Data-Driven Anomaly Detection for Smart Energy Meters

In electrical grids, anomalies in power measurement, such as drifts, have notable impacts on the accuracy of electric meters, negatively impacting load balancing and pricing schemes. Conducting manual checks for compliance across the extensive array of energy meters is impractical, making it crucial for researchers to find alternative strategies. This study suggests employing data-driven techniques and using smart meters as IoT gateways for early-stage anomaly detection. To achieve this goal, the study begins with a review of existing anomaly detection methods, emphasizing their impact on the grid. Following this, a cloud infrastructure has been introduced for gathering data from second-generation smart meters used in Italy. Additionally, the paper presents a numerical method for detecting anomalies in smart meters that may happen over time based on machine learning techniques and the results are provided to validate the effectiveness of this approach. Finally, the study concludes with a discussion of the results and potential future aspects.