Digital twins have emerged as a promising concept for improving building energy efficiency, but their implementation faces challenges in interoperability and adaptability. This paper presents a large-scale field demonstration of an interoperable energy modeling framework for building digital twins, using ontology-based semantic models as data sources for automated model generation and calibration of data-driven component models. The study focuses on a single floor of a hospital building, comprising 12 conditioned zones and data from 45 measuring devices. Across the 45 sensors, the model achieved average mean absolute errors of 0.40∘C for temperature, 32 ppm for CO2 concentration, 0.06 for valve position, and 0.04 for damper position predictions. These results demonstrate the framework's ability to generate and calibrate accurate and flexible building energy models with reduced effort. The paper also showcases the framework's practical application in exploring system modifications to improve indoor comfort, highlighting its potential for scenario analysis and decision support. The proposed approach significantly streamlines the process of creating and maintaining accurate, up-to-date energy models, offering a robust foundation for digital twin applications in the built environment.