Optimization of Charge/Discharge Coordination to Satisfy Network Requirements Using Heuristic Algorithms in Vehicle-to-Grid Concept. (2024)

I. INTRODUCTION

Consumed energy at transportation sector is twenty five percent ofthe total consumed energy on the worldwide [1]. Therefore, fossil fuelsconsumption in transportation leads environmental pollution and highenergy cost. Along with that, there is a growing sensitivity on energyefficiency and environment. Electric Vehicles (EVs) are the importantoptions to reduce both fuel cost and green gas emissions. Therefore,many countries encourage people to purchase EV, and as a result, thenumber of EVs on the road is increasing day by day. According to themoderate scenario, it is estimated that 35% and 62% of total vehicleswill be in hybrid or electric form by 2020 and 2050 in US, respectively[2]. In 2016, the number of open access charge points has reached320.000 worldwide, growing by 72% since 2015 [3]. Such predictions andstatistics promise a bright future for EV.

The smart grid, regarded as the next generation power grid,includes distributed energy sources, intelligent control and advancedcommunication technologies. It uses bidirectional flows of power andinformation to create a widely distributed automated energy deliverynetwork [4]. EVs are expected to be an important part of future smartgrid with their opportunities as much as challenges. The main challengeis the huge loads from EV charging due to many of the EV owners arrivefrom work to home between 16.00 and 19.00 that corresponds the peaktimes for residential distribution network [5, 6]. If vehicles startcharging as soon as they arrive home it causes such problems asincreasing in peak load, overloading of transformers, degradation ofvoltage etc. [7-9]. That kind of problems can be solved with chargecoordination. Moreover, EV can perform more than a load in smart griddue to its bidirectional power transfer feature once the requirednecessary infrastructure is established [10]. The ability of EVs toinject power into the grid is called Vehicle-To-Grid (V2G) technology[11, 12]. In this concept, EVs can be used as a generation resource aswell as a storage device for certain periods of time to provide power tothe grid.

The current situation of V2G technology, the impact on distributionnetwork, challenges and opportunities are investigated in [10, 11]. V2Gsystem consists of 6 main subsystems. In this structure, Energy SupplyProvider (ESP) provides energy to customers through the distributionnetwork (DN). Independent System Operator/ Regional TransmissionOrganization (ISO/RTO) provide the power system operation and control.Aggregator determines the charge/discharge status of EVs and provides aninterface between EV, ESP and ISO/RTOs. Charging infrastructure, two wayelectrical power and communication, smart metering and control are othersubsystems of V2G. As the penetration of EVs grows, auxiliary servicessuch as frequency regulation, load shaving, spinning reserve, andvoltage support can be provided by EVs [13-21].

In [22], optimization methodologies of charge/discharge arereviewed and numerical applications are carried out in [23].Comprehensive objective function subject to constraints should bedefined for optimal charging strategies. Objective functions aregenerally based on minimizing cost [24], power loss [13] and maximizingvoltage profile [21], welfare [25], V2G revenue [26] etc. Constraintsindicate the bounds of physical limit of system and EV owners'specifications. After defining objective function subject to constraint,optimization method is applied to reach best solution. Though numerousmathematical optimization methods such as linear programing [27],non-linear programming [28], dynamic programming [29], game theory [30]etc. have been used to solve the optimum charge/discharge problem, theyhave some draw-backs. Simplification may be required for mathematicalmethods due to difficulties in solving highly non-linear and non-convexterms of objective function. This may lead to the loss of accuracy andnon-optimum results. Also, mathematical methods cannot address the highdimension problem in a reasonable amount of time [31]. By contrast,heuristic algorithms naturally immune to non-linear, non-convex andhigh-dimensional systems, computational time can also be limited. Hence,heuristic algorithms are generally preferred for solving optimumcharging coordination which is high dimensional and complex problem[32].

In [33], charging coordination of EV is provided using GA withoutpeak load mitigation in their formulation. In [34], Particle SwarmOptimization approach is proposed in order to maximize EV owner incomeand mitigate power losses in distribution system. Peak load and networklosses are decrease 9.76% and 2.62%, respectively. However, systemvoltage is not considered as constraints. In [35], Ant-based swarmalgorithm is performed for charging coordination of EVs with loadfluctuation and the transformer capacity constraints. While peak valleydifferences were 504.54 kW in free charging mode, it is decreased to127.49 kW with charging coordination. However, EVs are not located in adistribution system. Hence, system losses and constraints are ignored insimulations. In [25], proposed algorithm aims to maximize the totalutility considering EV charge demands. The method provides chargingprofits 3.4 times much more earned by the no-control strategy. However,only transformer capacity is taken into account as network constraints.In [36], Hybrid particle swarm optimization Gravitational SearchAlgorithm based optimization is used in order to optimally allocatepower to each of the EVs. As increase in EV penetration, fitness valueis increased from 144.838 to 183.094. However, no systems levelrealistic assessments have been performed.

In [37], Genetic Algorithm based solution is proposed foroptimizing EV coordination in order to flatten load profile. However,the method does not guarantee fully charged battery at departure time.In [38], a heuristic algorithm is proposed to solve problem ofscheduling EV charging with storage units. The aggregator's revenuecan be improved by 80.1% using optimal charging scheduling. However, EVowners' benefits and EV constraints are not mentioned. In [39] and[40], EVs charging load is not considered individually, total load isassigned for system improvement. Hence, EV owner satisfaction is notprovided. In [41], charging power of EVs in a fixed period is maximized.However, behavior of EV owner is not taken into account.

In [42], an EV charging coordination strategy is proposed withobjective function of charging cost. Also system constraints areconsidered. In [43], Tabu Search algorithm is used to minimize the totaloperational costs of the distribution system. In [44], proposed methoddetermines optimal schedule for the charging of each EV consideringsystem requirements and individual EV owners. However, V2G strategy isnot involved in [42-44].

On these bases, optimization unit should make optimal coordinatedcharge/discharge decisions in order to satisfy system constraints, meetpower demand, maximize aggregator profit and owner comfort level.Nevertheless, most papers fail to simulate EV coordination considering adistribution network or ignore some system constraints. Hence, systemreliability is not guaranteed. Also, charging freedom has higherpriority than financial income for EV owners in reality. Namely,delaying charging or discharging to grid negatively affects EV ownerscomfort. However, many of researches have deficiency in terms of EVowners satisfaction.

This paper addresses the charge/discharge coordination problem ofEV for supporting system in a V2G concept. Our contributions are asfollows;

- The methodology that considers, the uniform randomness of arrivaland departure times, initial state of charge and EVs are located inIEEE-33 bus system to achieve realistic results.

- We proposed an objective function which minimizes cost ofa*ggregator, guarantees maximum charging level of battery at departuretime and satisfy network constraints in V2G systems simultaneously.Also, EV batteries reach to full as soon as possible. Hence, comfortlevel of EV owner is maximized.

