Evaluation of the physical interpretability of calibrated building model parameters

Syracuse, NY

Identifying building envelope thermal properties from the calibration of a lumped model raises identifiability issues. Not only needs the simplified model to be structurally identifiable, i.e. deliver unique estimates after calibration, but also the data used might not be informative enough to result in either or both accurate estimates and physically interpretable values. This could particularly be the case when data is extracted from non intrusive in situ measurements, in the sense not disturbing potential occupancy. In this frame, this paper develops a method to investigate the physical interpretation of the parameters of lumped models through a numerical tests procedure. Each test runs a simulation of a comprehensive thermal model of a building, with variations in thermal resistance properties of the envelope. Each simulation delivers data used to calibrate a lumped model. The parameters of the lumped model are then physically interpretable if their value vary according to the variations of the comprehensive model. The overall test procedure is applied to the study of a 2R2C model. Results show that the calibration of this model delivers robust calibration results for all parameters but one and also shows satisfactory robustness of the estimation of the overall thermal resistance. This paper concludes that the numerical test procedure does allow to evaluate practical identifiability of lumped models, and will in future work be used to examine more complex lumped models

https://surface.syr.edu/ibpc/2018/MS7/3