Modeling on Outdoor Thermal Comfort in Traditional Residential Neighborhoods in Beijing Based on GAN

With the support of new urban science and technology, the bottom-up and human-centered space quality research has become the key to delicacy urban governance, of which the Universal Thermal Climate Index (UTCI) have a severe influence. However, in the studies of actual UTCI, datasets are mostly obtained from on-site measurement data or simulation data, which is costly and ineffective. So, how to efficiently and rapidly conduct a large-scale and fine-grained outdoor environmental comfort evaluation based on the outdoor environment is the prob lem to be solved in this study. Compared to the conventional qualitative analysis methods, the rapidly developing algorithm-supported data acquisition and ma chine learning modelling are more efficient and accurate. Goodfellow proposed Generative Adversarial Nets (GANs) in 2014, which can successfully be applied to image generation with insufficient training data. In this paper, we propose an approach based on a generative adversarial network (GAN) to predict UTCI in traditional blocks. 36000 data samples were obtained from the simulations, to train a pix2pix model based on the TensorFlow framework. After more than 300 thou sand iterations, the model gradually converges, where the loss of the function gradually decreases with the increase of the number of iterations. Overall, the model has been able to understand the overall semantic information behind the UTCI graphs to a high degree. Study in this paper deeply integrates the method of data augmentation based on GAN and machine learning modeling, which can be integrated into the workflow of detailed urban design and sustainable construction in the future.