Detection of energy theft and metering defects in advanced metering infrastructure using analytics

Non-technical losses including electricity theft and anomalies in meter readings are estimated to cost the utility providers tremendous losses of approximately $96 billion per annum. The adoption of smart meter has encouraged utility providers to use analytics to identify theft. To curb nontechnical losses, they are increasingly leveraging on real-time smart metering and analytics to identify energy theft and irregularities in meter readings. We have previously put forward linear regression-based and linear programming-based anomaly detection frameworks to study consumers' energy consumption behavior for detecting the localities of metering defects as well as energy thefts. In this work, we design and construct an advanced metering infrastructure test rig in the laboratory to perform comparison studies on our previously proposed anomaly detection frameworks in smart grid environment. Results from both test rig and simulations show that linear regression-based anomaly detection framework is able to identify the positions of energy thieves and faulty smart meters without requiring large volume of data samples. However, linear programming-based framework is more robust as compared to linear regression-based because the former is capable of detecting more sophisticated types of energy theft/meter irregularities accurately even in the presence of technical losses/calibration errors.