Energy Disaggregation for SMEs using Recurrence Quantification Analysis

Abstract

Energy disaggregation determines the energy consumption of individual appliances from the total demand signal, which is recorded using a single monitoring device. There are varied approaches to this problem, which are applied to different settings. Here, we focus on small and medium enterprises (SMEs) and explore useful applications for energy disaggregation from the perspective of SMEs. More precisely, we use recurrence quantification analysis (RQA) of the aggregate and the individual device signals to create a two-dimensional map, which is an outlined region in a reduced information space that corresponds to 'normal' energy demand. Then, this map is used to monitor and control future energy consumption within the example business so to improve their energy efficiency practices. In particular, our proposed method is shown to detect when an appliance may be faulty and if an unexpected, additional device is in use.

Optimising Parameters in Recurrence Quantification Analysis of Smart Energy Systems

“…Recently, recurrence quantification analysis (RQA) was applied to supervised disaggregation of small business load in [6]. RQA, suitable for non-linearity and complex time-dependencies, was combined with principal component analysis to create an easy visualisation of disaggregation for users.…”

“…Following [6], we concentrate on several RQA variables, all obtained from the recurrence plot matrix R:…”

“…We apply the proposed methods to two systems, the Lorenz attractor and to real data from electricity demand from an SME [6]. For all systems, the timedelay is identified through the first local minimum of the mutual information and the embedding dimension using Cao's algorithm [22].…”

“…The energy data corresponds to total electric current at five minute resolution from a dry cleaners, a small UK business [6], for a period of 6 weeks from 11-th of September to 22-nd of October 2017. Figure 2 shows a typical weekly profile.…”

“…The applications include heart rate [5], financial time-series [3], etc. Only recently, RQA was applied to energy data obtained from high resolution monitoring of electricity usage [6] and was used to map different appliances, to detect faulty devices and identify unexpected usage patterns. The benefits of RQA include relatively simple calculations, powerful investigative characteristics, and applicability to a wide range of systems.…”