routing ev users towards an optimal charging plan

EVS26Los Angeles, California, May 6 - 9, 2012

Routing EV Users Towards an Optimal Charging Plan

Sandford Bessler1, Jesper Grønbæk1

1FTW Telecommunication Research Center, Donau-City 1, 1220 Vienna, Austria, [email protected]

Abstract

In this work we address the efficient operation of public charging stations. Matching energy supply anddemand requires an interdisciplinary understanding of both the mobility of electric vehicle (EV) usersand the load balancing mechanisms.As a result of a mobility study, we propose in this work a routing service for searching and reservingpublic charging spots in the neighborhood of the destination. When comparing the search results fordirect drive with those for a multimodal route (using driving, walking and public transport) in an urbanenvironment, we obtain for the latter significantly more charging options in particular at low e-mobilitypenetration levels, at a cost of slightly longer trip duration.Further contributions address the schedule optimization, that, due to the proposed distributed architec-ture, can be performed independently at each public charging station. We formulate an integer programfor the controlled charging and compare results obtained both with the exact and with a greedy heuristicmethod.

Keywords: e-mobility, EV routing service, multimodal route optimization, charging station, controlled charging,power flow calculation

1 IntroductionA well studied scenario for charging electric ve-hicles (EV) is charging overnight at home. Thisscenario alone, however, fails to address certainsignificant user groups such as residents of ur-ban areas without own garage, vehicle fleets, ve-hicles with higher mileage, etc. These users areall dependent of the existence of public charg-ing stations (PCS). Using a PCS poses howevertwo problems for the user: a) to find a PCS thatmatches the mobility needs and b) the found PCSmust be available in terms of energy and parkingspace in the desired charging period. We addressboth interdependent problems in this work, thathas been done within the KOFLA project [1]As the vehicular mobility can shift energy charg-ing energy very quickly from place to place, themain idea we followed in this work is to plan re-sources in advance that contribute both to usersatisfaction and to the service performance of thegrid operator.Recent mobility studies [2],[3] reveal that EVcharging must be subordinated to the mobilitygoal or activity and not viceversa. As a conse-quence, users should plug-in their EVs in walk-

ing distance of their destination. Partial chargingis acceptable, if the stay duration is limited.Based on these assumptions, we have defined aquery and reservation protocol between an EVand a routing service that enables the user to:a) query availability of charging stations anytimein advance, b) reserve a time slot for chargingat an available charging station, and c) be noti-fied, when a charging point becomes available.The protocol runs over a wireless channel avail-able in a cellular network GSM/UMTS/LTE, butin the future it could be integrated in the ITS ser-vice ecosystem, as proposed currently in ETSIEV notification draft ([18]).The routing service is designed to serve a geo-graphical region and represents a broker betweenEV users and energy providers. The broker en-tails the best PCS match in terms of availabil-ity of resources required for charging. It fur-ther considers user preferences such as locationconvenience, price importance, preferred energyprovider, etc.There are two parameters in the user provided in-formation (see Table 1) that help to route the userand reserve resources at a selected PCS: the timewindow, i.e. estimated arrival and leave time ,

EVS26 International Battery, Hybrid and Fuel Cell Electric Vehicle Symposium, 2012 1

and demand, i.e. the maximum amount of en-ergy needed. Whereas the arrival time and en-ergy can be estimated by the vehicle navigationand the battery management system, the stay du-ration has to be provided solely by the user.

