With the rapid development of generative AI-driven technologies, their applications in the field of architectural design are continuously expanding. From the generation of architectural conceptual ideas to design optimization and performance evaluation, artificial intelligence is reshaping traditional design processes. How to achieve automated integration from conceptual design to building energy performance simulation, so as to improve design efficiency and support low-carbon building decision-making, has become an important research direction in the fields of architectural design, the built environment, and intelligent construction.

Journal of Building Design and Environment (JBDE) Frontier Forum. This issue is honored to invite Prof. Qingyan Chen from The Hong Kong Polytechnic University to deliver a presentation on the cutting-edge theme Generative AI Empowers Architectural Design: Automated Workflow from Conceptual Design to Building Energy Performance Simulation. He will deeply explore the innovative applications of generative AI in architectural design, as well as how to achieve the full-process intelligence and automation from design conception to building energy performance analysis.

Meeting Format

1. Meeting Theme

Generative AI-Driven Architectural Design: Automating the Transition from Concept to Building Energy Simulation

2. Meeting Time

3:00 PM, July 22 (Beijing Time)

3. Format

Zoom Meeting Meeting ID: 862 6753 2033 Passcode: 340429

Presentation Content

The early stage of architectural design has a decisive impact on overall building performance. However, translating creative conceptual design processes into structured building energy consumption models remains a major bottleneck for the industry. To address this issue, this study proposes a generative AI‑based workflow that can automatically convert design intent into building energy simulation input files compatible with EnergyPlus. In the first workflow, ChatGPT‑Image and Hunyuan3D‑2.5 are used to automatically convert 2D conceptual designs into 3D mesh models, which are further applied to building energy performance simulation. In the second workflow, Claude 3.7 Sonnet acts as an “automated architect” and constructs geometric building models by generating Python code. Both approaches enable direct performance evaluation of design alternatives within the EnergyPlus platform. Tests were conducted using a case study of an office building in Hong Kong. The results show that optimization of building form and HVAC systems can significantly reduce energy consumption, while largely eliminating the extensive manual modeling time required in traditional building energy modeling. This study demonstrates the great potential of generative AI in architectural design, enabling seamless integration and automated coordination between creative design exploration and building energy performance optimization.