Anticipatory heating systems are a relatively recent change in residences — houses and apartments — because they are not cost-effective. However, large buildings where hundreds or even thousands of people gather to work, play, or study have benefited from this technology for some time. These systems work because they can be programmed to provide heat only when people are expected to be in the building. Furthermore, they can switch to cooling when seasons change. But these systems are about to become even smarter thanks to machine learning.
Current automated systems require someone to program them and make manual changes when necessary. This could be the days during the twice-yearly time changes. Future systems will be able to program themselves after “studying” previous data to provide more affordable heating and cooling, according to one study. When you remove the need for programmers to control the system, it becomes much easier to implement in residences.
A group of researchers has performed one current experiment from EMPA, the Swiss Federal Laboratories for Materials Science and Technology, fed an automated system data from the past ten months as well as upcoming weather forecasts and compared the results to a thermostatic valve. Both were set to maintain a comfortable temperature.
The smart setup was able to adhere to comfortable temperatures more closely and to use about 25% less energy during the warm months. How was this possible? The AI anticipated the need to cool a room before it became uncomfortably warm by anticipating the weather changes throughout the day, requiring less energy than the traditional room, which would begin to cool the area once it had reached a high temperature.
Researchers repeated the experiment as the weather cooled, and the rooms needed to be heated. The team lead was confident that the cool-weather test would be just as successful as the one that took place during warmer weather. But the technology is not quite ready to go to market.
The experiment will be replicated on a larger scale — a building containing 60 apartments. Scientists will outfit four of the units with their self-learning systems to learn how they perform in real-world situations. Should the experiment prove effective, it may be worthwhile to retrofit existing residences with this technology.
Many electricity markets have been deregulated and have successfully integrated renewable energy. Market pricing for electricity provide a clear signal when there is energy abundance, i.e. during daytime when solar panels are generating more electricity than might be needed. Heating, cooling, chillers, hot water tanks, boilers and of course batteries can instantly turn excess electricity into thermal energy and store it.