【摘要】 住区整体规模形态特征对碳排放会产生长期的影响。当前对低碳住区的研究多集中在单体建筑方面，对住区整体规模尺度和碳排放关系的量化研究并不多见。以四川金堂县杨柳北片区为例，利用实地调研和软件模拟的方式获取63处住区碳排放清单数据，选取容积率、空间紧凑度和街廓面积三个规模特征指标，基于XGBoost机器学习算法建立“住区规模——碳排放”预测模型，并据此模型对住区规模进行优化。研究发现，住区规模与碳排放之间并非简单线性关系，而是多指标综合作用下的非线性相关关系。通过机器学习方法可以建立量化的预测模型，并合理预测低碳住区的最优规模区间。就金堂县而言，住区街廓尺度在150～180 m,面积15 000～25 000 m~2,紧凑度在20～23最为低碳。本研究对控规层面住区碳排放预测，低碳住区规划设计、评价和规划管控的理论方法建立具有借鉴意义。更多还原

【Abstract】 The scale of residential area refers to the land area and development intensity of residential area. The overall scale and morphological characteristics of residential area will have a long-term impact on carbon emission. At present, there are many studies on the carbon emission of residential buildings, but the quantitative research on the relationship between the residential scale and carbon emission is rare. Taking Yangliubei area of Jintang County, Sichuan Province as an example, this study obtains the carbon emission data of residential areas by means of field investigation and software simulation, selects three core scale index elements of plot ratio, spatial compactness and street area, establishes the residential carbon emission prediction model of residential areas based on XGBoost machine-learning model, and optimizes the residential scale based on the model. It is found that there is a nonlinear correlation between residential scale and residential building carbon emission, and the machine-learning model can predict residential carbon emission through residential scale index. In terms of the conclusion of this study, the carbon emission of residential area in Jintang county is lowest when the street scale is 150～180 m, the total area is 15 000～25 000 m~2 and compact ratio is 20～23. This study has reference significance for the prediction of plot carbon emission and the establishment of theoretical methods of low-carbon planning, design, evaluation and planning management and control.更多还原