| Title |
Whole Building Energy Simulation using Bayesian Stochastic Calibration |
| Authors |
Lee, Dong-Hyun ; Kim, Young-Jin ; Park, Cheol-Soo ; Kim, In-Han |
| Keywords |
Building Simulation ; Bayesian Calibration ; Markov Chain Monte Carlo ; Uncertainty Analysis |
| Abstract |
Building simulation has become increasingly important in assessing potential energy savings in buildings. It has been widely acknowledged that many inputs are under strong uncertainty and this causes significant differences between the simulation prediction and the reality. For calibrating the model, there are generally three approaches: manual (trial and error), deterministic, stochastic. This paper reports the last approach so called Bayesian calibration technique. The technique has been widely accepted in other domains as a powerful tool to estimate the posterior distribution of uncertain inputs based on the measured outputs. In this study, an building was selected and unknown parameters were identified. The Bayesian calibration was conducted in four steps: (1) determination of prior probability distributions for uncertain parameters. (2) Markov Chain Monte Carlo method for estimating posterior distributions, (3) validation of the model. It is concluded that the Bayesian calibration can be successfully used to improve accuracy of simulation prediction and reduce uncertainty of the model. |