基于人工神经网络的严寒地区可持续办公建筑多目标优化设计方法研究

With the rapid growth of the office buildings located in severe cold region of China, its energy issues have become increasingly acute. The existing sustainable design methods for the office building located in severe cold region are limited by two problems. First, the existing sustainable design decision making processes rely on the experience judgment, which decrease the accuracy and scientific of sustainable design; besides, the performance evaluation consumed a lot of time which restricts the efficiency of sustainable building design process. These limitations have become the bottle necks in the development of sustainable office building located in severe cold region. This study will introduce artificial neural networks and multi-objective optimization technique to explore the next generation sustainable building design platform and method based on the energy consumption characters and configuration features of the office buildings located in severe cold region. The study will couple artificial neural networks, building information model and multi-objective optimization model through parametric programming technique to achieve the sustainable design decision making process guided by building performance data, which will reduce the building performance evaluation time and increase the accuracy and efficiency of sustainable design. This study will upgrade the information level in the sustainable office building design process and promote the development of sustainable office building in severe cold region.

我国严寒地区办公建筑建设量迅猛增长，能耗问题日益突出。既有严寒地区可持续办公建筑设计决策制定过程多依赖于建筑师的主观经验，信息化水平不足，降低了可持续设计决策的科学性和准确性，且建筑性能评价周期长，降低了可持续建筑设计效率，制约了我国可持续办公建筑发展。课题研究立足严寒地区办公建筑能耗构成、形态空间与构造材料特征，在充分调研分析既有建筑样本的基础上，综合应用神经网络建模与多目标优化搜索等前沿技术，展开新一代可持续建筑设计模型研发与方法探索。. 课题深入挖掘严寒地区地域特质，应用实测调查统计数据训练神经网络模型，采用参数化编程技术耦合神经网络模型、建筑信息模型和多目标优化搜索模型，旨在实现数据指导下的可持续设计决策制定，大幅缩减建筑性能评价时间，提高可持续设计决策精确性和设计效率，促进严寒地区可持续办公建筑设计过程的信息化转型，推动严寒地区可持续办公建筑发展。

严寒地区气候恶劣，建筑室内光热环境优化难度高，节能减排潜力大。既有严寒地区办公建筑设计缺乏局地数据体系支持、性能预测精度效率较低，且既有办公建筑设计决策过程依赖设计者主观经验，对建筑多性能的权衡能力和建筑设计可能性探索广度均有限。研究旨在立足严寒地区地域特征，通过建构严寒地区办公建筑数据体系，研发办公建筑参数化建筑环境信息模型、性能预测神经网络模型和多目标优化设计模型，提出基于神经网络的严寒地区可持续办公建筑多目标优化设计方法，突破既有技术瓶颈，提高严寒地区办公建筑设计精度与效率。.课题主要研究内容和重要结果包括：通过抽样调查与实地测量，综合应用气象站、全天空扫描仪、室内温湿度记录仪、热流计、照度计和开窗计数器等设备，建构了涵盖建筑信息、环境信息、性能信息和使用者行为信息的严寒地区办公建筑数据体系；根据严寒地区办公建筑数据体系，综合应用参数化编程、机器学习建模等技术，研发了能够基于参数逻辑生成、编辑建筑环境信息的严寒地区可持续办公建筑参数化建筑环境信息模型，为人工神经网络训练数据生成奠定基础；应用机器学习建模技术，依托参数化建筑环境信息模型生成神经网络模型训练数据，建构了严寒地区办公建筑性能预测神经网络模型，实践验证表明其具有较高的预测精度，并能显著降低办公建筑性能预测耗时；耦合办公建筑性能预测神经网络模型与多目标优化算法，研发了基于神经网络的严寒地区可持续办公建筑多目标优化设计模型和方法，实现了多性能导向下的严寒地区办公建筑优化设计。.研究成果深化了设计者对严寒地区办公建筑地域特征的考虑，突破了严寒地区办公建筑性能预测与设计决策支持技术瓶颈，改善了严寒地区办公建筑设计精度和效率，提升了我国严寒地区可持续办公建筑设计的智能化水平，深化了人工智能技术与建筑设计理论方法的融合，将推动建筑设计行业的信息化与工业化转型。