This article studies the impact of urban heat wave events on residents' health and quality of life in the context of global climate warming. The purpose of the research is to solve the high cost and low efficiency problems of traditional numerical simulation methods in calculating outdoor thermal comfort (UTCI). By proposing a UTCI prediction method based on a generative adversarial network (GAN), a fast and accurate outdoor thermal comfort assessment is achieved. Research methods include using the Rhino/Grasshopper platform to build city models, using Ladybug and Eddy3D plug-ins for climate simulation, training the pix2pix adversarial network model based on the TensorFlow framework, and verifying the prediction performance of the model through experiments. Research results show that the pix2pix model performs well in predicting urban environmental comfort, significantly improves prediction speed and reduces data collection costs, providing a scientific basis for urban design and sustainable construction.
研究内容
Research contents
（1）研究背景：随着全球气候变暖的加剧，城市热浪事件日益频繁，对城市居民的健康和生活质量构成了严重威胁。特别是在夏季，高温天气不仅增加了热应力天数，还可能对孕妇、婴儿等特定人群的健康产生严重负面影响。因此，深入研究户外热舒适度，特别是在传统居住区空间中的热舒适度，对于提升居民的生活质量、保障公众健康以及促进城市的可持续发展具有至关重要的意义。
（2）研究目的：本文旨在解决传统数值模拟方法在计算户外热舒适度（UTCI）时面临的高成本、低效率问题。通过提出一种基于生成对抗网络（GAN）的UTCI预测方法，期望实现快速、准确的户外热舒适度评估，为城市设计和可持续建设提供科学、可靠的依据。
(1) Research background: With the intensification of global climate warming, urban heat wave events have become increasingly frequent, posing a serious threat to the
health and quality of life of urban residents. Especially in summer, high temperature weather not only increases the number of heat stress days, but may also have serious
negative effects on the health of specific groups such as pregnant women and infants. Therefore, in-depth research on outdoor thermal comfort, especially in traditional
residential areas, is of vital significance to improving residents' quality of life, protecting public health, and promoting sustainable urban development.
(2) Research purpose: This paper aims to solve the high cost and low efficiency problems faced by traditional numerical simulation methods when calculating outdoor
thermal comfort (UTCI). By proposing a UTCI prediction method based on generative adversarial networks (GAN), it is expected to achieve rapid and accurate outdoor
thermal comfort assessment and provide a scientific and reliable basis for urban design and sustainable construction.
研究方法
Research methods
（1）城市模型构建：借助Rhino/Grasshopper平台，构建了精细的城市模型，以模拟真实的城市环境，为后续的气候模拟提供基础。
（2）气候模拟与数据生成：使用Ladybug和Eddy3D插件进行微环境气候模拟，生成关键的UTCI图像。随后对模拟结果进行图像分割和数据增强，构建出丰富且多样的数据集，以支持GAN模型的训练。
（3）GAN模型训练：基于TensorFlow框架，训练了pix2pix对抗网络模型。通过大量的迭代训练，使模型能够学习并预测UTCI值。
（4）模型性能验证与比较：通过实验验证了pix2pix模型的预测性能，并与其他GAN模型（如cycleGAN）进行了对比分析，以评估模型的优劣。
(1) Urban model construction: With the help of the Rhino/Grasshopper platform, a detailed urban model was constructed to simulate the real urban environment and
provide a basis for subsequent climate simulation.
(2) Climate simulation and data generation: Use Ladybug and Eddy3D plug-ins to perform micro-environment climate simulation and generate key UTCI images. The
simulation results were then subjected to image segmentation and data enhancement to build a rich and diverse data set to support the training of the GAN model.
(3) GAN model training: Based on the TensorFlow framework, the pix2pix adversarial network model was trained. Through a large number of iterative training, the model
is able to learn and predict UTCI values.
(4) Model performance verification and comparison: The prediction performance of the pix2pix model was verified through experiments, and a comparative analysis was
conducted with other GAN models (such as cycleGAN) to evaluate the advantages and disadvantages of the model.
研究结果
Research conclusions
（1）模型预测性能：pix2pix模型在预测城市环境舒适度方面展现出了出色的性能。经过训练，模型能够准确捕捉建筑布局与UTCI之间的复杂关系，预测的UTCI图像几乎满足项目的性能和精细度要求。特别是在封闭建筑群落的布局中，模型表现尤为出色。
（2）与传统方法比较：与传统数值模拟方法相比，pix2pix模型显著提高了预测速度，降低了数据收集的成本和时间。同时，通过合理设置迭代次数和数据增强策略，模型在保证预测精度的同时，显著提升了计算效率。
（3）城市设计与可持续建设应用：pix2pix模型为城市设计和可持续建设提供了有力的支持。通过快速预测城市不同区域的户外热舒适度，有助于城市规划者和设计师快速识别出过热和不适区域，并制定相应的改善措施。此外，模型还可以集成到精细化城市设计的工作流程中，为设计师提供实时的环境舒适度反馈，优化建筑布局和微气候环境，从而推动绿色建筑和低碳城市的发展。
(1) Model prediction performance: The pix2pix model has shown excellent performance in predicting urban environmental comfort. After training, the model is able to
accurately capture the complex relationship between building layout and UTCI, and the predicted UTCI images almost meet the project's performance and precision
requirements. The model performs especially well in layouts with enclosed building complexes.
(2) Compared with traditional methods: Compared with traditional numerical simulation methods, the pix2pix model significantly improves the prediction speed and
reduces the cost and time of data collection. At the same time, by reasonably setting the number of iterations and data enhancement strategies, the model significantly
improves computational efficiency while ensuring prediction accuracy.
(3) Urban design and sustainable construction applications: The pix2pix model provides strong support for urban design and sustainable construction. By quickly predicting
outdoor thermal comfort in different areas of the city, it helps urban planners and designers to quickly identify areas of overheating and discomfort and formulate
corresponding improvement measures. In addition, the model can also be integrated into the workflow of refined urban design, providing designers with real-time
feedback on environmental comfort, optimizing building layout and microclimate environment, thereby promoting the development of green buildings and low-carbon
cities.
