基于MOEA和ANN的严寒地区办公建筑形态节能设计决策支持模型研究

Building form is the interaction interface of indoor and outdoor environment. The building form energy-efficiency design decisions significantly affect the building energy consumption and indoor environmental performance. In recent years, the rising of energy-saving demands, carbon emission reduction requirement and environmental comfort standards required higher accuracy and efficiency for building form energy-efficiency design decision-making processes. The stern climate of severe cold region leads to natural deficiency of building form energy-efficiency design. However, the large number of office buildings and high level energy consumption give high energy-efficiency potential and wide social benefits to building form energy-efficiency design in severe cold regions. The study aims to achieve the multi-disciplinary information adaptive coordination through the parametric modeling of the building environment and performance information based on the regional characteristics of the office buildings in the severe cold regions, and to break through the bottleneck of green performance prediction efficiency through modeling the neural network by building form and green performance mapping. Through coupling multi-objective evolutionary algorithm (MOEA) and Artificial Neural Network (ANN) model, the study will develop office building energy saving design decision supporting model to improve the precision and efficiency of office building energy saving design decision-making in severe cold regions, and to reduce the energy consumption and carbon emission, and improve the daylighting and thermal performance of office buildings in severe cold regions.

建筑形态是室内外环境交互界面，其节能设计决策显著影响建筑能耗与室内环境性能。近年来，节能减排与环境舒适标准提升对建筑形态节能设计决策提出了更高的精度与效率要求。严寒地区气候严酷，建筑形态节能设计“先天不足”，而办公建筑总体基数大、单体能耗高，故其节能潜力大，社会效益广。研究旨在立足严寒地区办公建筑形态地域特征，通过建筑环境与性能信息参数化建模，实现多学科信息自适应协同；通过建筑形态与绿色性能映射关系神经网络建模，突破绿色性能预测效率瓶颈；通过耦合多目标进化算法（MOEA）和人工神经网络（ANN）模型研发办公建筑形态节能设计决策支持模型，提高严寒地区办公建筑形态节能设计决策制定精度与效率，降低严寒地区办公建筑能耗与碳排放水平，改善建筑自然采光与热舒适环境性能。

项目面向节能减排国家战略，立足寒地办公建筑地域特征，针对建筑形态节能设计决策中的瓶颈问题，基于机器学习、进化计算和性能驱动设计理论，融合建筑设计流程与人工智能技术流程，展开了建筑环境信息建模、绿色性能神经网络预测和设计决策支持三方面研究。.课题展开了建筑围护结构与环境交互机制研究，解析了影响建筑能耗和光热环境舒适度等绿色性能的建筑形态设计参量，探索了基于低空摄影测量技术的寒地建筑环境信息建模与模拟技术，综合应用建筑信息建模、建筑性能模拟和参数化编程技术，探索了建筑信息、地域环境信息和绿色性能信息的参数化关联关系，研发了建筑环境与性能信息参数化模型。针对局地气候对建筑性能仿真精度的影响，通过参数化编程，探索了基于局地光气候的严寒地区建筑光环境仿真方法，进而基于机器学习理论，综合应用拉丁超立方体抽样、建筑性能模拟技术、人工神经网络建模技术，探索了寒地建筑绿色性能映射神经网络预测模型。围绕建筑形态节能设计决策支持瓶颈问题，基于性能驱动设计理论，应用多目标进化算法整合建筑环境信息模型和绿色性能神经网络预测模型，以建筑绿色性能为适应度函数，探索了寒地建筑设计决策支持工具。.项目以建筑多绿色性能预测为技术基础，通过多绿色性能导向形态优化，提高了建筑形态设计决策对绿色性能提升多方面需求的权衡能力；通过研发建筑形态节能设计决策支持模型，将建筑形态探索方式由既有节能设计实践采用的穷举试错比较转换为种群进化迭代搜索，通过种群进化迭代搜索拓展形态节能设计决策过程对形态可能性探索的广度；通过建构建筑环境与性能信息参数化模型，为优化设计过程融入了寒地建筑与环境地域特征信息流，从而能够在方案阶段更准确地反映寒地城乡地域特征，从而提升形态节能设计决策过程对地域特征回应精度。研究成果将有助于推动“节能减排”战略落实，促进建筑设计产业的信息化转型。