Semantic digital twin for building performance analysis

During the Mad@Work projects, we developed a platform for building performance analysis. The solution is based on a semantic digital twin. Scaling machine-learning solutions in buildings is difficult since each building is unique and would require a lot of manual work to set up even simple machine-learning applications. Having accurate knowledge graphs of the building and its systems enables us to scale machine learning applications in buildings. Furthermore, a machine learning solution for estimating room occupancy was developed, and fault detection checks for common HVAC problems.

Platform supporting knowledge graphs. Enables storage of relationships between parts of the building system and their metadata. With easy queries, it is possible to get the right data in the right format for a particular machine-learning application. For example deep learning application for estimating room occupancy based on indoor air quality data.

Necessary to create a knowledge graph of the building and its systems

Building owners and facility managers

To be negotiated