A Stochastic and Holistic Method to Support Decision-Making in Early Building Design

The use of holistic certification tools is increasing and requirements in legislation are continuously being tightened. This calls for a holistic simulation approach in the early design phase where input uncertainties are large and decisions are crucial to the performance. An iterative parametric method is proposed: 1) Assign uniform distributions to uncertain design inputs of interest; 2) Perform sensitivity analysis (SA) by the method of Morris to rank input by relative importance; 3) Run Monte Carlo simulations to explore the entire design domain; 4) Apply Monte Carlo filtering to identify preferable input domains for the most influential parameters.

To enable computationally fast simulations, we combined calculations of energy demand and thermal comfort based on ISO 13790 (CEN 2008) with a regression model for daylight factor. We constructed scoring functions for the three outputs and applied weighting to combine the three scores into a single holistic score ranging from 0 to 100.

The method was tested on a simple office building. An initial run of 3000 simulations was performed using a Quasi-Random LpTau sampling strategy for 22 variable inputs. A filter was applied to the holistic score to collect the 10 % best performing simulations. From this collection, histograms were used to identify favourable and adverse input spans for a selection of the most sensitive parameters. Subsequently, two runs of each 3000 simulations were performed – one using the favourable input spans and the other using the adverse spans. The results showed that the distribution related to favourable input spans was shifted significantly towards higher holistic scores. The authors conclude that the use of a stochastic, holistic method can guide decision-making by identifying favourable input regions, and thereby increase the remaining solution space and overall building performance.