The increasing digitalization of energyintensive infrastructures opens up new opportunities for the datadriven optimization of operational processes with the aim of improving energy efficiency and reducing emissions. In building automation in particular, there is considerable potential for optimizing heating, ventilation, airconditioning, and lighting systems, whose efficient operation depends crucially on the availability of consistent and contextualized data. While Building Information Modeling (BIM), based on the Industry Foundation Classes (IFC) standard, provides a structured description of static building properties, the Asset Administration Shell (AAS), originating from the Industry 4.0 domain, supports the integration of dynamic states and executable functions. Existing approaches for deriving AAS submodels from IFC data, however, are limited to structural information and do not take runtime data into account. This paper extends this approach by introducing a runtimecapable architecture for the integration of live data. An AAS gateway serves as the runtime environment and enables a bidirectional coupling between digital twins and physical assets. The connection is realized via an MQTT-based communication layer with KNX integration. The semantic linkage between structural models and runtime data is implemented through the concepts of Asset Interface Description (AID) and Asset Interface Mapping Configuration (AIMC). The applicability of the approach is demonstrated using the use case of constant illuminance control in accordance with VDI 3814 Part 3.1. The resulting digital twin integrates planning and operational data and provides the foundation for the optimization of energyintensive infrastructures.