- Swarm based (ABC, PSO) and evolutionary based (GA, DE) heuristicalgorithms are used to solve optimum charge/discharge coordination andcomparative results are presented.

II. PROBLEM FORMULATION

The increase in EV penetration will result in additional loads onthe electricity grid. Moreover, simultaneous charging of all EVs causesviolation in system limit. However, due to the V2G features, EVs cansupply energy to distribution grid. The aggregator collects individualEV data and coordinates the EV charge/discharge based on ancillaryservice signal. Charge/discharge coordination problem is optimizingcharge/discharge status of each EVs in order to obtain an economicaloperation of the distribution system and satisfying the systemrequirements of the system. Objective function of charge/dischargecoordination can be technical or economical. In this work, objectivefunction aims to minimize the cost which aggregator has to pay EV ownerin order to provide system constraints. When the network constraints areviolated, the aggregator decide status of each EVs to charge/dischargein order to reduce system load and improve system voltage. Thecharge/discharge tasks are assigned to EVs based on coordination cost.While control variables are charge/discharge status of EV, value ofobjective function depends on these variables.

The objective function f(x) is cost of charge dischargecoordination to be minimized. [psi](x,u) defines power flow equationsand [phi]p(x,u) indicates physical boundaries of the power system.

Min.f(x) subject to [psi](x, u) = 0

[phi](x,u)[less than or equal to]0 (1)

Network standard is satisfied with optimization of charge/dischargecoordination. State and control variables, constraints and objectivefunction are formulated below.

State and control variables:

State variables of DN and EV are described as follows;

[mathematical expression not reproducible] (2)

[P.sub.DN] (t) represents the total load of the DN at t. M is thetotal bus number of the network and each buses are indexed by m =1,2,..M. [V.sub.m](t) denotes the voltage of bus mat t. N is the numberof EVs connected to the network at t and each connected EV is indexed byn = 1,2,...N. SO[C.sub.n](t) is the state of charge of the EV battery att.

Charge/discharge coordination requirement is decided for each tbased onx(t). Coordination is not required if the system is withinlimits. In case the network is out of the limit, first the chargecoordination is applied. Charge coordination refers selection of EVs tostop charging process. If charge coordination fails to return to thenetwork limits, then discharge coordination will be applied. Dischargecoordination refers selection of EVs to be discharged. Status of each EVis considered as control variables at t as follows;

u(t) = {[u.sub.n](t)|n=1,2..N} (3)

where [u.sub.n] (t) is the control action on EV at t. [u.sub.n] (t)= 1 means EV n is allowed to charge. [u.sub.n] (t) = -1 means EV n isassigned for discharge. If [u.sub.n](t)= 0, EV n neither charges nordischarges. Naturally, The EVs, has not arrived yet, are not considerfor charging or discharging options.

Network Constraints:

The total distribution network load [P.sub.DN](t) includes[P.sub.H] (t) denotes total household load on the network, [P.sub.EV](t) indicates total EV load and [P.sub.L] (t) represents losses on thenetwork as shown in (4) and (5).

[P.sub.DN](t) = [P.sub.H] (t)+[P.sub.EV](t) + [P.sub.L](t) (4)

[mathematical expression not reproducible] (5)

Limit of [P.sub.DN] which is decided by the generation capacity andthe distribution transformer rating given as follows;

[P.sub.DN](t)<[P.sub.DN.sup.max] (6)

Limit of [V.sub.m](t) given as follows;

[V.sup.min]<[V.sub.m](t)<[V.sup.max] (7)

Electric Vehicle Constraints:

[E.sub.n](t-1) denotes the energy of EV n at (t-1), [DELTA]t is theminutes interval between t and (t-1). [E.sub.n.sup.miss](t) and[T.sub.req](t) calculates the amount of the missing energy and requiredtime to reach maximum allowed capacity at t as represented in (8) and(9).

[mathematical expression not reproducible] (8)

[mathematical expression not reproducible] (9)

ae[t.sub.n](t) is availability end time for charge/dischargecoordination. [t.sub.n.sup.dep] is departure time of EV nandae[t.sub.n](t) is calculated subtracting [T.sub.req](t) from[t.sub.n.sup.dep] as given in (10).

ae[t.sub.n](t) = [t.sub.n.sup.dep]-[T.sub.req](t) (10)

The EVs only charge if the current time is equal or later thanae[t.sub.n](t )as given in (11).

if t[greater than or equal to]ae[t.sub.n](t) [u.sub.m](t) = 1 (11)

Namely, EVs cannot be used for charge/discharge coordination laterae[t.sub.n](t). Otherwise, battery of the EV n would not be in maximumcapacity at departure time. ae[t.sub.n](t) constraint is applied toensure the maximum capacity of EV n at departure time. The EV users mayprovide a departure time or it can be estimated by probabilistic methodsbased on history of EV usage path [12].

Availability of EV n for charge/discharge coordination is alsodepending on user preferences and battery SO[C.sub.n](t) which is therate of current energy ([E.sub.n](t)) to energy capacity of EVs(E[C.sub.n](t)) as given in (12).

SO[C.sub.n](t) = [E.sub.n] (t)/[EC.sub.n] (t) (12)

SO[C.sub.n.sup.min] and SO[C.sub.n.sup.max], minimum and maximumlimit of SO[C.sub.n](t) as given in (13).

[mathematical expression not reproducible] (13)

In case of SO[C.sub.n] (t) [greater than or equal to]SO[C.sub.n.sup.max] (t), EV n stop charge. Similarly, EV n is notavailable for discharging in caseofSO[C.sub.n](t)<SO[C.sub.n.sup.min](t).

Objective Function:

The cost of charge/discharge coordination is considered as theobjective function f(x) to be minimized as given in (14). Therefore theaggregator and EV owner are affected as less as possible fromcoordination process. [P.sub.n,dcc](t), denotes the purchased dischargepower from EVn and [[alpha].sub.dcc] is the cost of discharge per kW.Similarly, [P.sub.n,cc] (t) indicates the delayed charge power of EVn.[[alpha].sub.cc] and [[alpha].sub.dcc] are the costs of delayed chargeand discharged power, respectively. [P.sub.pen], [O.sub.pen] andae[t.sub.pen] are penalty functions which occur in case of violation ofmaximum load, voltage and availability end time at t as presented (15),(16) and (17). [c.sub.pen] is penalty coefficient.

