Analysis of predicted vs monitored space heat energy use in 120 homes. Residential Construction Demonstration Project Cycle 2

摘要

The SUNDAY thermal simulation program was used to predict space heat energy consumption for 120 energy efficient homes. The predicted data were found to explain 43.8 percent of the variation in monitored space heat consumption. Using a paired Student's to test, no statistically significant difference could be found between mean predicted space heat and monitored space heat for the entire sample of homes. The homes were grouped into seven classes, sub-samples by total heat loss coefficient. An intermediate class (UA = 300--350 Btu/(degrees)F) was found to significantly over-predict space heat by 25 percent. The same class was over-predicted by 16 percent in the analogous Cycle 1 research, but the sample size was smaller and this was not found to be statistically significant. Several variables that were not directly included as inputs to the simulation were examined with an analysis of covariance model for their ability to improve the simulation's prediction of space heat. The variables having the greatest effect were conditioned floor area, heating system type, and foundation type. The model was able to increase the coefficient of determination from 0.438 to 0.670; a 54 percent increase. While the SUNDAY simulation program to aggregate is able to predict space heat consumption, it should be noted that there is a considerable amount of variation in both the monitored space heat consumption and the SUNDAY predictions. The ability of the program to accurately model an individual house will be constrained by both the quality of input variables and the range of occupant behavior. These constraints apply to any building model.