In response to the growing demand for living environment, enhancing the physical environment of residential buildings has become an imperative priority. This study proposes an optimization and low-carbon decision-making framework. Interestingly, a multi-objective optimization model was developed by integrating a backpropagation neural network with the NSGA-II algorithm, and a carbon emission model was incorporated into the decision-making process to obtain optimal design parameters. To substantiate the applicability of this methodology, it will be applied to a prototypical model of residential buildings, which consists of multiple physical environmental units. Five types of design variables were identified, including layout dimension, window-to-wall ratio, building orientation, building envelope, and openable window area ratio. And the correlation analysis was conducted on three optimization objectives of useful daylight illuminance, percentage of predicted dissatisfaction and indoor CO2 concentration. The results indicate that all design variables have been retained as key design variables, and the three objectives meet the necessity of multi-objective optimization. Moreover, the constructed neural network prediction model has high accuracy. Compared with the prototypical model, layout dimensions of the optimal solution exhibited the greatest changes in the width and depth of kitchen, with respective increases by 25.38 % and 21.46 %. The unit modules of the residential buildings with the worst performance have been effectively optimized, the useful daylight illuminance has increased by 2.23 %, the percentage of predicted dissatisfaction has decreased by 11.5 %, the indoor CO2 concentration has decreased by 48 %, and the operational carbon emissions per unit area has decreased by 14 %.

中文翻译：

---

低碳决策兼顾采光、热舒适和室内空气质量的住宅建筑布局尺寸优化设计


为了满足日益增长的生活环境需求，改善住宅建筑的物理环境已成为当务之急。本研究提出了一个优化和低碳决策框架。有趣的是，通过将反向传播神经网络与 NSGA-II 算法集成，开发了多目标优化模型，并将碳排放模型纳入决策过程以获得最优设计参数。为了证实此方法的适用性，它将应用于住宅建筑的原型模型，该模型由多个物理环境单元组成。确定了 5 种类型的设计变量，包括布局尺寸、窗墙比、建筑朝向、建筑围护结构和可打开窗户面积比。并对有效日照度、预测不满意百分比和室内 CO2 浓度 3 个优化目标进行相关性分析。结果表明，所有设计变量都被保留为关键设计变量，三个目标满足多目标优化的必要性。此外，所构建的神经网络预测模型具有较高的准确率。与原型模型相比，最优解的布局尺寸在厨房的宽度和深度上表现出最大的变化，分别增加了 25.38 % 和 21.46 %。性能最差的住宅楼的单元模块得到了有效优化，有效日照度增加了 2.23 %，预测不满意的百分比降低了 11 %。5 %，室内 CO2 浓度下降了 48 %，单位面积的运营碳排放量下降了 14 %。