A probabilistic hot water usage model and simulator for use in residential energy management

Abstract

Water heating contributes up to 40% of a household’s total electricity usage and places a substantial burden on the electricity grid due to high power ratings and users’ largely simultaneous hot water usage. The main determinants of its electricity draw are physical properties such as set temperature, insulation, and plumbing configuration; environmental conditions such as ambient temperature and inlet temperature; and the hot water usage profiles. These profiles include the usage volumes, the times of usage and the outlet temperatures. The efficacy of energy management techniques that model water heaters and the accuracy of their simulation results therefore rely on representative hot water usage profiles. Existing models for household hot water usage neglect differences between users, and temporal variations such as the season and the day of the week, and are not fully autonomous. We propose a probabilistic data-driven model for modelling individualised hot water profiles and an accompanying hot water usage simulator that includes all these factors. We gathered data from 77 residential households over a period of one year to train and evaluate the model for all four seasons. The results show that the simulated hot water usage profiles match the statistical properties of the measured data. Moreover, the individual hot water usage modelling and the resulting aggregated energy load on the grid closely match the measured data, improving on the existing hot water usage by halving the modelling error.