The increasing integration of smart meters in modern power grids has enabled large-scale data collection for the detection of electricity theft, a major source of non-technical losses worldwide. Recent advances in quantum machine learning (QML) have introduced new opportunities for analyzing high-dimensional consumption patterns, although their practical applicability remains limited by hardware constraints. Ensuring reliable and efficient energy distribution is critical for modern societies, both globally and in Canada, where electricity represents a key economic and social resource. We present a quantum-classical hybrid model for energy theft detection, combining classical methods with quantum-inspired techniques. The framework is implemented in PennyLane and evaluated on environments ranging from a local workstation to Calcul Quebec HPC clusters to assess scalability and execution feasibility. Experimental results indicate that classical machine learning models remain fast and reliable, while hybrid quantum-classical models offer competitive accuracy with richer feature representations.
Charnaux et al. (Thu,) studied this question.