확률적 에뮬레이터를 이용한 외피 시스템 최적설계

Young Jin Kim, 안기언, 박철수 · 한국건축친환경설비학회 학술발표대회 논문집 · 2012

This paper addresses development of a simple and quick prediction model based on the Gaussian Stochastic Process (GASP) with a far less computation time than the dynamic whole-building simulation tools. The emulator model, which can be regarded as a surrogate model of the dynamic BPS tools, is developed using Gaussian process regression model and Bayesian approach. In this study, EnergyPlus was used to derive the emulator model. In the paper, the authors exemplify the use of the emulator for optimal design of glazing system for a given office building. In general, it requires a significant simulation run-time for stochastic optimization. In this paper, it was presented coupling of the Gaussian process regression model and an optimization routine out of MATLAB optimization toolbox. It was reported that the emulator reduces computation time significantly, produces almost identical results compared to the BPS tools.

Cited by 0 works.

Authors