Electricity Theft Detection: A Review of Methods, Performance, and Applicability

Electricity theft remains one of the major challenges faced by utility companies, especially in emerging countries, where socio-economic factors intensify the occurrence of nontechnical losses. This paper presents a comprehensive review of current electricity theft detection methods, classified into three main categories: conventional, regulatory, and technology-based. Emphasis is given to technology-based approaches, including machine learning (ML), deep learning (DL), wireless sensor networks, IoT systems, and game-theoretic models. A total of 15 recent studies (2020–2025) were selected and analyzed based on performance metrics such as Accuracy, Precision, Recall, F1-Score, and AUC-ROC. The results show that ML/DL models achieve superior performance but are often designed for Smart Grid or AMI infrastructures, limiting their applicability in conventional networks. Meanwhile, IoT-based and regulatory-enhanced models offer more practical alternatives for systems lacking advanced metering infrastructure. This study highlights the gap between the high performance of existing models and their real-world feasibility, especially in regions with limited technological resources, and discusses the potential for hybrid models to bridge this divide.