Component model calibration using typical AHU data for improved prediction of daily heat source energy consumption

Model predictive control (MPC) has the potential to reduce energy consumption during building operations by determining an optimal operating strategy in advance. However, the data acquired from operational buildings are insufficient to fit the models, leading to difficulties in calibrating the entire target system. Therefore, this study evaluates the impact of local calibration on the performance of heat source energy consumption prediction models using only AHU data typically collected in an operational building. This load-side calibrated model showed an accuracy of within 30% of the CVRMSE. Daily heat source consumption was predicted using the proposed model, which demonstrated superior explanatory power (R2 ≈ 0.95) compared to the initial uncalibrated model. Therefore, such a locally calibrated model developed with insufficient data has the potential to be used for MPC without installing additional sensors. •A full HVAC model was developed using typical AHU data.•Partial calibration based on gray-box models and full system modeling are combined.•Heating consumption prediction for source systems is improved by the proposed method.•The proposed model demonstrated a high coefficient of determination (R2 ≈ 0.95).