ECO DRIVING FOR TRANSIT - Home - NCST · ECO-DRIVING FOR TRANSIT ... Literature Review ... report “pump-to-wheel” (occurring at the vehicle) fuel consumption, greenhouse gas ...

ECO-DRIVINGFORTRANSIT

April2016 AWhitePaperfromtheNationalCenterforSustainableTransportation

YanzhiAnnXu,HanyanAnnLi,HaobingLiu,MichaelO.Rodgers,andRandallGuensler

GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

AbouttheNationalCenterforSustainableTransportationTheNationalCenterforSustainableTransportationisaconsortiumofleadinguniversitiescommittedtoadvancinganenvironmentallysustainabletransportationsystemthroughcutting-edgeresearch,directpolicyengagement,andeducationofourfutureleaders.Consortiummembersinclude:UniversityofCalifornia,Davis;UniversityofCalifornia,Riverside;UniversityofSouthernCalifornia;CaliforniaStateUniversity,LongBeach;GeorgiaInstituteofTechnology;andUniversityofVermont.Moreinformationcanbefoundat:ncst.ucdavis.edu.DisclaimerThecontentsofthisreportreflecttheviewsoftheauthors,whoareresponsibleforthefactsandtheaccuracyoftheinformationpresentedherein.ThisdocumentisdisseminatedunderthesponsorshipoftheUnitedStatesDepartmentofTransportation’sUniversityTransportationCentersprogram,intheinterestofinformationexchange.TheU.S.Governmentassumesnoliabilityforthecontentsorusethereof.NordoesthecontentnecessarilyreflecttheofficialviewsorpoliciesoftheU.S.Government.Thisreportdoesnotconstituteastandard,specification,orregulation.AcknowledgmentsThisstudywasfundedbyagrantfromtheNationalCenterforSustainableTransportation(NCST),supportedbyUSDOTthroughtheUniversityTransportationCentersprogram.TheauthorwouldliketothanktheNCSTandUSDOTfortheirsupportofuniversity-basedresearchintransportation,andespeciallyforthefundingprovidedinsupportofthisproject.TheauthorswouldliketothankreviewersattheCaliforniaAirResourcesBoard,GeorgiaRegionalTransportationAuthority,andPedalLogicLPforprovidingcommentsonpreliminaryversionsofthiswhitepaper.

Eco-DrivingforTransitANationalCenterforSustainableTransportationWhitePaper

April2016

Yanzhi(Ann)Xu,GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

Hanyan(Ann)Li,GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

HaobingLiu,GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

MichaelO.Rodgers,GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

RandallGuensler,GeorgiaInstituteofTechnology,SchoolofCivilandEnvironmentalEngineering

[pageleftintentionallyblank]

i

TABLEOFCONTENTSEXECUTIVESUMMARY....................................................................................................................ii

Introduction....................................................................................................................................1

LiteratureReview...........................................................................................................................3

Data................................................................................................................................................7

Methodology..................................................................................................................................8

Eco-drivingCycleDevelopment..................................................................................................9

ImplementationofEco-drivingCycleModification..................................................................12

FuelandEmissionsAnalysis.....................................................................................................17

Results..........................................................................................................................................20

Eco-drivingScenario.................................................................................................................20

CNGFleetPurchaseScenario...................................................................................................22

Eco-drivingwithCNGFleetPurchaseScenario........................................................................23

OverallComparisonandDiscussion.........................................................................................24

Conclusions...................................................................................................................................29

AppendixI.....................................................................................................................................31

KalmanFiltering........................................................................................................................31

SplineDataInfillProcess..........................................................................................................31

GISNetworkMapping..............................................................................................................32

Engine-offIdentification...........................................................................................................33

BreakingGPSDataintoTripsforEco-DrivingAnalysis.............................................................35

DataProcessingResults............................................................................................................35

AppendixII....................................................................................................................................40

AppendixIII...................................................................................................................................42

References....................................................................................................................................44

ii

Eco-drivingforTransit

EXECUTIVESUMMARYEco-drivinghassignificantpotentialtoreducefuelconsumptionandemissionsfromtransitoperations.Analyseswereconductedof68thousandmilesofreal-worldoperationsdatafrom26buses,collectedfromlocaltransitserviceprovidedbytheMetropolitanAtlantaRapidTransitAuthority(MARTA),andexpressbusserviceprovidedbytheGeorgiaRegionalTransportationAuthority(GRTA).Theanalysisutilizedsecond-by-secondoperationsdatacollectedviaglobalpositioningsystem(GPS)devicesfrombusesoperatedbythesetransitagencies.Theresearcherssimulatedtheimplementationoftransiteco-drivingstrategies,basedonthemodalemissionsmodelingframeworkemployedbytheMOtorVehicleEmissionSimulator(MOVES)designedtoreduceengineloadandemissions.Thisalgorithmseekstominimizefuelconsumptionbylimitinginstantaneousvehiclespecificpower(VSP),whilemaintainingaveragespeedandconservingtotaldistance.Fuelconsumptionandfuel-cycleemissionswerecomparedacrossthemonitoreddrivingcyclesandtheirmodifiedeco-drivingcycles.Thesavingsfromeco-drivingwerealsocomparedagainstexpectedfuelandemissionsreductionsviaconversionofthetransitfleetstocompressednaturalgas(CNG),whichisanotherpopularfuelconservationstrategy.Thetransiteco-drivingstrategyshoweda5%reductioninfuelconsumptionandfuelcyclegreenhousegas(GHG)emissionsforMARTA’s508-busfleet(~35%diesel/65%CNG),anda7%reductioninfuelconsumptionforGRTA’s166-busdieselfleet.Thefuelsavingstranslatetoabout300,000gallonsofdieselfuelequivalentperyearforMARTAand55,000gallonsofdieselperyearforGRTA.Eco-drivingwasalsoshowntoreducefueluseandemissionsforCNGfleets.Eco-drivingtrainingcanreadilybeimplementedifspeed/accelerationactivityismonitored.Becauseeco-drivingdoesnotrequiresignificantcapitalinvestmentitisapotentiallyverycost-effectivestrategyforlocalandexpressbustransitoperations.

1

IntroductionTransitagenciesarealwaysseekingopportunitiestoconservefuel(whichtypicallyprovidessimultaneousemissionsreductions)toloweroperatingcosts.Strategiesrangefrommakingwisenewvehiclepurchasedecisions,suchasalternativepropulsion/fuelbuses,tomakingoperationalimprovements,suchasimplementinganti-idlepoliciesandeco-drivingtraining.Eachemissionsreductionalternativeoffersdifferentreturn-on-investment(ROI),dependinguponthelocalconditionsandoperationalcharacteristicsofeachagency.Furthercomplicatingtheevaluationisthefactthatemissionsreductionsfromstrategiesarenotnecessarilyadditive.Inselectingasetofemissionsreductionstrategiestoimplement,transitagenciesneedtoevaluatemultipleoptionssimultaneously,underagency-specificoperatingcharacteristics.Thispaperfocusesontransitfuelandemissionssavingsfromeco-drivingfortwotransitagencies.Theanalysesinthisreportarebaseduponreal-worldoperationsdatacollectedfromtheMetropolitanAtlantaRapidTransitAuthority(MARTA),alocaltransitagency,andtheGeorgiaRegionalTransportationAuthority(GRTA),whichprovidesregionalexpressbusservices.Thepotentialreductionsinfuelconsumptionarederivedfromoperationalimprovementsachievedthroughdriverbehaviormodification,predominantlylimitingvehicleaccelerationratesandtopspeeds.Thepotentialbenefitsarequantifiedusinganeweco-drivingalgorithmdevelopedforthisproject.Theanalysesextendbeyondfuelconsumptionandtailpipeemissions.Anyreductioninfuelconsumptionatthevehiclealsoreducesfuelconsumptionandemissionsalongtheentirefuelchain:harvestingfuelfeedstocks,refiningandprocessingthefeedstocksintofuels,anddistributingthefuels.Theanalysesthatfollowwillreport“pump-to-wheel”(occurringatthevehicle)fuelconsumption,greenhousegas(GHG)emissions,andcriteriapollutantemissionsand“well-to-wheel”GHGandcriteriaairpollutantemissions(associatedwiththeentirefuelchain).Inadditiontooperationalimprovements,suchaseco-driving,transitagencieshavealsoshownincreasinginterestinthedeploymentofalternativefuelbusesasastrategytolowertotalfuelcosts(TCRP,2010).Compressednaturalgas(CNG)isaparticularlypopularchoiceofalternativefuel,especiallyinlightofrecentdecreasesinCNGpricesduetoincreasedfrackingactivity.Asof2014,morethan10,000busesintheUnitedStatesarerunningonCNG,comparedtoabout4,000hybriddieselbuses(NationalTransitDatabase,2014).Therefore,thisprojectevaluateseco-drivingasastand-alonestrategy,butthesavingsfromeco-drivingisalsoputintoperspectivebyindependentlyandsimultaneouslyevaluatingfuelandemissionssavingsfromconvertingtheexitingfleetstoCNG.Thepaperfirstprovidesaliteraturereviewoneco-driving,asafuelconsumptionandemissionscontrolstrategyfortransitoperations.Thecollectionofthedataemployedinthisstudyisthendescribedandsummarystatisticsofthedataarepresented.Thedevelopmentoftheeco-drivingalgorithmusedintheanalysisofpotentialbenefitsisthenoutlined.Thisalgorithmwouldbeusedtotraindriversandassesstheironroadperformanceofeco-drivinginterventions

2

(i.e.afterintervention,istherestillroomforimprovementforthedriver).Thecomparativefuelconsumptionandemissionreductionresultsthatcouldbeachievedwitheco-drivinginterventionforthemonitoreddataarethensummarized,andthencomparedtothebenefitsthatcouldbeobtainedfromfleetconversiontoCNG.Assumingthatthemonitoreddataareroughlyrepresentativeoffleetoperations,transiteco-drivingcouldyielda5%reductioninfuelconsumptionforMARTA’sfleetanda7%reductioninfuelconsumptionforGRTA’sXpressbusfleet(morefreewayoperations).Thereductionstranslatetoabout300,000gallonsofdieselfuelperyearforMARTAand55,000gallonsofdieselperyearforGRTA.Eco-drivingcanalsoreducefueluseandemissionsfromCNGfleets.Becauseeco-drivingtrainingisrelativelyeasytoimplementwhenspeed/accelerationactivityismonitored,andbecausemonitoringcanbepaidforthroughfuelsavings,theresearchteamconcludesthateco-drivingstrategiesareareasonableapproachtoreducingfleetemissionsinlocalandexpressbustransitoperations.

