Embedded NILM as Home Energy Management System: A Heterogeneous Computing Approach

Abstract

This paper presents an embedded NILM engine to enable load disaggregation intelligence and explore its potential application as an energy management system. In this sense, the power meter is upgraded to a novel category called cognitive power meter. Therefore, this paper discloses a heterogeneous multiprocessing approach to attend NILM prerequisites and increase household interactivity. The proposed NILM performs the microscopic analysis using the Conservative Power Theory (CPT) for feature extraction; k-Nearest Neighbors (k-NN) for the appliance classification; and the Power Signature Blob (PSB) for energy disaggregation. Results show NILM can be performed on-site, embedded into modern cognitive power meters, and it may support households on providing valuable information concerning appliances' usage for energy management systems.