Multi-stage multi-level calibration of a school building energy model

The calibration of the building input parameters is the process aimed at minimizing the difference between actual and simulated performance. It is of paramount importance to implement a reliable model of an existing building, as this enables the study of its behaviour and the evaluation of improvement actions. However, when the number of unknown or uncertain parameters (such as thermo-physical properties of components and materials, infiltration and ventilation rates, internal thermal capacitances, system characteristics, etc.) is large, manual calibration methods require unacceptably long trial-and-error cycles and do not always ensure a significant improvement, as the complexity of the simulation increases. This paper explores the potential of calibrating an entire building simulation model by means of a stepwise approach and automated calibration of the model (optimization-based calibration). The approach is multi-stage since it considers different reference periods in order to calibrate different parameters, and multi-level as it starts from a room level, in order to apply the calibrated parameters to the entire building, and perform calibration to refine the estimation of the missing parameters. The described approach is shown to be effective in reducing the number of initial unknown inputs at each step as well as in validating the previous calibration results when moving to the multi-zone level. The application of the proposed calibration method to a case study aims at demonstrating the details of its implementation and its efficacy, using the available limited number of measurement sensors and short observation periods.