Entropy-based electricity theft detection in AMI network

Abstract

Advanced metering infrastructure (AMI), one of the prime components of the smart grid, has many benefits like demand response and load management. Electricity theft, a key concern in AMI security since smart meters used in AMI are vulnerable to cyber attacks, causes millions of dollar in financial losses to utilities every year. In light of this problem, the authors propose an entropy-based electricity theft detection scheme to detect electricity theft by tracking the dynamics of consumption variations of the consumers. Relative entropy is used to compute the distance between probability distributions obtained from consumption variations. When electricity theft attacks are launched against AMI, the probability distribution of consumption variations deviates from historical consumption, thus leading to a larger relative entropy. The proposed method is tested on different attack scenarios using real smart-meter data. The results show that the proposed method detects electricity theft attacks with high detection probability.