Intelligent optimal design for near-zero energy buildings performance: A case study of five climate zones in China

To enrich the methods used to analyze the sustainable performance of near zero energy consumption buildings (NZEBs), this paper introduces a hybrid intelligent algorithm combining the peak hedgehog optimizer (CPO), natural gradient boost (NGBoost) algorithm, and multi-objective gray wolf optimizer (MOGWO). Meanwhile, a knowledge and data-driven analysis framework is constructed. By using building information modeling-design builder (BIM-DB) simulation software and Monte Carlo simulation (MCS), a building data sample set of five climate regions in China is obtained. The input and output indices are determined according to a dual knowledge‒data–driven framework, and a performance prediction model is constructed on the basis of the CPO–NGBoost machine learning algorithm. The objective function is determined by the relationship between the input and output indices evaluated by the prediction model, and the Pareto frontier is derived using the MOGWO. Finally, the optimal scheme of the multiobjective optimization problem of building performance is obtained by using the ideal point method. Using a case study, we found that (1) the parameters and thresholds for the building are defined using knowledge-driven approaches. Through BIM–DB simulation and Monte Carlo simulation, a dataset of architectural samples can be effectively generated. (2) The CPO–NGBoost machine learning algorithm has good predictive performance. The R2 of all the prediction targets exceeds 0.9. (3) The CPO–NGBoost–MOGWO hybrid algorithm optimizes the building energy consumption to 17.59 %–34.25 % in five climate zones in China and optimizes the other performance metrics of NZEBs. Our approach can inform the sustainable performance of NZEBs in different climatic regions worldwide.