An approach for identifying sources of inadequacy and upgrades in models with high-dimensional outputs and boundary conditions

TL;DR

This paper proposes a Bayesian calibration-based approach to identify model inadequacies and guide upgrades in high-dimensional output models, demonstrated on building energy models for improved accuracy and extrapolation.

Contribution

It introduces a robust method for model improvement by analyzing discrepancies and comparing variants using likelihood criteria, enhancing current practices.

Findings

Effective identification of model inadequacies.

Improved model upgrades based on discrepancy analysis.

Validated approach on building energy models.

Abstract

The construction of computer models (mathematical models implemented in computer codes), with respect to observed phenomena, is usually undertaken by building different variants depending on modeller sensibility, and choosing the one yielding the best fit of the field data, according to Root Mean Squared Error (RMSE) based measures. Usually a particular model is chosen because of its marginally lower RMSE, and not because of its actual higher adequacy, risking that its capability of extrapolating predictions is poor. This work aims at improving the current practice in the creation of computer models by proposing an approach similar to those employed in statistical modelling, wherein starting from the simplest hypothesis, effective model upgrades are identified by analysing discrepancies between observations and predictions, and different model variants are compared according to robust likelihood based criteria. The method, focused on models with high dimensional outputs and boundary conditions and centred on Bayesian calibration, is demonstrated on numerical experiments considering a series of building energy models. The object of the modelling is a test facility used for round robin tests in the context of the International Energy Agency (IEA), Energy Building and Communities (EBC), Annex 58.