Modelling of Non-Residential Building Stocks Based on End-Use Consumptions and Bayesian Calibration: A Case Study of South Korea

Abstract

This paper introduces the development of reference building energy models to represent non-residential building stocks in South Korea. Eight representative stock models were developed: retail, restaurant, accommodation, hospital, non-public office, public office, and education (K-12 and university). In all models, historical revisions of the relevant codes/regulations are reflected in terms of building vintage. Each model was calibrated using a Bayesian technique and compared with end uses (heating, cooling, lighting, service water, electric load, etc.) derived based on national energy consumption statistics. It was shown that percentage differences between the stock models and national statistics in cooling, heating, and total energy consumption were roughly -4.2 - 2.3 %, -6 - 1.8 %, and -2.2- 0.8 % respectively. The lessons learned from the modeling processes are that not only sensitivity analysis (at the post processing), but also well-prepared national statistics are important, as it helps the modeler infer uncertain inputs at an initial stage, such as equipment or lighting power density. It can alleviate a burden of sensitivity analysis and calibration efforts.