Smart Building Reference Architecture with Linked Data

Key words. Data integration, BMS, IoT, BIM, Semantic Web, Brick Schema

In the rapidly evolving world of building technology, the concept of smart buildings has moved from visionary prototypes to operational reality. Yet, as systems multiply from Building Management Systems (BMS) and Energy Management Systems (EMS) to IoT devices, BIM models, and digital twins, the challenge has shifted from data collection and data integration to system integration.

This article discusses a service-oriented and linked-data approach proposed in the Brains for Buildings’ Energy Systems project (https://brains4buildings.org/) that is used as the foundation for a reference architecture that enables interoperability, scalability, and innovation across the smart building ecosystem.

Why do smart buildings need a reference architecture?

Each building today is a network of systems, often designed and installed by different vendors using diverse data formats and communication protocols. These systems are installed without a common architecture. Data-driven applications such as digital twins, predictive maintenance, and model predictive control remain siloed and difficult to scale. A reference architecture acts as a reusable blueprint, defining how various systems, services, and data sources interact within a common framework.

Such an architecture must support both operational data (e.g., sensor readings, control signals) and contextual data (e.g., system configuration, spatial relationships, and component metadata). Linked data technologies, supported by ontologies like Brick (https://brickschema.org/), REC(https://www.realestatecore.io/), BOT(https://w3id.org/bot#), and SSN(https://www.w3.org/TR/vocab-ssn/), enable this by standardising the vocabulary used to express the components, as well as the relationships between devices, systems, and their data streams. Integration of data, however, also needs to be supported with the appropriate infrastructure, as the datasets are of different kinds and reside in different systems with different owners.

From fragmented systems to interoperable ecosystems

Traditional building automation systems were designed as closed environments, optimized for specific vendors or functions. Modern smart buildings, however, must interact with edge devices, cloud platforms, energy markets, and other similar systems. The proposed service-oriented architecture (SoA) approach addresses this challenge by decoupling components and allowing them to communicate through standardized APIs.

A service-oriented reference architecture for smart buildings includes seven categories of components: existing business applications, new microservice-based applications, databases, integration software, infrastructure services, shared services, and user interfaces [1]. A conceptual layout of the architecture is shown in Figure 1.

Figure 1. Conceptual layout of a service-oriented smart building architecture. [1]

The role of Linked Data

Linked data introduces a semantic layer that connects diverse data sources through ontologies and graph databases. Rather than relying on rigid data models, linked data describes relationships, allowing applications to query, combine, and reason across multiple domains. For example, a digital twin can link a sensor’s real-time CO₂ readings to its physical location in a BIM model and its equipment hierarchy in the BMS.

In practice, this is achieved through metadata integration services that transform existing contextual data into semantic triples; subject, predicate, and object, stored in a graph database. APIs then expose these relationships to other services using SPARQL queries, enabling interoperability across the building’s digital ecosystem. An illustration of how linked data connects disparate data systems is shown in Figure 2.

Figure 2. How linked data connects BIM, BMS, and IoT through semantic graphs.

Of course, in the common case where these data resides in multiple systems, simply following a data integration approach through linked data does not work, at least not in a modular or scalable fashion, nor in a secure way. Integration is needed at the system level (Figure 1), and the linked data approach thus needs to be considered spread over multiple systems as well.

Implementing the architecture in practice

Practical implementations of the proposed architecture with a Linked Data layer have been demonstrated across several living labs [2,3]. These case studies demonstrate how a shared service-oriented infrastructure facilitates semantic data exploration using open standards and reusable APIs, enabling digital twin applications and real-time building control applications based on Model Predictive Control (MPC).

For instance, the digital twin of the Atlas building at Eindhoven University of Technology (TUe) combines BIM geometry with historical and real-time sensor data, enabling facility managers to visualize environmental conditions within the 3D model. The backend system utilises semantic models for linking the timeseries data with BIM contextual data [2]. By storing time-series data in optimized databases, and metadata in RDF-based graph databases, the system ensures that each data type resides in its most efficient environment. This separation enhances scalability while maintaining the system's modularity and reusability.

A different application implemented the proposed architecture to create a smart charging controller based on MPC for buildings integrated with PV and EV systems [3]. As depicted in Figure 3, the MPC system was designed as a blend of multiple modular services, each configured via the building's semantic models. This approach allows the algorithms to be developed independently of the data modelling conventions used by various buildings.

Figure 4 shows an example of a SPARQL (https://www.w3.org/TR/sparql11-query/) query aimed at obtaining the necessary information about a dynamic data stream of a temperature sensor required by an algorithm. Here, the query is made using the vocabulary of the Brick and Ref schemas. This type of metadata configuration was done with an MPC-based energy demand optimisation algorithm for the Kropman – Breda office building where the controller aims to optimise the energy demand in an PV- and EV- integrated building microgrid [4].

Figure 3. Example implementation showing data flows across microservices and APIs for MPC system, [3]

Figure 4. An example query leading to metadata configuration within an algorithm.

Benefits for the smart building industry

Building owners gain from greater flexibility, easier system upgrades, and improved data governance. Another key advantage of adopting a linked data–based reference architecture is reusability. Once the data model and integration framework are defined, new applications such as fault detection, predictive control, or energy optimization can be added. This lowers development costs and accelerates innovation. Moreover, a modular, service-oriented infrastructure aligns well with the diverse ecosystem of building technologies. Vendors can develop independent microservices that plug into the common architecture, ensuring interoperability while preserving specialised client-specific services.

Conclusion

Standardized ontologies and linked data models are essential for creating a common digital language across different systems and buildings. The integration of semantic web technologies within a reference architecture provides a foundation not only for technical interoperability but also for collaboration among stakeholders from designers and facility managers to software developers and energy service providers. As smart buildings become active participants in wider energy and data ecosystems, linked data is expected to be one of the key enablers of improved inter-connectivity of systems.

Smart buildings are complex systems that depend on collaboration between diverse technologies and disciplines. A linked data–enabled, service-oriented reference architecture offers a practical path forward, transforming fragmented systems into interoperable ecosystems. By adopting open standards, semantic models, and modular microservices, the building industry can accelerate digital transformation while maintaining flexibility and vendor independence.

Acknowledgements

The Brains for Buildings’ Energy Systems project received funding from the Dutch Ministry of Economic Affairs and Climate Policy and Ministry of the Interior and Kingdom Relations under the MOOI program.

References

[1]     Chamari L, Petrova E, Pauwels P. An End-to-End Implementation of a Service-Oriented Architecture for Data-Driven Smart Buildings. IEEE Access. 2023:117261–81. Available from: https://ieeexplore.ieee.org/document/10287934/.

[2]     Chamari L, Petrova E, Pauwels P. A web-based approach to BMS, BIM and IoT integration: a case study. In: REHVA 14th HVAC World Congress. 2022. p. 1–8. Available from: https://proceedings.open.tudelft.nl/clima2022/article/view/228.

[3]     Chamari L, Walker S, Petrova E, Pauwels P. Towards portable model predictive control-based applications for demand side management in buildings. Energy Buildings. 2025;347(PA):116257. Available from: https://doi.org/10.1016/j.enbuild.2025.116257.

[4]     Chamari, L., Walker, S., Petrova, E., Pauwels, P. Energy Demand Optimisation in PV- and EV-Integrated Buildings: A Supervisory Controller Based on MPC. (Manuscript submitted for publication). Preprint at http://dx.doi.org/10.2139/ssrn.5565107.

Please find the complete list of references in the original article available at https://ieeexplore.ieee.org/document/10287934.