基于偏最小二乘回归的高速公路运营期碳排放计量模型

Measurement Model of Carbon Emission in Expressway Operation Period Based on Partial Least Squares Regression

为解决交通碳排放软测量建模时存在的自变量间相关关系损害模型精度和稳定性的问题,文章提出基于偏最小二乘回归改进高速公路运营期碳排放计量模型的方法,通过成分提取和综合筛选获得对碳排放解释能力更强的新综合变量,消除自变量间相关关系的不良影响。针对主要问题,文章选取存在相关关系的碳排放影响因子平均车速和道路坡度为自变量,碳排放为因变量,以长深高速公路小汽车的试验数据为样本集,运用偏最小二乘回归方法建立高速公路运营期碳排放计量模型,交叉校验模型的有效性。对比模型预测结果与最小二乘回归模型的拟合结果,结果表明:模型的最大相对误差为8.7%,仅为最小二乘回归模型最大相对误差的1/3,可见应用偏最小二乘回归法消除了自变量间相关关系影响的碳排放计量模型精度更高,稳定性更好。研究对交通碳排放软测量建模技术的发展有重要的指导意义。

In order to solve the problem,correlation between variables damage model accuracy and stability when the traffic carbon emission soft measurement modeling,a method based on partial least squares regression to improve the measurement model of the carbon emission during the operation period of the expressway is proposed. Through the composition extraction and comprehensive screening,the new comprehensive variables of stronger explanation for carbon emissions are obtained,and the adverse effects of the correlation between the independent variables are eliminated. In view of the main problem,the paper selects related carbon emissions impact factor average vehicle speed and road grade exist as independent variables,and carbon emissions as the depen？dent variable,with Changchun-Shenzhen expressway car test data as sample set,partial least squares regression method is used to establish the highway operation and management of carbon emissions measurement model and cross validation model effectiveness.The model prediction results and the least square regression model fitting results are compared. The results show that the maximum relative error of the model is 8.7%,only for the least squares regression model has the maximal relative error of 1/3. Therefore,the application of partial least squares regression method of econometric model of carbon emission to eliminate the self correlation between variables,model has higher accuracy and better stability. The research has important guiding significance to the development of soft sensing modeling technology of traffic carbon emission.