Arc Fault Detection in Electrical Systems based on Multi-Feature Fusion with Graph Neural Network

In this era, Arc fault detection in electrical systems is a serious phase of safe guarding and preventing electrical fires. Moreover, arc faults occur because of accidental electrical discharges among the conductors that lead to sparking, overheating and severe damage if undetected. The main motive of this research is detecting the faults from the reliable and safe electrical systems. An effective detection system is critical because of non-linear, high-frequency and stochastic features of arc fault signals with traditional Machine Learning (ML) and Deep Learning (DL) methods. In order for overcoming these issues Multi-Feature Fusion (MFF) with Graph Neural Network (GNN) namely (MFF-GNN) is proposed to detect the Arc faults in electrical system. Initially, Discrete Wavelet Transform (DWT) is utilized to decompose the signals into time frequency components further Temporal Domain Visualization Convolutional Neural Network (TDV-CNN) utilized to extract complex patterns as well as anomalies indicative of faults. Then the MFF-GNN integrates various features including complex patterns, spatial relationships and signal characteristics into a graph structure for detecting the arc faults effectively in the terms of F1-Score compared to the existing Genetic Algorithm (GA).