[mathematical expression not reproducible] (14)

[mathematical expression not reproducible] (15)

[mathematical expression not reproducible] (16)

[mathematical expression not reproducible] (17)

III. OPTIMIZATION OF EV CHARGE/DISCHARGE COORDINATION WITHHEURISTIC ALGORITHMS

Heuristic Algorithms are effective in solution of power systems dueto ability of scanning wide range of solution quickly and approachingglobal optimum although the solution is generally reached locallyoptimum with classical methods [45]. Some optimization algorithms arepresented for optimum coordination of charge/discharge process in theliterature [46-48]. In this work, evolutionary and swarm basedalgorithms are used for EV selection in order to optimizecharge/discharge coordination in the network structure of V2G. In Fig.1,flowchart of optimum charge/discharge coordination with heuristicalgorithms is given. First, the charge/discharge coordinationrequirement is determined running Backward/Forward (B/F) Sweep powerflow considering the voltage and maximum load limit in the system.Status of each available EV is determined randomly. If any violationoccurs, penalty function is applied. Best solution is selectedconsidering fitness values of solutions. The solution is updated usingoperators of algorithms in each iterations. Iteration is stopped when itreaches the maximum iteration number.

Whereas chromosomes represent the potential solutions in GA and DE,quality of food sources and distance of the particle to the foodrepresent potential solutions for ABC and PSO, respectively. In [49],new optimization criteria is defined that can be used of fuzzycontroller with dynamics.

The solution vector of heuristic algorithms[X.sub.i]=([X.sub.i1],[X.sub.i2],[X.sub.ij],...,[X.sub.iD]) correspondsto control variable vector [??} in EV charge/discharge coordinationproblem. Number of variables in the solution vector equals the number ofcontrol variable elements of U. mi is the total number of potentialsolution. i is the number of potential solution, i = 1..mi. Eachpotential solution include D dimensional control variable vector. j isthe number of parameter in variable vector, j = 1...D [50, 51]. jthparameter of ith solution vector [x.sub.ij] represent the chargingaction of EV n at t. Control variables vector which demonstrates thecharging status of each EV for each time and it can be arranged as shownin (18). Therefore, the ith solution of heuristic algorithms has beenencoded by a control variable. Number of rows equal to the dimension ofthe total number of aggregated EV (N) and a number of columns depend onthe number of time interval between arrival ([t.sub.a]) and departure([t.sub.d]) of the each EV. Although parameters are defined as a matrixin [52], control parameters are not mentioned in a matrix due to plug indurations of each EV is different here.

[mathematical expression not reproducible] (18)

Each EV is indexed by n = 1,2,..., N. When EV arrives the home andplugged in, SOC of EV read and departure is provided or it is estimatedbased on historic. The problem is solved and charge/discharge statues ofEV are updated for each time interval. The EVs charging schedule isgenerated between arrival and departure of EV [34]. [??] is controlparameter of n th EV in the tth time for ith solution. Value of [??]defines the charging status of EV [5]. As mentioned detailed inSection-2, charging status can be 1, -1 and 0 expressing the charging,discharging and no action, respectively. The presented methodology findsthe optimal charging schedule in order to solve EV charge/dischargecoordination. The aggregator has to pay to EV owner for discharge anddelayed charging process. Objective function is used to minimize cost ofcharge/discharge coordination. Charge and discharge times of EVs aredecided using heuristic algorithms. While the cost of coordinationminimizes, EV owner and system constraints are satisfied. Common processof the algorithms is similar but operators of each algorithms changesolutions. In the algorithms, j th parameter of i th solution isinitialized randomly considering upper and lower limits of the parameteras follows;

[mathematical expression not reproducible] (19)

Best solution is selected considering their fitness values asfollows;

[mathematical expression not reproducible] (20

The algorithms develop quality of solutions using their own uniqueoperators which explained detailed under their headings.

A. Genetic Algorithm

New population is created using gene of chromosomes of previouspopulation [53]. Best fitness of the chromosomes is selected andtransfer to next population. Firstly, chromosomes which refer thesolution set are encoded. After random initialization, operator ofreproduction, crossover and mutation are used. Chromosomes are selectedfrom previous population for reproduction. Crossover refers gene changesbetween chromosomes as shown in Fig. 2.

Mutation is random changing between genes of a chromosome as shownin Fig. 3. It creates individuals in the solution space but those arenot in the population. Different mutation methods are availableaccording to the coding types of individuals.

Selection process is applied after genetic operators and thecurrent population is updated with selected population.

B. Differential Evaluation

DE is a population based algorithm. Each chromosome is exposed tomutation, cross over and selection operators in order to create a newindividual [54]. In mutation Donor vector ([v.sub.i,j](t)) is created bymultiplying with scaling factor (F) difference of two chromosomes(r1,r2) and added to the third one (r3) as follows;

[v.sub.i,,j] (t + 1) = [x.sub.r1,j] (t) + F ([x.sub.r2, j] (t) -[x.sub.r3, j] (t)) (21)

In cross over, the trial vector [U.sub.i](t) is created mixingcurrent vector [X.sub.i](t) and donor vector [V.sub.i](t) with CrossoverRate (CR).

[mathematical expression not reproducible] (22)

In selection, the chromosome which has the highest fitness degreeis transferred to the next generation regarding to comparison of[X.sub.i](t) and trial vector [U.sub.i](t).

[mathematical expression not reproducible] (23)

C. Particle Swarm Optimization

Searching for food of particles is simulated as searching solutionfor an optimization problem [55]. Fitness value of the particle refersits distance to food. Main operators are velocity and the position ofthe particle. Velocity of the i th particle updated as follows;

[mathematical expression not reproducible] (24)

where pbest = ([p.sub.i 1],[p.sub.i2],...,[p.sub.iD]) is bestprevious solution, and gbest is best global solution in the memory. t isnumber of current generation, r1,r2 are uniform random value in therange [0, 1], w is inertia weight factor, c1,c2 are accelerationconstant of pbesti and gbesti. Position of i th particle is updatedsumming its previous position and current velocity as follows;

[mathematical expression not reproducible] (25)

D. Artificial Bee Colony

A bee colony consists of three group bees; employed, onlookers andscouts [56]. Employed bee is on the food source in advance and it sharesthe quality of the source with onlookers bee. Onlookers bees select foodsources considering source's nectar quality. Scouts scatterrandomly to explore new food sources.

Producing new food sources: It is the operator of ABC. Neighborhoodprinciple is considered by employed bees in order to decide the new foodsources. Neighbors of quality food sources are selected as new sourcesas follows;

[mathematical expression not reproducible] (26)

[v.sub.i] i represents new food source. More quality sources havemore probability to be selected.

