洛阳市客车行驶工况构建  

Bus driving cycle construction in Luoyang City

作　　者：高爱云 张兴源 刘少华 孟宇飞 GAO Aiyun;ZHANG Xingyuan;LIU Shaohua;MENG Yufei(College of Vehicle and Traffic Engineering,Henan University of Science and Technology,Luoyang,Henan 471003,China)

机构地区：[1]河南科技大学车辆与交通工程学院,河南洛阳471003

出　　处：《公路交通科技》2025年第4期171-178,共8页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(62371182)。

摘　　要：【目标】为解决传统K-means算法初始质心随机选取造成所构建行驶工况精度不高的问题,提出了一种基于改进K-means算法的工况构建方法——K*-means。【方法】首先,采集洛阳市3条典型线路上9辆客车的实车数据,并对所得原始数据进行数据预处理和运动学片段划分;然后,提取数据特征参数,并采用主成分分析法对特征参数进行降维;最后,设计K*-means算法对运动学片段进行聚类分析,按类中心距离选取最具代表性的运动学片段构建出洛阳市客车典型行驶工况。【结果】将所构建工况与基于传统K-means算法构建的工况以及国内外其他工况进行对比,结果表明,LY_BDC工况与原始数据的特征参数值最为接近,各特征参数值的平均相对误差为1.97%,其中怠速时间比的相对误差最大,仅有3.9%;基于传统K-means算法构建的行驶工况与原始数据的特征参数值平均相对误差为4.6%,其中平均速度的相对误差最大,达到了12.9%;中国典型城市公交循环与原始数据的加速时间比、怠速时间比和平均速度差异较大;新欧洲测试循环与原始数据的减速时间比、匀速时间比差异也很明显。【结论】所设计的K*-means算法提高了数据点数少的类获得质心的概率,并降低质心随机选取造成的影响,避免了孤立点的问题;基于K*-means算法所构建的工况精度更高,更能够准确地反映洛阳市客车的运行特征。[Objective]To solve the problem of constructed driving cycle low accuracy caused by the initial centroid random selection with traditional K-means algorithm,based on the improved K-means algorithm,a driving cycle construction method,K-means,was proposed.[Method]First,the real vehicle data from 9 buses on 3 typical lines were collected in Luoyang City.The raw data were preprocessed,and the kinematic segments were divided.Then,the data feature parameters were extracted.The feature parameters were reduced in dimensionality by using the principal component analysis.Finally,K-means algorithm was designed to cluster the kinematic segments.According to the distance among cluster centers,the most representative kinematic segments were selected to construct the typical bus driving cycles in Luoyang City.[Result]The constructed driving cycles were compared with the driving cycles constructed with traditional K-means algorithm and other driving cycles at home and abroad.The result indicates that the feature parameters of LY_BDC are the closest to the raw data.The average relative error of each feature parameter value is 1.97%,among which the relative error of idle time ratio is the largest,reaching 3.9%.The average relative error between the driving cycle based on traditional K-means and the raw data is 4.6%,and the relative error of average speed is the largest,reaching 12.9%.The acceleration time ratio,idle time ratio and average speed of China City Bus Cycle are quite different from the raw data.The difference of deceleration time ratio and uniform speed time ratio between New European Driving Cycle and raw data is also obvious.[Conclusion]The proposed K-means algorithm can improve the probability of obtaining centroid for clusters with few data points.It will reduce the influence of centroid random selection,and avoid the problem caused by isolated points.The driving cycle accuracy with K-means algorithm is higher,which can more accurately reflect the operating characteristics of buses in Luoyang City.

关 键 词：汽车工程 行驶工况 K~*-means算法 洛阳市客车 主成分分析法 

分 类 号：U491.2[交通运输工程—交通运输规划与管理]