Table 1: Parameters of the query message

Parameter valuetimestampdestination (next stop) coordinatesexpected arrival timeexpected departure timeexpected State of charge (SoC)may use public transport yes/nocharging rates supported slow/normal/quickprice importance high, lowwaiting importance high, lowwalking distance importance high, lowrenewable importance high, lowpayment means card name, null

For this scenario, we introduce a decentralizedsystem architecture, where each PCS is able toschedule and control the charging of individualEVs allocatd to it. In this way the architecturesupports a new stakeholder: the charging stationowner. This provider has the freedom and incen-tive to setup charging stations and sell the park &charge service in conjunction with added valueservices like loyalty programs, public transporta-tion (see discussion in [4]).Figure 1 shows schematically the proposed archi-tecture: the routing server dispatches the reser-vations on the basis of the rough availability ofresources at certain selected public charging sta-tions. Each PCS sends load status updates both tothe routing service and to a LV-grid agent, whichhas the task to calculate periodically the feasiblecharging load (available power) for each PCS, us-ing a power flow model. In the example in Figure1 PCS A and B are in the same LV-grid, whereasPCS C is in a different LV-grid.

PCS A

Routing

Server

PCS C

PCS B

Low

voltage

grid

Low

voltagee

grid

Reservation

Protocol

Load

Update

Reserve

Available

Power

Load

Update

Figure 1: System architecture: charging load balanc-ing is performed through routing the requests to dif-ferent PCSs and through optimized activity schedul-ing in the PCS.

In the rest of the paper we focus on two mech-anisms that would drastically improve the avail-ability of charging resources at a PCS. In section2 we evaluate the park, charge & ride option and

in Section 3 we study the optimal allocation ofresources at a specific PCS.

2 Multimodal RoutingMultimodal routing covers the aspect of includ-ing different means of transportation in a route.Multimodality is relevant for park-and-ride sce-narios where drivers park in the perimeter of thecity and continue to their destination using pub-lic transport. Multimodality with public trans-port can also be combined with e-mobility charg-ing to increase the access to more charging sta-tions while still ensuring end-to-end mobility. In-creased charging station availability is importantin early low penetration scenarios. Further, iden-tifying several charging station options can bebeneficial to increase the possibility of identify-ing a charging station with free resources.In this section, we will shortly revisit our previ-ous work from [19], which introduces a multi-modal routing scheme and a preliminary analysison gains in charging station availability. In thiswork, the preliminary analysis is extended to as-sess the travel time impact of multimodal routesunder different road traffic conditions.

2.1 Multimodal Routing HeuristicThe multimodal routing heuristic proposed in ourprevious work [19] involves the modalities ofEV-driving, public transport and walking. Theobjective of the heuristic is to identify a set ofcharging stations which are adjacent to publictransport stops that can provide user-mobility tothe final destination. The heuristic is based on aset of presumptions: 1) An EV-user can acceptto walk maximally Twalk seconds between twopoints on a route i.e. from charging station topublic transport stop and from a public transportstop to the destination, 2) The user can accept tospend maximally Tpt seconds in public transporttransit, 3) The desired EV driving range is limitedto EVrange which enables for instance to satisfylow battery EV range constraints. The heuristicis realized as follows:Off-line initialization: All charging stationswithin Twalk seconds of a public transport stopare identified.Online processing (at route request):

A. Identify the public transport stop closest tothe destination Pds (destination stop), whichis within Twalk walking distance. If noneare found, return an empty set.

B. Identify public transport stops P icss, i =

0...N (charging station stop). N corre-sponds to the amount of stops that canbe reached within Tpt seconds from Pds(assuming symmetric of public transportroutes). The travel time to each station canbe calculated from a shortest-path-tree [14].Based on P i

css, look-up all charging stationsCSj

all, j = 0...M where M are the amount

EVS26 Electric Vehicle Symposium, 2012 2

of charging stations within walking distanceof the stops in P i

css.

C. Of the charging stations CSjall filter

out irrelevant charging stations based onEVrange. In the following evaluations,charging stations are considered in a radiusdefined by the distance from the EV to thedestination stop + 2000 meter to avoid ex-cessively long routes.

D. For each remaining charging station, calcu-late two routes: 1) driving from the currentlocation of the EV to the charging stationand 2) public transport from the chargingstation to the destination. These routes aremerged to a joint multimodal route.

