Download pdf - 392 IEEE TRANSACTIONS ON TRANSPORTATION …kcleung/papers/journals/... · power system due to power system constraints, such as the lack of capacity to transport renewable energy

392 IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION, VOL. 3, NO. 2, JUNE 2017

Vehicular Energy NetworkAlbert Y.S. Lam, Senior Member, IEEE, Ka-Cheong Leung, Member, IEEE, and Victor O.K. Li, Fellow, IEEE

Abstract— The smart grid spawns many innovative ideas, butmany of them cannot be easily integrated into the existingpower system due to power system constraints, such as thelack of capacity to transport renewable energy in remote areasto the urban centers. An energy delivery system can be builtupon the traffic network and electric vehicles (EVs) utilized asenergy carriers to transport energy over a large geographicalregion. A generalized architecture called the vehicular energynetwork (VEN) is constructed and a mathematically tractableframework is developed. Dynamic wireless (dis)charging allowselectric energy, as an energy packet, to be added and subtractedfrom EV batteries seamlessly. With proper routing, energy can betransported from the sources to destinations through EVs alongappropriate vehicular routes. This paper gives a preliminarystudy of VEN. Models are developed to study its operationaland economic feasibilities with real traffic data in U. K. Thispaper shows that a substantial amount of renewable energy canbe transported from some remote wind farms to London undersome reasonable settings and VEN is likely to be profitable inthe near future. VEN can complement the power network andenhance its power delivery capability.

Index Terms— Dynamic charging, electric vehicle (EV), energydelivery, energy routing.

I. INTRODUCTION

THE smart grid spawns many innovative ideas and appli-cations, potentially revolutionizing the power system.

Unfortunately, many of these new ideas cannot be easilyintegrated into the existing power system due to power sys-tem constraints. One such constraint is the lack of capacityto transport renewable energy, typically in remote areas, tothe urban centers. Any new smart grid design needs to bethoroughly tested and demonstrated to be fool-proof beforebeing integrated into the real power system. Some examplesof complications are given in the following.

A. Renewable Energy

There is tremendous amount of renewable energy genera-tion, but only a small portion is available and utilized at theloads. As of 2009, China had less than one-third of wind farmsconnected to the grid due to the difficulty of intermittent powerdispatch and transmission network limitations [2]. In 2013, the

Manuscript received June 27, 2016; revised September 24, 2016 andNovember 24, 2016; accepted January 1, 2017. Date of publication January 9,2017; date of current version June 14, 2017. This work was supportedin part by the Theme-Based Research Scheme of the Research GrantsCouncil of Hong Kong under Grant T23-701/14-N. The work of A.Y.S. Lamwas supported by the Seed Funding Programme for Basic Research ofThe University of Hong Kong under Grant 201601159009. This paperwas presented at the 6th IEEE International Conference on Smart GridCommunications, Miami, FL, USA, 2015 [1]. (Corresponding author:Albert Y.S. Lam.)

The authors are with the Department of Electrical and Electronic Engineer-ing, The University of Hong Kong, Hong Kong (e-mail: [email protected];[email protected]; [email protected]).

Digital Object Identifier 10.1109/TTE.2017.2649887

ISO New England cut back power from wind and hydroelectricplants a few times because too much electricity was producedand transmission lines with robust carrying capacity to connectthe wind farms located in remote areas were missing [3].However, cutting renewable output may lead to failure tomeet the mandates of renewable energy generation in somecountries [4].

B. Ancillary Services

When discrepancies occur between supply and demand inthe power system, ancillary services are required. Operatingreserves can be provided in diverse locations of the grid anddistributed energy resources (DERs), such as the renewablesand aggregations of electric vehicles (EVs) [5], may be prefer-able over conventional generators in some situations. Furtherstudy is still required before DERs may be brought into thepower system effectively.

C. New Energy Markets

In the power system, the demand is connected with thesupply through the power networks. The power infrastructureis generally managed or owned by multiple parties. However,the smart grid can introduce many variations and uncertain-ties to both supply and demand, e.g., from DERs, demandresponses, and vehicle-to-grid systems [6]. These new playershave great potential to develop new energy markets withnonstandard operational models, but they still rely on thepower network to transmit power. The grid operators in generaloppose the integration of new operational models until theyhave undergone the reliability and security assurance. If thepower network can be supplemented for energy transfer, thenew energy businesses can thrive.

The common theme among the above-mentioned scenariosis that the integration of new smart grid elements cannotguarantee that the operations of the current power system willnot be affected. If there exists a power delivery infrastructureindependent of the conventional power network to connect thevarious (stable or unstable) types of power sources and loads,many of the brilliant ideas in smart grid can become a realityovernight.

There is some related works toward that direction. Ref-erence [7] proposed the EV energy network, which utilizesEVs for energy transmission and distribution. In this design,EVs charge up when they stop at some energy routers anddischarge the energy when stopping at other energy routers.In [8], a greedy algorithm for energy scheduling and alloca-tion was developed. Shortest path routing [9] and multiplepath routing [10] were also studied. All these efforts arebased on the model developed in [7] and a well-defined

2332-7782 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

LAM et al.: VEHICULAR ENERGY NETWORK 393

framework is required for more in-depth analysis. OnlineEVs (OLEVs) [11] are EVs which support dynamic charg-ing, i.e., wireless charging in motion, and they have beencommercialized in South Korea. Reference [12] designed awireless power transfer (WPT) system to facilitate dynamiccharging for OLEVs. Reference [13] analyzed the benefits ofdynamic charging with an economic model and it showed thatdynamic charging is beneficial to battery life. References [14]and [15] introduced mobile energy disseminators (MEDs).MED is a moving vehicle and it can wirelessly charge someother moving vehicles in the neighborhood.

In this paper, our preliminary work [1] is extended andwe aim to develop a framework to model an energy deliv-ery architecture for transporting energy from one place toanother by means of EVs. Our framework generalizes theabove-mentioned infrastructure and designs and the systemmodeled by this framework is called the vehicular energynetwork (VEN). This framework defines the necessary com-ponents to facilitate more in-depth research. The rest of thispaper is organized as follows. VEN with supporting argumentsand characteristics are defined in Section II. Section III givesthe system model and quantifies the transferable energy andloss. In Section IV, the system is analyzed by studyingenergy transfer maximization and energy loss minimization.An economic model is developed for VEN in Section V.Section VI studies the operational and economic feasibilitiesof VEN and this paper is concluded in Section VII.

II. SYSTEM DESIGN

There are a number of existing technologies facilitating theestablishment of VEN.

A. Renewable Energy

To confront global warming and climate change, manynations have established targets and mandates for renewableenergy use [16]. For example, California and Colorado inthe U.S. have mandated renewable targets of 33% and 30%by 2020, respectively. The European Union 2030 target is atleast 30% of energy coming from renewable sources [16].China sets the 15% renewable target with 500 GW renewableelectricity by 2020 [16]. Moreover, renewable power capacityis massive [17]. Therefore, an appropriate approach to managethe abundant renewable energy generation can help meetvarious nations’ energy mandates.

B. Electric Vehicles

EVs refer to a family of vehicles with batteries equippedto store energy for operations. An EV can be consideredas a movable energy storage for various applications, e.g.,facilitating frequency regulation [5] and demand response [18].Boosting the number of EVs is also included in the greenpolicies of many countries. For example, the U.S. sets the goalof having 1 million EVs on the road by 2015 [19] and Chinatargets to have a similar goal [20]. Many automotive compa-nies have already included EVs in their major production lines.As the related equipment (e.g., batteries) is improved and thefacilitates (e.g., charging stations [21]) become available, EVswill be prevalent in the near future.

C. Vehicular Ad Hoc Network

Vehicular ad hoc network (VANET) is a mature technologyallowing vehicles, as mobile nodes, to communicate witheach other or some fixed infrastructures [22]. Newly designedvehicles, especially EVs, are equipped with many sensorsand this allows VANET to support a variety of functions,such as surveillance and route planning. This nurtures newapplications where vehicles submit (parts of) their intendingtraveling paths for planning purposes.

