Unsupervised Disaggregation of Water Heater Load from Smart Meter Data Processing

In the residential sector, electric water heaters are appliances with a relatively high power consumption and a significant thermal inertia, which is particularly suitable for Demand Response schemes. The success of efficient DR schemes via the control of water heaters presupposes an accurate estimate of their power demand at each instant. Although the load of water heaters is rarely directly measured, a large penetration of Smart Meters (SMs) in distribution grids enables to indirectly infer this information on a large scale via load disaggregation. For that purpose, a considerable number of Non-Intrusive Load Monitoring (NILM) approaches are suggested in the literature. However, they require data streams at a time resolution in the range of one second or higher, which is not realistic for standard SMs. Hence, this paper proposes an unsupervised approach to detect and disaggregate the load profile of water heaters from standard SM data with a time resolution in the minute range. Evaluated on multiple real loads with sub-metering, the proposed approach achieves a Normalized Mean Absolute Error (NMAE) lower than 2% and a precision generally higher than 92% with time resolutions between 5 and 15 minutes.