Unsupervised approach for load disaggregation with devices interactions

Abstract

Energy savings is one of the hottest issues and concerns nowadays due to high oil prices and global warming as a result of CO2 emissions. Non-intrusive appliances load monitoring (NIALM) is a methodology that aim to breakdown the total power consumption measured by the smart meter in each household into the power consumed by the individual appliances. These detailed information on individual appliances consumptions can influence the users to follow better energy usage profiles so as to achieve energy savings. We introduce a novel energy disaggregation model that consider mutual devices interactions and embed the information on devices interactions into the Factorial Hidden Markov Models (FHMM) representations of the aggregated data. The hidden states in the FHMM were inferred by means of the Viterbi algorithm. Devices interactions is a power quality issue that affects the measured power consumed by a device when there are other devices connected to the network. We tested our model using 16 a selected house from the REDD public data set. Our proposed approach showed enhanced results when compared with the standard FHMM. Devices interactions, when observed, enabled us to disaggregate and assign energy consumption for individual devices more accurately