高速公路短时交通流量组合预测方法研究

高速公路短时交通流量组合预测方法研究重庆大学硕士学位论文（学术学位）学生姓名：魏方强指导教师：余楚中副教授专业：交通信息工程及控制学科门类：工学重庆大学自动化学院二O一三年四月StudyonCombinationForecastMethodofShort-termTrafficFlowforFreewayAThesisSubmittedtoChongqingUniversityinPartialFulfillmentoftheRequirementfortheMaster’sDegreeofEngineeringByWeiFangqiangSupervisedbyAssociateProf.YuChuZhongSpecialty:TrafficInformationEngineeringandControlCollegeofAutomationofChongqingUniversity,Chongqing,ChinaApril2013中文摘要I摘要对高速公路交通状态的整体变化趋势进行有效把握，是对高速公路运营进行有效管理，防止交通拥堵的必要前提。高速公路交通流量预测是其中难度较大的核心问题，但现有预测方法的精度普遍不足，因此，交通流量预测方法研究近年受到广泛关注。通过改进预测方法从而提高流量预测的精度，对于改善高速公路交通管理水平具有重要意义。本文在对高速公路交通流特性进行深入分析的基础上，分别从高速公路交通流分车型的车流规律性、相似性及多断面的相关性出发，对高速公路短时交通流量预测方法的改进进行研究，提出了一种短时交通流量的组合预测方法。具体研究工作围绕分车型的交通流量预测、基于多断面相似性的交通流量预测、组合预测及实证分析三个方面展开：分车型的交通流量预测方法。首先对分车型进行流量预测的出发点进行阐述，其次进行车辆折算系数的确定，然后根据不同车型每天的交通流量波动情况差异，将小客车、中型车、大型车和拖挂车这四种车型分成两类，采用改进的时间序列算法对大型车和拖挂车的流量进行预测，利用二次指数平滑法对小客车和中型车的流量进行预测，最后通过车辆折算系数将各车型流量预测结果进行加权求和，得到总车流量预测结果。基于多断面相似性的交通流量预测方法。首先利用路段的最高限速和车检器的采样周期并结合各断面之间的距离进行相邻多断面的选取，其次根据预测断面和选取的多个相邻断面以及它们之间的互通收费站的车流量数据，建立关于流量的多维时间序列，最后给出一种相似性的匹配模式和改进的相似性度量方法，并利用多维时间序列的相似性对总车流量进行预测。组合预测方法的建立及实证分析。首先通过最优组合预测方式对上述两种预测方法进行组合，其次选取几个常用的预测评价指标建立一个模糊综合评价模型，最后通过渝武高速公路微波车检器采集的实测交通流量数据对该组合预测方法与其它预测方法进行了对比实验，并对各种预测方法的预测效果进行了分析和模糊评价。实验结果表明，无论在工作日还是节假日，本文所提出的组合预测方法均具有更好的预测效果，能够进一步提高短时交通流量预测的精度。关键词：交通流，短时预测，分车型，相似性，模糊综合评价重庆大学硕士学位论文II英文摘要IIIABSTRACTTograspthewholechangetrendofthefreewaytrafficstateeffectively,istheprerequisiteoffreewaymanagementandpreventthetrafficgetworse.Thetrafficflowforecastoffreewayisoneofthecoreissue,buttheaccuracyoftheexistingforecastmethodisnotenough,sotheforecastmethodoftrafficflowreceivedwidespreadattention.Toincreasetheaccuracyoftrafficflowforecastthoughimprovetheforecastmethodisofgreatimportanceforamelioratetheleveloffreewaymanagement.Onthebaseofanalysethetrafficflowcharacteristicsoffreeway,researchontheimprovementofshort-termtrafficflowforecastmethodwascarriedoutfromtheregularityofdifferenttypesofcar,similarityandthecorrelationofmultiplesections,onthisbasis,acombinationforecastmethodisproposed.Thespecificresearchworkwillbecarriedoutaroundthefollowingthreeaspects:Trafficflowforecastingmethodbasedondifferenttypesofvehicle.Primarily,thestartingpointoftrafficflowforecastingmethodbasedondifferenttypesofvehicleisexpounded,thenthevehicleconversioncoefficientisdecided,andaccordingtothedifferenceofdailytrafficflowfluctuationconditionofdifferenttypesofvehicle,thepassengercar,middle-sizedvehicle,oversizevehicleandtrailerisdividedintotwocategories,improvedtimeseriesalgorithmisusedtoforecastthetrafficflowofoversizevehicleandtrailer,andsecondexponentialsmoothingmethodisusedtoforecastthetrafficflowofpassengercarandmiddle-sizedvehicle.Finallythetrafficflowforecastresultsofeachtypeofvehicleareweightedsumthroughthevehicleconversioncoefficient,thenthetotaltrafficflowforecastingresultsareobtained.Trafficflowforecastingmethodbasedonmultiplesectionsandsimilarity.Primarily,theadjacentmultiplesectionsisselectedwiththespeedlimitofroadandthesamplingperiodofvehicledetectoraswellasthedistancebetweeneachsection,thenthemultidimensionaltimeseriesisestablishedwiththetrafficflowdataofforecastingsectionandtheselectedadjacentmultiplesectionsaswellasthetollstationbetweenthem,finally,asimilaritymatchingmodeandimprovedsimilaritymeasuremethodisgiven,andthetotaltrafficflowisforecastedusingthesimilarityofmultidimensionaltimeseries.Establishmentofthecombinationforecastmethodandempiricalanalysis.Primarily,theaforementionedtwokindsofforecastingmethodiscombinedthroughtheoptimal重庆大学硕士学位论文IVcombination,thenafuzzycomprehensiveevaluationmodelisestablishedwithseveralevaluationindicatorswhicharecommonlyusedforforecasting,finally,empiricalanalysisiscarriedoutforthecombinationforecastingmethodwhichproposedinthisarticleandothercombinationforecastingmethodwiththeactualmeasureddataofYU-WUfreeway,andtheforecastingresultsofthetwocombinationforecastingmethodisevaluatedthroughthefuzzycomprehensiveevaluationmodel.Theexperimentalresultsshowthatnomatteronweekdaysorholidays,thecombinationforecastingmethodproposedinthisarticlehasbetterforecastingresult,theaccuracyofshort-termtrafficflowforecastingcanbefurtherimproved.Keywords：trafficflow,short-termforecasting,differenttypesofvehicles,similarity,fuzzycomprehensiveevaluation.目录V目录中文摘要....................................................................................................................I英文摘要..................................................................................................................III1绪论........................................................................................................................11.1问题的提出........................................................................................................11.2国内外研究历史与现状.......................................................................................11.2.1基于线性系统理论的预测方法................................................................................21.2.2基于非线性理论的预测方法...................................................................