A distributed intelligent framework for electricity theft detection using benford's law and stackelberg game

Abstract

Electricity theft is a major contributor of nontechnical losses in the distribution systems of the smart grid. However, owing to the resource-limitations of smart meters and the privacy requirement of electricity usage data, theft detection has become a challenging task for electric utilities. To address this problem, a Distributed Intelligent Framework for Electricity Theft Detection (DIFETD) is proposed and implemented in this paper. It is equipped with Benford's Analysis for initial but powerful diagnostics on smart meter big data. A Stackelberg game-theoretic model is formulated to analyze the strategic interactions between one utility and multiple electricity thieves, which is applied to data flagged suspicious by Benford's Analysis. The Stackelberg equilibrium provides sampling rate and threshold to conduct a Likelihood Ratio Test (LRT) to detect potentially fraudulent meters. The framework is validated on real interval electricity usage data from an electric utility in Florida to filter fraudulent meters in a community.