A Simplified Bayesian Approach for The Calibration of District-Building Energy Models

Abstract: Bayesian optimization with surrogate modeling is widely used to calibrate building energy models. However, complexities arise in surrogate modeling due to the variability in building morphology at the urban scale. Thus, maintaining dynamic simulation accuracy is crucial. This study presents a novel optimization framework for calibrating district-building energy models using Bayesian decision theory. Once tested on a case study district, the approach reduces monthly calibration error by approximately 45%. Future works could be employing more robust classifiers and handling imbalanced target variables. The proposed approach can minimize computational demands for optimizing dynamic models while ensuring reliability.