D. Wireless Power Transfer

Most near-field electromagnetic induction-based WPT tech-niques can be categorized into magnetic induction and elec-trostatic induction [23]. The former is suitable for applicationswith various power levels and gap separations, while thelatter is more restricted to those with small gap distances.Reference [24] reviewed various WPT technologies for EVwireless charging. An inductively coupled multiphase reson-nant converter was designed for wireless EV charging appli-cations [25]. Reference [26] focused on heavy duty vehiclesand discussed the required elements of WPT for high powercharging. Furthermore, batteries of EVs can also be chargedwirelessly on the move, i.e., dynamic charging. Dynamiccharging facilitates frequent charging so that the batteries canbe made smaller and the overhead costs due to batteries can belowered. Shallow and frequent charging can enhance batterylife [13]. A double-spiral repeater was proposed to designeffective dynamic WPT systems [27]. Reference [28] revealedthat inductive power transfer is promising to dynamic charg-ing. In [29], an efficient dynamic wireless charging system forEVs was designed based on magnetic-coupled resonant powertransmission. Some companies, such as Qualcomm [30], ABB,and Microsoft [31], are working toward improving the wirelesscharging technologies for EVs. A Stanford team proved thatpower up to 10 kW could be transferred effectively with amoving car [32]. OLEVs have been deployed in Gumi Cityof South Korea [13]. England demonstrated a real dynamicWPT system on some motorways [33]. In [34], a simulationmodel was developed to analyze the daily amount of energysupported by a dynamic charging system. These evidencesshow that wireless charging and discharging can take placewithout interfering the movements of EVs.

The aforementioned technologies and developments makeVEN a possible energy delivery infrastructure. EVs can beutilized as energy carriers. Consider an EV moving on aparticular route connecting Locations A and B. If the EVis wirelessly charged at A and discharged at B, energy willbe brought from A to B via the EV. Again, suppose thatafter a while, another EV passes through Location B andmoves toward Location C. Then that EV can be charged anddischarged at B and C, respectively, and this allows us to bringthe energy to Location C. In this way, if there are vehicularroutes intersecting one another and some pass through theenergy sources and destinations, the required energy can betransmitted from the sources to the destinations through VEN.VEN is formally defined as follows.


Fig. 1. Schematic of charging and discharging facilities with energy storageat a segment of road.

Definition 1 (Vehicular Energy Network): VEN is a vehic-ular network aiming to distribute energy in a region bymeans of EVs without tight time limitation. It is built upona road network where EVs run on certain routes and thereare dynamic (dis)charging facilities, with limited storage forenergy between each charge and discharge, installed alongthe roadway or at some road junctions. Nodes with energyto be delivered and received are denoted as the energysources and destinations, respectively. During charging, acertain portion of energy is transmitted to an EV froma charging facility, and similarly, a portion of energy istransmitted from an EV to a discharging facility duringdischarging. An EV carries a “packet” of energy along itsroute between its charge and discharge. By properly chargingand discharging certain EVs at selected locations along theroadway, energy can be brought from the energy sources to theenergy destinations.

We do not need to strictly enforce energy supply anddemand balance at all times in VEN. The design of VEN isnot tied to particular types of energy generation and load. It isrelatively easier to construct VEN than the traditional grid insome places, e.g., remote areas. It has an ability to store upenergy, as an energy storage, but in a global sense. Dependingon the applications, we have different ways of manipulating theenergy stored in the network. Moving energy around is still afundamental operation of VEN. To the best of our knowledge,we cannot find a counterpart which is directly comparable toVEN, in terms of characteristics and functionalities.

Fig. 1 shows an illustrative example of a segment of road,where charging and discharging tracks are buried undergroundin each designated lane. The charging and discharging coilscan be installed at different locations to accommodate theconfiguration of roadway. These tracks are connected via acommon bus with an energy storage attached. A (dis)chargingtrack is responsible for (dis)charging the appropriate EVsrunning over it dynamically. The wireless energy transferequipment can be designed similar to the one given in [12]. Atany moment, it is possible to have multiple EVs undergoingcharging and discharging simultaneously. Energy will be trans-ferred from discharging EVs to charging EVs directly. Thedeficit will be compensated by the storage while the excesswill be stored.

Note that we do not need to actively control the EVsparticipating in VEN. The EVs are not required to stop orslow down for (dis)charging at particular locations as energycan be transmitted while they are moving. The EVs are not

needed to follow any dedicated routes. Instead, those EVs areselected with favorable routes to carry energy so as to performenergy delivery.

VEN possesses the following characteristics which make itsuitable as a supplement to the power grid to deliver power.

1) Controllable Energy Transmission Rate: Energy is trans-mitted in the form of packets and this allows VEN to bemanaged like a packet-switched data network. Given theintended route of an EV,1 It is known where the energyit carries can be delivered. By controlling how many EVsare utilized to carry energy and when proper charging anddischarging events take place with the EVs, the energy flowcan be controlled and the energy transmission rate can bespecified on each road segment.

2) High Flexibility: A dedicated infrastructure does notneed to be built in order to realize VEN. VEN relies onthe existing road network, which covers almost all locationsinvolved in human activities. A road network can be trans-formed into a VEN when a certain number of (dis)chargingfacilities have been installed. Unlike the conventional powernetwork where the power sources and loads are generallyprespecified, the locations of energy sources and destinationson VEN can be modified from time to time.

3) Low Overhead: Equipment does not need to be specif-ically designed for VEN; even without VEN, WPT facilitieswill likely be found in the road network to serve the growingEV population in the future.

Existing technologies render VEN feasible. There are somenontechnological factors which may be important when VENis brought into reality. Due to space limitations, more detailscan be found in [1].

III. SYSTEM MODEL

A. Road Network

A road network is modeled with a directed graph G(N ,A),where N and A are the set of energy points and the set ofroads connecting the energy points, respectively. N includesall possible energy sources, destinations, and routing pointssetup at those locations with (dis)charging facilities installed.Depending on the application, G(N ,A) may cover a district, acity, or even a whole country. Let head(ai) and tail(ai) be thehead and tail of arc ai ∈ A. Each arc ai incurs a delay d(ai )time units for transferring energy from tail(ai ) to head(ai);when energy is carried by a vehicle, it takes d(ai ) units totraverse ai .

B. Vehicular Traffic

Assume that the traffic flows are static. Let R be the setof all possible vehicular routes in G, each of which is loop-free. Each ri ∈ R is a sequence of connected arcs, i.e.,ri = 〈a′

1, . . . , a′|ri |〉 with traffic flow fi , which is the numberof vehicles traveling on ri per-unit time. The nth arc of ri is

1For public transport, the exact schedules and routes can be acquired. Forprivate vehicles, we expect that the participants are willing to disclose partsof their intended routes to the system. At the least, as soon as a vehicle is ona particular road segment, it will stay until the end of that segment for surebefore it can exit.


Fig. 2. Energy paths.

denoted by ri (n). Let ri (n, m), n < m, be the subroute of ri

starting from the nth arc and ending with the mth arc.

C. Energy Path

Definition 2 (Energy Path): An energy path p(s, t) is thepath along which energy is transmitted from the energy sourcenode s to the destination node t . Each path is composed ofsegments of the vehicular routes.

