本文提出了一个新的月度能源使用回归模型，该模型根据商业建筑的入住率来解释不同的建筑能源使用率。当建筑物在节假日有不同的运行模式时，由于节假日期间的能耗较低，带有单一温度变量的月度三参数（3-P）冷却变点模型可能会低估基本负荷消耗。这可能导致净测定偏差（NDB）高于ASHRAE指南14-2002中要求的可接受水平（即，第5.2.10节中的0.005%），这将导致与使用单一温度变量的基线模型计算的天气标准化能源性能变化（即节约）相关的更高水平的不确定性。 为了解决这个问题，本研究开发了一个三参数多变量回归（3-P MVR）制冷模型，结合ASHRAE的逆建模工具包（IMT），使用变化点模型中的室外温度和假期数作为附加自变量。然后，使用从德克萨斯州中部的一座案例研究办公楼收集的多年月度和每日整栋建筑用电数据，将使用拟议的3-P MVR模型的优势与使用单一温度变量的月度3-P冷却模型以及工作日、周末和节假日的每日3-P冷却模型进行了对比。结果表明，使用建议的3P-MVR模型，通过解决节假日问题（i。 e、 ，3-P模型因节假日而低估的基本负荷消耗），与基准年相比，计算节省的不确定性水平较低。对于其他具有类似能源使用情况的建筑，预计也会有类似的结果。引用:ASHRAE论文:2015年ASHRAE年会，伊利诺伊州芝加哥

This paper proposes a new monthly energy use regression model that accounts for different building energy use rates by occupancy schedules for commercial buildings. When the building has a different operating mode for holidays, the monthly three-parameter (3-P) cooling change-point model with a single temperature variable is likely to under-predict the base-load consumption due to lower energy consumption during the holidays. This may yield a net determination bias (NDB) higher than the acceptable level required in the ASHRAE Guideline 14-2002 (i.e., 0.005% per Section 5.2.10), which will result in a higher level of uncertainties associated with the calculated weather-normalized energy performance changes (i.e., savings) using a baseline model with a single temperature variable. To resolve this issue, this study developed a combination threeparameter multi-variable regression (3-P MVR) cooling model with ASHRAE's Inverse Modeling Toolkit (IMT) using outdoor temperature in a change-point model and the number of holidays as an additional independent variable. The advantages of using the proposed 3-P MVR model were then examined compared to a monthly 3-P cooling model with a single temperature variable as well as the daily 3-P cooling models for weekdays, weekends, and holidays, using the multi-year monthly and daily whole-building electricity use data collected from a case-study office building in central Texas. The results show that the use of the proposed 3P-MVR model improves the accuracy of the monthly 3-P cooling model by resolving the holidays issue (i.e., the under-predicted base-load consumption of 3-P model due to holidays), with a lower level of uncertainty in the computed savings against the baseline year. Similar results are expected for other buildings with similar energy use profiles.