Paper Conference

Proceedings of BSO Conference 2018: Fourth Conference of IBPSA-England


An Evaluation of Computational Algorithms in the Automatic Generation of Zonal Reduced-Order Models from CFD Simulations

Thibault Marzullo, Marcus Keane, Marco Geron, Rory Monaghan

Abstract: The present study compares the accuracy and time to solution of three different algorithms used in automatic Reduced-Order Model (ROM) generation. This is part of an effort to develop a systematic tool chain for automatically extracting accurate ROMs from Computational Fluid Dynamics (CFD) simulations for use in the design and operation of near-zero energy buildings, with a higher accuracy than traditional zonal models but at a fraction of the computational cost of CFD. The results show that the Mean Values Segmentation (MVS) method is preferable to Coarse Grid Interpolation (CGI) and Classic Watershed (CW) for this application. Furthermore, it is possible to generate accurate models with mean absolute errors for temperature distributions as low as 0.3 K that can be generated in less than 70 seconds and solved in less than 1 second. Further work is needed to develop the flexibility of the tool chain to solve for Boundary Conditions (BC) different from the initial CFD BC’s. This will allow the generation of multiple ROMs from a reduced set of CFD simulations, allowing to accurately simulate indoor thermal conditions at low computational costs.
Pages: 119 - 123