Impact of Nighttime Shut Down on the Prediction Accuracy of Monthly Regression Models for Energy Consumption in Commercial Buildings

摘要

Regression models of measuredenergy use in buildings are widely used asbaseline models to determine retrofit savingsfrom measured energy consumption. It is lessexpensive to determine savings from monthlyutility bills when they are available than toinstall hourly metering equipment. However,little is known about the impact of nighttimeshut off on the accuracy of savingsdetermined from monthly data. This paperreports a preliminary investigation of thisquestion by comparing the heating andcooling energy use predicted by regressionmodels based on monthly data against thepredictions of calibrated hourly simulationmodels when applied to a medium-sizeduniversity building in Texas with (i) DDCAVsystem operating 24 hours per day, (ii)DDCAV system with nighttime shut down,(iii) DDVAV system operating 24 hours perday, and (iv) DDVAV system with nighttimeshut down.The results of the four cases studiedindicate : 1) when the AHUs are operated 24hours/day, the annual prediction error of thecooling regression models is less than 0.5%of the annual cooling energy consumption;however, 2) when the AHUs are operatedwith nighttime shut down, the annualprediction error of the cooling modelsbecomes as high as 6% of annual energyconsumption. It should be noted that thecases considered here include only singleend-uses of energy and have not investigatedenergy-use data which includes multiple end-uses.Modified regression models are thereforerecommended when AHUs are not operated24 hours per day and the temperature patternis significantly different between pre and postretrofit years.