The Impact of Occupancy-Related Input Data Uncertainty on the Distribution of Building Simulation Results

In building performance simulation, fixed input assumptions lead to fixed computed values for building performance indicators. This has been suggested to be misleading, as it does not express the uncertainty of simulation-based performance predictions. A counterargument to this position suggests that the empirical basis for the determination of the statistical uncertainty distribution of occupancy-related input assumptions is rather scant. Arbitrary assignment of uncertainty functions (distribution ranges and shapes) to input variables can indeed generate corresponding performance result distributions. However, this could be even more misleading than fixed values, as the resulting uncertainty impression is empirically ungrounded. To address this objection, it has been suggested that the computed uncertainty ranges for performance indicators may be, to a certain extent, resistant to the ranges and shapes of associated input data distributions and hence still useful. In the present contribution, we examine the above suggestion, namely the resilience of performance simulation output distribution to the assumed model input uncertainties. To this end, parametric simulations were conducted and processed to explore the implications of different input data assumptions for the values of computed performance indicator values for a sample building model.