For each pair of (s, t), the set of all possible pathsP(s, t) can be constructed. Each path p j (s, t) ∈ P(s, t)can be represented by p j (s, t) = 〈r j

1 (n1, m1), . . . ,

r ji (ni , mi ), . . . , r j

|p j |(n|p j |, m|p j |)〉, where r ji (ni , mi ) refers to

the i th segment of p j , and it is the subroute of r ji with the

starting arc ni and the ending arc mi . |p j | is the numberof subroutes used to construct p j . p j (s, t) also needs tosatisfy: 1) tail(r j

1 (n1)) = s; 2) head(r j|p j |(m|p j |)) = t ; and 3)

head(r ji (mi )) = tail(r j

i+1(ni+1)), for i = 1, 2, . . . , |p j | − 1.Consider the example given in Fig. 2, which contains

a subgraph of G. There are three vehicular routes, r1, r2,and r3, and each passes through some of the nodes in thesubgraph. Specifically, r1 passes through Nodes 1 and 3,r2 passes through Nodes 2–4, and r3 passes throughNodes 1, 2, 5, and 4. Suppose that we decide to trans-mit energy from Nodes 1–4. There are three energy pathsin P(1, 4) = {p1(1, 4), p2(1, 4), p3(1, 4)}. p1(1, 4) =〈(1, 3), (3, 4)〉 is constructed from two vehicular routes,where (1, 3) and (3, 4) come from r1 and r2, respec-tively. Hence, we have p1(1, 4) = 〈r1

1 ((1, 3)), r12 ((3, 4))〉.

p2(1, 4) = 〈(1, 2), (2, 3), (3, 4)〉 is constructed from r3 and r2.We have p2(1, 4) = 〈r2

3 ((1, 2)), r22 ((2, 3), (3, 4))〉. Similarly,

p3(1, 4) = 〈(1, 2), (2, 5), (5, 4)〉 is constructed from r3 only.Hence, we have p3 = 〈r3

3 ((1, 2), (5, 4))〉.In general, more nodes result in more edges in the net-

work. The denser the network, the more energy paths wecan establish between the energy sources and destinations.This results in higher flexibility in routing energy across thenetwork. Consider the above-mentioned example again. If weremove Node 3 and its associated arcs, there will be one energypath left connecting Nodes 1 and 4, i.e., P(1, 4) = {p3(1, 4)}.It can be easily seen that it is more flexible to route energywith three energy paths than one single path.

The maximum amount of energy to be transmitted in eachcharging or discharging event is standardized and it can becarried by each EV in each charging–discharging cycle, i.e.,

the “packet size,” denoted by w units.2 w is a system parameterand its value should be set subject to the adopted WPTtechnology and equipment, the lengths of the (dis)chargingfacilities, and the battery capacities of the participating vehi-cles. It should be small enough so that the transmission ofenergy packet can complete in every charging and dischargingevent and that an appendance or removal of an energy packetfrom a vehicle will not make its state-of-charge fluctuatingsignificantly. Let f j

i be the traffic flow of the i th segmentof p j . By assigning some EVs along the subroutes to carryenergy, the energy transfer rate of p j , denoted by g j units,satisfies

g j ≤ w f ji , i = 1, . . . , |p j |. (1)

D. Charging–Discharging Cycle

There is energy loss in both of the dynamic charging anddischarging processes. The energy efficiencies of charging anddischarging are given by zc and zd , respectively, and thus,the fractions of (1 − zc) and (1 − zd ) correspond to the por-tions of energy lost in charging and discharging, respectively.When a vehicle is employed to carry energy, there exists acharging–discharging cycle along each subroute. For example,in Fig. 2, when some energy is decided to be transported fromNodes 1–3, a vehicle can be charged along r1 at Node 1 andthen discharged at Node 3. Let z = zczd . Hence, at the end ofeach charging–discharging cycle, only a fraction z of energycan be retained.

If the designated routes of all EVs are known, the EVscan be used to transmit energy more than one hop in Gwithout discharging. Otherwise, they need to be dischargedand recharged at each node (i.e., the charging–dischargingcycle takes place at each hop). With the support of VANET,most vehicles are connected. Without loss of generality, theparticipating EVs are obligated to report (part of) their travelplans. In this way, for each route ri , it can be told which EVsare traveling along the route. Suppose that the routes of allEVs are known. If x units of energy are required to reach thedestination of the energy path p j , which is composed of |p j |subroutes, x

z|p j | units of energy need to be transmitted at thesource of p j . On the other hand, this incurs energy loss equalto ( 1

z|p j | − 1)x units.

E. Delays

Thanks to the opportunistic contacts of the vehicles via the(dis)charging facilities, VEN is not an instantaneous system.Time can become critical to energy transfer over VEN. In gen-eral, a delay characterizes how much time is required for aparticular event to happen and we evaluate all time-relatedissues in terms of delays. Due to its “packet switching”-likecharacteristics, we can analyze the delay incurred to routeenergy packets from a particular source to a destination interms of the following [35].

2For the ease of modeling and implementation and without loss of general-ity, the same packet size of w units is assumed for each charging–dischargingcycle on each vehicle.


1) Processing Delay: In computer networking, it refersto the time required to examine the packet’s header andto determine the outgoing path at a router. While data indifferent data packets are basically unique, energy in VENis a single commodity. As long as sufficient energy reachesthe destination as planned, it is not important if the energyactually comes from the designated source or is “borrowed”from other sources. Therefore, there is basically no header inan energy packet. Furthermore, the processing delay may alsobe considered to capture the time required to transfer energywirelessly between the vehicles and the charging facilities.As discussed in Section II, wireless energy transfer takes placewhen the vehicle is moving along its route. Energy transfercan be completed before the vehicle leaves the (dis)chargingtrack (see Fig. 1) if the packet size w is sufficiently smallor the tracks are sufficiently long. As vehicle travel time willbe included in the propagation delay, processing delay can beignored in VEN.

2) Queuing Delay: As seen in Fig. 1, wireless chargingand discharging events normally take place simultaneouslyalong the road segments installed with (dis)charging facilities.Moreover, it is not uncommon for a (dis)charging segmentto serve multiple transmissions of different source–destinationpairs at the same time. Together with the single commodityproperty, energy at the energy storage can be advanced to someextent that a charging event take place before a dischargingevent along an energy path. Therefore, energy packets do notneed to be queued at the storage and thus there is no queuingdelay.

3) Transmission Delay: It refers to the amount of timeneeded to push all the energy packets onto an energypath. This depends on how many vehicles available travel-ing along the vehicular routes on the path. This is relatedto the energy transfer rate g j constrained by the trafficflow f j

i [see (1)].4) Propagation Delay: It refers to the time required to

propagate an energy package from its source to its destina-tion. At the (dis)charging facilities, energy can be transferredwirelessly from and to the vehicles on the move. It takes timed(p j ) units for energy to “propagate” along p j , where

d(p j ) =|p j |∑

i=1

d(r ji (ni , mi )) =

|p j |∑

i=1

∑

ak∈r ji (ni ,mi )

d(ak). (2)

Therefore, the amount of time required to transport energy iscomposed of the total “propagation delay” and “transmissiondelay” along the transmission path. To transport x

z|p j | units ofenergy along p j from its source, it takes a duration of d(p j )+

xz|p j |g j

units [see (2)]. In other words, in a time window of T

units, the amount of energy transferable along p j , denotedby x j units, is governed by x j ≤ (T − d(p j ))z|p j |g j , andthe corresponding energy loss incurred is ( 1

z|p j | − 1)x j units.Therefore, the amount x(s, t) units of energy transferred toNode t from Node s in a time period T satisfies

x(s, t) =∑

j |p j∈P(s,t)

x j ≤∑

j |p j∈P(s,t)

(T − d(p j ))z|p j |g j . (3)

The corresponding amount of energy loss of L(s, t) units isgiven by

L(s, t) =∑

j |p j∈P(s,t)

(1

z|p j | − 1)x j . (4)

IV. SYSTEM ANALYSIS

To utilize VEN, the system needs to be configured by settingup the appropriate energy paths, each of which can deliverenough energy within a given time frame. Two key concernsare the transferable amount of energy and the energy loss.They will be studied as optimization problems and give someanalytical results.

A small packet size of w units is sufficient to facilitatea noticeably large amount of energy transfer over a largegeographical area and this will be verified in Section VI.Since a practical value of w is very small when comparedwith the battery capacity of typical EVs, for a single-batteryEV, only a small portion of the battery will be reserved forVEN, while the rest can still be used to support normal EVoperations. It can basically be assumed that an EV can servemultiple energy paths by carrying multiple energy packetssimultaneously. It is unlikely that an EV cannot serve due tobattery overflow. As discussed, w should be carefully chosenwith a sufficiently small value by considering the battery sizesof all participating EVs. Moreover, the longer a vehicle runs,the more energy it consumes and the more room availablefor it to store energy for VEN in its battery. Furthermore,as pointed out in [1], there exists an option of installing asecondary battery in a participating EV dedicated to carryingenergy in VEN. In this case, the variation of energy-carryingcapabilities due to different EVs as original battery capacitiescan be eliminated.

