prominent classifiers were found to be the conventional Artificial Neural Networks (ANN) for the detection and localization of power system faults. This shows that the researchers will prefer the predictive ML approaches that are based on the simple intelligent operations that occur in mammalian brains i.e. the conventional Back-Propagation Neural Networks (BPNNs). However, it will be desirous if researchers equally investigate other potential but less popular neural ML model schemes.