A comparison of methods for uncertainty and sensitivity analysis applied to the energy performance of new commercial buildings

Abstract

An Energy Performance Contracting (EPC) is a financing agreement offered by general contractors that enables cost savings from reduced energy consumption to building owners. To create such an offer, the contractor has to provide an energy consumption threshold and a measurement plan.This article aims to draw some recommendations to choose an appropriate approach to provide the information necessary to create the contract, regarding computation time budget, expected accuracy and type of information provided.To get these results, we couple thermal simulations to various uncertainty and sensitivity methods.We first compare screening and differential sensitivity to reduce the number of inputs of the statistical study. Then, we analyze various uncertainty analysis methods to set an appropriate energy consumption threshold, considering the input uncertainties and the study context (Quadratic combination, directional and importance sampling and reliability methods).Sensitivity analyses in various input spaces are then carried out to identify the most critical contributors to energy levels to create the measurement plan. Finally, two metamodeling approaches are tested to reduce the overall computational time: Kriging and sparse polynomial chaos.These methods are tested and compared on a 4000 m² office building in Nantes, France. The resulting recommendations can be applied to any building, depending on the model regularity, the number of uncertain parameters and the objective of the study.