AI-BASED FAULT DETECTION IN SMART POWER GRIDS

Abstract

The integration of information and communication technologies (ICT) into electrical power systems has ushered in the era of smart power grids, characterized by enhanced monitoring, two-way information flow, and improved reliability. However, increased complexity and distributed energy resources (DERs) also raise susceptibility to faults that degrade performance and stability. Conventional fault detection and diagnosis (FDD) mechanisms—such as threshold-based relays and impedance calculations—lack the adaptability and predictive insight required for modern grids. This research explores artificial intelligence (AI)–based methodologies for real-time fault detection, classification, and localization in smart grids. A comprehensive methodology utilizing machine learning (ML) and deep learning (DL) models is developed, incorporating real-time data from phasor measurement units and smart sensors. Models including support vector machines (SVM), random forests (RF), convolutional neural networks (CNN), and recurrent neural networks (RNN) are compared. Experimental results demonstrate that AI techniques significantly outperform traditional methods in accuracy, response time, and robustness across diverse fault types. Furthermore, integration with edge computing and federated learning improves scalability and data privacy. Key challenges—such as data scarcity, cybersecurity threats, and model interpretability—are discussed, and future research directions toward autonomous self-healing grids are proposed.