We first study how to convey energy over the energypaths by maximizing the total amount of transferred energysubject to a maximum tolerable energy loss. Given the source–destination pair (s, t), the set of energy paths P(s, t) can bedetermined based on the method discussed in [36]. Supposethat we have an energy loss requirement with an upper limit L(called the maximum tolerable energy loss). Based on SectionIII, the problem is formulated as follows.

Problem 1 (Energy Transfer Maximization):

maximize|P(s,t)|∑

j=1

x j (5a)

subject to x j ≤ (T − d(p j ))z|p j |g j , j = 1, . . . , |P(s, t)|

(5b)

g j ≤w f ji , i =1, . . . , |p j |, j =1, . . . , |P(s, t)|

(5c)|P(s,t)|∑

j=1

(1

z|p j | − 1)x j ≤ L (5d)

x j ≥ 0, j = 1, . . . , |P(s, t)|. (5e)The total energy transferred is maximized in (5a), subject to

the constraint of the transferred amount on each path (5b), theenergy transfer rate constraint (5c), the energy loss constraint


(5d), and the nonnegativity constraint (5e). Given P(s, t),|P(s, t)|, d(p j )’s, and |p j |’s are constants. z, w, T , f j

i ’s, andL are system parameters, while x j ’s and g j ’s are variables. Itcan be seen that Problem 1 is indeed a linear program (LP)and the solution can be easily computed with a standard LPsolver.

In some cases, we may just need to convey a given amountof energy and retain the flexibility on the energy loss. Thetotal energy loss incurred can be minimized provided that agiven minimum energy amount X (also called the minimumrequired transferable energy) can be achieved. The problem isformulated in the following.

Problem 2 (Energy Loss Minimization):

minimize|P(s,t)|∑

j=1

(1

z|p j | − 1)x j (6a)

subject to x j ≤ (T − d(p j ))z|p j |g j , j = 1, . . . , |P(s, t)|

(6b)

g j ≤ w f ji , i = 1, . . . , |p j |, j = 1, . . . , |P(s, t)|

(6c)|P(s,t)|∑

j=1

x j ≥ X (6d)

x j ≥ 0, j = 1, . . . , |P(s, t)|. (6e)The total energy loss is minimized in (6a), subject to theenergy transfer constraint (6d) and some similar constraintsas in Problem 1, where X is a system parameter. Similarto Problem 1, Problem 2 is also an LP and the solutioncan be easily determined. Once the energy paths have beenconstructed, the problems are relatively easy to solve. Notethat, if it happens that an EV cannot serve temporarily dueto whatever reason (e.g., with battery overflow), we may notallow that EV from further carrying more energy packets untilsome energy has been discharged. We can always do so as thevehicles are connected and can be tracked. If needed, we mayadjust the traffic flow f j

i to reflect the situation and resolveProblem 1 or 2, again. The recomputation will not induceserious problems as they are just LP and can be solved veryeffectively.

Instead of showing how the problems are solved withnumerical examples, some analytical results is given in thefollowing to obtain more insights. Consider that both energytransfer maximization and energy loss loss minimization needto be implemented simultaneously. Given the maximum tol-erable energy loss L and the minimum required transferableenergy X , the following biobjective optimization can be con-structed.

Problem 3 (Biobjective Optimization):

minimize [−|P(s,t)|∑

j=1

x j ,

|P(s,t)|∑

j=1

(1

z|p j | − 1)x j ] (7a)

subject to|P(s,t)|∑

j=1

x j ≥ X (7b)

|P(s,t)|∑

j=1

(1

z|p j | − 1)x j ≤ L (7c)

x j ≤ (T − d(p j ))z|p j |g j , j = 1, . . . , |P(s, t)|

(7d)

g j ≤w f ji , i =1, . . . , |p j |, j =1, . . . , |P(s, t)|

(7e)

x j ≥ 0, j = 1, . . . , |P(s, t)|. (7f)Theorem 1: Suppose

z|p j | ≤ 1

2, j = 1, . . . , |P(s, t)|. (8)

We have the following results.

1) X ≤ L is a necessary condition for the feasibility ofProblem 3.

2) Let f ∗1 (x) be the optimal objective function value of

Problem 1. When X is set larger than f ∗1 (x), Problem 3

is infeasible.3) Let f ∗

2 (x) be the optimal objective function value ofProblem 2. When L is set smaller than f ∗

2 (x), Problem 3is infeasible.

Proof:

1) By (7b), (7c), and (8), we have

X ≤ ∑|P(s,t)|j=1 x j ≤ ∑|P(s,t)|

j=1 ( 1z|p j | − 1)x j ≤ L. (9)

If X > L, there will not exist an x which can satisfy(9). Hence, Problem 3 is infeasible unless X ≤ L .

2) Let x = [x1, . . . , x|P(s,t)|]T , f1(x) = ∑|P(s,t)|j=1 x j , and

f2(x) = ∑|P(s,t)|j=1 ( 1

z|p j | − 1)x j . Suppose that Problem 3

is feasible. It has a set of optimal solutions X = {x}constituting a Perato frontier { f1(x), f2(x)} such thatδ1 ≤ f1(x) ≤ δ1 and δ2 ≤ f2(x) ≤ δ2,∀x ∈ X , whereδ1 = inf{ f1(x)}, δ1 = sup{ f1(x)}, δ2 = inf{ f2(x)}, andδ2 = sup{ f2(x)}. By introducing the X from Problem 3,Problem 1 can be rewritten as

minimize − f1(x) (10a)


j=1

x j ≥ X (10b)

(5b), (5c), (5d), (5e). (10c)

Since Problem 3 is feasible, (7b) holds, so as (10b).Then, (10) is equivalent to (5). Furthermore, with thesame feasible region, solving (10) is as solving (7)without f2(x) and this gives f ∗

1 (x) = δ1 for maximiza-tion. However, X > f ∗

1 (x) violates (9), which gives ancontradiction.

3) Similar to the above-mentioned proof in Part (2), byintroducing the L from Problem 3, Problem 2 can berewritten as

minimize f2(x) (11a)


j=1

(1

z|p j | − 1)x j ≤ L (11b)

(6b), (6c), (6d), (6e). (11c)


Fig. 3. Role of VEN between the energy sources and loads.

(11) is equivalent to (6). With the same feasibleregion, solving (11) is as solving (7) without f1(x)and this gives f ∗

2 (x) = δ2 for minimization. However,L < f ∗

2 (x) violates (9), which gives an contradiction.�Condition (8) is met when the energy efficiency z is low or theenergy paths are composed of many vehicular subroutes. Thelatter may be due to the source and destination being too farapart or limited vehicular routing information. Theorem 1 isuseful when the total transferred energy and energy loss needto be optimized simultaneously. Solving a simpler problem,i.e., Problem 1 or 2, gives us ideas on how to set the valuesof the parameters, X and L , for the harder Problem 3.

V. ECONOMIC MODEL

VEN is designed as an energy distribution system, in whichboth the sources and loads of energy are not parts of the system(see Fig. 3). It provides energy delivery services for the users,i.e., energy source–load pairs, and the service charge maydepend on many factors, e.g., distances between sources andloads, delivery time frames, amounts of deliverable energy,quality of service, and so on. The detailed service chargingmodel depends on the applications and it is out of the scope ofthis paper. In order to assess the economic feasibility, a simpleeconomic model is developed for renewable energy dispatch,which will be used to evaluate the system performance inSection VI. In this application, the “offline” renewable energyis delivered, due to the excessive generation which cannot bebrought to the grid, to appropriate destinations over VEN.

The economic value of VEN is evaluated by the annualrevenue (Rann) minus the annual cost (Cann). Suppose that therevenue comes from the annual amount of energy delivered atthe destinations, that is

Rann = Re Ed = Re Esε

where Re, Ed , Es , and ε stand for electricity rate, annualdelivered and generated amounts of energy, and overall systemefficiency, respectively.

