A grey-box temperature model for digital twins of building-integrated rooftop greenhouses

• Grey-box model predicts rooftop greenhouse temperature accurately. • Model integrates evapotranspiration effects from real sensor data. • Requires only 1-day data for 24-hour accurate temperature forecast. • Validated with real data from a greenhouse placed in Spain. • Suitable for real-time control in Digital Twin applications. Urban agriculture is increasingly recognized as a strategy for enhancing food security, sustainability, and climate resilience in cities. Building-Integrated Rooftop Greenhouses (BIRTGs) offer a compelling solution by combining food production with building energy synergies. However, accurately predicting indoor temperature in such systems remains challenging due to their low thermal inertia, highly variable heat transfer dynamics, and the complex, time-dependent thermal interactions introduced by crops. This study presents a grey-box temperature model tailored for BIRTGs and designed for integration into Digital Twin (DT) frameworks. The model captures heat exchanges with the outdoor environment, host building, solar radiation, ventilation and plant evapotranspiration. Unlike traditional models, the evapotranspiration component is calibrated using sensor data, bypassing the need for complex crop-specific inputs. The model was calibrated and validated using experimental data from a BIRTG at Universitat Politècnica de Catalunya, both with and without crops. Validation results showed high predictive accuracy, with RMSE values of 2.2 °C without plants and 2.0 °C with plants. Furthermore, a systematic evaluation of 70 dataset combinations revealed that a single day of calibration data is sufficient to accurately forecast temperature over a 24-hour period. These results highlight the suitability of the model for real-time climate control and automation in urban agriculture using DTs.