Hybrid quantum learning based fault detection method for an active distribution system

Abstract Modern distribution systems with integrated renewable energy sources exhibit complex fault behaviour due to inverter-based generation and bidirectional power flow, making conventional protection methods less reliable. This paper presents a PMU-based fault detection framework for active distribution networks using hybrid quantum machine learning (QML) techniques. A modified IEEE 34-bus system with solar photovoltaic and wind distributed generation units is modelled, and faults are systematically created near the DG locations with wide variation in fault type, location, and fault resistance. PMU-measured voltage and current phasors are processed using symmetrical component analysis to extract compact numerical features representing peak magnitudes and phase angles of sequence components. Two hybrid QML models, namely Quantum Random Forest (QRF) and Variational Quantum Classifier (VQC), are developed within a unified modular framework. In the QRF model, quantum computation is employed for nonlinear feature transformation, followed by classical ensemble learning for classification. In the VQC model, parameterized quantum circuits learn nonlinear decision boundaries through variational optimization. The performance results show a clear difference between QRF and VQC. QRF gives stable and reliable fault detection with high precision, but VQC performs better overall. The VQC testing accuracy reached 99.8%, with precision of 99.7%, recall of 100%, and an F1-score of 99.9%. These results were achieved through the consideration of SMOTE for class balancing and feature selection for enhanced model performance. When compared with classical and deep learning approaches, the advantage of using numerical PMU data with hybrid quantum learning becomes more evident, especially since no image-based signal conversion is required. Overall, hybrid QML can be considered as a useful technique for advanced protection system.