Non-intrusive load disaggregation with adaptive estimations of devices main power effects and two-way interactions

Abstract Energy management and savings in residential homes are emerging concerns nowadays because of several challenges facing the energy sector such as energy sources limitations and environmental impacts. Non-intrusive load monitoring (NILM) was introduced as a set of methods and techniques that aim to decompose the total aggregate consumption measured by the smart meter into the consumptions by individual appliances present in the household. The detailed information on energy usage for each device were found to be a good influencing method for the residents to adopt better devices usage profiles which lead eventually to noticeable energy savings. Recent research had shown that the Hidden Markov Models (HMMs) and its extensions are effective models in the load disaggregation problem. The authors had introduced a new unsupervised approach for load disaggregation that includes the mutual devices interactions information into the Factorial Hidden Markov Model (FHMM) representation of the aggregate signal in an earlier work. In this paper, we introduce an adaptive approach for estimating devices main power consumptions and their two-way interactions during the disaggregation process. The adaptive approach is used to mimic the changes in devices consumptions and two-way interactions. The adaptive estimation process was carried out only for cases when there are four devices or less that are operating/ON instantaneously. The proposed approach was tested with data from the REDD public data set and it showed better performance in terms of energy disaggregation accuracy compared with the standard FHMM. The adaptive estimating of main factors effects (primary power consumptions) and two-way interactions during the disaggregation process provided higher disaggregation accuracy results, in general, than those with fixed factors and two-way interactions values.