基于机器学习的京津冀城市群建筑业碳排放影响因素分析及碳达峰预测

Analysis of influencing factors and carbon peak prediction of construction-industry carbon emissions in the Beijing-Tianjin-Hebei Urban Agglomeration based on machine learning

  • 摘要: 在京津冀协同发展战略与建筑业绿色低碳转型背景下,识别该区域建筑业碳排放的影响因素并预测其达峰路径具有重要意义。基于2003—2023年京津冀城市群面板数据,采用政府间气候变化专门委员会推荐的排放因子法测算建筑业施工阶段直接碳排放量;在方差膨胀因子检验基础上筛选9项核心影响因素,构建GWO-GRU-XGBoost复合预测模型,并运用Shapley加性解释方法解析各因素的贡献特征;据此设定基准、低碳与高碳三种情景,预测各城市碳达峰路径。结果表明:1)京津冀建筑业碳排放总量由8 664.03万t升至19 424.44万t,整体呈波动上升态势,空间上呈现以唐山、北京、天津和石家庄为高值区、承德和秦皇岛等为低值区的“中间高、南北低”的分布特征;2)R&D内部经费支出、第一产业增加值、专利授权数、能源消耗总量、常住人口城镇化率、第二产业增加值、第三产业结构、公路客运量和建筑业从业人员数是影响建筑业碳排放的核心因素,且表现出显著的非线性特征,其边际效应在特定阈值前后存在阶段性突变及复杂的交互作用;3)低碳情景下,京津冀城市群整体于2023年达峰并平稳回落;高碳情景下,排放短期下降后持续回升,截至2035年尚未形成稳定峰值。基于此,建议对唐山、北京、石家庄等高排放城市强化施工用能和清洁替代约束,对保定、廊坊等反弹风险城市加强高耗能项目准入,并推动京津绿色技术向河北重点城市扩散。

     

    Abstract: Against the background of the coordinated development of the Beijing-Tianjin-Hebei region and the green and low-carbon transition of the construction industry, it is of great significance to identify the influencing factors of construction-industry carbon emissions and predict carbon-peaking pathways. Based on panel data from cities in the Beijing-Tianjin-Hebei urban agglomeration from 2003 to 2023, this study measured the direct carbon emissions of the construction industry during the construction phase using the emission factor method recommended by the Intergovernmental Panel on Climate Change. Nine core influencing factors were selected through variance inflation factor tests. A GWO-GRU-XGBoost hybrid prediction model was then constructed, and the Shapley additive explanations method was used to interpret the contribution features of each factor. On this basis, three scenarios, namely the baseline, low-carbon, and high-carbon scenarios, were designed to predict city-level carbon-peaking pathways. The results showed that: (1) Construction-stage direct carbon emissions of the construction industry in the Beijing-Tianjin-Hebei region increased from 8664.03×104 t to 19424.44×104 t, showing an overall fluctuating upward trend. Spatially, Tangshan, Beijing, Tianjin, and Shijiazhuang were high-emission areas, whereas Chengde and Qinhuangdao remained at relatively low emission levels, forming a spatial pattern of high emissions in the central region and lower emissions in the northern and southern regions. (2) Intramural R&D expenditure, added value of the primary industry, number of patent authorizations, total energy consumption, urbanization rate of permanent residents, added value of the secondary industry, tertiary-industry structure, road passenger traffic, and number of construction-industry employees were the core factors affecting construction-industry carbon emissions. These factors exhibited significant nonlinear characteristics, and their marginal effects showed abrupt stage-specific shifts around specific thresholds and complex interactions. (3) Under the low-carbon scenario, construction-industry carbon emissions in the Beijing-Tianjin-Hebei urban agglomeration peaked in 2023 and subsequently declined steadily. Under the high-carbon scenario, emissions rebounded after a short-term decline, and no stable peak is expected by 2035. Accordingly, Tangshan, Beijing, Shijiazhuang, and other high-emission cities should strengthen constraints on construction energy use and clean-energy substitution; Baoding, Langfang, and other cities facing emission-rebound risks should tighten access requirements for energy-intensive projects; and green technologies from Beijing and Tianjin should be promoted across key cities in Hebei Province.

     

/

返回文章
返回