Conference Paper

Proceedings of Building Simulation 2019: 16th Conference of IBPSA

A Robust Approach For The Calibration of the Material Properties in an Existing Wall

Alessandro Prada ¹, Andrea Gasparella ², Paolo Baggio ^1
1 University of Trento, Italy
² Free University of Bozen/Bolzano, Italy

DOI: https://doi.org/10.26868/25222708.2019.210570
Abstract: Calibration of building energy simulation model is essential to obtain an accurate model able to emulate the actual energy behavior of existing buildings. There are two different approaches in the literature that are deterministic and Bayesian calibrations. The deterministic formulation of the calibration problem is ill-posed and consequently the pursuit of model discrepancy reduction may result in an over-fitting. On the contrary, Bayesian calibration assumes uncertain inputs and tries to reduce the uncertainty consistently with observed data but its implementation is the first stumbling block in its adoption. This paper describes a new framework for systematically adjusting the parameters of a building simulation model using a deterministic approach but considering also the a-priori information about the model parameters. The new calibration procedure is tested on the calibration of a finite difference scheme applied to the solution of the heat equation in a timber wall. The calibrated model is compared against other calibrated models obtained by different procedures. The results highlight a better performance of the model and especially a higher reliability in emulating the wall behavior in other independent set of measurements.
Keywords: calibration, regularization, building envelope, optimization
Pages: 4554 - 4561
Paper: