Development of a Self-adaptive Evolutionary Model to Optimize Singlesided Ventilated Facade Design

Abstract

Residential buildings with single-sided ventilated (SSV) facade require greater energy consumption of heating, ventilation and air conditioning (HVAC) systems to achieve a comfortable indoor environment and to maintain the health and productivity of the occupants. SSV facade has been reported to increase the likelihood of a poor indoor environment and high energy consumption. Hence, optimizing SSV facade design to create an energy-efficient and comfortable indoor environment is a challenging task. Existing studies have employed computational fluid dynamics (CFD) methods to analyze natural ventilation of buildings with SSV facade or passive design strategies to reduce energy consumption. However, most existing studies have focused on addressing individual performance issues of SSV facade such as natural ventilation or energy consumption. There has been a lack of dealing with the integrated performance of SSV facade especially the integrated optimal performance across ventilation efficiency, energy efficiency and visual comfort. There has been also a lack of understanding which design variables and relationships may drive optimal performance of SSV facade design. This research aims to fill the research gap by developing an innovative self-adaptive evolutionary model to optimize SSV facade design, targeted at the integrated optimal performance across ventilation efficiency, energy efficiency and visual comfort. The model includes a self-adaptation and learning mechanism which integrates unsupervised machine learning with evolutionary algorithms. The self-adaptation and learning mechanism has the ability to discover emergent patterns named evolved genes which represent key design variables and relationships that lead to high-performance of ventilation efficiency, energy efficiency and visual comfort of SSV facade design. The discovered key design variables and relationships are then evolved from simple to complex in the self-adaptive evolutionary process until a set of optimal SSV facade design is obtained. The utility of the self-adaptive evolutionary model is demonstrated using multi-story residential buildings with SSV facade in Sydney. A set of optimal SSV facade design is obtained in the prototype implementation, which shows on average 20% improvement in ventilation efficiency, 40% energy saving on heating and cooling loads and improvement in daylight visual comfort compared to the baseline performance of a building with SSV facade v design. The evolved genes in different complexities are discovered and evolved over time, demonstrating the dynamic mapping between key SSV facade design variables and high-performance outcomes of ventilation efficiency, energy efficiency and visual comfort. The analysis results also prove the effectiveness of the self-adaptation and learning mechanism, which accelerates the process to enable high-performance of SSV facade design to be achieved at earlier generations and increases the integrated optimal performance of SSV facade design by 6 - 8% compared to a conventional evolutionary process. This research develops an innovative interdisciplinary approach which is built upon artificial intelligence, facade optimization and building sustainability to tackling the challenge faced in SSV facade design. Research outcomes will advance the interdisciplinary knowledge of utilizing artificial intelligence technologies to improve the indoor environment and reduce energy consumption of SSV facade design.