3

LiteratureReviewEco-drivingtrainingiswell-knownasafeasiblestrategytodecreasefuelconsumptionandemissions.Itisgenerallyacceptedthateco-drivingencompassesthefollowingdrivingtactics(IntelligentEnergyEurope,2013):anticipatingtraffic,limitinghighspeedoperations,avoidinghardacceleration,shiftingtothehighestavailablegearrpmwillallow,maintainingasteadyspeed,andlimitingidling.Existingstudieshaveevaluatedthebenefitsofeco-drivingthroughreal-worldimplementation,throughsimulatedvehicleactivitydata,orthroughacombinationofboth.Inreal-worldimplementations,theobservedfuelsavingsrangefrom2%to14%(BarthandBoriboonsomsin,2009;Beusen,etal.,2009;Dib,etal.,2014;Ho,etal.,2015;Rutty,etal.,2013;StrömbergandKarlsson,2013;TransportCanada,2004;Wåhlberg,2007;Zarkadoula,etal.,2007).Inaddition,Rolim,etal.(2014)reportedthatdriverswithinstantin-cabvoicefeedbackshowedmuchmorereductionsinhardaccelerationscomparedtodriverswhoonlyreceivedin-classeco-drivingtraining,althoughtheactualfuelsavingsfromthesetwoeco-drivingstrategiescomparedtoabaselineconditionwasnotreported.Estimatedeco-drivingbenefitsthroughsimulatedvehicledataexhibithighervariabilitythanthebenefitsobservedinreal-worldimplementation,rangingfrom8%toabout35%infuelsavingsandCO2reduction(BarthandBoriboonsomsin,2009;Mensing,etal.,2014;QianandChung,2011;SuzdalevaandNagy,2011).Eco-drivingstudiesbasedonsimulationshavedevisedarangeofdrivingstrategiestorepresenttheimplementationofeco-drivingobjectives.Inmoststudies,eco-drivingstrategiesarerealizedthroughmodifyingvehiclespeedand/oracceleration.BarthandBoriboonsomsin(2009)devisedadynamiceco-drivingsystemthroughwhichdriversareprovidedwithsuggestedspeedsbasedonaveragetrafficspeedandlevel-of-service(LOS)forthefreewaysectiononwhichthevehiclewasoperating.Mensing,etal.(2013)createdanumericalmodelofthevelocitytrajectoryofavehicleoperatingaccordingtoeco-drivingprinciplesandreal-lifetrafficconstraints.Usingsimulatedtrafficdata,QianandChung(2011)evaluatedfuelconsumptionandCO2emissionsofeco-drivingbyreducingthemaximumaccelerationratesby10%and20%insimulation.SuzdalevaandNagy(2011)developedadata-basedBayesianapproachtoidentifyandmodifythespeedtooptimizefuelconsumptionforconventionalvehicles.However,allofthesealgorithmsweredesignedforlight-dutyvehicles.Table1summarizestheresultsfromthevarietyofstudiesidentifiedandreviewedinthisresearcheffort.

4

Table1.SummaryofEco-DrivingBenefitResearch

Source VehicleType

BeforeData

AfterData Methodology TimeScope FuelSavings/CO2Reduction

/PollutantReduction

BarthandBoriboonsomsin,2009

Light-dutyvehicles

Real-worldvehicleactivitydata

Simulatedvehicleactivitydataandreal-worldvehicleactivitydata

Staticrecommendedspeedtodrivers;Simulationmodelingtoolsandreal-worldvehicleexperimentation

3probevehiclesonfreewaysSeptember2005,May2006,andMarch2007

Fuelsavings:13%(real-world),37%(simulated)CO2

reduction:12%(real-world),-35%(simulated)Savingsdependoncongestion

Beusen,etal.,2009

Light-dutyvehicles



Four-hourtraining;10drivers;Atleast100kmofdrivingpermonth

Twomonthsofdatacollection:10monthsfor10driversduringreal-lifeconditions,monitoredweekly

Fuelsaving:5.8%withlargedifferencesbetweenindividuals

Dib,etal.,2014 EV



ParticipantsdroveEVinfixedroute.Energycomparisonweremadebeforeandafterecotraining

N/A Fuelsavings:14%forEV

Ho,etal.,2015

Light-dutyvehicles



116participants;Classroomtraining

Pre-testofdrivers;30to45mintrainingsessions;Re-testdriversrightaftertraining

Fuelsavingandcarbonemissions:inexcessof10%

Mensing,etal.,2014

Light-dutyvehicles

Simulatedvehicleactivitydata

Simulatedvehicleactivitydatabasedontheoptimizationmethod

Simulatingaconventionalpassengervehicle;Applyingoptimizationmethodstoachieveecologicallyandeconomicallyoptimalvehicleoperations

N/A

EconomiccycleFuelsaving:2.5L/100kmCO2

reduction:31.9%NOXreduction:16.4%

EcologiccycleFuelsaving:2.3L/100kmCO2

reduction:26.8%NOXreduction:54.5%

5

Source VehicleType

BeforeData


/PollutantReductionHCreduction:7.4%

QianandChung,2011

Light-dutyvehicles



Trafficmicro-simulationmodel;Differenttrafficcondition,penetrationratesofeco-drivers,andaccelerationrates

N/A

Scenariosofheavycongestionand25%penetrationimpactstrafficandenvironmentalperformancenegatively;Moderateandsmoothaccelerationsaves11%fuelwithoutmajorincreaseintraveltime

Rutty,etal.,2013

Light-dutyvehicles



11gasolinevehicles,4hybridvehicles;40kmperdayGoal-directedfeedback

Post-trainingdatacollection:1monthTraining:1monthPost-trainingdatacollection:1month

Fuelsavings:0.48Lpergasolinevehicleperday;0.3LperhybridvehicleperdayCO2

reduction:1.1kgpergasolinevehicleperday;0.6kgperhybridvehicleperday.Baselinedatanotreported

StrömbergandKarlsson,2013 Buses



54busdrivers,dividedinto3groups:control,eco-drivingfeedbackonly,andeco-drivingfeedbacksupplementedwithtraining

Baseline:3weeksTestperiod:3weeks

6.8%reductioninfuelconsumptionbetweentheeco-drivinggroupsandcontrolgroup;nosignificantdifferencebetweenthetwoeco-drivinggroups

SuzdalevaandNagy,2011

Light-dutyvehicles



Bayesianapproachtoidentifyandmodifythespeedinordertooptimizefuelconsumption

N/A Fuelsavings:8.2%overall

TransportCanada,2004

Light-dutyvehicles



Approximately1,000corporateemployeeswereidentified;Training,2-hourclassroomand2-houron-road

Pre-testofdrivers;4-hourtraining;Re-testdriversafteroneyear

Fuelsavings:5.5%overall

Wåhlberg,2007 Buses Real-worldvehicleactivity

Real-worldvehicleactivity

Phase1:practicaleco-drivingtraining;247trained 2000to2003

Trainingprovided2%fuelreduction;

6

Source VehicleType

BeforeData


/PollutantReductiondata data driversvs.147untrained

drivers;Phase2:28buseswereequippedwithfeedbackdevices

feedbackprovidedanother2%fuelreduction

Zarkadoula,etal.,2007 Buses



3drivers,2buses;Fixed15kmrouteTrainingseminar

Pre-training1.5months;Post-training2months

Fuelsavings:10.2%duringtraining,4.35%inactualcondition

7

DataToevaluatethepotentialemissionsandfuelconsumptionsassociatedwitheco-drivingfortransitoperationsintheAtlantametropolitanarea,second-by-secondtransitoperationsdatawerecollectedfromlocaltransitoperationsandregionalexpressbuses.Forlocaltransitoperations,MetropolitanAtlantaRapidTransitAuthority(MARTA)operationsdatawerecollectedon13busesfor381days(June28,2004toOct24,2005)usingtheGeorgiaTech(GT)TripDataCollector(Ogle,etal.,2006).Forexpressbuses,datawerecollectedviaspotsampling(typicallytwotothreedaydeploymentsbetweenAugust6,2013andMarch3,2014).QstarzBT-Q1000eXGPSloggersweretemporarilyinstalledonGeorgiaRegionalTransportationAuthority(GRTA)expressbusesinthissamplingeffort.Inall,second-by-secondreal-worldtransitoperationsdatawerecollectedformorethan68thousandmilesfromfrom26buses.TheGPSdataunderwentqualityassurance/qualitycontrol(QA/QC)andpostprocessingbeforebeingusedintheanalyses.First,aninitialscreeningwasperformedtoremovetripsshorterthanone-minuteanddatapointswithinvalidlatitudeandlongitudeinformation.Second,speedvaluesweretreatedwithaKalmanfilteralgorithmtoreplacelow-validityGPSspeeds(typicallyatlowspeedsandinurbanstreetcanyons)withlocation-inferredspeed,usingasplinealgorithmtofillindatagaps.Afterthesedataprocessingsteps,thedatawereoverlaidonbaseGISmapstoidentifythetypeoffacility(i.e.freeway,non-freeway,oroff-network)onwhichabuswasoperatingforeachsecondofthedrivingrecord.Distinguishingbetweenfreewayfromnon-freewayoperationsisanimportantstepforsubsequentanalysisbecausetheeco-drivingstrategiesforfreewaysandnon-freewaysdiffersubstantially.Theonroadandoff-networkdistinctionisusedtoidentifyandeliminateextendedidle.BecauseneithertheGTTripDataCollectornortheQstarzGPSloggershadtheabilitytodetectwhethertheenginewason,whichwouldrequireanoilpressuresensor(Xu,etal.,2013a)oron-boarddiagnostics(OBD)connection,therewasnofeasiblewaytodeterminewhetherabuswasidlingwhenthespeedvalueswerenearzero.Assuch,theteamelectedtoignorethepotentialbenefitsofidlereductioninthispaper.Somejurisdictionsassumethatbusesshouldnotidleformorethan10minutesonroad,andshouldnotidleformorethan30secondsoff-network.DifferentidlespeedcutpointsaresetforMARTAandGRTAoperationsgiventhedifferencesindeviceprecisionlevels(Xu,etal.,2013b).Nevertheless,idlereductionisanotherviablestrategythatcanbeimplmenetedtoreduceemissions(Shancita,etal.,2014;Xu,etal.,2013a).Finally,tripfilesrecordedbythedatacollectiondeviceswerebrokenintotripsegments,separatedbygapsindataattripends,andwheregapsresultedfrommissingdata.Onlythosetripsegmentslongerthan30secondswithanaveragespeedof5milesperhour(mph)orgreaterwereretainedforsubsequenteco-drivinganalysis.Table2summarizesthefinalanalyticaldataset.Step-by-stepdescriptionsofthedataprocessingproceduresareprovidedinAppendixI.

8

Table2.SummaryofAnalyticalDataSet

TypeofOperation LocalTransit ExpressServiceAgency MARTA GRTANumberofBuses 13 13NumberofTrips 9,984 852TotalDistance(miles) 61,247 3,637TotalDuration(hours) 3,716 84AverageSpeed(mph) 16.5 43.3

MethodologyFollowingQA/QCprocessinganddatapreparation,theobserveddrivingcyclesfromMARTAandGRTAweremodifiedtoreflecttheimplementationofeco-drivingstrategies.Forcomparisonpurposes,themodifieddrivingcyclesarereferredtoaseco-cycles.Theobservedcyclesandeco-cycleswerethenemployedinfuelandemissionsanalysisinparalleltoassesstheenergyandenvironmentalbenefitsofeco-driving.Figure1depictsthegeneralprocessofthestudymethod.Afterinitialprocessingoftherawdataobtainedfromtransitmonitoringdevices,theobserveddrivingcyclesarelinkedtoemissionratesfromtheU.S.EnvironmentalProtectionAgency’sMOtorVehicleEmissionSimulator(MOVES)toestimatefuelconsumptionandemissions,andalsototheGreenhouseGases,RegulatedEmissions,andEnergyUseinTransportationModel(GREET)modeltoestimatewell-to-pumpfuelconsumptionandemissions(ArgonneNationalLaboratory2013;EPA,2014;Xu,etal.,2015;Guensler,etal.,2016;Guensler,etal.,2015).Theobserveddrivingcyclesarethenpost-processedtogeneratetheeco-cycles,asdescribedlater,andthenalsolinkedtoMOVESandGREETtoestimatecomparativefuelconsumptionandemissions.

9

Figure1.MethodologyFlowChart

Eco-drivingCycleDevelopmentThebasicapproachtoeco-drivingistolimitenginepowerdemandsoastoconservefuelandreduceemissions.Powerdemandisanon-linearfunctionofspeedandacceleration;hence,managingengineloadistypicallyaccomplishedbymanagingtopspeeds(forwindresistance)andaccelerationrates(forallloadparameters).Engineloadisalsohighduringluggingoperations(accelerationfromthestopline),sominimizingstopandgoactivityisagoalofeco-driving.However,becauseengineloadinvolvestheproductofspeedandacceleration,itisevenmoreimportanttoensurethathardaccelerationconditionsdonotoccuratmoderateandhighspeedoperations.Asindicatedintheliteraturereview,avarietyofemissionratemodelshavebeendevelopedtopredictemissionsfromheavy-dutyvehicleoperations.Modelsthatpredictemissionsasafunctionofoperatingmode(speed/accelerationconditions)arecommonlyknownas“modalmodels.”Thesemodalmodelsrangefromhigh-resolutionengineloadmodelsthatpredictsecond-by-secondemissionsasafunctionofpredictedinstantaneousengineload(Barth,etal.,1996;Feng,etal.,2007;Guensler,etal.,2005),tomodelsthatpredictsecond-by-secondemissionrates(oraverageemissionratesforaroadway)asafunctionofsomesurrogateforengineload.Thewiderangeinpotentialbenefitseco-drivingnotedintheliteraturearisesinpartfromtheapplicationofawiderangeofmodelingapproaches.Theeco-drivingstrategy(i.e.optimalchangeindrivingcycletoachieveemissionsreductions)isthereforeadirectfunctionofthemodelemployedintheanalysis.Theeco-drivinganalysesreportedinthisstudyemploythemodalemissionsmodelingframeworkintheU.S.EPA’sMOVESmodel.TheMOVESmodelusesscaledtractivepower(STP)asasurrogateforengineload,whereSTPisafunctionofvehiclespeed,acceleration,andvehiclemass.MOVESemploysabinningapproach,suchthathigherSTPvalueswithinspecific

10

operatingspeedbinsarelinkedtohigherfuelconsumption,CO2emissions,andcriteriapollutantemissions.Figure2presentsthefuelrateformodelyear(MY)2010transitbusesofeachoperatingmodebin(definedbyspeedandSTPranges)extractedfromMOVES.HighspeedsandhardaccelerationsatmoderateorhighspeedspushtheonroadactivityintohigherSTPvaluesandyieldhigherfuelconsumptionandemissions.Indevelopingthestrategytogenerateeco-cyclesforusewithMOVES,thegoalistomodifyeachvehicle’strajectorytominimizeactivityinhigherSTPbins,whilepreservingaveragespeedandtotaldistance.

Figure2.FuelRate(MJ/h)foreachOperatingModeBinfor2010MYTransitBuses(MOVES2014Output)

ThisstudyproposesanewmethodforoptimizingeachvehicletrajectoriesbaseduponthestructureoftheMOVESSTPoperatingmodebins.Themethodologyconservescycledistance,maintainsoverallaveragespeed,butpreventsinstantaneousSTPfromincreasingsignificantlybysettingaccelerationlimitswithineachMOVESspeedgrouping.STPiscalculatedas:

𝑆𝑇𝑃 = !!𝑣 + !

!𝑣! + !

!𝑣! + !

!𝑎𝑐𝑐 + 𝑔 ∗ sin𝜃 𝑣(1)

0 1000 2000 3000 4000 5000 6000

Brake Idle

STP <0 STP 0-3 STP 3-6 STP 6-9

STP 9-12 STP ≥12 STP <0

STP 0-3 STP 3-6 STP 6-9

STP 9-12 STP 12-18 STP 18-24 STP 24-30

STP ≥30 STP <6

STP 6-12 STP 12-18 STP 18-24 STP 24-30

STP ≥30

Spee

d 1-

25 m

ph

Spee

d 25

-50

mph

Sp

eed ≥5

0 m

ph

FuelRate(MJ/h)

11

where:A=therollingresistancecoefficient(kWs/m)B=therotationalresistancecoefficient(kWs2/m2)C=theaerodynamicdragcoefficient(kWs3/m3)m=vehiclemass(metrictonnes)M=fixedmassfactor(unitless)v=instantaneousvehiclevelocityattimet(m/s)a=instantaneousvehicleacceleration(m/s2)g=gravitationalaccelerationwiththevalue9.8(m/s2)θ=roadgrade(radiansordegrees,asrequiredbythesincalculationalgorithm)

A,B,C,andMarefixedparametersforeachvehicletypemodeledinMOVES.Thevaluescanbefoundin“sourceusetype”tableintheMOVESdatabase(providedinAppendixII).Forsimplification,alloftheanalysesinthisreportassumezeroroadgrade(sin𝜃 = 0).STPincreasesmonotonicallywithspeedandacceleration.Thefirststepintheeco-drivingprocessistosetaSTPlimitvalue(STPL).Foreachspeed𝑣,theaccelerationlimitaccLthatpreventSTPfromexceedingSTPLis:

𝑎𝑐𝑐! =!"#!!!"

− !!

− !!

𝑣 − !!

𝑣! (2)FromtheMOVESoperatingmodeclassification,eachspeedlevelincludesdifferentSTPlevels.BasedtheSTPcategories(seeTable11),wecansetSTPLtodifferentlevels:STPL-1=30,STPL-2=24,STPL-3=18,STPL-4=12,STPL-5=9,STPL-6=6,andSTPL-7=3.FromSTPL-1toSTPL-7,foragivenspeed,theaccelerationlimitbecomesmorestringent,asillustratedinFigure3.

12

Figure3.AccelerationLimitforeachSpeedLevelateachSTPLimitLevel

TheprincipleofthecomputationalmethodistoreadeachsecondofvehicleactivityandadjusttheaccelerationratedownwardwhentheSTPreachesorexceedstheSTPL.TheaccelerationrateisadjusteddownwardenoughtolowertheSTPofthenextdatapointtothemedianvalueoftheSTPrangethatmeetstheSTPlimit.Forexample,iftheSTPlimitissetasSTPL-4=12,whenSTPreachesorexceeds12,theaccelerationwillbeadjusteddownwarduntilthecalculatedSTPforthatdatapointequals10.5(themedianSTPvalueforthe9to12STPbin,whichmeetstheSTPL).ItisimportanttosetappropriateSTPlimitsbydrivingcycle.Iftherulesaretoolenient,theruleswillnotsignificantlyfuelconsumptionandemissions.However,iftherulesaretoostringent,theaveragespeedofthetracewillbesignificantlylower,whichmaybedifficultfordriverstoaccept.Furthermore,areductioninaveragespeedleadstoincreaseddrivingtime,offsettingsomeofthefuelandemissionssavings.Inthisstudy,theresearchteamestablisheddifferentSTPlimitsbyroadtypeandspeedafteriterativetesting.TheresultingSTPlimitsaresummarizedinTable3.

Table3.STPLimitsforLocalRoadsandFreewaysEmployedinThisAnalysis

RoadType SpeedLevel STPLimit

LocalRoad0~25mph <=625~50mph <=6>=50mph <=6

Freeway0~25mph <=625~50mph <=9>=50mph <=12

ImplementationofEco-drivingCycleModificationToimplementtheeco-drivingstrategy,threeiterativestepsappliedtoeachvehicletrajectory:

1 MaintainingStatusQuo:WhentheSTPoforiginalcycledoesn’treachorexceedSTPL,nomodificationofthecycleisrequired.Thenextdatapointintheeco-cycleisthesameasthedatapointfromtheoriginalcycle.

2 Smoothing:WhentheSTPoforiginalcyclereachesorexceedstheSTPL,theacceleration

rateisadjusteddownwardsuchthattheresultingSTPforthedatapointequalsthemedianvaluefortheSTPbinthatdoesnotexceedtheSTPlimit.Becausetheaccelerationratedecreases,thespeedofthenextdatapointintheeco-cyclewillbeslightlylowerthanthespeedforthatpointintheoriginalcycle.TheaccelerationratesforsubsequentpointsinthecyclearealsosettoachievethemedianSTPvalueforthat

13

STPbin.Smoothingofaccelerationcontinuesuntilthespeedofeco-cyclematchesthatoftheoriginalcycle.

3 ConservationofDistance:Oncethespeedofeco-cycleandoriginalcyclealign,the

distancecoveredbytheeco-cycleislessthanthatoftheoriginalcycle(duetotheimplementationofloweraccelerationrates).Toconservedistancetraveled,theeco-cyclecruisespeedisextendeduntilthedistancetraversedbytheeco-cyclematchesthatoftheoriginalcycle.Thisstepassumesthatthevehicleisnotlimitedbythepresenceofaslower-movingvehicleinitspath.

Figure4illustratestheresultsofthethreestepsappliedtoshortdrivingcycle.Inthisfigure,theinitialtrajectoriesofecoandoriginalcycleareexactlythesame(statusquo)becausetheearlyportionofthecycledoesnotexceedSTPL.OncetheSTPLisexceeded,smoothingbeginsandtheaccelerationratesoftheeco-cycleissetlowerthanthoseobserved.Smoothingisnormallyfollowedconservationofdistancetoensurethatthevehicletraversesthesamedistanceintheeco-cycleasintheobservedcycle.Anexampleofanobservedcycleanditscorrespondingeco-cycleispresentedinFigure5.Themodifiedeco-cyclesmoothedthesharpacceleration,especiallyduringhighspeedoperations.Figure6providesthefullflowchartdescribingtheimplementationoftheeco-drivingalgorithmanditerationprocesses.

Figure4.Eco-cycleExample

0

20

40

60

80

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100 103 106 109 112 115 118 121 124 127 130

Spee

d (m

ph)

Time (second)

Original Cycle Eco Cycle

Smoothing

Maintaining Status Quo

Conservation of Distance

Maintaining Status Quo (time not in sync)

Start point STP>=STP L

Speed(eco)= Speed(original)

Distance(eco)= Distance(original)

14

Figure5.ExampleofanObservedCycleandCorrespondingEco-cycle

15

Figure6.AlgorithmFlowChart

Cycle Optimization

1-steady status: STP<STPLEcoSpeed(i)=OriginalSpeed(i)EcoAcc(i)=OriginalSpeed(i)

STP (i)>=STPL?

next second(i=i+1)No

2-smoothing status:EcoAcc = Acc L

EcoSpeed(i)=EcoSpeed(i-1)+EcoAcc(i-1)

Yes

EcoSpeed (i) = OriginalSpeed (i)next second

(i=i+1)

Yes

No

3-conserving distance status:EcoSpeed(i) = EcoSpeed(i-1)EcoAcc = 0

EcoDistance(i)= OriginalDistance(i)?

next second(i=i+1)

Yes

No

Selection STP Limit Level

STPL-1=30 STPL-2=24 STPL-3=18 STPL-4=12 STPL-5=9 STPL-6=6 STPL-7=3

Average speed of eco cycle significantly smaller than original cycle

Fuel/emission of interest from eco cycle does not change much compared with original cycle

No

Optimization Complete

Yes

, Cho

ose

a le

ssst

rict r

ule

Yes

, Cho

ose

a m

ore

stric

t rul

e

16

Figure7showsthespeed-acceleration(acceleration>0mph/sec)scatterplotsfromtheobservedcyclesandeco-cycles.ThehardaccelerationratesinobservedcycleshavebeenreducedtokeeptheSTPbelowtheSTPlimitintheeco-cycle.ThedashedlinesinFigure7correspondtotheaccelerationlimitsforeachspeedlevelatSTPthreshold6,9,and12inFigure3.Afterthemodification,theoveralldistanceincreasesby0.05%,andtheoverallspeedreductionis3.07%,withinwhichthehighwayandlocalspeedreductionis1.93%and3.13%respectively.

Figure7.Speed-AccelerationScatterPlot(10,000-secondSample)

STP limit-12: for highway (>=50 mph)

STP limit-9: for highway (25~50 mph)

17

FuelandEmissionsAnalysisThegoalofthefuelandemissionsanalysisistoassesstheimpactoftheeco-drivingandotherfuelsavingstrategiesatthetransitagencylevel.Todoso,real-worldoperationsdatacollectedfromMARTAandGRTAbusesareusedandapplicablefuelconsumptionandemissionratesareappliedasiftheentirefleetexperiencestheobservedoperatingconditions.ToprovideafaircomparisonbetweendieselandCNG,weevaluatedthefullfuelcycle(i.e.well-to-wheelemissions)followingtheapproachedadoptedintheFuelandEmissionsCalculatorforTransitFleets(Xu,etal.,2015).Pump-to-wheelemissionswereestimatedusingMOVES-Matrix(Guensler,etal.,2015),amulti-dimensionalemissionratelookuptablederiveddirectlyfrommillionsofMOVESemissionraterunsforAtlanta.ThedetailedsettingsfortheMOVESrunsarelistedinAppendixIII.Well-to-pumpemissionswereestimatedusingtheGREETmodel(ArgonneNationalLaboratory,2015).Toestimatetheemissionsandfuelconsumptionsfortheentirefleet,fleetsizeandannualmileageinformationwastakenfromtheNationalTransitDatabase(NTD)(2014).TheNTDdoesnotprovideinformationofoperatingmileageondifferentroadtypes,butdoesdifferentiatebetweenrevenueandnon-revenue(alsoknownasdeadhead)mileage.Therefore,proportionsoffreewayandnon-freewaymileagewereestimatedseparatelyforrevenueanddeadheadoperations,usingspatialanalysisinArcGIS(detailsareprovidedinAppendixI).Table4summarizesthepump-to-wheelemissionratesforlocaltransitserviceandTable5summarizesthepump-to-wheelemissionratesforexpressbusservice.Well-to-pumpemissions(emissionstoharvestfeedstock,processfeedstockintofuel,anddeliverfueltothepump)wereestimatedforthepredictedpump-to-wheelfuelconsumption.

18

Table4.Pump-to-wheelEmissionRatesforLocalTransitBasedforMARTABuses

Fuel DutyCycle Revenue/Deadhead RoadType Duration

(seconds)

Avg.Speed(mph)

Distance(miles)

EmissionRate(g/mile/vehicle;MJ/mile/vehicleforfuel)

HC CO NOx PM2.5 CO2 GHGs Fuel

Diesel(E

xisting)

Observed Deadhead

Local 4,545,040 16.9 21,390 1.72 6.27 16.54 1.07 2,142 2,142 29.1Freeway 301,994 54.7 4,591 0.68 3.42 9.25 0.46 1,341 1,341 18.2

RevenueLocal 19,090,668 16.5 87,424 1.72 6.27 16.54 1.07 2,142 2,142 29.1Freeway 172,750 54 2,589 0.68 3.42 9.25 0.46 1,341 1,341 18.2

Eco

DeadheadLocal 4,686,687 16.4 21,344 1.82 6.49 16.02 1.01 2,032 2,032 27.6Freeway 310,210 53.8 4,636 0.69 3.53 9.07 0.47 1,327 1,327 18.0

RevenueLocal 19,711,963 16.0 87,463 1.82 6.49 16.02 1.01 2,032 2,032 27.6Freeway 178,599 52.7 2,613 0.69 3.53 9.07 0.47 1,327 1,327 18.0

CNG(E

xisting)

Observed Deadhead

Local 4,545,040 16.9 21,390 24.80 15.45 11.28 0.05 2,104 2,682 35.6Freeway 301,994 54.7 4,591 7.33 9.52 7.71 0.08 1,237 1,407 21.0


Eco



CNG(N

ew)

Observed Deadhead

Local 4,545,040 16.9 21,390 3.32 6.00 1.93 0.00 1,975 2,054 33.5Freeway 301,994 54.7 4,591 0.73 2.99 1.31 0.00 1,161 1,178 19.7


Eco



19

Table5.Pump-to-wheelEmissionRatesforExpressServiceBasedforGRTABuses

Fuel DutyCycle

Revenue/Deadhead

RoadType

Duration(seconds)

Avg.Speed(mph)

Distance(miles)

EmissionRate(g/mile/vehicle;MJ/mile/vehicleforfuel)

