A Framework for Non Intrusive Load Monitoring Using Bayesian Inference

Abstract

Non-Intrusive Load Monitoring (NILM) refers to the disaggregation of electric appliances from a single point measurement. The problem is gaining a lot of attention recently, primary due to the promising energy savings as well as potential business prospects such a solution brings. However, in a large scale deployment, the digital meter is unlikely to have multiple electrical parameters which most existing NILM research rely on. In this paper, we report the results of using a Bayesian approach to obtain the disaggregation of the loads where only active power measurements are available at a sampling rate of a few seconds. The proposed method requires the prior availability of appliance information (i.e., the prior probability and appliance ratings). To obtain the appliance information for the disaggregation algorithm, we adopt an unsupervised learning approach. Further, we present the results of these algorithms on a simulated and an open household electric consumption data set.