Novel Consumer Classification Scheme for Smart Grids

Abstract

Classifying different type of consumers (households, office buildings and industrial plants) is an important task in Smart Grids. In this paper, we propose a novel classification scheme based on nonlinear prediction for consumption timeseries obtained from a smart meter. The candidate predictors were tested under different assumptions regarding the statistical behavior of the underlying consumption time-series. As a result a feedforward neural network based predictor has been shown to be the most promising solution. In order to demonstrate the power of the proposed method simulations have been carried out. The consumption data came from a bottom up model, where Markov model of individual appliances and real measurements of photo-voltaic generators have been applied. The numerical results prove that our method is capable of distinguishing an office-building with installed photo voltaic mini power plant from an office-building which is lack of such power plant.