IV. SIMULATION RESULTS

Optimization of charge/discharge coordination using heuristicmethods are implemented on 33 bus residential distribution networksserving 1000 houses with nominal voltage of 12.66 kV and base power of100 MVA. Houses are separated to 33 bus distribution networkproportional with the load data of the network as given in APPENDIX A[57]. The load profile is generated in GridLAB-D which is developed byPacific Northwest National Laboratory as a modeling and simulation toolof electric network [58, 59]. Load flow and optimization is simulated inMATLAB [60]. GridLAB-D allows detailed modelling of end use technologiesbased on users' behaviors and control of appliances. Multi-stateappliances models were used to obtain realistic load profiles in houseand distribution system. Total load profile of distribution systemdepends on many factors such as set points of thermostatic loads, outputtemperature, appliance usage frequency etc. In this study, simulatedappliances are clothes washers, dishwashers, clothes dryers,refrigerators, plug loads, lighting loads HVAC units (heating,ventilating, and air conditioning), water heaters, and ranges [9, 61,62]. House area is assumed to vary from 140 [m.sup.2] to 230 [m.sup.2].Meteorological data of Yakima, WA, USA is used as outdoor temperatureand simulation is carried out in July. Cooling and heating set points ofhouses are selected between 21.1-23.8[degrees]C and 18.3-20.5[degrees]C,respectively. The set point of water heater is 48.8[degrees]C. Usagefrequency of random pulsed appliances such as dryer, clothes washerdishwasher, range are varied by GridLAB-D based on calibrated End-UseLoad and Consumer Assessment Program (ELCAP) residential load data [63].The voltage magnitude at the substation was fixed at 1.0 p.u. (perunit). Maximum and minimum voltage magnitude limits are defined 1.00p.u. and 0.9 p.u., respectively [64]. Load capacity is specified as 5000kW. Three models of EVs placed randomly in 33 bus distribution networkas given in APPENDIX A. Specifications of connected EVs are given inTABLE I. EV may be used for commuting or longer trips with highercapacity batteries. We considered daily commuting purpose and featuresof EVs were chosen to suit this purpose.

EV owners mostly departure from home between 06.00-09.00 and arrivehome between 16.00-19.00. Hence, departure and arrival time distributionof EV created according to a normal probability distribution functionwith the mean at 07:30 and 17.30 of the variance of 1 h, respectively[65]. Daily average trip distance is 33 mile according to [66]. Tripdistances of each EVs are calculated using probability distributionfunction with the mean at 33 mile of the variance of 4 mile.SO[C.sub.n.sup.arr] is battery state of charge at arrival time andcalculated as follows;

[mathematical expression not reproducible] (27)

SO[C.sub.n.sup.arr] depends on roundtrip distance (di[s.sub.n]),range (rang[e.sub.n]) and energy capacity of EV. Due to deep dischargeand full charge decrease the battery life, EV n is allowed to be chargedand discharged between SO[C.sub.n.sup.min] and SO[C.sub.n.sup.max] [67].These are randomly selected as follow; SO[C.sub.n.sup.max] =rand[0.9-0.99], SOSO[C.sub.n.sup.min] = rand[0.3-0.4].

It is assumed that the aggregator has to pay 1.5 $/kWh and 5 $/kWhfor delayed charging power and for discharged power to EV owner.

B/F Sweep method is used to perform a load flow analysis [68]. Linecurrents and bus voltages are calculated for each iteration to determineEV states to provide optimum charge/discharge coordination. At theinitialization step, voltage of each bus assumed 1.0 p.u. and voltagedeviations of buses are calculated. In the backward step, the currentsare computed considering voltages of the previous iteration. In forwardstep, the node voltages are updated using voltage drops on thedistribution network lines. The currents and voltages are updatediteratively until nodal voltage criterion satisfied.

Charge/discharge coordination is required when the DN constraintsare violated. If the network does not turn to its limit although all EVstop charging, then discharge coordination is applied. Optimumcharge/discharge coordination process determines the charge/dischargestates of each EV to minimize the cost of coordination and disturbanceof EV owners.

V2G system consists of three main components; grid, aggregator andEVs. There is a bidirectional communication and power flow betweencomponents. That can be foreseen to be available in the future smartgrid. Also, smart meters play important role in order to send andreceive data. The aggregator receives the support signal from operator,if the system limits are violated. Then, the aggregator of EVs starts tocoordinate charge/discharge schedules to meet system requirements.Charging status of each plugged EV are decided to minimize objectivefunctions. Following consideration is assumed in the application:

-EVs have ability of bidirectional load flow and grid has therequired infrastructure for communication between EV and aggregator.

- The EV coordination is controlled for each time interval whichtime period is divided into.

-Departure time of each EV is notified by user or it is estimatedfrom historic driving patters.

-The EV owner permits the aggregator to determine charging statusof EV.

In Fig. 4, EV load, DN load w/ and w/o EV are given. Uncoordinatedcharging process starts at 14.30 and finish around 24.00. Peak load isincreased from 4760 kW to 5593 kW with integration of EVs at 18.00.Although total EV load is 748.5 kW, increase at total load is higherwith increase of losses. Total load of the network is 4997 kW at 18.30.After that, total load is also lower than the maximum load limit of thenetwork despite EV penetration.

In Fig. 5, minimum bus voltage magnitudes of DN are presented. Theminimum bus voltage magnitude is lower than the 0.9 p.u. w/o EVpenetration only at 18.00. That means, if all EV stopped chargingprocess the network would be still out of limits. Therefore, dischargecoordination of EV is required at that time. Because charge coordinationis not adequate to satisfy network limits. On the other hand, minimumbus voltage decrease below 0.9 p.u. from 17.30 to 19.30 with theuncoordinated charging process of EV. Charging coordination is needed atthose times except for 18.00.

In Fig. 6, total network losses which highly increase withintegration of EV are given. Total network losses increased from 300 kWto 383 kW at peak time.

In this paper optimization of charge/discharge coordination isprovided using heuristic methods, GA, PSO, DE and ABC. The results ofeach algorithm are compared with each other. For each algorithm,population size and the iteration numbers are selected 20 and 100,respectively.

The best values of specific parameters of each algorithm are chosenbased on experiment as follow;

PSO: [w,c1,c2] = [0.5, 1.2, 1.5]. w affects the search abilitysignificantly but [c.sub.1] and [c.sub.2] decided the final values ofposition expectation and position variance. If w is selected too small,particles may not search sufficiently. Low values of [c.sub.1] and[c.sub.2] may lead particles to search far from target region beforetugged back. High values of the weighting factors [c.sub.1] and[c.sub.2] may cause excessive motions or overshooting in the targetregion [69].

DE: [(F),(CR)] = [0.6, 0.4]. Small values of CR result in gradualand small exploratory moves in search space, while large values of CRproduce rapid moves at angles to the search space's axes. Using toosmall a value of F leads to premature convergence, while high value highvalues slow down the search [70, 71].

ABC: Limit= [100]. Limit is the number of trials which bees toleave the food source. If limit is too low, sufficient search cannot beperformed. If it is high, too much search is performed on one foodsource. Although time consumption is increased, the solution may not beincreased [72].

GA: [Crossover, Mutation, Selection] = [Scatted, Constraintdependent, Roulette]

In Fig.7, Converge curves of the algorithms which are employed forcharge coordination cost at 17.30 are shown due to network load and EVload level are same for each algorithm at that time. GA has the bestresults with $0.014. Initialization value of GA is also better thanother algorithms. Although, initialization values of PSO, DE, ABC arealmost same, ABC has the highest value with $0.089.

