Multi-atrium configuration design for energy efficiency in shopping malls: an ANN-based metamodel for sensitivity analysis and design optimization

Abstract

This paper explores the design approach of multi-atrium configurations to improve the energy efficiency of shopping malls in cold climates. A typical building prototype in northern China was considered, and an artificial neural network (ANN)-based metamodel was developed with four design variables: the atrium area ratio (r), atrium number (n), distance between adjacent atriums (d 0), and distance between atrium and façade wall (d 1). The proposed metamodel was used for sensitivity analysis (SA) and building design optimization (BDO) to understand the performance of multi-atrium configuration. The SA indicates that the contribution of the design variables to the building energy efficiency follows the order r > d 0 > n > d 1, and there are interactions between variables r and n. The BDO demonstrates that it is beneficial for energy efficiency to incorporate more atrium areas, appropriate atrium sizes, larger distances between adjacent atriums, and more west-facing atriums.