HC CO NOx PM2.5 CO2 GHGs Fuel

Diesel(E

xisting)

Observed Deadhead

Local 143,802 32.2 1,285 0.59 2.41 9.56 0.50 1,891 1,892 25.7Freeway 123,346 59.6 2,041 0.37 1.82 7.33 0.29 1,523 1,523 20.7

RevenueLocal 87,694 20.3 494 0.89 3.04 11.36 0.60 2,139 2,140 29.0Freeway 225,294 55.6 3,482 0.39 1.88 7.41 0.30 1,525 1,526 20.7

Eco

DeadheadLocal 160,670 28.5 1,271 0.72 2.86 8.31 0.43 1,543 1,544 20.9Freeway 126,231 58.2 2,042 0.37 1.92 6.78 0.27 1,430 1,430 19.4


CNG(N

ew)

Observed Deadhead

Local 143,802 32.2 1,285 2.23 5.89 1.86 0.00 1,738 1,791 29.4Freeway 123,346 59.6 2,041 0.74 2.87 1.37 0.00 1,270 1,287 21.5


Eco

DeadheadLocal 160,670 28.5 1,271 2.00 4.65 1.33 0.00 1,478 1,525 25.0Freeway 126,231 58.2 2,042 0.72 2.72 1.25 0.00 1,160 1,177 19.7


20

ResultsInthissection,wepresentfuelandemissionsresultsforthreescenarios.Scenario1evaluatestheimplementationoftheeco-drivingcycleswiththeexistingfleet.Scenario2evaluatesthepurchaseofnewCNGvehiclestoreplacetheexistingfleet.Scenario3combinesthetwostrategiesandimplementseco-drivingwithanewCNGfleetpurchase.Thesescenariosarecomparedagainstthebaselinescenariocomprisedofobserveddrivingbehavior,asrevealedthroughtheGPSdatasamples,andexistingfleetandannualmileage(seeTable6).Table7summarizestheannualfuelusageandfuelcycleemissionsofMARTAandGRTAgiventheexistingfleetanddrivingbehavior.Thisisthebasescenarioagainstwhichtheaforementionedthreescenarioswillbecompared.Amongairpollutants,weonlypresentNOxandPM2.5becauseheavy-dutyvehicleshaverelativelylowHCandCOemissions.

Table6.MileageandFleetInformation(NationalTransitDatabase,2014;GRTA,2015)

TransitAgency

AnnualMileage

(1,000Miles)

DeadheadingPercent(%)

NumberofBuses

CNGFleetPercent(%)

MARTA 25,850 12 508 69GRTA 4,701 44 166 0

Table7.AnnualOn-roadFuelConsumptionandFuelCycleEmissionsofBaseScenario

TransitAgency

Fuel(1,000GGE)

GHGs(metrictons)

NOx(metrictons)

PM2.5(metrictons)

MARTA 7,226 81,233 371 10.5GRTA 859 9,511 41 1.9

Eco-drivingScenarioThespeedandaccelerationmodificationsdescribedintheMethodologysectionresultedinashiftofoperationmodebindistributions.AsshowninFigure8fortheMARTAsampleandFigure9fortheGRTAsample,mostofoperationpointswithhighSTPvalueshavebeenadjusteddownwardtolowerSTPvaluesbylimitingtheaccelerationrate.

21

Figure8.OperatingModeBinDistributionsofObservedandEco-cyclesintheMARTASample

Figure9.OperatingModeBinDistributionsofObservedandEco-cyclesintheGRTASample

Table8summarizesannualtotalfuelconsumptionandemissionsresultsundertheeco-drivingscenario,inwhichalldriversareassumedtofollowtheeco-cycleforallmilestraveled.Inthisscenario,fleetcompositionisthesameastheexistingfleet.Whentheseoperatingmodebindistributionswereappliedtotheentire508-busMARTAfleet,ecodrivingwouldreducefuelconsumptionbyabout351,000gallonsperyear,measuredinstandardgasolinegallonequivalent(GGE).Hence,MARTAcouldsaveabout309,000gallonsperyearofdieselfuel

22

equivalentgiventhehigherenergycontentofdiesel.TheMARTAfuelsavingstranslatestoanannualreductionofabout3,930metrictons(5%)infuelcyclecarbondioxideequivalent(CO2e)emissions.Intermsofcriteriaairpollutants,eco-drivingimplementationintheMARTAfleetwouldreduceannualNOxemissionsby14metrictons(4%),andannualPM2.5emissionsby0.8metrictons(7%).FortheGRTAfleet,annualfuelsavingsamountedtoabout63,000GGEs.Thistranslatestoasavingsofabout55,000gallonsofdieselfuelperyeara7%reduction.GreenhousegasCO2eemissionsarealsoreducedbyabout700metrictonsperyear.NOxreductionswouldamountto2metrictons(5%),andtheannualPM2.5reductionswouldbe0.1metrictons(7%)peryear.

Table8.Eco-drivingScenarioAnnualFuelConsumptionandFuelCycleEmissions

TransitAgency

Fuel(1,000GGE)

GHGs(metrictons)

NOx(metrictons)

PM2.5(metrictons)

MARTA 6,875 77,304 356 9.8GRTA 796 8,809 39 1.7

CNGFleetPurchaseScenarioInthishypotheticalscenario,MARTAandGRTAareassumedtoreplacetheirexistingdieselbuseswithnewCNGbuses(modelyear2015).MARTAisassumedtoretaintheirexistingCNGbuses,sothisstrategyaffects31%ofthefleet(seeFigure10andTable6).TheagedistributionsoftheexistingfleetsaresummarizedinFigure10forMARTAandFigure11forGRTA.TheCNGscenarioassumednochangesintheexistingdrivingstyle.

Figure10.AgeDistributionofCurrentMARTAMixedDieselandCNGFleet

26%

1%

15%

8%

16%

23%

11%

0%

10%

20%

30%

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

CNG

34%

60%

6%

0%

20%

40%

60%

80%

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

Diesel

23

Figure11.AgeDistributionofCurrentGRTADieselFleet

Table9summarizesfuelandemissionsresultsoftheCNGfleetpurchasescenario.Comparedtothebasescenario,newCNGbusesslightlyincreasedon-roadenergyconsumption.Annualtotalfuelconsumptionincreasedby289,000GGEsforMARTA,and53,000GGEsforGRTA.However,duetothelowerwell-to-pumpCO2eemissionrateCNGascomparedtodiesel,thewell-to-wheelCO2eemissionsdidnotincrease,despitetheincreaseinfuelconsumption.TheannualtotalCO2eemissionsstayedaboutthesameforMARTA,anddecreasedbyabout800metrictonsforGRTA.ACNGfleetwouldsignificantlyreduceNOxandPM2.5emissions.AfterMARTA’sassumedreplacementthe158existingdieselbuseswithnewCNGbuses,thefuelcycleNOxemissionsreducedby112metrictons(30%)peryear,andPM2.5emissionsreducedby9metrictons(85%)peryear.IfGRTAreplacedallofits166dieselbuseswithCNGbuses,itsannualfuelcycleNOxemissionswoulddecreaseby29metrictons(70%),andannualfuelcyclePM2.5emissionswoulddecreaseby2metrictons(95%).

Table9.CNGFleetPurchaseScenarioAnnualFuelConsumptionandFuelCycleEmissions

TransitAgency

Fuel(1,000GGE)

GHGs(metrictons)

NOx(metrictons)

PM2.5(metrictons)

MARTA 7,515 81,349 258 1.5GRTA 912 8,704 12 0.1

Eco-drivingwithCNGFleetPurchaseScenarioInthisscenario,wecombinedthechangesindrivingstylewiththechangesinfleetcomposition.ForbothMARTAandGRTAfleets,allexistingdieselbuseswereassumedtobereplacedwithnewCNGbuses.Eco-cycleswereappliedtotheagencies’entireannualmileage.Table10summarizestheresultsforthecombinedeco-drivingandCNGfleetscenario.ForMARTA,thecombinedstrategyreducedannualfuelconsumptionby70GGEs(1%).FuelcycleCO2eemissionsdecreasedby3,780metrictons(5%)peryear.Theall-CNGfleetwitheco-cyclesshowedsignificantreductionsinfuelcycleNOxandPM2.5reduction,by124metrictons(34%)and9metrictons(87%),respectively.TheGRTAfleetexhibitedevenmorefuelsavingsandemissionreductions.Annualfuelconsumptionfellby4%,amountingto33GGEs.Thefuel

13%17%

10%

32%29%

0%5%10%15%20%25%30%35%

Diesel

24

cycleemissionsinCO2e,NOx,andPM2.5decreasedby1,628(17%),30(73%),and2(96%)metrictons,respectively.

Table10.CNGandEco-drivingScenarioAnnualFuelConsumptionandFuelCycleEmissions

TransitAgency

Fuel(1,000GGE)

GHGs(metrictons)

NOx(metrictons)

PM2.5(metrictons)

MARTA 7,156 77,453 246 1.3GRTA 826 7,884 11 0.1

OverallComparisonandDiscussionThereductionsinfuelconsumptionandfuelcycleemissionspresentedinthispaperreflecteachagency’sfleetsizeandextentofoperations.Inthissection,theresultsarepresentedonapermilebasis,whichwillshedlightongeneralizedfuelandemissionsimpactsforlocalandexpressservices.Figure12throughFigure15providecomparisonsacrossscenariosandoperationtypesforfueleconomy,fuelcycleCO2e,NOx,andPM2.5emissionrates,respectively.Ingeneral,eco-drivingismoreeffectiveinexpressservice,improvingfueleconomyby8%,thanlocaltransitservicewherethefueleconomyimprovementis5%.Eco-drivingreducesmorefuelcycleCO2eemissionsthantheCNGfleetpurchase.Combiningeco-drivingwithnewCNGfleetpurchasecanprovideaddedbenefitsinfuelcycleenergysavingsandemissionsreduction.InthecaseoffuelcycleCO2eemissionsinexpressbusservice,thereductionachievedbythecombinedCNGandeco-drivingstrategyismorethanthesumofreductionsachievedbytheeco-drivingscenarioandCNGfleetpurchasescenario.Thisshowsthateco-drivingcanbeespeciallyeffectiveasafuelconservingstrategyforagenciesthatprovideexpressservicewithaCNGfleet.IntermsofNOxandPM2.5,theadditionalemissionsreductionfromeco-drivinginaCNGfleetismarginal,sinceaCNGfleetalreadyhasverylowNOxandPM2.5emissions.

