Table Of ContentThèse de Doctorat
Alejandro M
ONTOYA
Mémoire présenté en vue de l’obtention du
gradedeDocteurdel’Université d’Angers
Label européen
sous le label de l’Université de Nantes Angers Le Mans
Écoledoctorale: Sciencesettechnologiesdel’information,etmathématiques
Discipline: Informatiqueetapplications,sectionCNU27
Unitéderecherche: LaboratoireAngevindeRechercheenIngenieriedesSystèmes(LARIS)
Soutenuele9décembre2016
Electric Vehicle Routing Problems: models and
solution approaches
JURY
Président: M.Emmanuel NÉRON,Professeur,PolytechTours
Rapporteurs: M.Dominique FEILLET,Professeur,EcoledesMinesdeSaint-Etienne
M.Daniele VIGO,Professeur,UniversitàdiBologna,Bologna,Italy
Examinateurs: M.Emmanuel NÉRON,Professeur,PolytechTours
M.Fabien LEHUÉDÉ,Maîtredeconférences,EcoledesMinesdeNantes
Directricedethèse: Mme Christelle GUÉRET,Professeur,Universitéd’Angers
Co-directeursdethèse: M.JorgeE. MENDOZA,Maîtredeconférences,PolytechTours
M.JuanG. VILLEGAS,Professeur,UniversidaddeAntioquia,Medellin,Colombie
Acknowledgement
Thisresearchprojecthasbeenpossiblewiththehelpandsupportofmanypeople. Iwouldliketoexpress
mysinceregratitudetoallpeoplewhowereveryhelpfulduringmyPh.D.experience. Thisexperiencegave
me the opportunity to know different countries and cultures, to meet a lot of friendly people, to do new
friends,andtoimprovemyacademicandresearchskills.
I first thank to my wife, Diana, for her support, sacrifice, company, and love. She was always there, no
matter the place, the weather or the language. I am fortunate to have her by my side. I would also like to
thank my mother, Angela, my stepfather Orlando and my grandmother Blanca, for their company. They
havesupportedmeallthewayinthedevelopmentofthisthesisandtheyalwaysbelievedIcouldfollowthis
dream.
I like to thank my thesis advisors, Christelle Guéret for her patience, support and advice in all this
process; Jorge Mendoza I have learned from his many things that go far beyond optimization, vehicle
routing problems, and operations research; and Juan Guillermo Villegas for his intellectual generosity and
tobewithmeingreatpartofmyacademicformation(mymasterandPh.D.).
IwouldliketothankthejurymembersDanieleVigo,DominiqueFeillet,EmmanuelNéron,andFabien
Lehuédéforacceptingtheinvitationtobepartofthejury.
I want to thank the University EAFIT for the support in this project, especially to Gabriel Arango,
DirectorofTeaching;AlbertoRodriguez,DeanoftheFacultyofEngineering;andSergioAugustoRamirez,
Chief of Production Engineering department for their trust and support. I also want to thank Mario Velez
forbelievinginmesinceIwasdoingmycareer;JairoMayaforsharehistimeandknowledge;andGabriel
Hincapié and Nora Cadavid for his support and helping. And I would like to thank the Universidad EAFIT
scientificcomputingcenter(APOLO)fortheirsupportinthecomputationalexperiments
IthanktopeopleofUniversitéd’Angers,especiallytoMichelLandronforhissupportandforintroduc-
ing me to French culture; Simone Rees for her support and help; and my partners at LARIS, Achraf, Fally,
Khadim,Ibrahim,AminandKhaoulafortheirfriendship.
IwouldliketothankeveryonewhosupportedmeduringmydoctorateinAngers. PedroandNataliamy
"paisasbrothers",whoofferedmetheircompany,andsupportedmeinthefirsttwoyears. Ialsowouldthank
Alexis,Evelin,Cristhope,JuanPablo,Clemence,Ana,SilvanandVictoriafortheunforgettablemoments.
I want to thank the people who supported me in Colombia. To my in-laws and unconditional friends
(Joa, Marc, James, Naty, David and Naty) for being with Diana when I was in France; Pauline for her
FrenchclassesinMedellin;andPabloMayaforhisnetworkflowcourseattheUniversityofAntioquia.