E. The routes calculated for each chargingstation in D. are returned in a final setCSk

options, k = 0...P to the EV-user for a fi-nal selection. P is the total amount of reach-able charging stations returned.

To evaluate the proposed mechanism, it hasbeen realized using open source route calcula-tion tools. Graphserver [20] is applied to derivethe public transport and walking routes whereasGosmore [21] provides vehicle routes. An evalu-ation scenario has been constructed from the Vi-enna region which represents a large city with astrong public transport network. For the evalu-ation, only the subway system is considered. Itcovers the major parts of the city region and alsoa majority of all public transport trips conductedin the region (64% [24]). Geographical data areobtained from OpenStreetMap.org [12] and pub-lic transport data from “Wiener Linien”.Results on charging station reachability and im-pact on total travelling time have been obtainedby generating trips in a 16 km by 16 km areadepicted in Figure 2. Note, the term reachabil-ity is used instead of availability as the resultsdo not consider if a changing station is free oroccupied, but only if it can be reached. A tripconsists of a starting point and a destination. Foreach trip three routing schemes are calculated: 1)a direct route which defines the driving time di-rectly to the destination, 2) a charging in desti-nation vicinity route, where a charging station iswithin Twalk seconds from the destination, and3) a multimodal route via a charging station andpublic transport. A multimodal and a direct routefor an example trip is provided in Figure 2. Togenerate and compare different trips, indepen-dent start and destination locations are generated.Start locations are randomly generated (uniform)all over the region to simulate trips starting in thecity as well as the surroundings. The destina-tions are randomly chosen from a database withpoints-of-interest (from data.wien.gv.at contain-ing schools, restaurants, hotels, sport centers,etc.) to ensure commonly relevant destinations.To assess the impact of multimodality in differ-ent degrees of electrified public parking places,three different EV penetration levels have beencompared: 10%, 50% and 100%. The pen-etration levels are controlled by the parameter

16.28 16.3 16.32 16.34 16.36 16.38 16.4 16.42 16.44 16.46

48.14

48.16

48.18

48.2

48.22

48.24

48.26

Parking with or without charging

Driving directly

Driving partially

Walking

Public transport

Start

Charging Station

Destination

Figure 2: Example trip from start to destination bydirect driving without charging, driving to a chargingstation in vicinity of the destination and travelling viapublic transport.

pCS , which describes the probability that a publicparking place has been electrified. For the eval-uation scenario, public parking places have beenobtained from the OpenStreetMaps data leadingto a total of 568 parking places.

Table 2: Main parameters used in the simulation ofrouting schemes.

Parameter ValueTrips evaluated 484Walking speed 5 km/hTrip start time 09:00:00 (weekday)Tpt unlimitedTwalk 300 s

From the analysis presented in [19], significantgains have been shown in terms of charging sta-tion reachability. The previously obtained resultsare summarized in Table 3 using the values of Ta-ble 2 comparing the destination vicinity chargingscheme to the multimodel routing scheme. Theresults show the fraction of trips where at leastone (P > 0) or at least 3 (P ≥ 3) charging sta-tions are reachable. Identifying several reachablecharging stations is clearly necessary when con-sidering that for a given charging station no freecapacity may be available.

Table 3: Results of reachability of one or more charg-ing stations on a trip.

CS set pCS = 0.1 pCS = 0.5 pCS = 1P > 0 8 .7% 39 .3% 62 .6%

40.5% 61.0% 74.2%P > 3 0 .0% 1 .9% 11 .0%

32.9% 35.7% 41.11%Destination vicinity charging, Multimodal


For both results sets, significant benefits of themultimodal scheme can be demonstrated; espe-cially in low penetration scenarios. For moredetails on this analysis interested readers are re-ferred to [19].

