基于建筑碳排放与室内舒适度双目标优化方法的研究  

Research on Building Carbon Emissions and Indoor Thermal Comfort Based on Multi-objective Optimization Method

作　　者：朱泓宇 肖敏 张嘉敏 ZHU Hong-yu;XIAO Min;ZHANG Jia-min(a.College of Architecture,Changsha University of Science and Technology,Changsha 410114,China;College of Electrical and Information Engineering,Changsha University of Science and Technology,Changsha 410114,China)

机构地区：[1]长沙理工大学建筑学院,长沙410114 [2]长沙理工大学电气与信息工程学院,长沙410114

出　　处：《湖南师范大学自然科学学报》2023年第6期93-102,共10页Journal of Natural Science of Hunan Normal University

基　　金：住房和城乡建设部科学技术计划项目(2021-K-105);长沙理工大学研究生创新项目(CLSJCX22147)。

摘　　要：为实现建筑碳排放与室内热舒适度的高精度预测与性能优化,建立了基于反向传播神经网络与优化算法相结合的综合优化框架。首先,构造以实际调研数据为基础的长沙市居住建筑模型,使用5种采样方法随机生成输入参数,并模拟生成数据库。其次,基于综合灵敏度分析方法筛选出对碳排放量与热舒适度具有重要影响的决策参数。然后,使用筛选后的数据集训练反向传播神经网络(BPNN)模型,并多指标验证模型的可行性。最后,基于Pareto前沿集结果选择非支配排序遗传算法-Ⅲ与BPNN相结合求解多目标优化问题。结果表明,拉丁超立方采样(LHS)方法生成的训练数据集在进行综合灵敏度分析和优化后,其BPNN的回归系数(R 2)可达到0.977。相较于基础模型案例,最优碳排放方案的碳排放量降低了27.3%,室内不舒适度时间减少了3.5%,证明该方法在建筑节能优化领域的有效性。In order to realize the high-precision prediction and high-performance optimization of building carbon emissions and indoor thermal comfort,a comprehensive optimization framework based on the combination of backpropagation neural network(BPNN)and optimization algorithm was established.Firstly,a residential building model of Changsha based on actual survey data was constructed,the input parameters were randomly generated using five sampling methods,and the database was generated by simulation.Secondly,based on the comprehensive sensitivity analysis method,the decision parameters with important effects on the carbon emissions and thermal comfort were screened out.Then,the BPNN model was trained using the filtered dataset,and the feasibility of the model was verified by multiple metrics.Finally,the Pareto frontier set result selection nondominated ranking genetic algorithm-Ⅲ(NSGA-Ⅲ)combined with BPNN were used to solve the multi-objective optimization problem.Results show that the regression coefficient(R 2)of BPNN can reach 0.977 after the comprehensive sensitivity analysis of the training dataset generated by the Latin hypercubic sampling(LHS)method.Compared with the basic model case,the carbon emissions of the optimal carbon emission scheme are reduced by 27.3%,and the indoor discomfort time is reduced by 3.5%,which proves the effectiveness of the proposed method in the field of building energy efficiency optimization.

关 键 词：多目标优化 建筑碳排放 室内热舒适性 优化算法 

分 类 号：TU375.4[建筑科学—结构工程] TU201.5[自动化与计算机技术—控制理论与控制工程] TP18[自动化与计算机技术—控制科学与工程]