I like to thank VeRoLog group for allowing me participating to the summer schools of 2014 and 2015.
Inthoseschools,Icouldimprovemyresearchskillsandmadealotoffriends.
Finally, I gratefully acknowledge the financial support provided by Programa de Movilidad Doctoral
hacia Francia (Colfuturo - Emb. de Francia - ASCUN - Colciencias - Min. de Educación), and Programa
EnlazaMundos(AlcaldíadeMedellín).
3
Contents
1 Introduction 11
2 MSHfortheGreenVRP 15
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Literaturereview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Multi-spacesamplingheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.1 Generalstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.2 Samplingheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.3 Split . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.4 Repairprocedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.5 Setpartitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Computationalexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.6 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6.1 Notationforproblemdefinition . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6.2 Notationformulti-spacesamplingheuristic . . . . . . . . . . . . . . . . . . . . . 31
2.6.3 Notationforrepairprocedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3 AcomparativestudyofchargingassumptionsineVRPs 33
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Chargingasumptionintheliterature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 Settingupthestudy: problemandformulations . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.1 Problemdescription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 MILPformulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Computationalexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4.1 Experimentalsettings&Testinstances . . . . . . . . . . . . . . . . . . . . . . . 41
3.4.2 Experimentalenvironment&Parametersetting . . . . . . . . . . . . . . . . . . . 41
3.4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.6 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.6.1 NotationfortheMILPformulation . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4 ILS+HCforeVRP-PNL 45
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Hybridmetaheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2.1 Initialsolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2.2 Split . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.3 Variableneighborhooddescent . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.4 Perturb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.5 Heuristicconcentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5
6 CONTENTS
4.3 Thefixed-routevehicle-chargingproblem . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3.1 Mixed-integerlinearprogrammingformulation . . . . . . . . . . . . . . . . . . . 50
4.3.2 SolvingtheFRVCP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.4 Computationalexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.4.1 TestinstancesfortheeVRP-NL . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.4.2 Parametersettings&experimentalenvironment . . . . . . . . . . . . . . . . . . . 56
4.4.3 Solutionaccuracy: optimalvs. heuristicchargingdecisions . . . . . . . . . . . . . 57
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.6.1 Notationforproblemdescription . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.6.2 Notationforhybridmetaheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.6.3 Notationforthefixed-routevehicle-chargingproblem . . . . . . . . . . . . . . . 62
5 TRP-CEV 63
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.2 Literaturereview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3 Problemdescription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.3.1 Mixed-integerlinearprogrammingformulation . . . . . . . . . . . . . . . . . . . 67
5.4 Parallelmatheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.4.1 Identifyingfeasiblerequests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.2 GRASP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.4.3 Setcovering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5 Thefixed-routevehiclechargingproblemwithtimewindows . . . . . . . . . . . . . . . . 73
5.5.1 Greedyheuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.6 Computationalexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.6.1 TestinstancesfortheTRP-CEV . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.6.2 Parametersettings&experimentalenvironment . . . . . . . . . . . . . . . . . . . 77
5.6.3 PerformanceofPMa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.6.4 Managerialinsight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.8 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.8.1 Notationfortheproblemdescription . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.8.2 NotationfortheMILPformulation . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.8.3 NotationforthePMa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.8.4 NotationfortheFRVCP-TW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6 Generalconclusions&perspectives 87
