Exploring the possibility of calibrating a whole-building model from the short-term monitoring of selected reference rooms

Accurate simulation models of existing buildings provide a reliable picture of building and system behaviour, useful for diagnostic purposes, for designing retrofit interventions or improving the control strategies. Reliable building models can be obtained through monitoring and calibration, aiming at minimizing the discrepancy between predicted and actual performance by fine-tuning the values of the simulation parameters. Analytical calibration methods may become a complex and time expensive process, especially when a large number of parameters has to be estimated as in the case of the largest buildings. In addition, overfitting issues can undermine the reliability of the calibration process. This work explores the possibility of carrying out a model calibration based on low-cost and short-term measurements in order to avoid overfitting issues. In particular, the proposed approach is based on the selection of representative spaces in the buildings, and the identification of multiple monitoring periods during which only a subset of building parameters needs to be calibrated at a time. This calibration approach is extended to the entire building in a multi-stage and multi-level approach