Energy disaggregation of overlapping home appliances consumptions using a cluster splitting approach

Abstract

Non-intrusive load monitoring (NILM) is a set of techniques that aims to decompose the aggregate energy consumptions of a household into the energy consumed by the respective individual appliances. When some of the home appliances have power consumptions levels that overlap with each other, it becomes a challenging problem to disaggregate the energy consumed by each of these appliances. In this work, we present an approach that split the clusters of the overlapping energy consumptions into the respective energy consumed by the individual appliances. The proposed approach involves firstly to analyze the cohesion of devices clusters to determine if a cluster should be split into two clusters. The proposed cluster splitting approach was tested on cases of overlapping devices clusters from six real houses available from the REDD public data sets. The results showed that the performance of the proposed approach depends on the degree of overlapping of the devices clusters, on whether the clusters are tight or loose and on the sizes of the clusters. The proposed approach can be applied to a clustering-based load disaggregation method as a subsequent step to deal with situations of overlapping appliances consumptions, so as to improve the overall energy disaggregation accuracy.