GeoAI4Retrofit

Background

In order to meet the climate targets set by the German government and the Paris Agreement, Germany’s building stock must become climate-neutral by 2045. In recent years, the construction and property sector has focused primarily on energy-efficient new-build residential properties. The focus is now shifting towards the refurbishment of the existing building stock, with the aim of achieving the highest possible energy efficiency in the refurbished buildings. If we compare new construction with refurbishment and also take into account the embodied carbon emissions from the production of building materials in the life-cycle analysis, refurbishment generates only around 50% of the carbon footprint of a new build.

Managers of large property portfolios are seeking efficient and, where possible, automated solutions that enable them to take stock of their portfolios and thus gain an overview of their renovation status. According to Dena, the diverse non-residential building stock in Germany also requires solutions to harness the decarbonisation potential in this sector.

Objectives

In the non-residential building sector, there has been a complete lack of automated solutions for assessing renovation potential to date. This is primarily due to a lack of reliable information on the condition of non-residential buildings and the resulting inability to classify them into age and renovation condition categories. This makes it very difficult to assess renovation potential. To close this knowledge gap, the GeoAI4Retrofit project is developing an automated, AI-based solution that automatically analyses the renovation potential of entire property portfolios and identifies suitable renovation options.

Approach

The role of the IOER specifically encompasses the sub-area of predicting building age class, building typology and renovation condition through the use of modern AI algorithms from the fields of computer vision and GeoAI. A novel approach is being developed that, for the first time, links image and geodata to achieve the highest possible classification accuracy for the complex age categories and building types. Image data from the ENOB:dataNWG project plays a central role here, serving as the primary data basis for model development. Similarly, 3D building models are of great importance to the process.

Results

The models developed serve as a key foundation for the building and portfolio analysis system. An automated potential analysis of property portfolios or buildings at a company site can thus be carried out in a matter of hours, and for the first time, non-residential buildings in particular can be analysed and appropriate refurbishment options identified.