Quantum Convolutional Neural Network Approach for Fault Detection and Localization in Smart Power Distribution Systems

Abstract

The reliability of electric power distribution systems is essential for maintaining a continuous energy supply, especially with the growing integration of renewable energy and distributed generation. The complexity of modern systems necessitates fast and accurate fault detection methods in order to prevent systemic losses. This study introduces a Quantum Convolutional Neural Network (QCNN) approach to identify and locate three-phase faults in smart distribution systems. The model features a hybrid quantum-classical architecture that utilizes superposition and entanglement in order to enhance fault signal feature extraction. Data have been generated through simulations of the IEEE 33-Bus distribution system under various fault scenarios and processed with normalization and training. The results show that the QCNN achieves 99.21% accuracy, a 99.00% F1-score, and an inference time of 4.7 ms, outperforming CNN, SVM, and Random Forest. This method proves effective for real-time fault detection and localization and can help improve the reliability of future smart power distribution systems.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint