Object-oriented metamodelling for dynamic semantic validation of building systems

Buildings comprise complex systems, such as heating, ventilation, and air conditioning (HVAC), that work together to maintain occupants’ comfort and safety. These systems often consume and waste significant amounts of energy, even with building management systems (BMS) designed to be efficient. Researchers working to improve building energy efficiency need access to building system data to develop and deploy effective solutions. BMS provide APIs and representations of building systems and their data, but solutions deployed based on BMS representations are tightly coupled to specific BMS, limiting generalisability and scalability. Ontologies have been used to abstract building systems and their interactions, thereby decoupling solutions from specific BMS implementations. However, RDF/OWL-based ontologies lack built-in constraint enforcement and validation, leading to faulty models. They also struggle to model system behaviour and fail to support the seamless integration of control strategies due to their static RDF outputs. We propose MetamEnTh , an object-oriented metamodel for modelling the operational aspects of built environments. It utilises classes with predefined attributes (properties) and methods, as well as relationships, constraints, and validation rules, enabling behaviour modelling and the direct integration of control strategies through defined interfaces. We conducted a needs assessment to gather practical modelling requirements, followed by an empirical comparison with the Brick ontology and three expert-assessed use cases. Our evaluation indicates that MetamEnTh tends to produce more accurate models than ontology-based approaches (mean accuracy: 4.2 vs 3.9), although it introduces higher modelling complexity (mean difficulty: 3.6 vs 2.5). While these results show consistent improvement trends, not all differences are statistically significant. We also demonstrate how MetamEnTh supports seamless integration of control strategies through defined interfaces and implemented behaviours.