"At NEST we were able to show that our algorithm can save 25 percent energy"
Empa researchers analyzed how a building can learn to save energy.
They used NEST as a research platform to test and further develop an innovative, self-learning heating control system under real conditions.
The result: the "smart" control system was able to evaluate the behavior of the building better and act with foresight. This not only saved energy, but also achieved a higher level of comfort.
Factories, airport terminals or office buildings are often already equipped with automated "predictive" heating systems. These work with pre-programmed scenarios calculated specifically for the building, thus saving the operators a lot of heating energy. However, such individual programming is too expensive for individual apartments and private houses.
In summer 2019, a group of Empa researchers succeeded for the first time in proving that there is a simpler way: the intelligent heating and cooling control system does not necessarily have to be programmed; it can easily learn how to save energy from the data of previous weeks and months.
Experiment at NEST
The decisive experiment took place at NEST. "Since the NEST unit Urban Mining and Recycling is inhabited, it offered us the ideal, real-life environment", says Benjamin Huber, Research Associate ehub. The layout of the apartment is also optimal: A large kitchen is symmetrically framed by two student rooms. Both are 18 square meters each. The entire window front faces east – towards the morning sun. In the UMAR unit, water runs through a stainless steel ceiling paneling and provides the desired room temperature. The heating and cooling capacity can be calculated for each room using the respective valve position.
Smarter cooling – thanks to weather forecast
Since project manager Felix Bünning and Benjamin Huber did not want to wait for the heating period, they started a cooling experiment already in June 2019. The aim was to keep the temperature in the two bedrooms below 25 degrees during the day and 23 degrees at night. A conventional thermostatic valve provided cooling in one room, while the experimental control system designed by Bünning, Huber and their team was operating in the other. The artificial intelligence had been fed with data from the last ten months - and it knew the current weather forecast from MeteoSwiss.
More comfort with ¼ less energy
The result was very significant: The intelligent heating and cooling control complied much more precisely with the comfort specifications and required around ¼ less energy to do so. This was mainly due to the fact that in the morning, as the sun started shining through the windows, the intelligent system started carrying out the cooling with foresight. In comparison: The mechanical thermostat in the other room only reacted once the temperature was already too high – too late, too hectically and at full power.
Field test in an apartment building
Thanks to the successful tests at NEST, Felix Bünning and Benjamin Huber are now able to put their innovation to the test in a larger field experiment. To this end, they will equip four of a total of 60 flats in an apartment building with the intelligent heating and cooling control system. But before that, a further developmental step is necessary: the software, which was previously running in the cloud, must be implemented on a small hardware control unit. "We will then replace the room thermostat in the test apartments with this control unit," says Felix Bünning. He is already looking forward to the results: "I believe that new controllers based on machine learning are a huge opportunity. Using this method, we can construct a good, energy-saving upgrade solution for existing heating systems using relatively simple means and the data collected."