A Performance Comprehension of Various Numerical Estimators for Variance-Based Sensitivity Analysis in Building Energy Simulations

Abstract

In the building energy performance simulation, the uncertainty analysis (UA) couples to the sensitivity analysis (SA) to handle ever-existing uncertainties; induced by the sources of uncertainty including random occupants behavior and degradation of building materials over time. As a building simulation tool reaches to a high level of complexity, it becomes more challenging for the sensitivity analysis to deliver reliable outputs; thus the accuracy of the SA results substantially depends upon the number of sample sets or the type of analysis performed. This paper describes a variance-based SA tool integrated into a building Resistance-Capacitance (RC) thermal model. Then, for a hypothetical residential building test case, three distinct first-order sensitivity index simulators and three total sensitivity index simulators are implemented and compared in terms of the dependency of results on the sample size, i.e., the demand for the computational cost.