Artificial intelligence reduces energy consumption of buildings

In the coming years, Denmark will undergo a comprehensive green transformation. A political majority supports a 70 per cent carbon reduction by 2030, and Denmark has joined the Paris Agreement, the international climate agreement.

The solutions are still unknown, but energy efficiency measures and renovations have enormous potential. The heating of buildings represents 40 per cent of our collective energy consumption, and especially in older buildings there is a lot to be gained through renovation. More than 80 per cent of buildings are over 20 years old.

However, the current methods of planning and carrying out energy efficiency measures, including the methods used to select the relevant buildings, are a result of habitual thinking and lack of new technology, and they need an overhaul, believes Henrik Madsen, Professor and Manager of the Centre for IT-Intelligent Energy Systems (CITIES) at DTU Compute:

 “If we want to get our money’s worth of carbon savings, we should utilize artificial intelligence—it provides a form of X-ray vision, enabling you to look into the hidden layers of buildings and ‘see’ whether a building ‘performs’ well or not.”

Based on Smart City projects, DTU researchers know that data from frequent readings of heat meters provide evidence-based knowledge that indeed reflects reality.

Two identical houses are not identical
Today, energy renovations and energy labelling related to, e.g., house sales are based on simulation-based methods, in which the insulation thickness, walling structure, etc. are taken into account.

This is a problem, because two standard houses or two housing blocks may look identical on the construction drawing yet could be completely different in reality. Heating accounts can fluctuate enormously, both because the residents behave differently, and because of differences in the quality and execution of the construction. For example, concrete may accidentally have been dropped into the insulation in some buildings, which degrades the insulation, while that is not the case in other buildings.

“With our method, we can create an ‘X-ray vision’ where we can zoom in and see that some buildings are up to three times draughtier than others, even if the construction drawings are identical. This means that if we have DKK 100 million for energy renovation, we can focus our resources on areas where we will gain the most, such as window replacement, loft insulation, or draught sealing. Finally, the method can be used for an improved and digitized energy labelling scheme,” explains Henrik Madsen, who has helped develop the ‘X-ray method.’

Artificial intelligence chewing through data
The method consists of analyses of big data and artificial intelligence, where algorithms chew through frequently read data on heat consumption and frequent measurements of the outdoor climate. In the model, the energy consumption for heating is a function of weather data with sun, wind, and temperature. If the temperature is low, heat consumption is high—and vice versa.

“Weather phenomena also influence each other, e.g., wind and temperature. If the outdoor temperature is high, it does not matter if it’s windy, because the wind consists of hot air. But if it’s cold outside and it’s also windy, cold air will be blown into the building, resulting in a higher heat consumption. These influences have been incorporated as parameters in the model, and the same goes for how well-insulated a building is,” explains Henrik Madsen.

The ‘X-ray vision’ separates user behaviour and building performance in the equation, since behaviour influences heat consumption. It is assumed, among other things, that user behaviour is the same within a certain range, e.g., at two degrees above or below zero. If any additional consumption occurs in this context, it is therefore due to the climate screen, i.e., how well-insulated the building is.

Data-based management is the future
In developing the algorithm for the ‘X-ray vision’, DTU Compute draws on its section for dynamic systems where researchers are experts in analysing time series data, i.e., numbers that are registered every second, every hour, etc.

The researchers have been working on the method for several years, and it has been tested on terraced houses and housing blocks in Belgium and England. The method can be programmed into software that can be used on a PC or mobile phone, in principle enabling both energy consultants and ordinary citizens to use it. But the best solution would be to store the method in the cloud so that homeowners can monitor the energy consumption in their homes on a continuous basis.

With Henrik Madsen leading the project, CITIES has continuously attempted to get particularly the Danish Energy Agency to introduce the method. And now the times are on the ‘X-ray vision’s’ side, too.

By the end of 2020, for example, all households must have remotely read electricity meters, the data from which can be accessed by individual households on an hourly basis. This allows for the dissemination of the ‘X-ray’ method to the whole of Europe.

In the smart cities of the future, buildings should be heated at times when there is plenty of green solar and wind energy in the electricity grid. And this can be managed by using the new building performance models.

“This means that we will be able to carry out an efficient and accelerated green transformation through data-driven energy renovation. Ten years ago, you only had the construction drawings. Now we’re beginning to be able to use data for things other than the electricity and heating bills,” Henrik Madsen pointed out.