A Novel Multivariate and Accurate Detection Scheme for Electricity Theft Attacks in Smart Grids

Abstract

In advanced metering infrastructure (AMI), smart meters (SMs) are deployed to periodically forward accurate power consumption readings from the client side to the electric utility companies/operators. Such readings are crucial for load monitoring, grid management, and billing. However, malicious clients or manipulated SMs may initiate electricity theft cyberat-tacks by reporting false/manipulated readings to deteriorate the grid performance or decrease their bills illegally. To identify these attacks, this paper proposes a novel multivariate electricity theft detector that considers not only the power consumption readings, like most existing techniques in the literature, but also the grid voltage and power losses. The proposed detector allows the electric utilities to accurately detect the electricity theft incidence and monitor diverse clients' loads. The proposed model was evaluated using real-world data, where it could outperform the baseline detector, that relies only on power consumption readings of different clients, by achieving around 5-15% enhancement in the detection rate of different, considered attacks.