[Back]

M. Taheri:
"Optimization-Based Calibration of a Building Thermal Performance Simulation Model";
Supervisor: A. Mahdavi; Abteilung Bauphysik und Bauökologie, Institut für Architekturwissenschaften, 2013; final examination: 2013-03-20.

Building performance simulation is being increasingly deployed beyond the buidling deisgn phase and into the building operation phase. Specifically, the predictive feature of the simulation-assisted building systems control strategy provides distinct advantages in view of building systems with high latency and inertia. Needless to say, such advantages could be exploited only if model predicitons could be relied upon. Hence, it is essentially important to calibrate simulation models based on monitored data. As such, whole-building simulation applications require extensive input data to accurately model the thermal performance. It would be thus beneficial to conduct the model calibration in an efficient manner. In the optimization-aided calibration approach, some input parameters of the model are adjusted through an optimization process so that the difference between the model outputs and the monitored data is minimized.

This master thesis reports on the use of optimization-aided model calibration in the context of an existin university building. Thereby, the main objective was to deploy data obtained via the monitoring system to both populate the initial simulation model and to maintain its fidelity through an ongoing optimization-based calibration process. The initial simulation model uses, asides from basic physical building information (geometry, layout, materials, etc.), monitoring data to define assumptions pertaining occupancy, state of devices such as luminaires and windows, and energy output of heating terminals. By doing so, one of the sources of inaccuracy in building model can be addressed. The calibration and validation process will be performed in different summer and winter conditions in order to analyze the building model accuracy in different environmental conditions. To judge the quality of the implemented calibration, the model predictions were examined using long-term monitored data. The results suggest that the calibration can significantly and sustainably improve the predictive performance of the thermal simulation model.

[English Version] Building performance simulation is being increasingly deployed beyond the buidling deisgn phase and into the building operation phase. Specifically, the predictive feature of the simulation-assisted building systems control strategy provides distinct advantages in view of building systems with high latency and inertia. Needless to say, such advantages could be exploited only if model predicitons could be relied upon. Hence, it is essentially important to calibrate simulation models based on monitored data. As such, whole-building simulation applications require extensive input data to accurately model the thermal performance. It would be thus beneficial to conduct the model calibration in an efficient manner. In the optimization-aided calibration approach, some input parameters of the model are adjusted through an optimization process so that the difference between the model outputs and the monitored data is minimized.

This master thesis reports on the use of optimization-aided model calibration in the context of an existin university building. Thereby, the main objective was to deploy data obtained via the monitoring system to both populate the initial simulation model and to maintain its fidelity through an ongoing optimization-based calibration process. The initial simulation model uses, asides from basic physical building information (geometry, layout, materials, etc.), monitoring data to define assumptions pertaining occupancy, state of devices such as luminaires and windows, and energy output of heating terminals. By doing so, one of the sources of inaccuracy in building model can be addressed. The calibration and validation process will be performed in different summer and winter conditions in order to analyze the building model accuracy in different environmental conditions. To judge the quality of the implemented calibration, the model predictions were examined using long-term monitored data. The results suggest that the calibration can significantly and sustainably improve the predictive performance of the thermal simulation model.