DC Series Arc Fault Detection Using Machine Learning in Photovoltaic Systems: Recent Developments and Challenges
Shibo Lu, Animesh Sahoo, Rui Ma, B.T. Phung
Abstract
DC arc faults, especially series arc faults, are becoming more common in photovoltaic (PV) systems. Without timely detection and interruption, such dangerous events can cause catastrophic fires, posing severe threat to human safety and properties. This paper presents a review on DC series arc fault detection using machine learning (ML) in PV systems. Technical details of applied ML methods, including conventional ML and deep learning (DL), in recent published paper are summarized and discussed. In addition, several popular ML methods are evaluated and compared using the same experimental datasets collected in laboratory to examine their effectiveness in DC series arc fault detection. Finally, practical challenges are identified, potential solutions are provided, and future research directions are recommended.