Nonintrusive power estimation of residential appliances under voltage variation

Nonintrusive Load Monitoring (NILM) systems aim to estimate the individual appliance operation and consumption in a household from measurements at a single point, in order to motivate energy conservation behaviors. Power estimation based on appliance rated powers is not always accurate because of the voltage supply variations, thus misleading load disaggregation algorithms. In this paper we propose models of power as a function of voltage for correcting the power estimation. Measurements in the laboratory of residential loads are performed under multiple voltage levels. These data are processed to find regression models. Moreover, some analytic expressions to calculate the powers are proposed. Proposed models result more accurate than the rated powers and other models in the scientific literature.