In Fig. 8, daily costs of the charge/discharge coordination processusing heuristic algorithms are shown. Optimization of coordinationprocess starts at 17.30 and finish at 19.30 for each algorithm.Charge/discharge coordination is not required due to voltage and load ofthe network are within the limits at 20.00. Daily cost of coordinationprocess are $833.10, $2871.90, $3288.30, $4169.10 using GA, PSO, DE,ABC, respectively. Cost of charge/discharge coordination changesdepending on selected EV. Because connected bus of selected EV issignificant for voltage support. Minimum cost is obtained using GA dueto optimum selection of EV for coordination process.

While the coordination process creates a cost for the aggregator,it also discomforts the EV owners by discharging or delaying chargingprocess. Amount of discharged or delayed load are given in Fig. 9. V2Gdischarge period is shown in a box. Discharged powers are 93.5 kW, 146.1kW, 180.8 kW, 254.2 kW using GA, PSO, DE and ABC at 18.00, respectively.EVs fully charged earlier with GA due to amount of delayed chargingprocess is lower. Therefore number of charging EV is lower with GA at19.30. Also, GA had the best convergence performance. Hence, delayedcharging loads are 22.6 kW, 76.1 kW, 166.3 kW, 336.1 kW using GA, PSO,DE and ABC at 19.30, respectively. Delayed or injected load is absent at20.00.

In Fig. 10, number of the EVs which affected from optimumcharge/discharge coordination is presented. 12, 25, 41 and 78 EVs areselected to stop charging at 17.30 and 30, 57, 67, 92 EVs are selectedfor discharge at 18.00 by GA, PSO, DE and ABC, respectively. GA presentsthe minimum number of selected EVs for coordination as well as the costof coordination. The difference between algorithms is more evident at19.30. While GA selects only 10 EVs, ABC selects 123 EVs to satisfynetwork requirement at 19.30.

In Fig. 11, total load of DN is presented for given hours. Totalload of DN with GA is higher than others during charge/dischargecoordination due to allowing using maximum capacity of the network.Total loads of DN are 4652.1 kW, 4595.6 kW, 4558.3 kW, 4479.7 kW usingGA, PSO, DE and ABC at 18.00, respectively. However, total network loadwith GA is lower at 20.00 because of EVs which ended charging processbefore 20.00. Total loads of DN are 3948.7 kW, 4089.7 kW, 4136.8 kW, and4184 kW using GA, PSO, DE and ABC at 20.00, respectively.

In Fig.12, voltage values of each bus are given for 18.00. Minimumvoltage magnitude w/o EV is lower than 0.9 p.u. only at 18.00.Therefore, discharge coordination is applied only at 18.00, while chargecoordination is applied at other given times. Voltage magnitude isincreased from 0.8768 p.u. to 0.9 p.u. with the given algorithms duringdischarge period.

Also, minimum voltage magnitudes are increased to minimum 0.9 p.u.with charging coordination at other given times as demonstrated in TABLEII. The coordination is not required at 20.00 due to any of the busvoltages are not lower than 0.9 p.u. or total load of the system is nothigher than 5000 kW.

V. CONCLUSION

The power consumption which is already continuously increasingstressed on network much more with the impact of charging EVs in peakhours. This leads to problems such as overloading, voltage drops etc. indistribution network. These problems can be solved by deferring chargingand injecting power from EV using V2G feature in the smart gridinfrastructure. However charge/discharge process should be welloptimized considering system requirements, cost and comfort of EV owner.In this study, charge/discharge coordination cost is minimized toincrease each bus voltage to EN50160 standards and reduce the total loadbelow the maximum network load capacity using GA, PSO, DE and ABC. Alsoit is provided that EVs to be charged maximum at departure time.Therefore charging process and EV owner are affected as less as possiblefrom coordination process. In case of comparison of the algorithms, GAprovided both minimum cost and maximum convenience for EV owners. Whilethe cost is minimized, network capacity is optimally used not todiscomfort EV owner by reducing the number of selected vehicles forcharge/discharge coordination. This allows both vehicles to be chargedas quickly as possible, as well as longer average battery life due tominimizing number of switched EV.

REFERENCES

[1] U.S. Energy Administration Office, International EnergyOutlook, Washington, DC, USA, DOE/EIA-0484(2016), May 2016. [Online].Available: https://www.eia.gov/outlooks/ieo/pdf/0484(2016).pdf

[2] Electric Power Research Institute, Environmental assessment ofplug-in hybrid electric vehicles. Volume 1: Nationwide Greenhouse GasEmissions, CA, USA, 1015325, July 2007. [Online]. Available:https://energy.gov/sites/prod/files/oeprod/DocumentsandMedia/EPRI-NRDC_PHEV_GHG_report.pdf

[3] International Energy Agency. Global EV Outlook 2017 Two millionand counting. Paris, FR, OECD/IEA 2017, July 2017. [Online]. Available:https://www.iea.org/media/topics/transport/Global_EV_Outlook_2017_Leaflet.pdf

[4] X. Fang, S. Misra, G. Xue, and D. Yang, "Smart Grid--TheNew and Improved Power Grid: A Survey," IEEE Commun. Surveys Tuts.,vol. 14, no. 4, pp. 944-980, 2012. doi:10.1109/SURV.2011.101911.00087

[5] S. Xie, W. Zhong, K. Xie, R. Yu, and Y. Zhang, "FairEnergy Scheduling for Vehicle-to-Grid Networks Using Adaptive DynamicProgramming," IEEE Trans. Neural Netw. Learn. Syst., vol. 27, no.8, pp. 1697-1707, 2016. doi:10.1109/TNNLS.2016.2526615

[6] E. De Caluwe, Grid-supportive charging infrastructure forplug-in electric vehicles, PhD thesis, K.U.Leuven,-Faculty ofEngineering Science, 2015. [Online]. Available:https://lirias.kuleuven.be/bitstream/123456789/511457/1/Phd_NLeemput_Final.pdf

[7] J. Taylor, A. Maitra, M. Alexander, D. Brooks, and M. Duvall,"Evaluation of the impact of plug-in electric vehicle loading ondistribution system operations," in Proc. IEEE Power Energy Soc.Gen. Meet., Calgary, Canada, 26-30 July 2009 pp. 1-6.doi:10.1109/PES.2009.5275317

[8] M. J. Scott, M. K. Meyers, D. B. Elliott, W. M. Warwick,"Impacts Assessment of Plug-in Hybrid Vehicles on ElectricUtilities and Regional US Power Grids Part 2: Economic Assessment,"Pacific Northwest Nat. Lab., Richland, WA., DE-AC05-76RL01830, Nov. 2017[Online]. Available:http://energyenvironment.pnnl.gov/ei/pdf/PHEV_Economic_Analysis_Part2_Final.pdf

