Artificial Neural Network-Based Fault Detection in Transmission Lines with Real-Time Hardware-in-the-Loop Validation
Transmission lines are highly susceptible to faults, making rapid and reliable fault detection essential for maintaining power system stability. Although conventional protection schemes are widely used, their performance may deteriorate under challenging operating conditions. To address these limitations, researchers have increasingly explored Artificial Intelligence (AI)-based approaches. However, many existing AI methods rely on computationally intensive feature extraction and are typically validated only on a single transmission network. This dissertation proposes a computationally efficient methodology based on Artificial Neural Networks (ANNs) that is applicable to power systems with different network topologies without requiring retraining, while maintaining suitability for real-time implementation. The ANN was trained and evaluated using data from a 500 kV transmission line, validated through cross-topology testing on the IEEE 9-Bus system without retraining, implemented on NI CompactRIO-9073 hardware, and tested under Hardware-in-the-Loop (HIL) simulations using a Real-Time Digital Simulator (RTDS). The results demonstrate high fault detection performance, successful cross-topology generalization, and reliable real-time operation, highlighting the potential of the proposed ANN as a computationally efficient backup or supervisory protection scheme for transmission lines.