The artificial intelligence reformation of sustainable building design approach: A systematic review on building design optimization methods using surrogate models
Ibrahim Elwy, Aya Hagishima
Abstract
• Surrogate-assisted sustainable building design optimization studies are discussed. • Global disparities in surrogate modeling and optimization practices are revealed. • Efficient surrogates foster high-dimensional passive design optimization approaches. • Recent advances in surrogate modeling facilitate context-diversified optimizations. • Contextual variations prompt versatile pre-experimentation for suitable techniques. Artificial Intelligence (AI) applications in building performance prediction for environmental sustainability outcomes play a significant role in compensating for computationally incompetent and expensive approaches to solving increasingly complex optimization problems. Although machine-learning-based surrogate models (SMs), one of the many AI approximation strategies for higher-order models, have long been utilized in sophisticated optimization studies in various fields, their reliability in the field of sustainable and low-energy architectural design is now being extensively explored. To comprehend the effectiveness of applying state-of-the-art surrogate-assisted optimization methodologies for building energy performance, this study presents a comprehensive review of recent studies aimed at identifying innovative theories and critical factors for each component of the optimization process. Considering passive form and envelope design variables, seventy-two relevant studies from the Scopus database were selected for review after screening. The results, including the current trends for surrogate modeling and optimization methods, accuracy levels of SMs, extent of building performance improvement, and decisive building design variables, are thoroughly discussed according to the varying conditions of each study. Finally, a further discussion of recent methodological advancements and limitations can help identify potential approaches and challenges for future research.