Using semantic data models for active building operation: a proof-of-concept implementation in a real-life office building

Semantic data models are increasingly recognized as a key enabler for scalable, interoperable, and data-driven building operation. However, practical methodologies for integrating ontologies with live operational systems remain insufficiently demonstrated in real buildings. This paper presents a proof-of-concept implementation of a Brick-based knowledge graph deployed in the ZEB Laboratory, a highly instrumented office building in Trondheim, Norway. The proposed workflow covers structured data collection, graph construction, linkage to real-time measurements and integration with an LLM-supported reasoning layer for decision support. The approach is validated through a semi-automated use case. Elevated indoor CO₂ levels are detected, semantically contextualized through the knowledge graph, and translated into a supervised control recommendation targeting a motorized window actuator via MQTT and BACnet. By embedding both semantic relationships and operational metadata in the graph, the system enables interpretable, portable, and context-aware supervisory control. The results demonstrate how semantic modeling can bridge heterogeneous building data sources and support active building operation beyond static metadata management.