As usually only a portion of renewable generation can bebrought “online” and utilized in the grid, it is considered thatVEN can improve the utilization of renewables by manipulat-ing the “offline” energy. Thus, it is assumed that the “offline”renewable energy is free. Then, the cost mainly stems from theenergy storage for temporary repository (Cs ), the (dis)chargingfacilities (C f ), and EVs (Cv ). Cs is estimated based on thelevelized cost of energy (LCOE) [37], which is the ratio oflifetime costs to lifetime delivered energy. It takes into account

factors, such as capital cost, financing cost, fixed and variableoperations and maintenance cost, installation cost (componentsand labor), and so on. As seen from Fig. 1, the storagemanages temporary energy imbalance among the passing-byEVs. Total amount of energy delivered by the storage dependson the real dynamics on the roads. This amount is modeledas a fraction δs of total generated energy and thus we haveCs = CLCOE

s Esδs , where CLCOEs is LCOE of the storage.

For the dynamic (dis)charging facilities, the technologyis still in the research and development stage and thus itslevelized cost is not available. The capital cost of such afacility is estimated by considering a similar technology fordynamic charging reported in [13]. The cost model givenin [13] is adopted to estimate the lifetime cost of (dis)chargingfacilities, which is then annualized by multiplying with thecapital recovery factor (RCRF) [38], which is defined asRCRF = τ f

1−(1+τ f )−η f

, where τ f and η f are the discount rate

of the equipment and the number of years the equipment lasts,respectively. Hence, we have C f = (c f

f +cvf l f )n f RCRF, where

c ff , cv

f , l f , and n f , are the fixed cost, the per-unit lengthvariable cost, average track length, and the required quantity,respectively.

In VEN, EVs are responsible for transporting energy butauxiliary in nature; they are generally not owned by the systembut incentivized to carry additional energy for the system ontheir own. It is simply assumed that a part of the revenue isused as the incentive and thus we have Cv = Rv Rann, whereRv is the incentive rate. Therefore, the total annual cost isgiven by

Cann = Cs + C f + Cv

= CLCOEs Esδs + (c f

f + cvf l f )n f RCRF + Rv Rann.

VI. FEASIBILITY STUDY

In this section, the operational and economic feasibilitiesof VEN are evaluated. Wind energy dispatch is selected as anapplication example of VEN. For the former, it is demonstratedthat a significant amount of energy can be transmitted throughVEN in a reasonable period of time. For the latter, theeconomic model discussed in Section V is evaluated for windenergy dispatch.

A. Operational Feasibility

Consider that the renewable energy produced from the windfarms located in remote areas of U.K. is transmitted to the cityof London. A VEN is constructed by adopting the existingU.K. road networks. Based on [39], the network is createdwith 998 nodes and 2470 arcs (ai ’s). One set of real traffic dataacquired in June 2013 is adopted to specify the journey time,distance, and vehicular flow for each of the road segments(i.e., arcs). 4788 vehicular routes (ri ’s) are randomly createdin the road network, each of which is composed of severalconnected road segments no longer than 200 km. The trafficflow and total journey time of each ri are set with the minimumof vehicular flows of the corresponding composite ai ’s andthe sum of individual journey times of the composite ai ’s,respectively.


Fig. 4. Locations of road junctions, and onshore and offshore wind farms(adopted from Google earth).

U.K. is the sixth largest nations producing wind energy, withan annual energy production of more than 26×106 MWh [40],but only 26% was brought online in 2013 [41]. Accordingto [42] and [43], there are 203 onshore and 20 offshore windfarms. Fig. 4 shows the locations of the road junctions and thewind farms. 67 nodes are selected in the road network, whichare close to the wind farms, as the energy sources (s’s) in VEN.One junction near London is selected as the energy destina-tion t for demonstrative purpose. Energy paths p j (s, t)’s areconstructed by augmenting subroutes of the 4788 ri ’s. Energyis transmitted from the sources to the destination along someof the energy paths.

Consider the nominal setting of 0.1% EV penetration rate,3

packet size w of 0.1 kWh,4 efficiency rate z of 0.9,5 andtime period T of 5 h. As most popular EVs available in themarket are equipped with batteries of more than 20 kWh [47],our energy packet size is rather conservative. In the follow-ing, the total transferred energy (reaching London from thewind farms) and the corresponding energy loss are studiedby varying one parameter in each case. Each data point iscomputed from the right-hand side of either (3) or (4), inwhich, for simplicity, only some paths are considered insteadof all possible energy paths.

Fig. 5(a) shows the impact of total transferred energy andenergy loss by changing z from the nominal setting. In 5 h,the amounts of energy transferred are in the order of MWh.Note that the allowed period has already included the timerequired for the EVs to move from one place to another.The more efficient the wireless energy transfer technology,

3The EV penetration rate refers to the percentage of participating EVs inthe vehicle population. When the EV penetration rate is 0.1%, one out of1000 cars supports VEN.

4An existing dynamic charging prototype, OLEV, can support charging rateof 100 kW [44], and an energy packet of 0.1 kWh will take 3.6 s. Thus, thepacket size of 0.1 kWh is feasible.

5The industry expects that the efficiency of wireless energy transfer onmoving vehicles can exceed 90% [45]. NASA has developed a prototypecalled EVWireless with over 90% energy efficiency [46].

the more energy can be transferred with smaller energy loss.As an extreme case, taking z as one gives no energy loss. Thiscan be achieved only when the system is perfect; the energyefficiency is 100%. It is just a speculative result. When z issmaller than 0.83,6 the amount of energy loss can overshootthe transferred energy. Hence, z is an important factor whenVEN is designed. However, if there is no VEN or othersimilar energy transmission architecture available, the hugeamount of renewable energy may need to be stored at theoriginal generation locations and eventually wasted due tosurplus renewable production. As the technology of wirelessenergy transfer advances, the overall efficiency will improve.Fig. 5(b) studies the impact of the allowed time period. WithT ≤ 1 h, the transferred energy is very small as most of theEVs carrying the energy have not reached the destination. Witha larger T , more transferred energy and energy loss occur. Thegap between transferred energy and energy loss grows with Tshowing that the energy transfer is more efficient in a longertime period. Fig. 5(c) shows the impact of the packet size.More energy will reach the destination and more energy isalso lost when a larger amount of energy can be (dis)chargedeach time. A larger packet size leads to a higher efficiencyas more energy can be transmitted than lost. Finally, Fig. 5(d)shows that the transferred energy and energy loss grow linearlywith the EV penetration rate. The penetration rate affects thenumber of EVs providing the service only. The more EVsavailable, the more energy can be transferred. At the sametime, the more frequent charging and discharging events resultin higher energy loss. So both lines have similar slopes.

To summarize, a VEN is proposed to transport renewableenergy from remote wind farms to London. Note that the totaltransferable energy and loss shown above are not optimal, assome of the energy paths are just selected to transport theenergy. In fact, there may be more efficient paths experiencingfewer charging–discharging cycles. Nevertheless, the amountof energy transferable through VEN in our illustration is stillvery significant. Although there is energy loss in VEN, mostof the energy to be transported in VEN would be lost anywaywithout VEN. This shows that VEN can complement thepower network and enhance the overall power transmissionrate. Besides pure energy transfer, there exist other applica-tions in which an energy source and destination do not havedirect energy transfer relationship. We may also consider VENas an energy storage system spanning a large geographical areawith many interfaces (i.e., nodes in VEN). It can be used toabsorb energy excessively generated at the “energy sources”and release energy at the “energy destinations” where energyis in deficit. Moreover, we may also use the energy “stored” inVEN to charge EVs for their energy consumption or for otherpurposes. This implicitly creates more loads in the networkand may improve the renewable energy utilization.

B. Economic Feasibility

The total possible profit is estimated when implementingwind energy dispatch illustrated in Section VI-A for a year.

6This threshold actually depends on the energy paths chosen for energytransmission.


Fig. 5. Transferred energy and energy loss. (a) At different efficiencies. (b) In various periods of time. (c) With various buffer sizes. (d) With different EVpenetration rates.

