BOSS - Building Energy Systems on causal reasoning
Short Description
Starting point / motivation
Buildings are responsible for 40% of energy consumption and 36% of CO₂ emissions in Europe. Optimization of building technology is necessary to achieve the goals of the European Green Deal. The actual energy consumption of buildings often deviates by up to 30% from the planned values due to operating errors and unforeseen user behavior.
Traditional methods for fault detection and diagnosis (FDD) require a great deal of manual effort and are unsuitable for widespread use due to a lack of scalability and explainability. The lack of availability and consistency of semantic building data makes it difficult to implement automated solutions.
Contents and goals
In the project, Causal AI methods are developed that derive semantic data based on time series on the one hand and are used for FDD applications on the other. Causal methods offer the advantage that they aim to identify cause-and-effect relationships instead of only statistical correlations, which increases the explainability of the models and significantly improves their performance and reliability compared to conventional approaches.
Methods and expected results
The consortium combines knowledge in the theoretical foundations of Causal AI with applied research in the field of AI for building and energy systems, complemented by the practical expertise of industry partners. Together, the consortium covers the entire innovation chain, from basic research to the implementation of market-ready technologies, and aims to create the foundation for scalable FDD solutions.
Project Partners
Project management
TU Wien, Intelligent Buildings and Systems Research Group
Project or cooperation partners
- Institute of Science and Technology Austria (ISTA)
- DiLT Analytics
- EAM Systems
Contact Address
Johannes Exenberger
Karlsplatz 13
A-1040 Vienna
E-mail: Johannes.exenberger@tuwien.ac.at
Web: github.com/tuw-isab