The “million” housing program built in the sixties and seventies is now in urgent need of renovation and energy savings. By adding machine learning to existing property data, the aim is to identify patterns that show which measures needs to be taken and where. In the long term, the findings will contribute to the government's goal of Sweden as a fossil-free welfare country in 2045.
According to the report "Renoveringskompetens" (2018) there are currently 800,000 apartments in Sweden that are in need of renovation. The main reasons why renovations have been postponed are lack of competence and lack of profitability for the property owner. In the 1960’s, when the buildings were built, there were low knowledge and understanding of how different materials and substances can affect our health, making renovation of the buildings harmful to our health.
Using available data
Sweden has an extensive database that covers housing stock (including the million program). By adding machine learning to this data, priorities for renovation may be identified.
"There is a lot of information about buildings, such as construction year, ownership form, heating system and building type. But there is also information that is important in energy efficiency, which is not found in any registers and which often relates to the type of building, such as cold winds and facade materials. We will develop that information and combine it with existing information, explains Claes Sandels," researcher at RISE.
"Through machine learning, we can find connections in the housing stock that are not yet described. We will thus create new information that can lead to great benefit when deciding on improvement measures," continues Claes Sandels.
With the use of artificial intelligence (AI) and machine learning, both known and unknown patterns in the buildings can be identified. Patterns that in turn can provide better cost estimates for energy efficiency, and hopefully be linked to a prioritized action plan.
Known and unknown patterns
The AI tools will identify different categories of buildings by combining building-specific information from national databases together with analysis of images on buildings.
"Through this, we can get an idea of what renovation concepts are suitable for which buildings and which instruments can help property owners to renovate their buildings in an efficient manner," says Kristina Mjörnell, Business and Innovation Area Manager at RISE.
In addition to creating a new basis for authorities, there is also a hope that the machine learning methods can be applied in fields other than energy in building stock research. A major problem area in the building stock is the presence of various hazardous substances. It becomes expensive, cumbersome and even dangerous when you encounter hazardous substances during renovations. The hope is to use the same methods for identifying the renovation potential to also be able to predict risks.
In 2019, the National Board of Housing, Building and Planning and the Swedish Energy Agency shall produce a proposal for an updated national renovation strategy for Sweden. The strategy will include a roadmap that ensures a high degree of energy efficiency and the phasing out of fossil fuels, and in the long term contribute to the government's goal of fossil-free energy 2045.