面向碳中和的设计路径：一种人工智能驱动的住宅平面生成与能耗优化框架

打开文本图片集

关键词：多模态输入；智能生成设计；建筑平面图；碳中和；人工智能

Abstract：Against the backdrop of the “dual-carbon"strategyand the Sustainable Development Goals，the building sector facesa dual mandate of energy reduction and intelligent transformation.Residential buildings account fora substantial share of end-use energy，and spatial layout decisions made in early design directly shape daylighting，ventilation，and thermal performance-making this stage pivotal for carbon neutrality. This paper proposes an Al-driven framework for residential floor-plan generationand energy optimization that integrates generativemodels， anenergy-use predictor，and evolutionaryoptimization intoaclosed loop driven bymultimodal inputs.Specifically：（1） an energyprediction model istrained on floor-plan simulation data to provide accurate monthly energy estimates;（2）adiffusion-based floor-plan generator takes building outlines and room blocksas inputs to produce diverse，spatially

1.引言

1.1研究背景

1.1.1以人为本的建筑平面设计流程

平面设计是建筑设计过程中的基础和核心环节[1]，它不仅决定了建筑的功能性和实用性，而且直接影响到建筑的安全性、经济性和可持续性[2-4]。（剩余19123字）