组合主成分分析和模糊c均值聚类的车辆行驶工况制定方法  

A Method of Vehicle Driving Cycle Development Based on Combined Principal Component Analysis and Fuzzy C Means Clustering

作　　者：刘应吉[1,2] 夏鸿文 姚羽[1,2] 刘梦雅 

机构地区：[1]交通运输部公路科学研究院,北京100088 [2]运输车辆运行安全技术交通行业重点实验室,北京100088

出　　处：《公路交通科技》2018年第3期79-85,共7页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51405213);交通运输部应用基础研究计划项目(2014319223240);交通运输部公路科学研究院基本科研业务费项目(2015-9011);运输车辆运行安全技术交通行业重点实验室开放课题项目(KFKT2015-03);交通运输标准(定额)项目(2017-3302)

摘　　要：目前国内不同规模城市和地区的机动车行驶特征,与欧美采用的标准工况相比存在着较大的差异。为完善车辆行驶工况制定方法,开发了适合我国城市和地区交通特点的车辆行驶工况制定方法,研究采用道路运输企业营运车辆运行时产生的海量卫星定位数据,作为工况建立使用的源数据,这些数据能够真实反映拟建工况地区的各类型道路交通状况。方法首先将源数据计算划分为运动学片段即微行程,针对每一微行程进行平均速度和怠速时间占比等特征值的计算,利用改进后的主成分分析方法对特征矩阵进行降维处理,然后在特征空间利用模糊c均值聚类方法将这些微行程聚类成不同组。每一组聚类代表和生成一类交通状况,然后将各类子工况合成为初始合成工况。利用带有双权平滑核函数的滤波器来平滑初始工况,使其易于在实验室设备上开展车辆工况试验的遵循测试。最后选取在某地区行驶且时间段和路段较具代表性的道路运输车辆,利用其正常运行过程中产生的卫星定位数据,对所提出的工况制定方法进行了验证,并将所得工况与通过传统的V-A矩阵法所得工况进行结果的比较分析。试验数据的对比结果表明,应用该方法给出的合成工况能够反映出所建工况地区的交通状况,可以为开发符合当地交通特点的行驶工况提供依据。At present, the driving characteristics of vehicles in different cities and regions of China are different from those driving cycle standards of Europe and America. In order to improve the development method of our vehicle driving cycle, a method for developing the driving cycle which is suitable for the actual traffic characteristics of cities and regions in China is worked out. Mass commercial vehicle satellite positioning data of road transport enterprise are chosen as the source data used in this driving cycle development method, and the data is a real reflection of various types of road traffic condition in the proposed area. First, the source data is calculated and divided into some kinematics sequences, namely short micro trips, and driving characteristic values such as average speed and proportion of idle time are calculated for each micro trip, dimension reduction treatment is performed on characteristic matrix by improved principal component analysis, and the micro trips are then classified by using fuzzy c means clustering technique in eigen space. Each cluster can represent and generate a type of sub traffic condition, and then all kinds of sub driving cycle are synthesized as an initial synthesis driving cycle. A filter with biweight smoothing kernel function is applied to smooth initial driving cycle, which makes a driving cycle suitable for a laboratory test. Finally, the road transport vehicles which are traveling in a certain area with representative travel routes and time periods are selected as test vehicles, so the proposed method is verified by using the satellite positioning data generated during the driving process of these vehicles, and the given driving cycle is compared with the result obtained by the traditional V-A matrix method. The comparison result of the test data shows that the treated driving cycle obtained by the proposed method can reflect the traffic conditions in the proposed area, it can provide a reference for developing the driving cycle that accord with local traffic char

关 键 词：交通工程 行驶工况制定 主成分分析 模糊C均值聚类 平滑滤波 卫星定位数据 

分 类 号：U491[交通运输工程—交通运输规划与管理]