25

Figure12.FuelEconomyComparisonacrossScenariosandTypesofOperation

3.6 3.8 3.4 3.6

5.55.9

5.25.7

0

1

2

3

4

5

6

7

Base

Eco-driving

CNG

Eco-driving+CN

G

Base

Eco-driving

CNG

Eco-driving+CN

G

LocalTransit ExpressService

FuelEcono

my(m

iles/GG

E)

26

Figure13.CO2eEmissionRateComparison

2.47 2.35 2.44 2.33

1.59 1.48 1.38 1.25

0.670.64

0.700.67

0.430.40 0.47

0.43

0

0.5

1

1.5

2

2.5

3

3.5

Base

Eco-driving

CNG

Eco-driving+CN

G

Base

Eco-driving

CNG

Eco-driving+CN

G


EmissionRa

te(kg/mile)

Well-to-Pump

Pump-to-Wheel

27

Figure14.NOxEmissionRateComparison

12.74 12.26

8.27 7.89 7.82 7.23

1.45 1.29

1.601.52

1.721.64

0.961.15

1.15 1.04

0

2

4

6

8

10

12

14

16

Base

Eco-driving

CNG

Eco-driving+CN

G

Base

Eco-driving

CNG

Eco-driving+CN

G


EmissionRa

te(g/m

ile)

Well-to-Pump

Pump-to-Wheel

28

Figure15.PM2.5EmissionRateComparison

0.360.33

0.03 0.03

0.330.30

0.05

0.05

0.03 0.02

0.070.07

0.02 0.020

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Base

Eco-driving

CNG

Eco-driving+CN

G

Base

Eco-driving

CNG

Eco-driving+CN

G


EmissionRa

te(g/m

ile)

Well-to-Pump

Pump-to-Wheel

29

ConclusionsThispaperevaluatedpotentialfuelandemissionssavingsfromtheimplementationofeco-drivinginAtlanta’sMARTAlocaltransitandGRTAexpressbusfleets.Theanalysesemployedreal-worldoperationsdatacollectedfromthesetwofleetsasbaselineoperatingconditions,andeco-drivingdutycyclesdevelopedthroughaspeedandaccelerationmodificationalgorithm.Theeco-drivingalgorithmsreducefuelconsumptionandemissionsbylimitingengineload,asindicatedbySTPintheMOVESmodelingscheme,whilestillconservingtotaldistanceandaveragespeed.Thebenefitsoftheeco-drivingstrategywerecomparedtoCNGfleetconversion,anotherpopulartransitfuelreductionstrategy.Thesimultaneouseffectsofeco-drivingandCNGfleetpurchasewerealsoassessed.Changesintotalannualfuelconsumptionandemissionsforthethreestrategieswerecomparedforthetwoagencies,aswellasfuelandemissionratesonaper-milebasis.AssumingtheexistingfleetcompositionofMARTAandGRTA,eco-drivingcanreducefuelconsumptionby5%inlocaltransitservice,and7%inexpressbusservice.Althoughthepercentagedecreaseislargerfortheexpressbusfleet(freewaybenefitsarelarge),theactualfuelsavingsperyearisgreaterforbusesinthelocalMARTAfleetgiventhenumberofmilesdrivenperbuseachday.Bycomparison,anewCNGfleetwouldslightlyincreasefuelconsumption,albeitkeepingthefuelcycleCO2eemissionsaboutthesameasthebaselineconditions.Eco-drivingwasalsofoundtobeaneffectivestrategyforreducingfuelconsumptionandemissionsforCNGfleets.FortheGRTAexpressbusservice,eco-drivingconservedalargerpercentageoffuelinthehypotheticalCNGfleetthanintheexistingdieselfleet.Eco-drivingcanproveaverycost-effectivestrategyfortransitagenciesseekingtoreducefuelconsumptionandemissions.Forexample,thefuelsavingsthatGRTAcanachieveamounttoabout55,000gallonsofdiesel,translatingtoabout$132,000inannualfuelsavings(about$800/bus/year),assumingadieselfuelpriceof$2.40/gallon.ForMARTA’smixedCNGanddieselfleet,fuelsavingsfromeco-drivingamountedtoabout300,000gallonsofdieselfuelequivalent(85,200gallonsofdieselfuelplus252,500gasolinegallonsequivalentofCNG)peryear.Assumingadieselpriceof$2.40/gallon($0.63/liter)andaCNGpriceof$1.20pergasolinegallonequivalent(Skelton,2015),thecostsavingsfortheMARTAfleetamounttotranslatingtoabout$720,000inannualfuelsavings,or$1,000/bus/year.Unlikethepurchaseofanalternativefuelbusfleet,eco-drivingdoesnotrequiresignificantcapitalinvestment.Oncebusesarebeingmonitored,eco-drivingiseasytoimplement,requiringonlydevelopmentofdriverreports,trainingandfeedback.Basedontheresearchteam’spriorexperiencewithfleetmonitoring(Xuetal.,2013a),preliminarycostestimatesshowthatimplementingeco-drivingwouldcostanagencyabout$650/bus/year,inclusiveofequipment,communications,driverincentives,anddataanalysis.Forfleetsthatarenotcurrentlymonitoringtransitspeed/accelerationactivity,thefuelsavingsissufficienttopayforsuchmonitoring.Notonlywillfleetmonitoringenablereal-timefeedbacktodrivers,whichhasbeenshowntoprovideaddedfuelsavingsthanin-classtraining(Rolim,etal.,2014),butitwill

30

alsoprovideancillarybenefits,suchasassetmanagement,on-timeperformanceassessment,anddriversafetyassessment.Theeco-drivingalgorithmdevelopedforthisstudyutilizesthemodalmodelingframeworkofU.S.EPA’sMOVESmodel.Theadvantagesarethree-fold.First,MOVESisnotcomputationallydemanding,andthereforecanbeusedinreal-timeornear-real-timedrivingadvisingforfutureapplications.Second,usingtheMOVESframeworkallowsaunifiedplatformforfuelandemissionsestimation.Third,MOVESistheU.S.EPA’sapprovedmodelforregulatoryuse.However,thedisadvantageintheanalyticalapproachisthatthealgorithmislimitedbyuncertaintiesandemissionsaveraginginherentintheMOVESmodel,especiallythoserelatedtothelackofresolutionforhigh-speed,high-poweroperatingmodebinsforheavy-dutyvehicles.Theanalysespresentedinthispaperserveasapointofdeparturefordebatingthebenefitsofeco-drivingfortransitoperations,andtheinitialassessmentofbenefitsappearsignificantandarelikelytobeverycost-effective.Infuturework,theauthorsplantoexpanddatacollectiontoallMARTAandGRTAroutes,andrefinetheeco-drivingalgorithmusinghigh-fidelityvehiclesimulationmodels.

31

AppendixIDataQAQCProceduresGPSpositionandspeedtracescollectedfrommovingtransitvehiclesunderwentaseriesofQA/QCandpostprocessingroutinestoensuredatavalidity,accuracy,andcontinuity,beforeusingthedataintheemissionsanalysisdescribedinthispaper.ThedataprocessingproceduresincludedKalmanfilteringtoeliminateproblematicdatapoints,splinefittingtoinfillmissingdata,mappingofdatatoroadways,identificationofoff-networkactivity,eliminationofparkedvehicledata(parkedvehicleinnon-idleconditions),andtreatmentoflargedatagaps.Eachprocessisdescribedindetailbelow,andsummaryfiguresdescribeanydatalossesateachprocessingstep.

KalmanFilteringThequalityofanyGPSdatastronglydependsonGPSsignalcondition,whichisafunctionofnumberofsatellitesandpositionaldilutionofprecision(PDOP)values.AlthoughGPSreceiversemployproprietarydatafilteringalgorithmsintheirembeddedchipsetfirmwaretohelpcorrectdataon-the-flypriortodeliverytotheuser,itisstillnecessarytofurtherprocessthedata.TheproprietaryGPSchipsetfilteringalgorithmsdonotidentifyandeliminatealldataoutliers,ascanbeseeninrandomerrorsintheGPSoutputdatastream.ThemodifieddiscreteKalmanfilteralgorithmisprovedtoeffectivelyenhanceitscapabilityofcontrollingGPSrandomerrors(Jun,etal.,2005).TheKalmanfilterisusedtocorrecttheGPSspeedwiththeKalmanGainMatrixandthedifferencebetweentheestimatedandthemeasuredspeeds.TheKalmanGainMatrixisgeneratedbasedontheGPSqualitycriteria,thenumberofsatellitesandPDOPvalues.Ifthenumberofsatellitesisbelow4andthePDOPisabove8,thequalityofthespeedisdeterminedtobepoor;therefore,estimatedspeedvaluesareusedinplaceofmeasuredGPSspeedvalues.Consideringthedataavailability,themodifieddiscreteKalmanfilteralgorithmisimplementedtoMARTAoperationsdatatominimizetherandomerrorsfromGPSloggers.TheKalmanfilterroutinewasnotimplementedonGRTAdataduetothelackofGPSparametersprovidedbytheQstarzdataloggers.

