Hybrid Quantum-Classical Convolutional Neural Network for Detection and Identification of Power Quality Disturbance

Abstract

Power quality disturbances (PQDs) pose significant challenges to modern power systems, necessitating precise detection and identification to mitigate their impacts and enhance grid robustness. In this paper, we propose a hybrid quantum-classical convolutional neural network model (PQDs-QC-CNN) for detecting and identifying power quality disturbances with high efficiency. The model employs a hierarchical framework consisting of quantum convolutional layers, fully connected layers and softmax regression, which can effectively extract multiscale features from disturbance data while mitigating overfitting. Utilising $N$ quantum bits, the model achieves a time complexity of $O\left(\text{poly}\left(N\right)\right)$ and a space complexity of $O\left(N\right)$, ensuring scalability and efficiency. By conducting experiments on the datasets generated in compliance with IEEE Std 1159–2019, the results show a 100% detection accuracy and 99.56% identification accuracy, even with minimal quantum bits and simple configurations. Additionally, the model demonstrates robust noise resistance, maintaining approximately 98% identification accuracy across various noise scenarios. PQDs-QC-CNN not only shows promise for power system applications but also explores new avenues for quantum algorithm integration in smart grid technologies.