DL BASED ANOMALY CLASSIFICATION IN POWER DISTRIBUTION NETWORK

Abstract

Deep learning (DL)-based classification techniques offer robust and scalable solutions for detecting anomalies and failures in Intelligent Electronic Devices (IEDs), which are critical components of modern smart power grids. These devices play a key role in efficient energy management and maintaining a stable electricity supply. However, the growing complexity and interconnectivity of smart grid infrastructures make them increasingly vulnerable to both operational failures and cyberattacks. Traditional fault detection methods, such as rule-based systems and manual inspections, are often time-consuming, error-prone, and may fail to detect subtle signs of impending issues. Moreover, they frequently struggle to distinguish between genuine faults and normal system variations, leading to false alarms and unnecessary maintenance. To overcome these challenges, this work proposes a novel DL-based anomaly classification framework that leverages supervised learning algorithms trained on labeled datasets, including simulated power system attack scenarios. By extracting and analyzing features such as voltage variations, current measurements, and communication patterns from IED data, the proposed system can accurately classify various types of failures, including equipment malfunctions, operational anomalies, and cyber-induced disruptions. Furthermore, DL models enhance cybersecurity by detecting suspicious network activities, such as unauthorized access or configuration tampering, thereby enabling proactive threat mitigation. This approach significantly strengthens the resilience of smart grids, reduces downtime, and ensures a secure and continuous power supply to consumers. The proposed system represents a substantial advancement over conventional fault detection techniques, offering higher accuracy, faster response times, and greater adaptability to the evolving threats in smart grid environments.