2.2 Impact of Traffic ConditionsIn the following, a quantification of the impactof the multimodal routes is made compared todestination vicinity charging routes in the Vi-enna city scenario. The quantification focuseson the change in average trip travel time betweenthe two schemes considering different road trafficconditions. This enables to clarify the potentialadvantages of the subway based transport, whichis clearly not affected by varying road traffic con-ditions. For the study, two extremes are consid-ered: perfect low traffic conditions and duringrush hour.To model the road traffic conditions a simplisticmodel is applied. The model separates the roadsegments of the studied region into the classesof city roads and motorways. Each class isparametrized by an average speed. This sim-plistic model is in correspondence with existingtraffic information systems for the Vienna regionallowing for a direct mapping of existing statis-tics [23]. The traffic conditions are modelled asaverage speed conditions for all vehicles in theroad network. As such the analysis does not takeinto consideration potential congestion scenariosin individual trips. Average speed parameters forcity roads have been obtained from the work in[22]. The authors describe an ITS system for on-line collection of traffic data based on taxis withGPS. The system has been deployed in Vienna toenable realistic average speed estimates under thedifferent conditions. Also based on GPS data, thework of [23] similarly presents average values formotorways in the Vienna region. A summary ofthe utilized parameters is provided in Table 4.

Table 4: Average speed according to traffic conditionsand road type.

Trafficconditions

City roads Motorway

Ideal 30 km/h 85 km/hRush hour 20 km/h 58 km/h

The classification of road segments into respec-tively city roads and motorways in the regionof Figure 4 has been made from maxspeed val-ues available in the OpenStreetMap data. Roadsegments with a maximum speed of 80 km/h orabove are classified as motorway.Utilizing the evaluation approach introduced inSection 2.1 and the values from Table 4 completeroute traveling times have been calculated. Fig-ure 3 presents the mean increase in travel timefor the two routing approaches under differentcharging station penetration levels. The increaseis calculated in relation to the direct route to thedestination (without charging). The results sug-gest under ideal road conditions that the differ-

0

5

10

15

Mean Incre

ase [m

in]

City Roads=30 km/h, Motorway=85 km/h

p =0.1 (53) p =0.5 (283) p =1 (568)0

5

10

15

Charging Station Penetration (Number of charging stations)

Mean Incre

ase [m

in]

City Roads=20 km/h, Motorway=58 km/h

Destination vicinity charging

Multi−modal only (>0)

Multi−modal only (≥3)

CS CS CS

Figure 3: Comparison of the overall trip time increasecompared to driving directly to the destination.

ence in increase between the destination vicinitycharging and the fastest multimodal route (greybar) are between 4 and 7 minutes depending onthe penetration level. Also, it can be observedthat the mean travel time does not increase sig-nificantly when considering going via one of thethree fastest multimodal routes (the white bar).Under rush hour traffic conditions, it can be ob-served that this difference is decreased to be-tween 1 minute (for PCS = 1) and 5 minutesfor low penetration scenarios.Which increase in travel time EV owners subjec-tively are willing to accept is up to future usabil-ity studies to clarify. It is, however, clear that themultimodal scheme offers a significant improve-ment in the reachability of charging stations. Fur-ther, our results suggest that in medium to highpenetration scenarios and rush hour conditionsthe mean travel time increase is in the order of afew minutes. Overall, the results are encouragingand future work must establish how the increasedreachability of charging stations can be applied toimprove overall charging availability and poten-tially geographical balancing of grid load.

3 Controlled Charging at theCharging Station

With the increase of EV penetration, controlledcharging, i.e. the coordination of charging timeslots will become indispensable both in residen-tial areas such as apartment blocks with elec-trified parking lots, and public charging facili-ties. A controlled charging strategy schedulesthe charging jobs in such a way that it reducesload peaks caused by users that plug-in their carsapproximately in the same time. In contrast toother works on recharge scheduling [8],[9],[10],our model uses as key parameter the time windowcorresponding to user’s estimated parking startand end time. Further required are the amount


of energy needed to fill up the battery and thecharging rates (in kW) supported by both the ve-hicle and the charging station.

