Linked Geodata: Improving Rooftop Photovoltaic Production Estimates through BIM-GIS Integration using Semantic Web Technologies

Sammanfattning

Accurately predicting rooftop photovoltaic (PV) electricity output is crucial for sustainable urban energy planning, yet current workflows that rely on 3D city models in CityGML Level of Detail 2 (LOD2) tend to inflate roof areas and, consequently, PV potential. This study develops and evaluates a proof-of-concept framework that integrates detailed Building Information Modelling (BIM) data in Industry Foundation Classes (IFC) format, with Geographic Information Systems (GIS) based 3D city models in CityGML format. It uses Semantic Web technologies to achieve interoperability. The approach enriches a CityGML knowledge graph (KG) by transferring precise roof area value from an IFC KG. The CityGML model for the test building, KTH Undervisningshuset, Stockholm, is stored in a relational database (3DCityDB), exposed as a virtual KG using R2RML mappings with Ontop plug-in in Protégé, and then materialized. The IFC model is converted to an IFC KG with the IFCtoRDF tool. A custom alignment ontology links equivalent IFC and CityGML classes and properties, enabling inference and the transfer of roof area data to the CityGML KG through SPARQL queries executed in GraphDB graph database. Using the IFC-derived roof area within the enriched CityGML KG reduces the estimated yearly PV electricity production by 16% (28.7 MWh/year) compared with the native CityGML LOD2 value, thereby eliminating the overestimation caused by geometric generalization. Benchmarking against the Photovoltaic Geographic Information System (PVGIS) shows that the knowledge graph-based estimate is 27% lower than the CityGML result and 12% lower than the IFC result. Although demonstrated on a single building and older schema versions (IFC2X3 TC1 and CityGML 2.0), the findings indicated that the method could reduce errors in PV estimates and support better decision-making in urban energy planning. As more IFC models become available, the workflow can be scaled at neighborhood and city levels, making Semantic Web technologies a viable alternative for BIM-GIS integration.