Fault Detection in Power Systems Using Convolutional Neural Networks: A Review of Voltage and Current Data Visualization Techniques

Abstract

Fault detection in power systems is critical for ensuring reliability, safety, and operational efficiency. Recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have revolutionized fault detection by leveraging voltage and current data visualization. This review paper synthesizes findings from 28 recent studies, exploring CNN-based approaches for detecting and classifying faults in high-voltage equipment, transmission lines, and electrical components. Techniques such as Faster R-CNN, self-attention CNNs, and hybrid CNNLSTM models are analyzed, alongside data preprocessing methods like wavelet transforms and Park’s transformation. The review highlights the efficacy of CNNs in achieving high accuracy (up to 100%) and rapid fault detection (as low as 8 ms), addressing challenges posed by complex backgrounds, noise, and intermittent distributed generation. This paper provides insights into current methodologies, their applications, and future directions for intelligent fault detection in power systems.