SplineDataInfillProcessMissingsegmentsexistsinalmostallthetripfilesduetovariousfactors,suchasobstructionandsignalinterference.Missingsegmentsofshortdurationcanbereasonablyinterpolatedtogenerateacontinuousspeedprofileforemissionmodeling.Acubicsplinealgorithmisimplementedforallofthetripfilestointerpolatemissingsegmentsofnolongerthanthreeseconds.Ifthegapofthemissingsegmentisnolongerthanthreeseconds,i.e.,singlesecondortwoorthreeconsecutivesecondsaremissing,thissegmentwillbecandidatesegmentstobesplined;otherwise,segmentslongerthan3secondsarenotsplinedorfurtherusedformodelingbecauseinterpolatingsuchlongmissingsegmentsmaybeinaccurateorunreliableifthevehicleisnotinasteadycruisemode.Tosplineinfillamissingsegment,sixgoodspeeds(i.e.,speedwascollectedsuccessfully)areneeded;threesecondsbeforethestartmissingsegmentbutnoearlierthan10seconds,andthreesecondsafterendofthemissingsegmentbutnolaterthan10seconds.Ifsufficientgoodspeeddatacannotbefoundwithinthe20secondswrappingthecandidatesegment,thissegmentisidentifiedasmissingandisnotsplineinfilled.

32

GISNetworkMappingFacility Type Toidentifythefacilitytypeonwhichthevehicleisoperating,theGISshapefileofrestrictedroadways(highwaysandfreeways)isused.Twostepsareconductedtoidentifywhetherthetracesareonrestrictedhighwayorunrestrictedarterial:Step1:OverlayGPSDatawiththeRestrictedRoadwayLayerA72-footbufferiscreatedforeachcenterlineoftherestrictedroadwayoneachdirection.Thisbufferlayeristhenoverlaidwiththeoperationsdatatoidentifytheportionsoftracesrunningonrestrictedhighways.Figure16illustratesthefacilitytypeidentificationprocess.

Figure16.ExampleofFacilityTypeIdentification

Step2:FilterRoadwayClassificationResultsThedirectprocessingofGPStracescanleadtoincorrectidentificationoffacilitytypeundercertainconditions.Forexamplewhenabusonanarterialpassesbeneathafreeway,viaan

33

arterialunderpass,aportionofthetripmaybeidentifiedasoperatingontherestrictedhighwayforashortduration..Therefore,post-processingandfilteringisneededtoensurecorrectroadwayclassificationidentification.Becauseconsecutivefreewayexitsarenormallysetoneormoremilesapart,theresearchersimplementedafilteringrulethatemploysduration(consecutiveseconds)ofoperationonrestrictedhighway.Ifthevehicledoesnotoperateformorethanoneminuteonarestrictedhighwayfacility,thedataarelinkedbacktotheunrestrictedarterial.TwoscenariosaredescribedbelowandillustratedinFigure17.

1) SomeGPSpointsfallwithinthebufferandaredeterminedas“restricted”inthe“spatialoverlay,”asthebuspassesalongorbelowtheroadsthatareintersectingwiththefreeway.Thesepointsarechangedbackto“unrestricted”inthisstep

2) Somepointsfalloutofthebufferandaredeterminedas“unrestricted”inthe“spatialoverlay.Thesepointsarechangedbackto“restricted”inthisstep

Figure17.ExampleofFilteringFacilityTypeResults

Engine-offIdentificationOnceabusdriverstopstheirvehicleandturnsofftheengine,fuelconsumptionandemissionsnolongeroccur.Theequipmentemployedbytheresearchteamdoesnotmonitorengineonstatus.Hence,theresearcherswearenotabletoconfirmwhethertheengineisoffunderwhatwouldnormallybedefinedasidleconditions(speed<1mphforMARTAoperations,andspeed<3mphforGRTAoperations,giventheequipmentsensitivity)fromGPSdatadirectly.Forthesefleets,driversaresupposedtoturnoffengineiftheyneedtostopthebusforanextendedperiod.Forthepurposesofthisstudy,engineidleidlingactivityisdefinedbaseduponidlingtimeandlocation.Theengineistreatedasoff(i.e.noengineidle)underthefollowingtwocircumstancesandexcludedfromfuelconsumptionandemissionsanalysis:

3) Onroad:Whenthelengthofcontinuousidlingspeed(speed<=1mphforMARTAoperations,andspeed<=3mphforGRTAoperations)exceeds10minutes.

34

4) OffNetwork:Whenthelengthofcontinuousidlingspeed(speed<=1mphforMARTAoperations,andspeed<=3mphforGRTAoperations)exceeds30seconds.

Toassesstheengineoffthreshold,theresearchersexamined35engine-offoperationswithidlingdurationbetween10and15minutes,and35identifiedengine-onoperationswithidlingdurationbetween5and10minutes.Theseengine-offoperationstypicallyoccurredinremoteparkinglotlocations(notMARTAmaintenanceyards)orinthemiddleofaroadway.IntheexampleshowninFigure18(b),thebusstayedatthelocationoftheleftandrightredmarkerfor845and690seconds,respectively,andthenheadedintorevenueservice.ThereisnoreasontobelievethatthebusidledforthatperiodinviolationofMARTAoperationspolicy(althoughadrivermayhavedoneso),sotheactivitywasconsideredtobeanenginestopandtripendlocation.IntheexampleshowninFigure18(b),thebusstoppedinthemiddleofthestreetfor855seconds.Thison-streetlocationwasneitheratabusstopnoranintersection.Thereisnowaytoknowwhatthebuswasdoingatthismid-streetlocation.Thedelaymayhaveresultedfromroadconstruction,thedrivermayhaveparkedthevehicletousearestroom,orthedrivermayhavestoppedforsomeotherreason.Inanycase,thereisnoreasontobelievethatthebuswasidlingfor855secondsinviolationoftransitagencypolicyan.Forthepurposesoftheanalysis,thelocationwasdeemedatripendandengineofflocation.Neitherofthetwoscenariosdescribedabovearepartregularbusserviceactivity.Withoutevidencetothecontrary(monitoringofengineonstatus),activityattheselocationsarenotidentifiedasextendedidleactivityandarethereforeexcludedfromtheassessmentofeco-drivingforMARTArevenueoperations.Ontheotherhand,forthelocationsadjacenttoexistingbusstops,asshowninFigure19,,andwherethestopdurationsareshorterthan10minutes,thestopisdeemedpartofregularrevenueoperationsandbecomespartoftheeco-drivingassessment.

(a) (b)

Figure18.ExampleofIdentifiedEngine-OffConditions(ParkingLotsandMidblock)

35

(a) (b)

Figure19.ExampleofEngine-OnCondition(TransitStops)

BreakingGPSDataintoTripsforEco-DrivingAnalysisThedataprocessingroutinesdescribedaboveidentifytripendlocations,wheretheengineispresumedtohavebeenturnedoff.Inaddition,therearestillgapsinthedatastreamwheretheresearchteamdidnotallowthesplinefunctiontoinfilldata.Thesestoplocationsanddatagapswereusedtobreakthemonitoreddataintotripsforeco-drivinganalysis.However,someofthetripsthatresultfromthedataprocessingareveryshort,typicallyassociatedwithchoppyperiodsofGPSdataloss.Toensuretripqualityandapplicability,tripswereincludedintheanalysisonlyif:1)thelengthofcycleislongerthan30seconds;and2)theaveragespeedofthecycleisgreaterthan5mph.

DataProcessingResultsFigure20andFigure21summarizethedataprocessingproceduresfortheMARTAandGRTAdata,respectively.Figure22andFigure23showthegeographiccoverageofMARTAandGRTAoperationsasreflectedinthefinalanalyticaldataset.

36

Figure20.ProcessingstepsforMARTAdata

37

Figure21.ProcessingstepsforGRTAdata

38

Figure22.GeographicCoverageofMARTABusRoutes

39

Figure23.GeographicCoverageofGRTABusRoutes

40

AppendixIIMOVESBackground

Table11.DefinitionsofSTPOperatingModeBinsinMOVES

Operating Mode ID Operating Mode Scaled Tractive

Power Vehicle Speed Vehicle Acceleration

Description (STPt, skW) (vt, mph) (a, mph/sec)

0 Deceleration/Braking at ≤ -2.0 OR (at < -1.0

AND at-1 <-1.0 AND at-2 <-1.0)

1 Idle -1.0 ≤ vt < 1.0 Any

11 Coast STPt< 0 0 ≤ vt < 25 Any

12 Cruise/Acceleration 0 ≤ STPt< 3 0 ≤ vt < 25 Any




16 Cruise/Acceleration 12 ≤ STPt 0 ≤ vt < 25 Any

21 Coast STPt< 0 25 ≤ vt < 50 Any








30 Cruise/Acceleration 30 ≤ STPt 25 ≤ vt < 50 Any

33 Cruise/Acceleration STPt< 6 50 ≤ vt Any

35 Cruise/Acceleration 6 ≤ STPt< 12 50 ≤ vt Any

37 Cruise/Acceleration 12 ≤ STPt<18 50 ≤ vt Any



40 Cruise/Acceleration 30 ≤ STPt 50 ≤ vt Any

41

Table12.STPParametersfromMOVES2014

Source Rolling Rotating Drag SourceMass FixedMassSourceType TypeID Term Term Term Metric Factor

Name ID A B C Tonnes MIntercityBus 41 1.29515 0 0.00371491 19.5937 17.1TransitBus 42 1.0944 0 0.00358702 16.556 17.1

42

AppendixIIIMOVESInputSettingsGeneralsettingsarelistedbelow,andthespecificsettingsarelistedinTable13.• Region:

o FultonCounty,Georgia• CalendarYear:

o 2015• Month:

o January

• DateandTime:o Weekday,7:00-8:00AM

• I/MStrategy:o Default2015I/MstrategyfromMOVES2014

• Meteorology(defaultvaluedeterminedbytimeandregionfromMOVES):o Temperature:30Fo Humidity:75%

