Paper Conference

ASHRAE & IBPSA-USA SimBuild 2016: Building Performance Modeling Conference


Simulation Using in Situ Adaptive Tabulation and Fast Fluid Dynamics

Dan Li, Wei Tian, Wangda Zuo, Michael Wetter
University of Miami, Coral Gables, FL
Lawrence Berkeley National Laboratory, Berkeley, CA

Abstract: Ventilation with stratified air distribution is commonly used to improve building energy efficiency and indoor environment quality. A fast indoor airflow simulation can be useful for the ventilation design, performance evaluation, and model predictive ventilation control. As an intermediate model between computational fluid dynamics (CFD) and multizone models, a fast fluid dynamics (FFD) was proposed to balance simulation accuracy and computing speed. In this paper, we propose to further speed up the FFD simulation by using a computation reduction technique called in situ adaptive tabulation (ISAT). ISAT is a general function approximation method and was first proposed to speed up CFD simulation for combustion. Using ISAT, we are able to store the key FFD simulation data in a table and retrieve the data from the table for simulations with similar boundary conditions. This paper presents our ISAT-FFD implementation and some preliminary results. In a parametric study with 60,000 simulations, we showed that the ISAT-FFD simulation could compute the key indoor environment data for a natural convection flow at a speed up of 50 times faster than the FFD simulation and the prediction errors are within 1K. In the other case study, we showed that a trained ISAT-FFD model can predict the key environmental data by simply retrieving from the data table with controllable accuracy and little computing time. The ISAT-FFD can also properly handle the scenario when the independent variables are outside pre-trained data range.
Pages: 65 - 71