本文选取湖南省的人口数、人均GDP、基础设施固定资产投资、单位产值能耗比和单位能耗碳排放量作为高速公路基础设施的碳排放影响因素,选用湖南省2003—2021年相关数据并采用支持向量回归(SVR)机器学习法,建立了湖南省高速公路基础设施碳排放预测模型,预测在基准、低碳和超低碳情景下的碳排放数据。结果表明:训练样本交叉验证均方误差为0.007011,模型的预测值和真实值的拟合回归效果良好,训练集和测试集的相关系数分别为0.9869和0.9870,即模型具有良好的学习和推广能力。本文识别了碳排放的影响因素,预测了未来碳排放趋势,对交通基础设施碳减排行动具有一定的参考意义。

The construction and operation of highway infrastructure is responsible for a large amount of CO2 emissions. The largeamount of carbon dioxide emitted causes serious environmental pollution. In order to achieve sustainable development and greenproduction, highways need to reduce carbon emissions. It is of guiding significance to clarify the current status of carbon emissionsfrom transportation infrastructure, identify the influencing factors of carbon emissions, and predict future carbon emission trends. Thepurpose of this paper is to establish a model for scientifically predicting the carbon emissions of highway infrastructure. Throughliterature analysis and expert interviews, the influencing factors of carbon emissions of highway infrastructure were identified,including population, GDP per capita, fixed asset investment in transportation infrastructure, energy consumption per unit oftransportation output value, and carbon emissions per unit of transportation energy consumption. Support vector regression is used toconstruct a prediction model of highway infrastructure carbon emissions. Taking Hunan Province as an example, the data from HunanProvince from 2003 to 2021 were selected as samples to train the model. Fourteen samples were randomly selected as the training set,and the model was trained on historical data. The remaining 5 samples form a test set to test the trained model. The results show thatthe mean square error of cross-validation of the training sample is 0.007011, the fitting regression effect of the predicted value and thereal value of the model is good, and the correlation coefficients of the training set and the test set are 0.9869 and 0.9870, the model hasgood learning and generalization ability. Then, using scenario analysis and consulting relevant documents of the 14th Five-Year Planof Hunan Province, the future values of the five influencing factors were predicted and analyzed, and finally three scenarios were setup: low-carbon, benchmark and ultra-low-carbon. The results show that under the baseline scenario, the peak of carbon emissions willoccur in 2035, lagging behind the planned carbon peak process of Hunan Province. Even under the low-carbon scenario, the peak ofcarbon emissions will not be achieved until 2032. Only under the ultra-low-carbon scenario can the goal of carbon peak be achievedby 2030. This shows that Hunan Province still has a long way to go to achieve its carbon emission reduction target, and in order toachieve the goal of carbon peaking, the government should forecast the annual carbon emissions in advance and actively take carbonreduction measures to deal with it. The government can continue to promote carbon emission reduction by reducing energyconsumption per unit of output value and carbon emissions per unit of energy consumption. The carbon emission prediction model ofhighway infrastructure constructs a carbon emission regression equation by selecting macro factors at the economic, demographic andenergy technology levels, so as to improve the calculation efficiency of carbon emissions. Grasping the future carbon emission trendof expressway infrastructure can plan the construction and operation of expressways in advance, provide a scientific basis forformulating carbon emission reduction plans for transportation infrastructure, and achieve the goal of carbon neutrality. At the sametime, it provides a reference scheme for other industries to conduct research on carbon emissions and carbon peaking, and realizesindustry-wide carbon emission reduction.