The right way to do building simulations? Using Monte Carlo simulations, sensitivity analysis, and metamodeling on a design case

Abstract

Monte Carlo simulations, sensitivity analysis and metamodeling are becoming popular in academia but are rarely applied in real building projects. In this case study, we demonstrate how a combined framework of these methods can aid decision-making in relation to building performance of nine 16-story residential buildings. We describe the processes before, during, and after a meeting between building engineers and the building owner. For preparation, BeDesigner was used to create, run, and analyse 5.000 Be18 simulations in roughly 4 hours. The meeting is initiated with a presentation of sensitivity analysis results to focus the attention towards the most influential design inputs. The 5.000 simulations are visualized with parallel coordinates plots in DataExplorer, which enable decision-makers to observe the consequences of different design choices and regulatory requirements. Real-time sensitivity analysis, TOR, highlights the parameters affected the most by the applied constraints, while histograms indicate favourable or disadvantageous design choices. However, no solutions exist among the 5.000 simulations, which is due to the vastness of the multi-dimensional input space and the decision-makers’ numerous requirements. Using metamodels, 500.000 additional input combinations are calculated and from this extensive dataset a variety of solutions are found. It becomes clear that a “norenewables” ambition necessitates costly countermeasures and makes it difficult to realize the architectural and indoor climate requirements. In conclusion, the combined framework improves the information quality for decision-making and significantly increase the likelihood of finding diverse, high-performing solutions within the same time-frame as traditional practice.