Uncertainty Analysis in Estimating Building Energy Retrofit Savings in the LoanSTAR Program

A crucial requirement to promote and sustain energy conservation measures in buildings is the ability to perform careful and reliable appraisals of exactly how much energy has been saved. This paper briefly describes the general issues involved and then focuses on commercial buildings where retrofits are made in order to reduce energy consumption in hot water, chilled water and in electricity consumed by air-handler and chilled water pumps. How retrofit energy savings in the framework of the Texas· LoanSTAR project are estimated is described, and the need to assess the accuracy in these estimates themselves is highlighted. The various sources of uncertainties in the retrofit savings estimates are itemized and discussed. We point out the need to consider the strong serial correlations present in daily (and other types of time series) data and how this impacts the model identification process as wen as the determination of uncertainty. Subsequently, equations to compute the prediction uncertainty associated with the use of a regression model involving serially correlated data streams and with error in the measurement data are presented. Use of these equations is illustrated means of a case study involving one large commercial LoanSTAR building. Attention has been drawn to areas in which additional work is needed in order to improve the estimates of retrofit savings.