Review of the Data-Driven Methods for Electricity Fraud Detection in Smart Metering Systems

In smart grids, homes are equipped with meters (SMs) to monitor electricity consumption and report fine-grained readings electric utility companies for billing energy management. However, malicious consumers tamper their SMs low reduce bills. This problem, known as fraud, causes tremendous financial losses worldwide threatens the power grid’s stability. To detect several methods have been proposed in literature. Among existing methods, data-driven achieve state-of-art performance. Therefore, this paper, we study main fraud detection emphasis on pros cons. We supervised including wide deep neural networks multi-data-source learning models, unsupervised clustering. Then, investigate how preserve consumers’ privacy, using encryption federated learning, while enabling because it has shown that can reveal sensitive information about activities. After that, design robust detectors against adversarial attacks ensemble model distillation they enable evade stealing electricity. Finally, provide a comprehensive comparison of works, followed by our recommendations future research directions enhance detection.