Abstract:
This paper reports on a case study of monitoring-based optimization-assisted calibration of a thermal
simulation model for an office building. Such a calibrated model could effectively support the operation of
the building. For example, it could be deployed toward diagnostics, fault detection, preventive maintenance,
and a model-based building systems control. To explore the potential of optimization-assisted calibration in a
realistic setting, we selected an actual office. This facility is equipped with a monitoring infrastructure, which
provides various streams of data about outdoor and building conditions. Our intention was to deploy data
obtained via the monitoring system to both populate the initial simulation model and to maintain its fidelity
through a systematic calibration process. The initial simulation model used, asides from static physical
building information, dynamic monitored data including electrical plug loads, occupancy, and state of
devices such as luminaires and windows. In the optimization-assisted calibration, a weighted cost function
was defined, which addressed the bias error between measured and simulated indoor temperature and the
goodness of fit of the model. A limited number of model input parameters were varied in the optimization
process toward minimizing the cost function. The resulting calibrated model showed noticeable accuracy
improvement and the optimization-assisted method displayed a promising potential as a systematic
calibration method in model-based predictive systems control.
