考虑数据缺失的高速公路交通参数计算方法  

Calculation Method of Expressway Traffic Parameter Considering Missing Data

作　　者：刘春生 曹蓉 王晓晗 贾健民 LIU Chun-sheng;CAO Rong;WANG Xiao-han;JIA Jian-min(School of Traffic Engineering,Shandong Jianzhu University,Jinan 250101,China;Shandong Expressway Co.Ltd.Jinan 250101,China)

机构地区：[1]山东建筑大学交通工程学院,济南市250101 [2]山东高速股份有限公司,济南市250101

出　　处：《公路》2023年第12期215-221,共7页Highway

基　　金：国家自然科学基金项目,项目编号41901396,42001396;山东省高等学校青创科技支持计划,项目编号2021KJ058。

摘　　要：为提高数据异常状态下的高速公路交通参数计算的精确度,提出了基于数据融合的交通参数计算方法,首先结合收费站数据对门架数据所缺失的行车记录进行填充,再根据每辆车的平均行程速度,利用两阶段的聚类算法剔除驶入服务区和行驶异常的车辆,最后计算各个路段的路段流量和路段平均行程速度,对缺失的交通参数利用融合时间特征的KNN算法进行填充。选取3d15个门架的行车数据作为实验数据。结果表明,门架数据行程记录缺失较大,平均缺失率为46.95%,其中门架数据记录最大缺失率为56.25%,最小缺失率为31.73%,并且速度较快的小客车相较于速度较慢的中大型货车的行车记录更容易缺失。两阶段的聚类算法可以有效地识别并剔除驶入服务区和行驶异常的车辆。在填充连续复杂缺失场景下的交通流时空数据集方面,KNN算法相较于最大似然估计、决策树、链式方程多重填补方法在RMSE指标上分别下降24.90、15.88、5.81,添加时间特征以后RMSE在原始KNN基础上下降3.03,证明了融合时间特征的KNN算法在填充连续复杂缺失的交通流场景下的可行性和有效性。数据异常情景下的交通流缺失值填充以及交通参数计算方法为管理部门在高速公路交通规划和管理方面提供了技术支持。In order to improve the accuracy of expressway traffic parameter calculation under abnormal data,a traffic parameter calculation method based on data fusion is proposed.Firstly,the missing traffic records are filled by combining toll station data with portal data.Then,according to the average travel speed of each vehicle,the two-stage clustering algorithm is used to eliminate vehicles entering the service area and driving abnormally.Finally,the traffic flow and average travel speed of each section are calculated,and the missing traffic parameters are filled by KNN algorithm with time characteristics.The driving data of 15gantry frames in 3days are selected as the experimental data.The results show that the travel records of gantry data are missing greatly,with an average missing rate of 46.95%,and the maximum missing rate of gantry data records is 56.25%and the minimum missing rate is 31.73%.Moreover,the fast passenger cars are more likely to miss the travel records than the slow medium and large trucks.The two-stage clustering algorithm can effectively identify and eliminate vehicles driving into the service area and abnormal driving.In terms of filling the spatio-temporal data set of traffic flow in continuous complex missing scenes,the RMSE index of KNN algorithm decrease by 24.90,15.88and 5.81,respectively,compared with the maximum likelihood estimation,decision tree and chain equation multiple filling method.After adding the time feature,RMSE decreases by 3.03on the basis of the original KNN.The feasibility and effectiveness of KNN algorithm based on time feature fusion are proved in the case of continuous complex missing traffic flow scenes.The filling of missing value of traffic flow and the calculation method of traffic parameters under the abnormal data scenario provide technical support for the management department in terms of expressway traffic planning and management.

关 键 词：智能交通 门架数据 交通参数 缺失值填充 

分 类 号：U491.11[交通运输工程—交通运输规划与管理]