期刊
SOLAR ENERGY
卷 211, 期 -, 页码 1184-1193出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.solener.2020.10.043
关键词
Daylighting; Energy; Principal component analysis; Multicollinearity; Standardized regression coefficients; Machine learning
The identification of Standardized Regression Coefficients (SRCs) can be used to aid designers in identifying the most influential design variables affecting building performance. Nevertheless, multicollinearity can nullify the reliability of SRCs. This study aims to develop a method for estimating reliable SRCs of design variables in daylighting and energy performance of buildings in the face of multicollinearity. A parametric model is developed, then, daylight and energy simulations are commenced, next, Principal Component Analysis (PCA) is performed and the extracted Principal Components (PCs) are interpreted to meaningful factors, finally, the SRCs of the extracted PCs in daylighting and energy performance were calculated. The method was applied to a standard office, where the quantitative performance of (3564) different cases were predicted in terms of twelve urban context configurations against nine variables. Four PCs out of nine variables were extracted, namely, PC1: window geometry, PC2: glazing properties, PC3: opposing building paint, and PC4: building paint. The generated SRCs charts of the extracted PCs described the behaviors of the PCs in daylighting and energy performance. The Horizontal Obstruction Angle (HOA) affected the PCs' behavior for all performance metrics, Annual Sunlight Exposure (ASE) was not affected by PC3 and PC4, PC1 had an inverse relationship with Lighting Energy (LE), whilst, the contribution of PC3 on LE was increased dramatically when a steep HOA existed. Total Energy Consumption (TEC) showed the same trend as Cooling Energy (CE). The resulted behaviors were consistent with the literature and confirmed the efficiency of the proposed method.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据