ElectroShield: XAI-assisted fault detection in transmission lines using multi-modal data

Abstract Transmission lines play a pivotal role in ensuring the seamless flow of electricity. They are crucial for ensuring that electricity generated at power plants can reach distant consumers. The reliable operation of transmission lines is pivotal to ensuring uninterrupted power supply and the overall stability of the electrical grid. However, transmission lines are vulnerable to faults and failures due to various factors, including extreme weather events, equipment aging, and accidents. Detecting these faults is essential for maintaining grid reliability and preventing potential hazards. Conventional methods for detecting faults in transmission lines often fall short in accuracy and robustness. This poses a serious problem, as undetected or misdiagnosed faults can lead to costly downtime, safety hazards, and disruptions for consumers and industries alike. In this study, we address the pressing issue of fault detection in transmission lines by proposing an innovative approach that leverages the fusion of information from vibration, current, and voltage signals. The proposed solution incorporates eXplainable AI, providing human-understandable explanations to assist engineers in efficiently identifying the causes of faults. The proposed method is validated using real-time data obtained from a reduced-scale laboratory model of a transmission line, demonstrating high diagnostic accuracy and robustness.