Building energy models: Quantifying uncertainties due to stochastic processes

Abstract

Energy efficient retrofits of existing buildings present an immediate and large opportunity to reduce the energy footprint of the built infrastructure, which consumes nearly 40% of primary energy consumption in the U.S. and worldwide. Whole building energy modeling and simulation tools are increasingly being used for detailed performance analysis and for evaluation of multiple retrofit design options. However, the models typically involve several hundreds of input parameters and processes (e.g. weather and occupancy schedules) that are uncertain in early stages of design, and are not fully understood until after retrofit installation and commissioning. We present tools for sensitivity analysis and uncertainty quantification of such building energy models that help designers understand the key drivers to energy consumption and estimate error bounds on predicted energy savings. The focus is on quantifying uncertainties due to stochastic processes, such as weather conditions and schedules of occupants, which are modeled using a Karhunen-Loève expansion.