An advanced generative framework for low-carbon urban morphology optimization
The PC-GAN model integrates Pix2pix with a two-step CycleGAN-inspired mechanism to automatically generate high-resolution 3D urban block designs optimized for specific energy-reduction targets. The framework was tested on Guangzhou's six main Local Climate Zone types (LCZ1, LCZ2, LCZ4, LCZ5, LCZ6, and LCZ8) using stepwise 10% reductions in baseline energy use to produce low-carbon spatial configurations.
Why it matters — It replaces slow, iterative post-evaluation workflows with a generative design tool that directly translates quantitative energy-reduction goals into 3D urban forms.
Caveat: The model's optimization pathways and design generations are demonstrated only on the specific climate and urban morphology conditions of Guangzhou.