TABLE I

PARAMETER SETTINGS

The parameter settings are given in Table I. U.K. generated28.1 TWh of wind energy in 2014 [41] and it is assumedthat 70% of the generation is delivered through VEN (afterexcluding the online wind power directly transmitted to thegrid). As observed in Fig. 5, around 2/3 of the total generatedenergy can reach the destination and thus it is assumed thatthe overall system efficiency equals 0.67.7 As discussed in

7The actual system efficiency highly depends on the settings of the real sit-uation, including locations of the sources and destinations, network structure,traffic, and energy routing. It cannot be easily quantified and thus the valueis estimated from our study.

Section II, the energy storage has relatively short (dis)chargingtime and the required capacity is generally small. Supercapac-itor is considered the best choice among all kinds of batterytechnologies for the energy distribution purpose and its LCOEis 0.51 USD/kWh [37]. The cost of the (dis)charging facilityis estimated based on the prototype given in [13] and the(dis)charging track is assumed to be installed at each roadjunction is 1000 m long. The discount rate for the capitalrecovery factor is set to 0.1 and each piece of equipment isassumed to last for ten years. Consider that one (dis)chargingfacility is installed at each of the 998 road junctions and10% of the revenue is allocated as the incentive cost forEVs.8 Fig. 6(a) shows the annualized total cost and revenueof the wind energy dispatch at different equipment discountrate, which accounts for how much discount of the cost whenthe relevant technologies advance and the equipment can bemassively produced in the future. It can be seen that storagerepresents most of the cost. Without any discount, the systemis not economical to operate. When the equipment cost dropsmore than 15%, the system becomes profitable. As forecast bythe World Energy Council, there will be 70% drop in energystorage costs by 2030 [49] and it is not hard to imagine that

8The amount is estimated on an annual basis. Each EV may contributedifferently during the year. How the incentive cost is distributed depends onthe business model, which is beyond the scope of this paper.


Fig. 6. Total profit. (a) At various equipment discount rates. (b) At variousenergy amounts in storage.

VEN will become economically feasible when the relevanttechnologies become mature. Moreover, as the amount ofenergy stored during operation depends on the EV conditionsat the road junctions, δs is a rough estimate. So δs is varied andhow δs affects the total profit is investigated. To account forfuture technological advances, The various equipment costsare considered to be dropped by half. Fig. 6(b) shows theannualized costs and total revenue corresponding to differentδs values. The total cost increases with δs and the systembecomes economically infeasible when δs gets large. Thissuggests that the system can produce higher profits when theEVs go in platoons such that charging and discharging can takeplace simultaneously and the energy storage is not requiredmost of the time.

C. Discussion

Although the above-mentioned feasibility studies aredemonstrated with wind energy dispatch, VEN can exerciseother applications with different kinds of energy sources andloads. In general, VEN can be considered as a mediumconnecting the energy sources to the loads spatially andtemporally. When an energy source is not physically connectedto a load (e.g., in a remote area without power infrastructure),VEN can help transport the required energy. When there isa time discrepancy between the energy supply and demand,we may advance energy from the VEN to serve the urgentloads or temporarily store energy in VEN to defer the delivery.However, energy transfer is a fundamental operation of VENand thus this capability is demonstrated in Section VI-A.

VEN and the power grid seemingly share some similarities

TABLE II

COMPARISON BETWEEN THE POWER GRID AND VEN

in terms of energy/power delivery capability. However, theyhave many subtle differences in functionality, cost, efficiency,practicality, flexibility, and marketability. They are summa-rized in Table II and further discussed in the following.

1) Functionality: The power grid is highly sophisticatedand its utilization is strictly regulated. It mainly serves fortraditional power delivery with real-time matching betweengenerations and loads. On the other hand, VEN is selfcon-tained and there are less stringent regulations on the energysources and loads. It is a good candidate to test, and serve,immature smart grid applications.

2) Cost: Constructing a power grid or even expanding anexisting power system is extremely complicated. It involvesmany issues related to network topology, facility ratings,siting, right-of-way, and visual and esthetic effects [50]–[52].It may need a decade or more for the planning and theassociated time cost can be considerable. On the other hand,VEN can be built by attaching to existing infrastructure, i.e.,the road network. The required equipment is smaller in scaleand VEN mostly takes advantages of the equipment which isdesigned and utilized for their original purposes (e.g., dynamiccharging is primarily designed for the convenience of batterycharging for basic EV operation). If the related technologieshave become mature, VEN can be deployed in relativelyshorter time.

3) Efficiency: The power grid is specially designed forpower transfer. The power lines are rated for low powerloss and implementing optimal power flow allows minimizingoverall power loss [53]. It imposes strict regulation to itsutilization, resulting in high efficiency. On the other hand,most parts of VEN are primarily designed for other purposesbut their integration can exercise energy transfer. It imposesfewer regulations allowing a broader range of applications.The frequent WPT may incur certain energy loss and itsefficiency is generally lower. From the simulation resultsgiven in Section VI-A, the system efficiency can be improvedthrough improvements in WPT efficiency (i.e., z), the packetsize (i.e., w), and EV penetration. With technological advance-ments, WPT efficiency will be enhanced and the packetsize can be made larger. Longer (dis)charging tracks andbigger EV battery reserve can also allow larger packet sizes.EV penetration can be improved through enforcing variousgovernment green policies and promoting the public’s envi-ronmental awareness. Moreover, dedicated known vehicularschedules (e.g., from public transport and autonomous vehicle


systems [54]) allow formation of energy paths with fewercharging and discharging events. This can reduce energy lossand thus improve the system efficiency.

4) Practicability: The power system was invented for morethan a century. It has been kept being modified and improvedand it is the most important means for power delivery. On theother hand, although VEN is primarily investigated in thispaper, the development of most required equipment is promis-ing and the corresponding technologies are likely to becomemature in a few years time. Hence, both the power grid andVEN are highly practical.

5) Flexibility: Following the above-mentioned discussionon cost, the power grid is once constructed, it is not easy tobe reconfigured. Moreover, new smart grid applications maynot be easily deployed in the power system. On the otherhand, the structure of VEN can be modified rather easily and itcan accommodate more applications. Hence, VEN has higherflexibility.

6) Marketability: The power grid is generally constructedand managed by a small number of parties (e.g., regionaltransmission organization and independent system operator).Energy markets over the power system are generally wellstructured and regulated. On the other hand, various scalesof VEN can be easily constructed by interested organi-zations. New energy markets, e.g., vehicle-to-grid energytrading [6], [55], can be deployed over VEN relatively easily.Hence, VEN achieves higher marketability.

From above, both systems have their own advantages andeach has its own role of energy/power delivery. VEN canaccomplish some tasks that the traditional power grid cannotdo easily, and vice versa. They are not replacing one another.As a whole, VEN may be considered as a feasible design tocomplement the power network and enhance the overall powersystem performance.

VII. CONCLUSION

The smart grid introduces many new elements which arenot easily incorporated into the existing power system dueto power reliability and security reasons. Many interestingsmart grid implementations can be more easily realized ifthere exists an independent infrastructure for power deliverywithout interference to the existing power network. In thispaper, VEN is introduced and it is capable of distributingenergy across a large geographical area with EVs. VEN isbuilt upon the traffic network and it relies on existing well-developed technologies. Small amounts of energy, as energypackets, are carried by EVs through multiple vehicular routes.By carefully designing the energy packet routing, energy canbe transported from energy sources to destinations. In thispaper, a preliminary study of VEN is given and a model forfurther analysis is developed. Its operational and economicfeasibilities are studied by setting up a VEN with real roadtraffic data in the U.K. It is shown that a considerable amountof renewable energy can be transported from some remotewind farms to London under some reasonable assumptions.It is also revealed that VEN can be profitable in the near future.This paper is not intended to posit VEN as a perfect solutionfor energy delivery. Instead, VEN may be considered as a

feasible design to complement the power network and enhancethe overall power system performance. Besides energy deliverydedicated to a source–destination pair, it can also serve as abig energy buffer with many interfaces across a region forenergy absorption and distraction. The stored energy can beutilized in a wiser way economically and operationally andthis may facilitate many applications which are hard to berealized in the power grid or stationary energy storage alone.It is likely that VEN can be realized and deployed in the nextdecade. With the energy packet-switching design, VEN hasmany potential future extensions by incorporating ideas andresults from the packet-switched data network.

