On Using Graph Signal Processing for Electrical Load Disaggregation

Abstract

Graph Signal Processing (GSP) is an emerging field in data science and is being increasingly used for different applications in various domains like vision, biomedical and sensor networks, etc. GSP analyses and transforms signals defined on the vertices of a graph. In this paper, GSP is applied to the problem of source separation, in particular, electrical load disaggregation where, given smart meter measurements, it is required to estimate the contribution of different loads which could have resulted in those measurements. The proposed strategy based on GSP is formulated as a regularization on the graph whereby appropriately maximizing the smoothness of the underlying graph signal, electrical loads are disaggregated iteratively. The proposed optimization problem is solved in a greedy way and a closed-form solution is presented. Experimental results using publicly available REDD are presented to demonstrate the potential of the technique for disaggregating loads from low rate aggregate power measurements sampled at 1 minute and 15 minutes intervals.