TO-22-C001 - Development and Application of BIR-BEM: A Bayesian Inference R Platform for Building Energy Model Calibration

The building sector accounts for nearly 40% of global energy consumption and plays a critical role in societal energy securityand sustainability. A building energy model (BEM) simulates complex building physics and provides insights into theperformance of various energy-saving measures. The analysis based on BEMs has thus become an essential approach toslowing down the process of increasing building energy consumption. The reliability and accuracy of BEMs have a high impacton decision-making. However, how to calibrate a building energy model has remained a challenge. Existing calibrations areoften deterministic without uncertainties quantified. In this study, a new automated multi-module calibration platform, BIRBEM(Bayesian Inference on R for Building Energy Model), is developed using an R programming language for calibratingbuilding energy models. The sensitivity analysis module determines the calibration parameters, and the building energy modelis replaced by the developed meta-model module for the Markov Chain Monte Carlo (MCMC) process to save computing time.An application of a high-rise residential building case in a hot and arid climate was demonstrated. The coefficient of variationwith a root-mean-square error (CVRMSE) value of the monthly total cooling energy consumption is 13.95%, which satisfiesthe monthly calibration tolerance of 15% required by ASHRAE Guideline 14.