Advanced Power Theft Detection in Smart Grids Using Entanglement-Driven Quantum Machine Learning

Abstract

Electricity theft in smart grids has become a critical issue due to the manipulation of smart meters and distributed generation systems such as photovoltaic (PV) units. Traditional machine learning methods often fail to effectively handle highdimensional, noisy, and imbalanced data, especially in distributed generation and net-metering domains. To address these limitations, this paper proposes an entanglement-enhanced Quantum Deep Learning (QDL) framework for accurate power theft detection. The model integrates a hybrid architecture combining Quantum Variational Circuit (QVC) and Data Reuploading Circuit (DRC), along with a novel entanglement layer to exploit quantum parallelism and strong feature correlations. The dataset is preprocessed using normalization and splitting techniques, with 80% used for training and 20% for testing. Comparative analysis with classical models such as XGBoost and LightGBM demonstrates improved performance. The proposed FH-QVCDRC-ENT model achieves superior accuracy of 91%, outperforming existing approaches. Experimental results validate that the incorporation of quantum entanglement significantly enhances classification capability, robustness, and generalization, making the model highly effective for detecting electricity theft across both consumption and distributed generation domains in smart grids.