3.1 Problem formulationFor the problem we want to solve there is givena time horizon of T discrete periods, a numberof M charging points and a number of A ac-tivities, so that each needs one of the M park-ing places for a duration time window (earli-est,latest), and 2) a power resource for a certainduration which depends on the selected charg-ing speed spk (from a discrete set). The solutionhas to determine the start charging time for eachactivity, the allocation to a charging point, suchthat the total power consumed in each time pe-riod does not exceed the planned value, and suchthat a certain profit objective function to be ex-plained below is maximized.Note that we have to deal with two types of re-sources: parking place during the whole timewindow, and power limitation during the charg-ing period. If we make the assumption that thecharging points are identical (machines), we cansplit the problem into two sub-problems:

1. since the machines are identical, we allo-cate first the activity time windows to thecharging points. This problem is similar toscheduling classes to classrooms, and canbe solved optimally by a greedy heuristic(increasing starting time rule). The intervalsare the activity time windows and cannot beshifted. The resulted allocation of activity tocharging point allows us to address the sec-ond subproblem as a one machine problem.

2. this is a bandwidth or resource allocationproblem over time intervals, or RAP) whichis NP-hard, since it can be reduced to aknapsack problem if the time windows areset to the intervaal [0,1]. The specificproblem which we denote EVRSTW (elec-tric vehicle recharge scheduling with timewindows) requires that the charging takesplace within the time window of the activ-ity, has different speeds (supported by bothcar and charging point), and is limited by to-tal power available in each of the T periods.

In case the machines are not identical, i.e. slowand fast charging, the machine allocation can bestill made under the following assumption: fastcharging spot is an expensive resource, thereforeis will not be used for slow charging, which leadsto two disjoint activity groups. In the most gen-eral case, instances have to be created for eachindividual machine, making the problem verylarge.In the rest of this section we assume the machinesare identical, so that the allocation activity-charging spot has been done and focus onlyon the second subproblem, the RAP. This andsimilar problems have been studied extensively.The latter problem is highly combinatorial, sincethe charging interval can be selected anywherewithin the time window, can have different dura-tions determined by the charging speed and the

DescriptionT set of time periods t ∈ TA Set of charging activitiesM set of charging pointsI instances generated from activities A

ej , lj time window (earliest, latest) j ∈ Asi, fi start and finish time of charging , i ∈ Idj energy demand of activity jP t total available power during t ∈ Twi charging speed i ∈ Ivti parking period of instance i ∈ I, t ∈ Tci completion degree of instance, one of Cpi profit generated by instance ixi variable: xi ∈ {0, 1}, i ∈ I .

Table 5: Summary of notation.

completeness criteria, leading to different prof-its for the same activity. In order to simplify theILP formulation, we generate from each parame-ter combination an instance i and use one binaryvariable xi to denote that that instance is selectedin the solution, see Bar-Noy et al [15] :for j ∈ A

for m ∈M (not identical m)for n ∈ C

for k ∈ Sfor l ∈ [ej , lj − dur], dur = ndi/k

generate instance inst(i,m, n, k, l)end

endend

endend

The profit pi is defined as a weighted sum oftwo objectives: the completion degree ci and thecost factor k[spi] of the charging speed (quickcharging requires more expensive equipment andshortens the battery life). We set α = 0.5

pi = αci + (1− α)k[spi] (1)

Problem EVRSTW:

Z = max∑i∈I

pixi (2)

s.t.∑

i∈I|a[i]=j

xi = 1, j ∈ A (3)

∑i∈I|t∈[si,fi]

wixi ≤ P t, t ∈ T (4)

∑i∈I|m[i]=m

vtixi ≤ 1,m ∈M, t ∈ T (5)

xi ∈ {0, 1}, i ∈ I (6)


∑i∈I

xi ≤ 1 (7)

The inequality (4) limits for each time slot t thetotal power due to simultaneous activities to theavailable power P t. The equality (3) makes surethat exactly one instance of each activity is se-lected.