[9] A. Dogan, M. Kuzlu, M. Pipattanasomp*rn, S. Rahman, and T.Yalcinoz, "Impact of EV charging strategies on peak demandreduction and load factor improvement," in Proc. Inter. Conf. onElect.l and Electronics Eng., Bursa, Turkey, 26-28 Nov. 2015, pp.374-378. doi:10.1109/ELECO.2015.7394559

[10] C. Guille and G. Gross, "A conceptual framework for thevehicle-to-grid ( V2G ) implementation," Energy Policy, vol. 37,no. 11, pp. 4379-4390, 2009. doi:10.1016/j.enpol.2009.05.053

[11] M. Yilmaz and P. T. Krein, "Review of the impact ofvehicle-to-grid technologies on distribution systems and utilityinterfaces," IEEE Trans. Power Electron., vol. 28, no. 12, pp.5673-5689, Dec. 2013. doi:10.1109/TPEL.2012.2227500

[12] C. S. Antunez, J. F. Franco, M. J. Rider, R. Romero, "ANew Methodology for the Optimal Charging Coordination of ElectricVehicles Considering Vehicle-to-Grid Technology," IEEE Trans.Sustain. Energy. vol. 7, no. 2, pp. 596-607, 2016.doi:10.1109/TSTE.2015.2505502

[13] K. Clement-nyns, E. Haesen, and J. Driesen, "The impactof vehicle-to-grid on the distribution grid," Electr. Power Syst.Res., vol. 81, no. 1, pp. 185-192, 2011. doi:10.1016/j.epsr.2010.08.007

[14] H. Liu, Z. Hu, Y. Song, and J. Lin, "Decentralizedvehicle-to-grid control for primary frequency regulation consideringcharging demands," IEEE Trans. Power Syst., vol. 28, no. 3, pp.3480-3489, Aug. 2013. doi:10.1109/TPWRS.2013.2252029

[15] C. D. White and K. M. Zhang, "Using vehicle-to-gridtechnology for frequency regulation and peak-load reduction," J.Power Sources, vol. 196, no. 8, pp. 3972-3980, 2011.doi:10.1016/j.jpowsour.2010.11.010

[16] Z. Wang and S. Wang, "Grid Power Peak Shaving and ValleyFilling Using Vehicle-to-Grid Systems," IEEE Trans. Power Del.,vol. 28, no. 3, pp. 1822-1829, 2013. doi:10.1109/TPWRD.2013.2264497

[17] M. Brenna, F. Foiadelli, and M. Longo, "The Exploitationof Vehicle-to-Grid Function for Power Quality Improvement in a SmartGrid," IEEE Intell. Transp. Syst. vol. 15, no. 5, pp. 2169-2177,2014. doi:10.1109/tit*.2014.2312206

[18] H. Liu, Z. Hu, Y. Song, J. Wang, and X. Xie,"Vehicle-to-Grid Control for Supplementary Frequency RegulationConsidering Charging Demands," IEEE Trans. Power Syst., vol. 30,no. 6, pp. 3110-3119, 2015. doi:10.1109/TPWRS.2014.2382979

[19] M. Kesler, M. C. Kisacikoglu, and L. M. Tolbert,"Vehicle-to-Grid Reactive Power Operation Using Plug-In ElectricVehicle Bidirectional Offboard Charger," IEEE Ind. Electron., vol.61, no. 12, pp. 6778-6784, 2014. doi:10.1109/TIE.2014.2314065

[20] J. Lin, S. Member, K. Leung, V. O. K. Li, and A. In,"Optimal Scheduling With Vehicle-to-Grid Regulation Service,"IEEE Internet Things J., vol. 1, no. 6, pp. 556-569, 2014.doi:10.1109/JIOT.2014.2361911

[21] X. Wu, L. Li, J. Zou, and G. Zhang, "EV-Based VoltageRegulation in Line Distribution Grid." IEEE Instr. and Meas. Tech.Conf. Taipei 2016. doi:10.1109/I2MTC.2016.7520568

[22] A. Andreotti, G. Carpinelli, F. Mottola, and D. Proto, "Areview of single-objective optimization models for plug-in vehiclesoperation in smart grid- Part I: Theoretical aspects," in Proc.Power and Energy Society General Meeting, San Diego, USA, 22-26 July2012, pp. 1-8 doi:10.1109/PESGM.2012.6345381.

[23] A. Andreotti, G. Carpinelli, F. Mottola, and D. Proto "Areview of single-objective optimization models for plug-in vehiclesoperation in smart grids part ii: Numerical applications to vehiclesfleets," in Proc. Power and Energy Society General Meeting, SanDiego, USA, 22-26 July 2012, pp. 1-8. doi:10.1109/PESGM.2012.6345390.

[24] X. Bai, W. Qiao, "Robust optimization for bidirectionaldispatch coordination of large-scale V2G," IEEE Trans Smart Grid,vol. 6, no. 4, pp. 1944-1954, 2015. doi:10.1109/TSG.2015.2396065

[25] W. Qi, Z. Xu, Z-J. Shen, Hu Z, Song Y. "Hierarchicalcoordinated control of plug- in electric vehicles charging inmultifamily dwellings,"IEEE Trans Smart Grid, vol. 5, no. 3, pp.1465-1474, 2014. doi:10.1109/TSG.2014.2308217

[26] C. Jin,, J. Tang, and P. Ghosh, "Optimizing ElectricVehicle Charging: A Customer's Perspective," IEEE Trans. Veh.Technol., vol. 62, no. 7, pp. 2919-2927, 2013.doi:10.1109/TVT.2013.2251023

[27] A. H. Hajimiragha, C. A. Canizares, M. W. Fowler, S. Moazeni,and A. Elkamel, "A robust optimization approach for planning thetransition to plug-in hybrid electric vehicles," IEEE Trans. PowerSyst, vol. 26, no. 4, pp. 2264-2274, 2011.doi:10.1109/TPWRS.2011.2108322

[28] K. Zhang, L. Xu, M. Ouyang, H. Wang, L. Lu, J. Li,"Optimal decentralized valley-filling charging strategy forelectric vehicles,"Energy Convers Manag., vol. 78, no. 57, pp.537-550, 2009. doi:10.1016/j.enconman.2013.11.011

[29] X. Wang, Q. Liang, "Energy management strategy forplug-in hybrid electric vehicles via bidirectionalvehicle-to-grid," IEEE Syst J, vol. 37, no. 3, pp. 1789-1798, 2017.doi:10.1109/JSYST.2015.2391284

[30] M. Shafie-khah, M. P. Moghaddam, M.K. Sheikh-El-Eslami, M.Rahmani- Andebili, "Modeling of interactions between marketregulations and behavior of plug-in electric vehicle aggregators in avirtual power market environment," Energy, vol. 40, no. 1, pp.139-150, 2012. doi:10.1016/j.energy.2012.02.019

[31] Z. Yang, K. Li, A. Foley, C. Zhang, "Optimal SchedulingMethods to Integrate Plug-in Electric Vehicles with the Power System: AReview," in Proc. 19th IFAC World Congress, Cape Town, SouthAfrica, 24-29 August 2014. doi:10.3182/20140824-6-ZA-1003.01804

