Scholarlink

<p>This paper investigates the thermal and energy performance of three types of urban courtyard blocks in hot and arid climates, focusing on their cooling potential, Average Outdoor Thermal Comfort Index (Av. OTCI), and Average Universal Thermal Comfort Index (Av. UTCI). A two-phase methodological framework is proposed, integrating Random Forest Regression for predictive modeling and the Non-dominated Sorting Genetic Algorithm III (NSGA-III) for multi-objective optimization. In the first phase, Random Forest models captured non-linear relationships between courtyard design parameters and performance metrics, achieving high predictive accuracy (e.g., R² = 0.976 for Av. OTCI and R² = 0.972 for Av. UTCI in Type A blocks). In the second phase, NSGA-III optimization was applied to maximize both cooling performance and thermal comfort indices, with the trained machine learning model embedded as the objective function. Pareto front analysis demonstrated that Type A blocks achieved the most favorable outcomes with minimal trade-offs, while Type B showed moderate variability and Type C exhibited the greatest challenges in approaching optimal solutions. These differences are mechanistically linked to geometry: Type A benefited from balanced height distribution and wide orientation improving shading and airflow, whereas Type C’s irregular heights and narrow orientation caused higher heat accumulation and reduced efficiency. Parallel coordinate plots further revealed the influence of block height, orientation, and distance on thermal performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-Objective optimization of courtyard block design: balancing thermal comfort and cooling efficiency using NSGA-III

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Hehong Ma,
  • Kangning Liu

摘要

This paper investigates the thermal and energy performance of three types of urban courtyard blocks in hot and arid climates, focusing on their cooling potential, Average Outdoor Thermal Comfort Index (Av. OTCI), and Average Universal Thermal Comfort Index (Av. UTCI). A two-phase methodological framework is proposed, integrating Random Forest Regression for predictive modeling and the Non-dominated Sorting Genetic Algorithm III (NSGA-III) for multi-objective optimization. In the first phase, Random Forest models captured non-linear relationships between courtyard design parameters and performance metrics, achieving high predictive accuracy (e.g., R² = 0.976 for Av. OTCI and R² = 0.972 for Av. UTCI in Type A blocks). In the second phase, NSGA-III optimization was applied to maximize both cooling performance and thermal comfort indices, with the trained machine learning model embedded as the objective function. Pareto front analysis demonstrated that Type A blocks achieved the most favorable outcomes with minimal trade-offs, while Type B showed moderate variability and Type C exhibited the greatest challenges in approaching optimal solutions. These differences are mechanistically linked to geometry: Type A benefited from balanced height distribution and wide orientation improving shading and airflow, whereas Type C’s irregular heights and narrow orientation caused higher heat accumulation and reduced efficiency. Parallel coordinate plots further revealed the influence of block height, orientation, and distance on thermal performance.