Detection of illegal consumers using pattern classification approach combined with Levenberg-Marquardt method in smart grid

Abstract

Detection of illegal consumers is an extremely challenging problem in smart grids as well as traditional environments. In a smart grid environment, electrical energy illegal customers can be divided into two types; (1) if the customer consumes all of its required energy in a portion of day illegally, (2) if the customer consumes a portion of its required energy illegally. Many methods about illegal consumption or electricity theft detection have been proposed but they are able to detect only one type of illegal consumptions. In this paper, a combined method is proposed to detect both two types of illegal consumptions. Customer energy consumption pattern classification method based on probabilistic neural network and mathematical model based on Levenberg-Marquardt method are used to detect the first and second type of illegal consumption, respectively. Moreover, the impact of Distributed Generation (DG) sources on illegal consumption of electricity is analyzed and proposed detection algorithm is modified to compensate it. Experimental results are presented to show the effectiveness of this method in detection of both two types of illegal consumption.