Seite - 98 - in Emerging Technologies for Electric and Hybrid Vehicles

Energies 2016,9, 86 Table2.Detailsof somenewVRPvariants relatedtoEVs. Variant DecisionVariables Constraints Objectives Fleet size andmix (1)Determine thenumberandtypeof EVstobepurchased. (2)Determinethe idealcompositionof the heterogeneousfleet. (1)Environmental standards andprice incentive to acquisitionofEVs. (2)Fixedandvariable chargingtimes. (3)Limitedbudget torenew thefleetofvehicles. (1)Minimize theacquisition andoperatingcostsofnewEVs. (2)Maximize thesatisfactionof customerneeds. (3)Minimize theenvironmental impact. Charging networks (1)Determinenumberandgeographical positionof rechargingstations. (2)Determinecapacityof rechargingstations. (3)Determine technologyof recharging stations (lowor fast recharge). (4)Decidebetweenswappingor rechargingofbatteries. (1)Limitedbudget to install newrechargingstations. (2)NeedsofEVstorechargeor exchangebatteries. (1)Minimize the investment andoperatingcostsofcharging networks. (2)Maximize the levelof service tocustomers. Routing (1)Determine thenumberofvisits to rechargingstations. (2)Determine the timingofvisits to rechargingstations. (3)Allocateavailablerechargingresources tovehicles inrechargingstations. (4)Select theoptionof rechargingor swappingbatteries. (1)Geographicalpositionof rechargingstations. (2)Capacityof recharging stations. (3)Fixedorvariable recharging/swappingtimes. (1)Minimizeroutingcost consideringrecharging operations. (2)Minimizeroutingtimes consideringrecharging operations. (3)Minimizerechargingand swappingcosts. 5. SolvingApproachesforVRPswithEVs Asdiscussed in theprevioussections, the introductionofEVs in freightfleets imposesanumber of strategic and operational challenges thatmust be efficiently addressedwith the use of novel methods and approaches. This is particularly true for VRPmodels, which can be classified into three levels according to their degree of realism (Figure 2). The classical-basic VRPmodels are mainly theoreticalmodels thatallowthedevelopmentofmathematicalapproaches, either if theyuse exactorapproximatesolvingmethods. Thesemodelsareusedto test solvingmethods incontrolled environments,whichallowsassessingtheirperformancebeforebeingusedinpracticalapplications. The classical-advancedVRPmodels are characterizedbyahigher level of realism, i.e., large-scale problems,multi-objective functions, integratedroutingandlogistics. Examplesof the latterare:VRPs combinedwithpacking[99],allocation, inventorymanagement[100], etc.Moreadvancedandcomplex VRPvariantsare includedin thiscategory.Usually, theseproblemshavebeensolvedbymetaheuristic approaches, suchasGeneticAlgorithms, IteratedLocalSearch,AntColonyOptimization,Simulated Annealing,etc.MostoftheexistingworkintheVRPliteraturesofardealswiththesetwoclassical types ofmodels.Recently,however,andduetoboththematurityofexistingexactandmetaheuristicmethods aswellas tonewbusinessneeds,newRichVRPmodelsarebeingconsidered. Thesolvingmethods for thesemodelscombinedifferentexactandmetaheuristicapproaches (matheuristics) [101]oreven simulationwithmetaheuristics (simheuristics) [102]. Simheuristics allowconsideringuncertainty both in theobjective functionandtheconstraintsofaVRPmodel, thusmakingthesemodelsamore accuraterepresentationofreal-life routingdistributionsystems. Thesehybridmethodsnotonlycan dealwithuncertaintyandreal-timedecisionmaking [103], but theycanalsoconsideraspects such as richerobjective functions (includingenvironmental costs), dynamism[104], diversityofvehicle driving ranges,multi-periodicity in the distribution activity, integrationwith other supply chain components, etc. 98