• FuelSupplyandFuelFormulation:o DefaultwinterfuelsupplyandfuelsharefromMOVES

• SourceType:o Transitbus(sourcetypeID=42)forMARTAlocaltransitbuseso Intercitybus(sourcetypeID=41)forGRTAexpressbuses

• 23Links:o Togenerateanemissionrateforamodeofoperation,eachlinkisassigned100%

fractionofoneoperatingmodebino Timeisscaledtoonehourofoperationusinglinklengthandlinkaveragespeed

43

Table13.EmissionRateswithSpecificSettings

TypeofOperation ScenarioSetting FuelType AgeDistribution Cycle

LocalTransit

Currentfleet,currentdrivingstyle

DieselandCNG

CurrentMARTAfleetagedistribution(Figure10) Observed

DrivingCycleNewfleet,currentdrivingstyle CNG Age=0

Currentfleet,eco-driving

DieselandCNG

CurrentMARTAfleetagedistribution(Figure10)

Eco-cycle

Newfleet,eco-driving CNG Age=0

ExpressBusService

Currentfleet,currentdrivingstyle Diesel

CurrentGRTAfleetagedistribution(Figure11) Observed

DrivingCycleNewfleet,currentdrivingstyle CNG Age=0

Currentfleet,eco-driving Diesel

CurrentGRTAfleetagedistribution(Figure11)

Eco-cycle

Newfleet,eco-driving CNG Age=0

44

ReferencesArgonneNationalLaboratory.(2013).TheGreenhouseGases,RegulatoryEmissions,andEnergyUseinTransportation(GREET)Model.RetrievedNovember13,2015from:https://greet.es.anl.gov/

Barth,M.,An,F.,Norbeck,J.,&Ross,M.(1996).ModalEmissionsModeling:APhysicalApproach.TransportationResearchRecord:JournaloftheTransportationResearchBoard,(1520),81-88.

Barth,M.,&Boriboonsomsin,K.(2009).EnergyandEmissionsImpactsofaFreeway-basedDynamicEco-drivingSystem.TransportationResearchPartD:TransportandEnvironment,14(6),400-410.

Beusen,B.,Broekx,S.,Denys,T.,Beckx,C.,Degraeuwe,B.,Gijsbers,M.,&Panis,L.I.(2009).Usingon-boardLoggingDevicestoStudytheLonger-termImpactofanEco-drivingCourse.TransportationResearchPartD:TransportandEnvironment,14(7),514-520.

Dib,W.,Chasse,A.,Moulin,P.,Sciarretta,A.,andCorde,G.(2014).OptimalEnergyManagementforanElectricVehicleinEco-drivingApplications.ControlEngineeringPractice,29,299-307.

Feng,C.,R.Guensler,&M.Rodgers(2007).“TransitBusLoad-BasedModalEmissionRateModelDevelopment.”U.S.EnvironmentalProtectionAgency,OfficeofResearchandDevelopment,Washington,DC,EPA/600/R-07/106.

Gallivan,F.,&Grant,M.(2010).CurrentPracticesinGreenhouseGasEmissionsSavingsfromTransit(No.ProjectJ-7,TopicSH-09).

Guensler,R.,Liu,H.,Xu,Y.,&Rodgers,M.(2015).MOVES-MATRIX:Setup,Implementation,andApplication.CRC-25:RealWorldEmissionsWorkshop.March22-25,2015.LongBeach,California.

Guensler,R.H.Liu,X.Xu,Y.Xu,&M.Rodgers(2016).“MOVES-Matrix:Setup,Implementation,andApplication.”95thAnnualMeetingoftheTransportationResearchBoard.Washington,DC.January2016.

Guensler,R.,S.Yoon,C.Feng,H.Li,andJ.Jun(2005)."Heavy-DutyDieselVehicleModalEmissionModel(HDDV-MEM):Volume1:ModalEmissionModelingFramework.Volume2:ModalComponentsandOutputs."U.S.EnvironmentalProtectionAgency,Washington,DC,EPA/600/R-05/090.

Ho,S.H.,Wong,Y.D.,&Chang,V.W.C.(2015).WhatCanEco-drivingDoforSustainableRoadTransport?PerspectivesfromaCity(Singapore)Eco-drivingProgramme.SustainableCitiesandSociety,14,82-88.

45

Jun,J.,Guensler,R.,&Ogle,J.(2006).SmoothingMethodsDesignedtoMinimizetheImpactofGPSRandomErroronTravelDistance,Speed,andAccelerationProfileEstimates.TransportationResearchRecord:JournaloftheTransportationResearchBoard,1972,141–150

Mensing,F.,Bideaux,E.,Trigui,R.,&Tattegrain,H.(2013).Trajectoryoptimizationforeco-DrivingTakingintoAccountTrafficConstraints.TransportationResearchPartD:TransportandEnvironment,18,55-61.

Mensing,F.,Bideaux,E.,Trigui,R.,Ribet,J.,&Jeanneret,B.(2014).Eco-driving:AnEconomicorEcologicDrivingStyle?TransportationResearchPartC:EmergingTechnologies,38,110-121.

NationalTransitDatabase.(2014).AnnualDatabase:RevenueVehicleInventory.Retrievedfrom:http://www.ntdprogram.gov/

Ogle,J.,Guensler,R.,&Elango,V.(2005).Georgia'sCommuteAtlantaValuePricingProgram:RecruitmentMethodsandTravelDiaryResponseRates.TransportationResearchRecord:JournaloftheTransportationResearchBoard,(1931),28-37.

Qian,G.&Chung,E.(2011).EvaluatingEffectsofEco-drivingatTrafficIntersectionsBasedonTrafficMicro-simulation.AustralianTransportResearchForum2011Proceedings,Adelaide,Australia.

Rolim,C.,Baptista,P.,Duarte,G.,Farias,T.,andShiftan,Y.(2014).QuantificationoftheImpactsofEco-drivingTrainingandReal-timeFeedbackonUrbanBusesDriver'sBehaviour.TransportationResearchProcedia,3,70-79.

Rutty,M.,Matthews,L.,Andrey,J.,&DelMatto,T.(2013).Eco-driverTrainingwithintheCityofCalgary’sMunicipalFleet:MonitoringtheImpact.TransportationResearchPartD:TransportandEnvironment,24,44-51.

Saynor,B.(2011).ECOWILLDeliverable6.1:GuidelinesforDeliveringShort-DurationEco-drivingTrainingCourses.IntelligentEnergyEurope.Retrievedfrom:http://www.ecodrive.org/download/internal_documents/Deliverables/ecowill_deliverable_61_guidelines_for_delivering_dhort_duration_ecodriving_courses.pdf

Shancita,I.,Masjuki,H.H.,Kalam,M.A.,Fattah,I.R.,Rashed,M.M.,&Rashedul,H.K.(2014).Areviewonidlingreductionstrategiestoimprovefueleconomyandreduceexhaustemissionsoftransportvehicles.EnergyConversionandManagement,88,794-807.

Skelton,I.(2015)NaturalGasVehicleOverview.Presentedatthe2015AirandWasteManagementAssociationSouthernSectionAnnualMeetingandTechnicalConference.Retrievedfromhttp://www.awma-ss2015conference.com/wp-content/uploads/2015/09/5-Skelton-AGL-NGV-Presentation-for-AWMA-August-2015.pdfonMarch8,2016.

46

Strömberg,H.K.,andKarlsson,I.M.(2013).Comparativeeffectsofeco-drivinginitiativesaimedaturbanbusdrivers–Resultsfromafieldtrial.TransportationResearchPartD:TransportandEnvironment,22,28-33.

Suzdaleva,E.,&Nagy,I.(2014).Data-basedSpeed-limit-respectingEco-drivingSystem.TransportationResearchPartC:EmergingTechnologies,44,253-264.

TransportCanada(2004).FuelSense:MakingFleetandTransitOperationsMoreEfficient(CaseStudy24).UrbanTransportationShowcaseProgram:CaseStudiesinSustainableTransportation.CityofEdmonton,Alberta,Canada.July2004.Retrievedfrom:http://data.tc.gc.ca/archive/media/documents/programs/cs24e_fuelsense.pdf

Xu,Y.,Gbologah,F.E.,Lee,D.Y.,Liu,H.,Rodgers,M.O.,&Guensler,R.L.(2015).AssessmentofAlternativeFuelandPowertrainTransitBusOptionsUsingReal-worldOperationsData:Life-cycleFuelandEmissionsModeling.AppliedEnergy,154,143-159.

Xu,Y.,Elango,V.,Guensler,R.,&Khoeini,S.(2013a).IdleMonitoring,Real-TimeIntervention,andEmissionReductionsfromCobbCounty,Georgia,SchoolBuses.TransportationResearchRecord:JournaloftheTransportationResearchBoard,(2340),59-65.

Xu,Y.,Guensler,R.,&Rodgers,M.(2013b).SensitivityofEmissionsPredictionsBasedonGlobalPositioningSystem-basedCycleDataandAssumptionsofIdleCutpoints.Air&WasteManagementAssociation’s106thAnnualConferenceandExhibition,Chicago,IL.June25-28,2013.

Wåhlberg,A.E.(2007).Long-termeffectsoftrainingineconomicaldriving:Fuelconsumption,accidents,driveraccelerationbehaviorandtechnicalfeedback.InternationalJournalofIndustrialErgonomics,37(4),333-343.

Zarkadoula,M.,Zoidis,G.,&Tritopoulou,E.(2007).TrainingUrbanBusDriverstoPromoteSmartDriving:AnoteonaGreekEco-drivingPilotProgram.TransportationResearchPartD:TransportandEnvironment,12(6),449-451.

ECO DRIVING FOR TRANSIT - Home - NCST · ECO-DRIVING FOR TRANSIT ... Literature Review ... report “pump-to-wheel” (occurring at the vehicle) fuel consumption, greenhouse gas ...

Documents