3.2 Heuristic algorithmSince problem EVRSTW is NP-complete, a nat-ural step would be to look for approximationheuristics that would performe satisfactorily in alarge charging station.A first selection has been the local ratio algo-rithm which belongs to a class of stack basedheuristics [16], and has been proposed in pack-ing problems to maximize bandwidth through-put [17], [15]. In our case the bandwidth corre-sponds to the charging power. For this type ofproblems, the algorithm can achieve a 1/3 ap-proximation, which gets better as the ratio be-tween the charging power (speed) of individualactivities and the total available charging poweris smaller. The algorithm works as follows: theinstances I are first sorted in increasing order oftheir end (charging) times. In the first sweep weselect an instance with minimum end time anddecrease the profits of all instances that a) belongto the same activity, b) overlap with the selectedinstance. Instances with non-positive profits areremoved, and the selected activity is pushed on astack. When no activities remain unconsidered,in the second sweep activities are popped fromthe stack and those who violate the total powerconstraint (4) are deleted. Denote the selected in-stance i, then the profits of conflicting instancespi are reduced as follows: pi = pi−βwipi, whereβ = 2. Other than in the exact algorithm, theamount of resource has to satisfy 0 ≤ wi ≤ 1.The problem arises at the calculation of if wi ifwi varies in the interval [si, fi]. The conservativerule would be

wi = mint∈[si,fi]

wi (8)

Another rule would be to use the average overthe interval. The algorithm and its implementa-tion are described in [15].

3.3 Computational resultsWe first have evaluated the integer programEVRSTW and the local ratio algorithm in termsof running time and accuracy. The time horizonwas 32 time slots of 15 minutes each. The sys-tem is initially empty, the activities are createdfrom randomly uniformly distributed time win-dows and energy demand values (earliest [0,20],duration [4,10], demand [5,10]).The main pa-rameter to determine the runtime performance isnumber of activities A. M and P are selected ac-cordingly to provide similar load conditions. Ta-ble 6 shows the results with the weighted objec-tive 1 with α = 0.5

|A| |M | Pi ZIP tIP ZH tH20 16 30 13,5 11 9,9 940 30 40 23,1 19 15,8 2780 48 70 40,8 37 35,4 132120 64 90 62 32 51,5 550160 80 150 93,9 20 74,7 126050 32 40 25,1 18 20,2 4550 32 50 30,0 17 24,5 4850 32 60 32,3 17 26,9 5550 32 70 33,2 22 29,0 6050 32 80 33,9 18 28,6 62

Table 6: Results for ILP and Heuristic

Furthermore, we have investigated the impact ofthe available power P on the solution for both theILP and heuristic algorithm. Relaxing the con-straint P both solutions slightly improve.The computations reveal that the solutions ob-tained with the ILP model, using the CPLEXsolver are superior both in accuracy and runtime,see Table 6. The computational complexity ofthe local ratio heuristic depends mainly on thenumber of instances which is proportional to thenumber of activities. The number of generatedinstances depends on the input, because a largetime window will create more instances than asmall one. The runtime depends also on theweight w of instances. The profit of an instancedecreases during the first sweep iterations, until itis eventually deleted. Therefore, for small weightvalues w(i, j) (i.e large P), the instances will per-sist for more iterations, increasing the runtime ofthe heuristic. The solution quality of the heuristicis in worst case 1/2 of the optimum (for ”small”bandwidth activities), as stated in [16] and [15].

