Implementation of Fast Fourier Transform and Artificial Neural Network in Series Arc Fault Identification and Protection System on DC Bus Microgrid

Abstract

A microgrid is a cluster of electrical sources and loads that are interconnected and synchronized. Microgrid operation is typically divided into two modes, isolated or connected to the grid with a single or standalone control system. In this context, it can enhance the reliability and quality of electricity supply for connected customers. When using a microgrid system, it is important to consider the risk of series arc faults. Series arc faults are sudden bursts of flames resulting from ionization of gas between two electrode gaps. These faults can occur due to manufacturing defects, installation Errors, aging, or corrosion on conductor rods, leading to imperfect connections. Detecting series arc faults in DC microgrid system operations can be challenging using standard protective devices. Failure in the protection system can pose risks of fire, electrical shock hazards, and power loss in the DC microgrid.Therefore, a device has been designed to detect series arc faults by utilizing the fast Fourier transform method and artificial neural network, which function to analyze DC signal and make decisions when faults occur by examining the average sum of current frequency values during normal and fault conditions. In this study, the average sum of current frequency values during normal conditions was found to range from 0.35437 to 0.36906 A, while during fault conditions, it ranged from 0.21450 to 0.22793 A, with an average protection identification time of 1087 ms and an ANN output accuracy of 99.98%.