Hardware Testing of Electric Hot Water Heaters Providing Energy Storage and Demand Response through Model Predictive Control

Abstract

This paper presents results from hardware testing which demonstrate that, 1) systems of water heaters under Model Predictive Control can be reliably dispatched to deliver set-point levels of power to within 2% error at very short timescales with minimal sensing requirements, and 2) a classical steady state model commonly used for simulation of electric hot water heaters can be inaccurate vs. results obtained on hardware. These results improve upon the current state of knowledge and show a promising pathway to control hot water heaters as energy storage systems capable of delivering flexible capacity and fast acting ancillary services on a firm basis. These energy products are shown to be deliverable without compromising the availability of hot water at the residence, even in control implementations which do not have sensors to monitor actual water use for the predictive optimization.