80

90

60

70

40

50

20

30

0

10

20

0

1 3 5 7 9 11 13 15 17 19 21 23 25 27

Available power P[i]

Cumulative load after ILP scheduling

Cumulative load heur min rule

Cumulative load heur average rule

Figure 4: Cumulative load of the optimized activityschedule versus the available power

Finally, we simulated varying values of the avail-able power (that can be obtained in reality fomthe metering values, and applying a grid flow cal-culation procedure [5]). The created schedule us-ing the ILP model produces a cumulative load,defined as L(t) =

∑i|t∈[si,fi] xidi that is tight to


the limiting available power profile, see Figure 4.The heuristic performance depends on the calcu-lation of wi according to 8 or average. Using theaverage, the cumulative load gets more closer tothe upper limit P t, however sometimes it exceedsthis limit. Using the minimum rule, the solutionis feasible but the performance is poor, see Figure4.

4 Conclusions and Future WorkIn this work we addressed two related aspects ofe-mobility: the logistic problem of finding a pub-lic charging station, and the charging energy bal-ancing problem that affects both the grid providerand the user. We proposed a distributed archi-tecture in which the charging station has somelocal control autonomy, and which achieves en-ergy balancing at three levels: at the charging sta-tion level through controlled charging, at the LVgrid level through grid flow calculation and at therouting level using the brokerage function.We proposed for the user logistic problem an on-line routing service that uses mobile communica-tion, and by including public city transportation,we have shown that the availability of chargingspots can be drastically increased, particularly atlow charging station densities.For the scheduling of charging intervals, we havegiven an integer linear formulation and a localratio heuristic and provided preliminary resultsthat show in practice that greedy is too slow andthat the accuracy has to be improved. Currently,research is being done to evaluate local searchheuristics to the scheduling problem.The system components, routing server, charg-ing station controllers and low voltage grid flowcalculation module are currently integrated witha discrete event simulator that will simulate therouting and energy related processes: activitybased mobility model in which EVs issue reser-vation requests on-trip, routing, arrival and plug-in, charging and leaving. Efficiency metrics suchas utilization of charging resources, overall per-formance of different charging strategies and in-teraction usability are going to be evaluated.

AcknowledgmentsThis work is performed within the Aus-trian project KOFLA ”Kooperative Fahrerun-terstutzung fur Lademanagement von elek-trischen Fahrzeugen”, which is funded by theprogramme ”ways2go” and the Austrian Fed-eral Ministry for Transport, Innovation and Tech-nology. Special thanks to our partners WienerStadtwerke and VKW for providing mobility andgrid data.

References[1] KOFLA Project website, online:

http://www.ftw.at/research-

innovation/projects/kofla

[2] M. Herry, R. Tomschy, Travel behaviour surveyin Lower Austria 2008, Final Report, Vienna2009.

[3] M. Herry, R. Tomschy, Travel behaviour surveyin Vorarlberg 2008, Final Report, Vienna 2009.

[4] Antti Rautiainen, Corentin Evens VTT, SamiRepo, Pertti Jarventausta, Requirements for anInterface Between a Plug-in Vehicle and an En-ergy System,IEEE PES PowerTech,Trondheim2011.

[5] A. Schuster, M. Litzlbauer, Easy grid analysismethod for a central observing and controllingsystem in the low voltage grid for E-Mobilityand renewable integration, 3rd European Con-ference SmartGrids and E-Mobility, Munich,October 17-18, 2011.

[6] J. A. Pecas Lopes, F. J. Soares, P. M. Almeida,M. Moreira da Silva, Smart Charging Strategiesfor Electric Vehicles: Enhancing Grid Perfor-mance and Maximizing the Use of Variable Re-newable Energy Resources, EVS-24, Stavanger,Norway, May 13-16, 2009.

[7] R.A.Verzijlbergh, Z.Lukszo, J.G.Slootweg,M.D.Ilic, The Impact of Controlled ElectricVehicle Charging on Residential Low Volt-age Networks, International Conference onNetworking, Sensing and Control Delft, theNetherlands, 11-13 April 2011.

[8] K. Clement, E. Haesen, and J. Driesen, Co-ordinated Charging of Multiple Plug-In Hy-brid Electric Vehicles in Residential Distribu-tion Grids, PES Power Systems Conference andExposition, 2009.

