Machine learning Techniques for Energy Theft Detection in AMI

Abstract

Advanced Metering Infrastructure (AMI and smart meter) is considered as the basic building block for the development of smart grid in the power distribution system. a Smart meter is one of the keys elements of Advanced Metering Infrastructure, it provides two-way communication between customer and electricity utility, Smart Meters send consumption data frequently (e.g., every 15 minutes) to the utility for monitoring and billing, therefore, a gold mine of data is generated for utilities. Smart meters have become a major focus for targeted attacks which lead to the energy theft, resulting in losses of billions of dollars per year in many countries. Therefore, multitude of papers have studied the energy theft detection by applying different disciplines on smart meter data. In this paper, we present an overview of machine learning research in energy theft detection using smart meter data. It then surveys these research efforts in a summary and comparison of learning models used, in terms of performance metrics, simulation and analysis environment, and data sets used. It finally highlights the challenges in energy theft detection. We approve that these challenges have not been adequately addressed and considered in previous contributions, also covering them, is necessary to advance the energy theft detection.