[32] Z. Yang, K. Li, A. Foley, "Computational schedulingmethods for integrating plug-in electric vehicles with power systems: Areview," Renewable and Sustainable Energy Reviews, vol. 51, no. 28,pp. 396-416, 2015. doi:10.1016/j.rser.2015.06.007

[33] Y. Sugii, K. Tsujino, T. Nagano, "A Genetic-Algorithmbased scheduling method of charging of electric vehicles," in Proc.IEEE Systems, Man, and Cybernetics, Conference Proceedings, Tokyo,Japan, 12-15 Oct. 1999, pp. 1-8. doi:10.1109/ICSMC.1999.812443

[34] G. Celli, E. Ghiani, F. Pilo, G. Pisano, G. G. Soma,"Particle Swarm Optimization for Minimizing the Burden of ElectricVehicles in Active Distribution Networks," in Proc. Power andEnergy Society General Meeting, in Proc. Power and Energy SocietyGeneral Meeting, San Diego, USA, 22-26 July 2012, pp. 1-7.doi:10.1109/PESGM.2012.6345458

[35] S. Xu, D. Feng, Z. Yan, L. Zhang, N. Li, L. Jing, J. Wang,"Ant-Based Swarm Algorithm for Charging Coordination of ElectricVehicles," Int. J. Dist. Sensor Network, vol. 9, no. 5, pp. 1-13,2013. doi:10.1155/2013/268942

[36] I. Rahman, P. Vasant, B. S. M. Singh, M.Abdullah-Al-WadudHybrid, "Swarm Intelligence-Based Optimization forCharging Plug-in Hybrid Electric Vehicle," In: Nguyen N., TrawinskiB., Kosala R. (eds) Intelligent Information and Database Systems. ACIIDS2015. Lecture Notes in Computer Science, vol 9012. Springer, Chamdoi:10.1007/978-3-319-15705-4_3

[37] M. Alonso, H. Amaris, J. G. Germain, J. M. Galan, OptimalCharging Scheduling of Electric Vehicles in Smart Grids by HeuristicAlgorithms," Energies, vol. 7, no. 4, pp. 2449-2475, 2014.doi:10.3390/en7042449

[38] C. Jin, J. Tang, P. Ghosh, "Optimizing electric vehiclecharging with energy storage in the electricity market," IEEE TransSmart Grid, vol. 4, no. 1, pp. 311-320, 2013. doi:10.1109/TSG.2012.2218834

[39] S. Shao, M. Pipattanasomp*rn, and S. Rahman, "Challengesof PHEV penetration to the residential distribution network," inProc. IEEE Power Energy Soc. Gen. Meeting, 2009, Calgary, Canada, 26-30July 2009, pp. 1-8. doi:10.1109/PES.2009.5275806

[40] C. D. White and K. M. Zhang, "Using vehicle-to-gridtechnology for frequency regulation and peak-load reduction," J.Power Sources, vol. 196, no. 8, pp. 3972-3980, 2011.doi:10.1016/j.jpowsour.2010.11.010

[41] P. Richardson, D. Flynn, A. Keane, "Optimal Charging ofElectric Vehicles in Low-Voltage Distribution Systems," IEEE Trans.Power Syst., vol. 27, no. 1, pp. 268 - 279, 2012.doi:10.1109/TPWRS.2011.2158247

[42] S. Deilami, A. S. Masoum, P. S. Moses, M. A. S. Masoum,"Real-Time Coordination of Plug-In Electric Vehicle Charging inSmart Grids to Minimize Power Losses and Improve Voltage Profile,"IEEE Trans. Smart Grid, vol. 2, no. 3, pp. 456 - 467,2011.doi:10.1109/TSG.2011.2159816

[43] N. Banol A., J. F. Franco, M. Lavorato, M. J. Rider, R.Romero, "Plug-In Electric Vehicle Charging Coordination inElectrical Distribution Systems Using a Tabu Search Algorithm,"IEEE 15th Int. Conf. Environment and Electrical Engineering (EEEIC),Rome, Italy, 10-13 June 2015, pp. 1-6. doi:10.1109/EEEIC.2015.7165395

[44] O. Sundstrom, C. Binding, "Flexible Charging Optimizationfor Electric Vehicles Considering Distribution Grid Constraints,"IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 26-37, 2011.doi:10.1109/TSG.2011.2168431

[45] A. Dogan, T. Yalcinoz, M. Alci, "A Comparison ofHeuristic Methods for Optimum Power Flow Considering Valve PointEffect", Elektronika Ir Elektrotechnika, vol. 22, no.5, pp.32-37,2016. doi:10.5755/j01.eie.22.5.16340

[46] P. Richardson, D. Flynn, and A. Keane, "Optimal chargingof electric vehicles in low-voltage distribution systems," IEEETrans. Power Syst., vol. 27, no. 1, pp. 268-279, 2012.doi:10.1109/TPWRS.2011.2158247

[47] C. Wu and H. Mohsenian-rad, "Vehicle-to-AggregatorInteraction Game," IEEE Trans. Smart Grid, vol. 3, no. 1, pp.434-442, 2012. doi:10.1109/TSG.2011.2166414

[48] H. Liang, B. J. Choi, and W. Zhuang, "Optimizing theEnergy Delivery via V2G Systems Based on Stochastic InventoryTheory," IEEE Trans. Smart Grid., vol. 4, no. 4, pp. 2230-2243,2013. doi:10.1109/TSG.2013.2272894

[49] R.-E. Precup, S. Preitl, "Optimisation criteria indevelopment of fuzzy controllers with dynamics," EngineeringApplications of Artificial Intelligence, vol. 17, no. 6, pp. 661-674,2004. doi:10.1016/j.engappai.2004.08.004 [50] T. S. Li, C. T. Su, T.L.Chiang, "Applying robust multi-response quality engineering forparameter selection using a novel neural--genetic algorithm,"Computers in Industry, vol. 50, no. 1, pp. 113-122, 2003. doi:10.1016/S0166-3615(02)00140-9.