[9] O. Sundstrom and C. Binding, PlanningElectric-Drive Vehicle Charging under Con-strained Grid Conditions, International Confer-ence on Power System Technology, 2010.

[10] P. Sanchez-Martin and Guillermo Sanchez, Op-timal Electric Vehicles Consumption Manage-ment at Parking Garages, IEEE PES TrondheimPowerTech 2011.

[11] Draft ETSI TS 101 556-1, Intelligent TransportSystem (ITS); I2V Application; Electric Ve-hicle Charging Spot Notification Specification,2011.

[12] CloudMade extract of Wien, Austria fromOpenStreetMap.org. Downloaded August 2011.


[13] Wienerlinien, Ein Blick auf Die Leistung -Jahresbericht 2010.

[14] Dijkstra, E. W. (1959). A note on two problemsin connexion with graphs. Numerische Mathe-matik 1: 269 271.

[15] A Bar-Noy, R Bar-Yehuda, A Freund, A Uni-fied Approach to Approximating Resource Al-location and Scheduling, Journal of ACM, 2001

[16] Allan Borodin, David Cashman, and Avner Ma-gen. 2011. How well can primal-dual and local-ratio algorithms perform?. ACM Trans. Algo-rithms 7, 3, Article 29 (July 2011), 26 pages.DOI=10.1145/1978782.1978784

[17] Reuven Bar-Yehuda, Keren Bendel, Ari Fre-und, and Dror Rawitz. 2004. Local ratio: A uni-fied framework for approximation algorithms.In Memoriam: Shimon Even 1935-2004. ACMComput. Surv. 36, 4 (December 2004), 422-463. DOI=10.1145/1041680.1041683

[18] ETSI TS 101 556-1, Intelligent Transport Sys-tem (ITS); I2V Application; Electric VehicleCharging Spot Notification Specification, 2011.

[19] Sandford Bessler, Jesper Grønbæk, Max Herry,Andreas Schuster, Rupert Tomschy. SupportingE-mobility Users and Utilities towards an Opti-mized Charging Service. European Electric Ve-hicle Congress, EEVC 2011.

[20] Graphserver: The open source multimodal tripplanner.http://graphserver.github.com/graphserver/.

[21] Gosmore:http://wiki.openstreetmap.org/wiki/Gosmore/.

[22] Schafer, R.P. and Thiessenhusen, K.U. andWagner, P. A traffic information system bymeans of real-time floating-car data. ITS worldcongress. Volume 2, 2002.

[23] Martin Peter and Daniel Suter. COMPETE An-nex 3. Country reports on congestion and de-lay measurement and their assessment in theEU and the US. Annex 3 section 16.1. Austria.2006.

[24] Wienerlinien, Ein Blick auf Die Leistung -Jahresbericht 2010.

AuthorsSandford Bessler has a M.Sc. inElectrical Engineering (1983) anda Ph.D.in Operations Research(1988), both from Vienna Univer-sity of Technology. From 1980till 2001 he held different R&Dpositions at Kapsch AG in thedevelopment of packet networks,later internet. Since 2001 Dr.Bessler works as a key researcherat The Telecommunications Re-search Center Vienna (FTW) andas a lecturer at the Vienna Uni-versity of Technology. His mainresearch interests are resourceoptimization problems in thee-mobility, and service conceptsin intelligent transportation andenergy distribution networks.

Jesper Grønbæk has anM.Sc.E.E. (2007) from Aal-borg University (AAU). In 2010he finished his Ph.D. from AAUunder the topic of dependableservice provisioning in futureubiquitous networks. Since 2010Dr. Grønbæk has worked as asenior researcher at The Telecom-munications Research CenterVienna (FTW). His main researchinterests are intelligent energynetworks, e-mobility and wirelessnetwork fault management.


routing ev users towards an optimal charging plan

Documents