Space layout automation and optimization for energy-efficient buildings: A multi-objective evolutionary approach with machine learning analytics

This research aims to explore a new method and implementation for the automated generation and optimization of energy-saving building three-dimensional space layouts for mixed-use buildings, focusing on the topological placement of spaces. A three-dimensional space layout generation method was developed in this study, integrating multi-objective evolutionary algorithms and energy consumption simulation. The proposed approach enables automated generation and evaluation of building space layouts, seeking to optimize building layouts that meet practical requirements and minimize energy consumption. In a case study, we applied our method to a high-rise office building in Shenzhen, China, generating more than seven thousand layout schemes and providing visual references for the optimized layouts, thereby validating the effectiveness of the proposed method. The experimental results revealed that the space layout, under fixed building contours and enclosures, influences building energy consumption, achieving up to 5.5% reduction in energy use for the case study building through topological optimization of thermal zone placement, with the maximum reduction primarily occurring in cooling consumption. Additionally, through analysis using random forest model and interpretable artificial intelligence, we summarized specific energy saving space layout strategies. Compared to traditional forward workflow energy prediction methods, the inverse workflow-based approach in this study can more efficiently search for optimal energy saving solutions. This method is suitable for building space layout design in multi-story and mixed-use buildings, offering adequate generalizability and applicability. © 2026 Elsevier B.V.