REFERENCES

[1] A. Y. S. Lam, K.-C. Leung, and V. O. K. Li, “An electric-vehicle-based supplementary power delivery system,” in Proc. 6th IEEEInt. Conf. Smart Grid Commun., Miami, FL, USA, Nov. 2015,pp. 307–312.

[2] Albert Y.S. Lam, “Evolution of the smart grid in China,” Shanghai,China, Tech. Rep., 2010.

[3] D. Cardwell. (Aug. 14, 2013). Intermittent Nature of Green Power isChallenge for Utilities. [Online]. Available: http://www.nytimes.com/2013/08/15/business/energy-environment/intermittent-nature-of-green-power-is-challenge-for-utilities.html

[4] M. Al-Muhaini and G. T. Heydt, “Evaluating future power distributionsystem reliability including distributed generation,” IEEE Trans. PowerDel., vol. 28, no. 4, pp. 2264–2272, Oct. 2013.

[5] A. Y. S. Lam, K.-C. Leung, and V. O. K. Li, “Capacity estimationfor vehicle-to-grid frequency regulation services with smart chargingmechanism,” IEEE Trans. Smart Grid, vol. 7, no. 1, pp. 156–166,Jan. 2016.

[6] A. Y. S. Lam, L. Huang, A. Silva, and W. Saad, “A multi-layer marketfor vehicle-to-grid energy trading in the smart grid,” in Proc. IEEEConf. Comput. Commun. Workshops (INFOCOM WKSHPS), Mar. 2012,pp. 85–90.

[7] P. Yi, T. Zhu, B. Jiang, B. Wang, and D. Towsley, “An energy trans-mission and distribution network using electric vehicles,” in Proc. IEEEInt. Conf. Commun., Ottawa, ON, Canada, Jun. 2012, pp. 3335–3339.

[8] P. Yi et al., “Energy scheduling and allocation in electric vehicleenergy distribution networks,” in Proc. IEEE PES Innov. Smart GridTechnol. (ISGT), Feb. 2013, pp. 1–6.

[9] P. Yi, T. Zhu, G. Lin, and Q. Zhang, “Routing renewable energy usingelectric vehicles in mobile electrical grid,” in Proc. IEEE 10th Int. Conf.Mobile Ad-Hoc Sensor Syst., Hangzhou, China, Oct. 2013, pp. 19–27.

[10] P. Yi et al., “Renewable energy transmission through multiple routes ina mobile electrical grid,” in Proc. IEEE PES Innov. Smart Grid Technol.,Washington, DC, USA, Feb. 2014, pp. 1–5.

[11] Y. D. Ko and Y. J. Jang, “The optimal system design of the onlineelectric vehicle utilizing wireless power transmission technology,” IEEETrans. Intell. Transp. Syst., vol. 14, no. 3, pp. 1255–1265, Sep. 2013.

[12] J. Shin et al., “Design and implementation of shaped magnetic-resonance-based wireless power transfer system for roadway-poweredmoving electric vehicles,” IEEE Trans. Ind. Electron., vol. 61, no. 3,pp. 1179–1192, Mar. 2014.

[13] S. Jeong, Y. J. Jang, and D. Kum, “Dynamic analysis of the dynamiccharging electric vehicle,” IEEE Trans. Power Electron., vol. 30, no. 11,pp. 6368–6377, Nov. 2015.

[14] L. A. Maglaras, J. Jiang, A. L. Maglaras, and F. V. Topalis, “Mobileenergy disseminators increase electrical vehicles range in a smart city,”in Proc. 5th IET Hybrid Electr. Vehicle Conf., London, U.K., Nov. 2014,pp. 1–6.

[15] L. A. Maglaras, F. V. Topalis, and A. L. Maglaras, “Cooperativeapproaches for dymanic wireless charging of electric vehicles in a smartcity,” in Proc. IEEE Int. Energy Conf., Dubrovnik, Croatia, May 2014,pp. 1365–1369.

[16] Wikipedia. (Aug. 2015). Renewable Portfolio Standard. [Online]. Avail-able: http://en.wikipedia.org/wiki/Renewable_Portfolio_Standard

[17] REN21. (2013). Renewables Global Status Report. [Online]. Available:http://www.ren21.net/

[18] H. N. T. Nguyen, C. Zhang, and J. Zhang, “Dynamic demand controlof electric vehicles to support power grid with high penetration levelof renewable energy,” IEEE Trans. Transport. Electrific., vol. 2, no. 1,pp. 66–75, Mar. 2016.


[19] J. Mitlitski. (Feb. 2011). Raising the Voltage: Is Obama’s Goal of1 Million Electric Vehicles on U.S. Highways by 2015 Realistic?[Online]. Available: http://www.scientificamerican.com/article/one-million-electric-vehicles-by-2015/

[20] Y. Liu. (Jun. 2011). China Begins Implementation of 12th Five-Year Planfor EVs. [Online]. Available: http://www.renewableenergyworld.com/rea/news/article/2011/06/china-begins-implementation-of-the-12th-five-year-plan-for-evs

[21] A. Y. S. Lam, Y.-W. Leung, and X. Chu, “Electric vehicle chargingstation placement: Formulation, complexity, and solutions,” IEEE Trans.Smart Grid, vol. 5, no. 6, pp. 2846–2856, Nov. 2014.

[22] G. Karagiannis et al., “Vehicular networking: A survey and tutorial onrequirements, architectures, challenges, standards and solutions,” IEEECommun. Surveys Tuts., vol. 13, no. 4, pp. 584–616, 4th Quart., 2011.

[23] J. Dai and D. C. Ludois, “A survey of wireless power transfer anda critical comparison of inductive and capacitive coupling for smallgap applications,” IEEE Trans. Power Electron., vol. 30, no. 11,pp. 6017–6029, Nov. 2015.

[24] S. Li and C. C. Mi, “Wireless power transfer for electric vehicleapplications,” IEEE J. Emerg. Sel. Topics Power Electron., vol. 3, no. 1,pp. 4–17, Mar. 2015.

[25] M. Bojarski, E. Asa, K. Colak, and D. Czarkowski, “Analy-sis and control of multi-phase inductively coupled resonant con-verter for wireless electric vehicle charger applications,” IEEETrans. Transport. Electrific., to be published. [Online]. Avilable:http://ieeexplore.ieee.org/document/7469373/

[26] J. M. Miller and A. Daga, “Elements of wireless power transfer essentialto high power charging of heavy duty vehicles,” IEEE Trans. Transport.Electrific., vol. 1, no. 1, pp. 26–39, Jun. 2015.

[27] J. P. K. Sampath, D. M. Vilathgamuwa, and A. Alphones, “Efficiencyenhancement for dynamic wireless power transfer system with seg-mented transmitter array,” IEEE Trans. Transport. Electrific., vol. 2,no. 1, pp. 76–85, Mar. 2016.

[28] S. Lukic and Z. Pantic, “Cutting the cord: Static and dynamic inductivewireless charging of electric vehicles,” IEEE Electrific. Mag., vol. 1,no. 1, pp. 57–64, Sep. 2013.

[29] X. Zhang, Z. Yuan, Q. Yang, Y. Li, J. Zhu, and Y. Li, “Coil designand efficiency analysis for dynamic wireless charging system for electricvehicles,” IEEE Trans. Magn., vol. 52, no. 7, Jul. 2016, Art. no. 8700404.

[30] Qualcomm Halo. (2014). The Future is Wireless EV Charging. [Online].Available: http://www.qualcommhalo.com/

[31] ABB. (Oct. 2015). ABB and Microsoft Join Forces toLaunch Next-Generation Electric Vehicle Charging ServicesPlatform. [Online]. Available: http://www.abb.com/cawp/seitp202/6d96ffb03d540739c1257ee40031c951.aspx

[32] X. Yu, S. Sandhu, S. Beiker, R. Sassoon, and S. Fan, “Wireless energytransfer with the presence of metallic planes,” Appl. Phys. Lett., vol. 99,no. 21, p. 214102, Nov. 2011.

