An integrated genetic algorithm-machine learning approach for morphological optimization of high-rise residential districts in Yulin

The pursuit of global carbon neutrality necessitates addressing the dual challenge of enhancing solar energy utilization while improving thermal comfort in high-rise residential areas, particularly in Yulin, northern Shaanxi, China, where abundant solar resources exist but maximizing solar acquisition often compromises summer thermal environment quality. This resource-comfort contradiction highlights the need for balanced architectural strategies in regions with pronounced seasonal variations. Building morphological parameter optimization is crucial for balancing annual solar energy capture against summer overheating risks, yet research remains insufficient. This study developed parametric layout models using Rhino-Grasshopper, considering key parameters including building length, width, height, density, floor area ratio, and south-facing angle deviation. Multi-objective optimization was conducted using NSGA-II genetic algorithm under regulatory constraints, while combining traditional regression analysis with convolutional neural networks (CNN) to investigate the influence mechanisms of these morphological parameters. Results indicate that the optimized building morphology can increase annual solar radiation acquisition (SRA) by 2.57% while maintaining comfortable summer Universal Thermal Climate Index (UTCI) values, effectively balancing solar energy capture and outdoor thermal comfort. Regression analysis revealed a positive correlation between building length and summer UTCI (r = 0.73), whereas CNN identified a negative correlation (−0.45). Both methods identified similar parameter combinations affecting SRA, with CNN demonstrating superior capability in capturing complex non-linear relationships. These findings provide evidence-based design guidelines specific to Yulin while offering implications for sustainable residential development in similar climates, advancing the integration of climate-adaptive design strategies.