Controllable cross-building multi-objective optimisation for NZEBs: A framework Utilising parametric generation and intelligent algorithms

Abstract

Balancing performance and cost are crucial in optimising nearly zero-energy buildings. This study proposes a method for efficient preliminary design tailored to diverse requirements. It starts with parametric modelling and performance simulation to compile a dataset of common spatial forms in almost zero-energy office buildings. Subsequently, a multilayer perceptron surrogate model is trained on this dataset to map the relationships between design parameters and performance objectives across various spatial forms. A controllable NSGA-II constraint method enables designers to manage design variables and performance objectives independently for cross-building optimisation. The framework integrates the surrogate model with the NSGA-II algorithm under precise constraints, using TOPSIS to finalise Pareto solutions, and includes visualisation software for architects. The framework is validated in a high-performance office building in northern China by measuring data and using an equivalent model under two constraints. The results indicated that the framework reduced the energy consumption of the building by 6.6% and 14.7%, increased the thermal comfort time ratios by 2.3% and 11.4%, and lowered the initial investment costs by 19.6% and 16%, respectively, under each condition.