[33] Highways England and A. Jones. (Aug. 2015). Off Road Trials for, ‘Elec-tric Highways’ Technology. [Online]. Available: https://www.gov.uk/government/news/off-road-trials-for-electric-highways-technology

[34] F. Deflorio, I. Pinna, L. Castello, and V. Cantello, “Dynamic chargingsystems for electric vehicles: Simulation for the daily energy estimationon motorways,” IET Intell. Transport. Syst., vol. 10, no. 4, pp. 258–269,2016.

[35] J. F. Kurose and K. W. Ross, Computer Networking: A Top-DownApproach, 7th ed. Boston, MA, USA: Pearson, 2016.

[36] A. Y. S. Lam and V. O. K. Li, “Energy loss minimization for vehicularenergy network routing,” in Proc. ACM Workshop Electr. Vehicle Syst.,Data Appl., Waterloo, ON, Canada, Jun. 2016, Art. no. 4.

[37] F. Díaz-González, A. Sumper, and O. Gomis-Bellmunt, Energy Storagein Power Systems. Chichester, U.K.: Wiley, 2016.

[38] W. Kempton and J. Tomic, “Vehicle-to-grid power fundamentals: Cal-culating capacity and net revenue,” J. Power Sour., vol. 144, no. 1,pp. 268–279, Jun. 2005.

[39] H. Agency. (Jun. 2013). Highways Agency Network Journey Time andTraffic Flow Data. [Online]. Available: http://data.gov.uk/dataset/dft-eng-srn-routes-journey-times

[40] RenewableUK. (2014). UK Wind Energy Database. [Online]. Available:http://www.renewableuk.com/en/renewable-energy/wind-energy/uk-wind-energy-database/index.cfm

[41] Wikipedia. (Aug. 2015). Wind Power in the United Kingdom.[Online]. Available: http://en.wikipedia.org/wiki/Wind_power_in_the_United_Kingdom

[42] Wikipedia. (Aug. 2015). List of Offshore Wind Farms in theUnited Kingdom. [Online]. Available: http://en.wikipedia.org/wiki/List_of_offshore_wind_farms_in_the_United_Kingdom

[43] Wikipedia. (Aug. 2015). List of Onshore Wind Farms in theUnited Kingdom. [Online]. Available: http://en.wikipedia.org/wiki/List_of_onshore_wind_farms_in_the_United_Kingdom

[44] I.-S. Suh and J. Kim, “Electric vehicle on-road dynamic charging systemwith wireless power transfer technology,” in Proc. IEEE Int. Electr.Mach. Drives Conf. (IEMDC), May 2013, pp. 234–240.

[45] S. Harris. (Mar. 2013). Your Questions Answered: Inductive Chargingfor Road Vehicles. [Online]. Available: http://www.theengineer.co.uk/automotive/in-depth/your-questions-answered-inductive-charging-for-road-vehicles/1015724.article

[46] NASA Ames Research Center. (Jul. 2014). EVWireless. [Online].Available: http://www.evwireless.net/

[47] Battery University. (Mar. 2016). BU-1003: Electric Vehicle (EV).[Online]. Available: http://batteryuniversity.com/learn/article/electric_vehicle_ev

[48] Worldatlas. (Oct. 2015). Electricity Rates Around the World.[Online]. Available: http://www.worldatlas.com/articles/electricity-rates-around-the-world.html

[49] P. Gardner, F. Jones, M. Rowe, A. Nouri, and H. van de Vegte,“E-storage: Shifting from cost to value: Wind and solar applications,”World Energy Council, London, U.K., Tech. Rep., Jan. 2016.

[50] A. Mills, R. Wiser, and K. Porter, “The cost of transmission for windenergy: A review of transmission planning studies,” Lawrence BerkeleyNat. Lab., Berkeley, CA, USA, Tech. Rep. LBNL-1471E, Feb. 2009.

[51] F. F. Wu, F. L. Zheng, and F. S. Wen, “Transmission investment andexpansion planning in a restructured electricity market,” Energy, vol. 31,nos. 6–7, pp. 954–966, May/Jun. 2006.

[52] S. Dehghan, N. Amjady, and A. J. Conejo, “Reliability-constrainedrobust power system expansion planning,” IEEE Trans. Power Syst.,vol. 31, no. 3, pp. 2383–2392, May 2016.

[53] B. Zhang, A. Y. S. Lam, A. Domínguez-García, and D. Tse, “An optimaland distributed method for voltage regulation in power distributionsystems,” IEEE Trans. Power Syst., vol. 30, no. 4, pp. 1714–1726,Jun. 2015.

[54] A. Y. S. Lam, Y.-W. Leung, and X. Chu, “Autonomous-vehicle publictransportation system: Scheduling and admission control,” IEEE Trans.Intell. Transp. Syst., vol. 17, no. 5, pp. 1210–1226, May 2016.

[55] J. J. Q. Yu, J. Lin, A. Y. S. Lam, and V. O. K. Li, “Maximizingaggregator profit through energy trading by coordinated electric vehiclecharging,” in Proc. IEEE Int. Conf. Smart Grid Commun., Sydney, NSW,Australia, Nov. 2016, pp. 497–502.

Albert Y.S. Lam (S’03–M’10–SM’16) receivedthe B.Eng. (First Class Hons.) degree in informa-tion engineering and the Ph.D. degree in electricaland electronic engineering from The University ofHong Kong (HKU), Hong Kong, in 2005 and 2010,respectively.

He was a Post-Doctoral Scholar with the Depart-ment of Electrical Engineering and Computer Sci-ences, University of California at Berkeley, Berkeley,CA, USA, from 2010 to 2012. He is currently aResearch Assistant Professor with the Department

of Electrical and Electronic Engineering, HKU. His current research interestsinclude optimization theory and algorithms, evolutionary computation, smartgrid, and smart city.

Dr. Lam is a Croucher Research Fellow.


Ka-Cheong Leung (S’95–M’01) received theB.Eng. degree in computer science from theHong Kong University of Science and Technology,Hong Kong, in 1994, the M.Sc. degree in electri-cal engineering (computer networks) and the Ph.D.degree in computer engineering from the Universityof Southern California, Los Angeles, CA, USA, in1997 and 2000, respectively.

He was a Senior Research Engineer with the NokiaResearch Center, Nokia Inc., Irving, TX, USA, from2001 to 2002. He was an Assistant Professor with the

Department of Computer Science, Texas Tech University, Lubbock, TX, USA,from 2002 to 2005. Since 2005, he has been with Department of Electricaland Electronic Engineering, The University of Hong Kong, Hong Kong, wherehe is currently an Assistant Professor. His current research interests includetransport layer protocol design, congestion control, smart grid, and wirelesspacket scheduling.

Victor O.K. Li (S’80–M’81–SM’86–F’92) receivedthe S.B., S.M., E.E., and Sc.D. degrees from theMassachusetts Institute of Technology, Cambridge,MA, USA, in 1977, 1979, 1980, and 1981, respec-tively, all in electrical engineering and computerscience.

He was a Professor of Electrical Engineering withthe University of Southern California, Los Angeles,CA, USA, and was the Director of the Communica-tion Sciences Institute. He was an Associate Dean ofEngineering and the Managing Director of Versitech

Ltd., Hong Kong, the technology transfer and commercial arm of The Univer-sity of Hong Kong (HKU), Hong Kong. He served on the board of China.comLtd., and currently serves on the board of Sunevision Holdings Ltd., listedon the Hong Kong Stock Exchange. Sought by government, industry, andacademic organizations, he has lectured and consulted extensively aroundthe world. He is currently the Chair Professor of Information Engineeringand the Head of the Department of Electrical and Electronic Engineering,HKU. His current research interests include information technology, artificialintelligence, deep learning, big data analytics, and their applications to cleanenergy and environment, Internet of Things, and smart city.

Dr. Li is a fellow of the Hong Kong Academy of Engineering Sciences, theIAE, and the HKIE. He was a recipient of numerous awards, including thePRC Ministry of Education Changjiang Chair Professorship at Tsinghua Uni-versity, the UK Royal Academy of Engineering Senior Visiting Fellowship inCommunications, the Croucher Foundation Senior Research Fellowship, andthe Order of the Bronze Bauhinia Star, Government of the Hong Kong SpecialAdministrative Region, China. He is a Registered Professional Engineer.