[51] S. Vrkalovic, T.-A. Teban, I.-D. Borlea, "StableTakagi-Sugeno fuzzy control designed by optimization,"International Journal of Artificial Intelligence, vol. 15, no. 2, pp.17-29, 2017. doi:

[52] R. D. Baruah, P. Angelov, "DEC: Dynamically EvolvingClustering and its application to structure identification of evolvingfuzzy models," IEEE Trans. Cybern., vol. 44, no. 9, pp. 1619-1631,2014. doi:10.1109/TCYB.2013.2291234

[53] D. E. Goldberg, Genetic Algorithms in Search, Optimization,and Machine Learning. Reading: Addison-Wesley Publishing Company, p.62,1989

[54] K. V. Price, "Differential evolution: a fast and simplenumerical optimizer," in Proc. Fuzzy Inf. Process. Soc. Conf. NorthAm., Berkeley, CA, USA, 19-22 June 1996 pp. 524-527.doi:10.1109/NAFIPS.1996.534790

[55] J. Kennedy, R. Eberhart, "Particle swarmoptimization", in Proc. IEEE Int. Conf Neural Networks, Perth,Austuralia, 27 Nov.-1 Dec. 1995, pp. 1942-1948.doi:10.1109/ICNN.1995.488968

[56] D. Karaboga, "An idea based on honey bee swarm fornumerical optimization". Technical Report TR06, Erciyes University,Eng. Faculty, Computer Engineering Department, Oct. 2005. [Online].Available: http://mf.erciyes.edu.tr/abc/pub/tr06_2005.pdf

[57] R. Ranjan, D Das, "Simple and efficient computeralgorithm to solve radial distribution networks," Electr PowerCompon Syst., vol. 31 pp.95-107, 2003 doi:10.1080/15325000390112099

[58] GridLAB-D, [Online]. Available: http://www.gridlabd.org/[59]D.P. Chassin, K. Schneider, C. Gerkensmeyer, "GridLABD: Anopen-source power systems modeling and simulation environment," inProc. Transmission and Distribution Conference and Exposition, Chicago,USA, 21-24 April 2008, pp.1-5. doi:10.1109/TDC.2008.4517260

[60] MATLAB, [Online]. Available:https://www.mathworks.com/products/matlab.html

[61] J. C. Fuller, B. Vyakaranam, N. Prakash Kumar, S. M.Leistritz, G. B. Parker, "Modeling of GE Appliances in GridLAB-D:Peak Demand Reduction," Technical Report-PNNL-21358 [Online].Available: https://www.pnnl.gov/main/publications/external/technical_reports/pnnl-21358.pdf

[62] Z. T. Taylor, K. Gowri, S. Katipamula, "GridLAB-DTechnical Support Document: Residential End-Use Module Version1.0," Technical Report-PNNL- 17694 [Online]. Available:https://www.pnnl.gov/main/publications/external/technical_reports/PNNL-17694.pdf

[63] R. G. Pratt, C. C. Conner, E. E. Richman, K. G. Ritland, W. F.Sandusky, and M. E. Taylor, "Description of Electric Energy Use inSingle Family Residences in the Pacific Northwest," DOE/BP 1379521, Bonneville Power Administration, Portland, OR, 1989. [Online].Available: https://elcap.nwcouncil.org/Documents/Electric%20Energy%20Use%20Single%20Family.pdf

[64] EN 50160, voltage characteristics of electricity supplied bypublic distribution systems, 1999.

[65] S. Shao, M. Pipattanasomp*rn, and S. Rahman, "GridIntegration of Electric Vehicles and Demand Response With CustomerChoice," IEEE Trans. Smart Grid., vol. 3, no. 1, pp. 543-550, 2012.doi:10.1109/TSG.2011.2164949

[66] FHA, "Summary of Travel Trends: 2009 National HouseholdTravel Survey," p. 82, 2011.

[67] J. D. Dogger, B. Roossien, and F. D. J. Nieuwenhout,"Characterization of Li-ion batteries for intelligent management ofdistributed grid connected storage," IEEE Trans. Energy Convers.,vol. 26, no. 1, pp. 256-263, 2011. doi:10.1109/TEC.2009.2032579E.

[68] E. Bompard, E. Carpaneto, G. Chicco, and R. Napoli,"Convergence of the backward / forward sweep method for theload-flow analysis of radial distribution systems," Int. J. ofElect. Power & Energy Syst., vol. 22, pp. 521-530, 2000.doi:10.1016/S0142-0615(00)00009-0.

[69] R.-E. Precup, R.-C. David, E. M. Petriu, M.-B. Radac, S.Preitl, J. Fodor, "Evolutionary optimization-based tuning oflow-cost fuzzy controllers for servo systems," Knowledge-BasedSystems, vol. 38, no. 9, pp. 74-84, 2013.doi:10.1016/j.knosys.2011.07.006

[70] D. Zaharie, "Influence of crossover on the behavior ofDifferential Evolution Algorithms," Applied Soft Computing, vol. 9,no. 3, pp. 1126-1138, 2009. doi: 10.1016/j.asoc.2009.02.012

[71] A. W.Mohamed, H. Z. Sabry, M. Khorshid, "An alternativedifferential evolution algorithm for global optimization," Journalof Advanced Research, vol. 3, no. 2, pp. 149-165, 2012.doi:10.1016/j.jare.2011.06.004

[72] D. Karaboga B. Basturk, "On the performance of artificialbee colony (ABC) algorithm," Applied Soft Computing, vol. 8, no. 1,pp. 687-697, 2008. doi: 10.1016/j.asoc.2007.05.007

APPENDIX A

Ahmet DOGAN, Serkan BAHCECI, Ferhat DALDABAN, Mustafa ALCI

Department of Electrical and Electronics Engineering, ErciyesUniversity, 38039 Kayseri, Turkey

[emailprotected]

Digital Object Identifier 10.4316/AECE.2018.01015

TABLE I. FEATURES OF EV IN 33 BUS DN EV1 EV2 EV3Number of 131 104 165Connected EVsBattery 16.5 17 24Capacity (kWh)Charge/Disch. 1.9 3.0 3.3Power(kW)Range (mile) 58.7 54.4 86.8TABLE II. MINIMUM VOLTAGE MAGNITUDES WITH GA, PSO, DE, ABC Minimum Bus Voltage (p.u,)Hour w/o w/ EV GA PSO DE ABC EV17.30 0.9129 0.8980 0.9000 0.9001 0.9007 0.902318.00 0.8942 0.8768 0.9000 0.9001 0.9003 0.900818.30 0.9082 0.8907 0.9000 0.9001 0.9002 0.900019.00 0.9113 0.8929 0.9000 0.9010 0.9002 0.900019.30 0.9168 0.8971 0.9000 0.9003 0.9005 0.903920.00 0.9335 0.9149 - - - -TABLE A. NUMBER OF HOUSES AND EV IN EACH BUS Number of EVBus No House EV1 EV2 EV3 Num. 1 0 0 0 0 2 27 0 0 0 3 24 7 3 3 4 32 6 2 6 5 16 0 6 0 6 16 7 0 0 7 54 0 0 0 8 54 0 0 0 9 16 4 4 610 16 7 1 511 12 2 2 712 16 9 5 013 16 7 3 314 32 21 0 015 16 7 3 316 16 1 6 717 16 7 0 018 24 0 7 719 24 0 0 1420 24 0 21 021 24 1 2 1022 24 10 7 323 24 6 4 1124 113 14 9 525 113 26 6 326 16 7 0 027 16 3 5 628 16 0 0 029 32 3 2 530 55 10 6 631 41 0 0 032 58 0 0 2133 17 0 0 0