A Appendices 89
A.1 DetailedresultsforGreenVRPinstances . . . . . . . . . . . . . . . . . . . . . . . . . . 89
A.2 DetailedresultsoftheeVRP-NL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
A.3 DetailedresultsfortheTRP-CEV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
A.4 DetailedresultsfortheE-FSMFTW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
List of Tables
2.1 SummaryresultsandcomparisonofourMSHwithothermethodsonthesmallinstancesof
Erdog˘an&Miller-Hooks(2012). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2 SummaryresultsandcomparisonofourMSHwithothermethodsonthelargeinstancesof
Erdog˘an&Miller-Hooks(2012). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Comparisonofourchargingassumptionswithchargingassumptionsfromtheliterature . . 42
4.1 Comparisonofthetwoversionsofthemetaheuristiconsmallinstanceswithprovenoptima 57
4.2 Comparisonofthetwoversionsofthemetaheuristiconlargeinstances . . . . . . . . . . . 58
4.3 Averagecomputingtime(inseconds)ofdifferentvariantsofthemetaheuristic . . . . . . . 59
5.1 ComparisonofthePMaonsmallinstanceswithprovenoptima . . . . . . . . . . . . . . . 78
5.2 ComparisonofthePMawiththeroutingsoftwareusedbyENEDIS . . . . . . . . . . . . 78
5.3 ComparisonofourPMawiththeALNSbyHiermannetal.(2016)onsmallinstances . . . 79
5.4 ComparisonofourPMawiththeALNSbyHiermannetal.(2016)onlargeinstances . . . 79
A.1 ResultsofMSHonsmallinstancesofErdog˘an&Miller-Hooks(2012). . . . . . . . . . . 90
A.2 ResultsofMSHonlargeinstancesofErdog˘an&Miller-Hooks(2012). . . . . . . . . . . . 91
A.3 ResultsofILS(H)+HCandILS(S)+HConthe20smallinstances . . . . . . . . . . . . . . 92
A.4 ResultsofILS(H)+HCandILS(S)+HConthe100largeinstances . . . . . . . . . . . . . 92
A.5 ResultsofPMaonsmallinstancesofTRP-CEV . . . . . . . . . . . . . . . . . . . . . . . 95
A.6 ResultsofPMaonlargeinstancesofTRP-CEV . . . . . . . . . . . . . . . . . . . . . . . 96
A.7 ResultsofPMaonsmallinstancesofHiermannetal.(2016). . . . . . . . . . . . . . . . . 98
A.8 ResultsofPMaonlargeinstancesofHiermannetal.(2016). . . . . . . . . . . . . . . . . 99
7
List of Figures
2.1 ExampleofafeasibleGreenVRPsolution . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 SplittingaTSPtourintoaGreenVRPsolution . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 OutlineofthestructureoftherepairgraphB = (Z,U) . . . . . . . . . . . . . . . . . . . 26
2.4 Optimalrepairexampleforthethree-customersequencer = {0,A,B,C,0},threeAFSs. . 27
2.5 Trade-offbetweensolutionqualityandCPUtime . . . . . . . . . . . . . . . . . . . . . . 29
2.6 PercentageshareofCPUtimebyMSHphases . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1 Typical charging curve, where i and u represent the current and terminal voltage respec-
tively. (SourceHõimojaetal.(2012)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Firstsegmentapproximation(FS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Linearapproximationsofchargingfunctions. . . . . . . . . . . . . . . . . . . . . . . . . 36
3.4 Approximationvsrealdata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5 ExampleofafeasibleeVRP-NLsolution . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.6 Piecewiselinearapproximationforthechargingfunction. . . . . . . . . . . . . . . . . . . 39
4.1 GeneralstructureofILS+HC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2 Piecewiselinearchargingfunctionandfixed-routefortheFRVCP . . . . . . . . . . . . . 51
4.3 Piecewise linear approximation for different types of CS charging an EV with a battery of
16kWh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.4 Percentageoftherouteswith/withoutvisitstoCSsbyinstancesize. . . . . . . . . . . . . 59
4.5 AnalysisofthenumberofvisitstoCSs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6 Histogram of the average battery level (in % of the total battery capacity) after a mid-route
charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.1 Average fixed (i.e., the fixed cost of each technician), variable (i.e., the sum of the total
travel cost, fixed charging cost and parking cost), and total cost for each instance for each
fleetcomposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Average emission (in Kg CO per Km) and maximum number of visited CSs in a solution
2
ofeachinstanceforeachfleetcomposition. . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.3 AveragegapbetweenthecostofthesolutionswithandwithouttheoptionofvisitingtheCSs 83
5.4 FactorsexplainingtheincrementonthecostwhenthevisitstotheCSsareforbidden . . . 83
9
Description:than 30 EVs in selected cities across the U.S (Priselac 2013). still hampered by technical constraints such as low driving ranges and long battery charging One of the fields in which the void is more critical is that of optimization.