Design optimization of climate-responsive rural residences in solar rich areas considering sustainability and occupant comfort

Abstract

Rural residences in solar rich areas (SRA) remain inadequately progress on the challenges of balancing sustainability and occupant comfort simultaneously. Recent advancements in achieving zero-carbon buildings have highlighted the growing demand on climate-responsive strategies that effectively address design considerations of building performances. This paper attempts to explore the optimal design solutions of a case building in Lhasa, a typical city located in SRA, through multi-objective optimization (MOO) based on machine learning in terms of energy, cost, carbon, thermal comfort, and daylight. Initiating comparative selection through representative surrogate models and metaheuristic algorithms, XGBoost achieves the most desirable predictive performances with the least prediction errors, and NSGA-II dominates the superior capability in this case. Based on the solutions discussed through entropy weight and biased weight TOPSIS of objectives, critical outcomes reveal the identification of the optimal building layout in addition to the potential interactions of design variables and building performances investigated. Offering a scalable solution-oriented climate-responsive design pattern, building envelope of the proposed building demonstrates greater potential in balancing building performances, while morphology tends to be comparatively constrained. Results also indicate that, in the most applicable scenario, energy use intensity (EUI) and life cycle carbon emission (LCCE) can be substantially reduced by 47.2% and 38.1%, while concurrently maintaining considerable levels of thermal comfort albeit with acceptable compromises in daylight availability. This paper presents a transferable approach to optimal climate-responsive design solutions, contributing notable insights of data-driven decision-making for rural residences in SRA.