Paper Conference

Proceedings of BSO Conference 2020: Fifth Conference of IBPSA-England

Back

---

Parameters identification of grey-box building energy model through Bayesian calibration

Victor Marty-Jourjon, Thomas Berthou, Pascal Stabat

Abstract: This study discusses the physical interpretability of the grey-box models parameters. This physical interpretability allows robust calibration of building models parameters with Bayesian methods, and could help to assess scenario of retrofitting, and detect building energy drift. However, since these models simplify the physical complexity, the link between the physical characteristics and the model parameters is not straightforward, and can significantly vary with structure of the RC model. Five RC models have been selected and calibrated using a Bayesian approach with a training data set issued from a typical office building simulated with TRNSYS. The calibrated parameters are compared to the theoretical physical values modeled in TRNSYS. The results discuss the parameters of thermal resistance and suggest that at this stage, only the global resistance can be identified. Issues of correlation between the parameters are highlighted and a set of robust models is proposed.
Pages: 253 - 260
Paper:
bso2020_Jourjon