Bayesian-Emulator based parameter identification for calibrating energy models for existing buildings

In this paper, an efficient Bayesian-Emulator approach is developed and applied to identify unknown input parameters for detailed energy models for existing buildings. The new approach improves on the traditional Bayesian approach for parameter identification, characterized by high computing requirements. Specifically, the developed approach utilizes a building energy model emulator to expedite the parameter identification process. Both discrete and continuous building energy model parameters can be identified using the developed Bayesian-Emulator approach. In the paper, the benefits of the developed Bayesian-Emulator parameter identification approach to calibrate whole-building energy models are discussed. Moreover, the proposed approach has been applied to a case study of an existing medium office building using monthly utility data. It is found that the Bayesian-Emulator is capable to calibrate the medium office building energy model in less than 1 minute using a personal computer. The Bayesian-Emulator approach presented in this paper can be an effective framework to automatically calibrate energy models for existing buildings.