基于高预测性能代理模型的建筑绿色性能优化设计研究——以寒地办公建筑采光与能耗性能为例  

Study on Building Green Performances Optimization Design Based on Surrogate Models with High Prediction Performances Taking Daylighting and Energy Performances of Office Buildings in Severe Cold Regions as an Example

作　　者：孙澄[1,2] 董禹含 梁静 SUN Cheng;DONG Yuhan;LIANG Jing

机构地区：[1]哈尔滨工业大学建筑与设计学院,哈尔滨150001 [2]寒地城乡人居环境科学与技术工业和信息化部重点实验室,哈尔滨150001

出　　处：《建筑学报》2024年第S2期112-117,共6页Architectural Journal

基　　金：国家自然科学基金重点项目(51938003);黑龙江省重点研发计划项目(GZ20210211);中央高校基本科研业务费专项资金项目(2022FRFK01012);“十三五”国家重点研发计划项目(2016YFC0700209)

摘　　要：广泛应用于绿色建筑设计的既有多目标优化方法优化结果质量不佳、运算耗时大。代理模型建构技术可在一定程度上代替高耗时的绿色性能模拟,为优化结果质量的提升与运算成本的降低提供了契机。针对常用代理模型的预测性能相对不佳的问题,建构高预测性能代理模型并将其嵌入多目标优化框架中;以寒地办公建筑的采光与能耗性能为例展开实践,有效改善了复合性能水平;优化结果显示,该方法较基于协同性能模拟、常用代理模型的两种既有多目标优化方法可获取更高质量的优化结果与更低的运算成本。Poor optimization results quality and high computation costs lie in the existing multi-objective optimization methods widely used in green building design.The technique of surrogate modeling is capable of replacing expensive simulations to a certain extent,which provides opportunities to improve the quality of optimization results and reduce the computation costs.Aiming at the weakness of the commonly used surrogates,the study constructs surrogates with high prediction performances and embeds the surrogates into the multi-objective optimization framework;a practical case is carried out with daylighting and energy performances of office buildings in severe cold regions as an example,which improves the composite performances level effectively;the optimization results possess better quality and the computation costs are much lower compared with two existing methods of co-simulation based and commonly used surrogate based multiobjective optimization.

关 键 词：建筑绿色性能 多目标优化 高预测性能代理模型 优化结果质量 模拟运算成本 

分 类 号：TU243[建筑科学—建筑设计及理论] TU201.5