Bayesian Calibration and Uncertainty Quantification for TRACE Based on PSBT Benchmark

摘要

This work applies the modular Bayesian calibration approach to the TRACE physical model parameters based on steady-state PSBT benchmark. The major uncertainty considered in this work is parametric uncertainty from physical model parameters that may not be measured accurately. The posterior uncertainty distributions for those parameters are calculated via the modular Bayesian calibration approach, and are then propagated through the computer code to obtain the uncertainty distribution of model responses. The comparison of the posteriors of physical model parameters obtained by different approaches shows the necessity of considering model discrepancy in Bayesian calibration. Future work will include sensitivity analysis on several model assumptions in the modular Bayesian approach and its application to high dimensional output.