NEWS
HOT WATER, COOL GRID
Demand-response initiatives for hot water heaters might not appeal to consumers if their comfort is compromised. A doctoral student at Purdue University has found an effective solution to the problem.
Written by Poornima Apte
CONSUMERS SPEND HUNDREDS OF DOLLARS ANNUALLY on water heating, making it the second largest energy expense in the average home. Saving even a small fraction would benefit consumers and reduce strain on energy infrastructure.
Levi Reyes Premer, a doctoral student at Purdue University, has developed a patent-pending algorithm for an intelligent demand response strategy that targets heat pump water heaters.
Preserving consumer comfort
Demand-response programs ease strain on the electric grid by encouraging customers to move energy-intensive tasks to off-peak hours. The problem with the basic version of such incentives is that they simply shift energy consumption away from peak hours without considering homeowners’ comfort.
Premer’s algorithm delivers an intelligent version of demand sensing. By recording when customers shower or use hot water, it creates a consumption pattern. Using historical data as a guide, the algorithm can predict future demand and keep the water hot closer to when it’s likely to be needed instead of having it ready to go all day and night.
Delivering hot water when it’s needed while also maintaining a demand response strategy is a win-win: Integrating with the utility to shift energy consumption loads works in such cases because the program is able to maintain a level of comfort. And consumers are more willing to participate in energy-saving initiatives if they save money without compromising comfort. Demand-response incentives usually save between 10–15 percent on energy bills.
“This work helps provide a pathway to reap benefits of heat pumps while still providing value toward the grid by shifting loads and making sure they’re within their thresholds of how they want to operate,” Premer said.
Temperature readings from the tank and inlet water line are first processed by the water usage estimation algorithm, which converts them into an estimated usage profile. This profile is then passed to the forecasting algorithm to predict water usage over the next 24 hours. The resulting forecast informs the predictive controller, which optimizes heating decisions based on grid signals and anticipated usage patterns. Image: Levi Rayes Premer/Purdue
How the algorithm works
Kevin Kircher, assistant professor of mechanical engineering at Purdue University, who is Premer’s advisor, breaks down how the process works.
Traditionally, a flow sensor gives information about how much hot water is flowing out of the tank at any given moment. But adding physical sensors can be expensive and disruptive. Instead, one can estimate hot water flow by using less expensive temperature sensors to track water temperatures.
When tank water temperatures drop, we know that hot water has been withdrawn. To estimate how much hot water was withdrawn, Premer uses a mathematical model of the tank’s energy dynamics:
Mass of hot water withdrawn = (Old tank water temperature - New tank water temperature) / (Old tank water temperature - Inlet water temperature).
Water heaters already measure most of these values; the only unknown is the inlet water temperature of the municipal supply. Since the municipal water pipes serve the entire neighborhood, measuring the inlet water temperature anywhere gives a pretty good estimate of the inlet water temperature everywhere. Premer uses a virtual sensor to measure water flow by using temperature measurements at the inlet lines as a proxy.
Comparison of estimated (blue) and measured (orange) water draws at one-minute intervals for a heat-pump water heater. The estimation closely matches real usage, accurately capturing event timing and peak magnitudes. Minor deviations arise from model simplifications and the thermal lag of the temperature sensors. Because the sensors are mounted on the exterior surface of the metal tank, the measured temperatures reflect both conduction and mixing effects, introducing delay relative to the actual water temperature changes. Overall, the estimation model achieves a monthly total volume error of −2.01 percent, underestimating water use by just more than one gallon. Image: Levi Rayes Premer/Purdue
Kircher pointed out that the algorithm is adaptive. “This means it’s continuously relearning the showering and water usage patterns based on recent history,” he said. A rental unit, for example, might house people who liked to shower in the mornings and train on such patterns. But it’s able to adapt when someone else moves in with maybe an evening shower pattern.
Even assuming that the water heater needs to be fully heated just before the showers start to ensure comfort “still leaves a lot of flexibility in terms of when the heating of the water needs to be done,” Kircher said. If the shower were set for 8 a.m., the heater could either aggressively heat the water from 7–8 a.m. or run at a much lower power from midnight to 7 a.m., and potentially reduce strain on the grid.
“This work helps provide a pathway to reap benefits of heat pumps while still providing value toward the grid by shifting loads and making sure they’re within their thresholds of how they want to operate.”
—Levi Reyes Premer, Doctoral Student at Purdue University
Electric resistance and heat pump water heaters
Two common types of water heaters include electric water heaters and heat pump heaters. Electric water heaters function by flowing current through heating elements which get hot and heat the water. Heat pump heaters use vapor compression to run hot gas through, which gets compressed, evaporates, and condenses in a cycle. The hot gas heats the water. The U.S. Department of Energy mandates that by 2029, most electric storage water heaters, particularly those over 35 gallons, will run on heat pump technology.
Most heat pump water heaters today are more efficient than the electric resistance equivalents, but they still have electric resistance heating elements for backup in case of malfunction or increased demand. These heating elements use a lot of current and have to plug into 240-volt outlets. Not all houses have a 240-volt outlet or circuit available, so installing a heat pump water heater for a house that has been using gas water heating could add $1,000 to the cost of an approximately $2,500 heater.
One promising application of Premer’s virtual sensing AI algorithm is that it could potentially allow a heat pump water heater to outgrow the backup electric resistance heating elements and function on its own.
“If you’re confident that just the heat pump by itself can keep people comfortable, then you don’t need the heating elements and you don’t need the 240-volt connection, and that makes the transition to the heat pump heater much more economically achievable for a lot of households,” Kircher said.
Poornima Apte is a technology writer based in Walpole, Mass.
© 2025 The American Society of Mechanical Engineers. All rights reserved.