Research ArticleSwarm Intelligence-Based Smart Energy Allocation Strategy forCharging Stations of Plug-In Hybrid Electric Vehicles
Imran Rahman,1 Pandian M. Vasant,1
Balbir Singh Mahinder Singh,1 and M. Abdullah-Al-Wadud2
1Department of Fundamental and Applied Sciences, Universiti Teknologi PETRONAS, Bandar Seri Iskandar,31750 Tronoh, Perak, Malaysia2Department of Software Engineering, College of Computer and Information Sciences, King Saud University,P.O. BOX 2454, Riyadh 11451, Saudi Arabia
Correspondence should be addressed to Pandian M. Vasant; [email protected]
Received 15 May 2014; Accepted 18 August 2014
Academic Editor: Baozhen Yao
Copyright Β© 2015 Imran Rahman et al.This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Recent researches towards the use of green technologies to reduce pollution and higher penetration of renewable energysources in the transportation sector have been gaining popularity. In this wake, extensive participation of plug-in hybrid electricvehicles (PHEVs) requires adequate charging allocation strategy using a combination of smart grid systems and smart charginginfrastructures. Daytime charging stations will be needed for daily usage of PHEVs due to the limited all-electric range. Intelligentenergy management is an important issue which has already drawn much attention of researchers. Most of these works requireformulation ofmathematical models with extensive use of computational intelligence-based optimization techniques to solvemanytechnical problems. In this paper, gravitational search algorithm (GSA) has been applied and compared with another memberof swarm family, particle swarm optimization (PSO), considering constraints such as energy price, remaining battery capacity,and remaining charging time. Simulation results obtained for maximizing the highly nonlinear objective function evaluate theperformance of both techniques in terms of best fitness.
1. Introduction
The vehicular network recently accounts for around 25% ofCO2emissions and over 55% of oil consumption around the
world [1]. Carbon dioxide (CO2) is the primary greenhouse
gas emitted through human activities like combustion of fos-sil fuels (coal, natural gas, and oil) for energy and transporta-tion. Several researchers have proved that a great amount ofreductions in greenhouse gas emissions and the increasingdependence on oil could be accomplished by electrificationof transport sector [2]. Indeed, the adoption of hybrid electricvehicles (HEVs) has brought significant market success overthe past decade. Vehicles can be classified into three groups:internal combustion engine vehicles (ICEVs), hybrid electricvehicles (HEVs), and all-electric vehicles (AEVs) [3]. Plug-in hybrid electric vehicles (PHEVs) which are very recentlyintroduced promise to boost up the overall fuel efficiency
by holding a higher capacity battery system, which can bedirectly charged from traditional power grid system thathelps the vehicles to operate continuously in βall-electricrangeβ (AER). All-electric vehicle or AEV is a vehicle usingelectric power as the only source to move the vehicle [4].Plug-in hybrid electric vehicles with a connection to thesmart grid can possess all of these strategies. Hence, thewidely extended adoption of PHEVs might play a significantrole in the alternative energy integration into traditionalgrid systems [5]. There is a need of efficient mechanismsand algorithms for smart grid technologies in order to solvehighly heterogeneous problems like energy management,cost reduction, efficient charging infrastructure, and so forthwith different objectives and system constraints [6].
According to a statistics of Electric Power ResearchInstitute (EPRI), about 62% of the entire United States (US)vehicle will comprise PHEVs within the year 2050 [7].
Hindawi Publishing CorporationMathematical Problems in EngineeringVolume 2015, Article ID 620425, 10 pageshttp://dx.doi.org/10.1155/2015/620425
2 Mathematical Problems in Engineering
Moreover, there is an increasing demand to implement thistechnology on the electric grid system. Large numbers ofPHEVs have the capability to threaten the stability of thepower system. For example, in order to avoid interruptionwhen several thousand PHEVs are introduced into thesystem over a short period of time, the load on the powergrid will need to be managed very carefully. One of themain targets is to facilitate the proper interaction betweenthe power grid and the PHEV. For the maximization ofcustomer satisfaction and minimization of burdens on thegrid, a complicated control mechanism will need to beaddressed in order to govern multiple battery loads froma number of PHEVs appropriately [8]. The total demandpattern will also have an important impact on the electricityindustry due to differences in the needs of the PHEVs parkedin the deck at certain time [9]. Proper management canensure strain minimization of the grid and enhance thetransmission and generation of electric power supply. Thecontrol of PHEV charging depending on the locations canbe classified into two groups: household charging and publiccharging. The proposed optimization focuses on the publiccharging station for plug-in vehicles because most of PHEVcharging is expected to take place in public charging locations[10].
Wide penetration of PHEVs in the market depends ona well efficient charging infrastructure. The power demandfrom this new load will put extra stress on the traditionalpower grid [11]. As a result, a good number of PHEV charginginfrastructures with appropriate facilities are essential to bebuilt for recharging electric vehicles; for this some strategieshave been proposed by the researchers [12, 13]. Charging sta-tions are needed to be built at workplaces, markets/shoppingmalls, and home. In [14], authors proposed the necessity ofbuilding new smart charging station with effective commu-nication among utilities along with substation control infras-tructure in viewof grid stability andproper energy utilization.Furthermore, assortment of charging stations with respect tocharging characteristics of different PHEVs traffic mobilitycharacteristics, sizeable energy storage, cost minimization,quality of services (QoS), and optimal power of intelligentcharging station are underway [15]. Thus, evolution of reli-able, efficient, robust, and economical charging infrastructureis underway. In this wake, numerous techniques andmethodshave been proposed for deployment of charging station forPHEVs [16, 17].
One of the important constraints for accurate chargingis state of charge (SoC). Charging algorithm can accuratelybe managed by the precise state of charge estimation [18].An approximate graph of a typical lithium-ion cell voltageversus SoC is shown in Figure 1. The figure indicates that theslope of the curve below 20% and above 90% is high enoughto result in a detectable voltage difference to be relied onby charge balancing control and measurement circuits [19].There is a need of in-depth study onmaximization of averageSoC in order to facilitate intelligent energy allocation forPHEVs in a charging station. Gravitational search algorithm(GSA) is one of the newest heuristic algorithms introduced byRashedi et al. [20]. GSA algorithm is also amember of swarmintelligence family which is inspired by the well-known law of
3.2
3.4
3.6
3.8
4
4.2
4.4
0 10 20 30 40 50 60 70 80 90 100
Volta
ge (V
)
State of charge (%)
Li-ion cell voltage versus SoC
Figure 1: Li-ion cell voltage versus state of charge [24].
gravity and interactions between the masses and implementsNewtonian gravity and the laws of motion [21β23].
GSA-based optimization has already been used by theresearchers for postoutage bus voltage magnitude calcula-tions, economic dispatch with valve-point effects, optimalsizing and suitable placement for distributed generation(DG) in distribution system, optimization of synthesis gasproduction [25], rectangular patch antenna [26], orthogonalarray based performance improvement [27], solving thermalunit commitment (UC) problem, and finding out optimalsolution for optimal power flow (OPF) problem in a powersystem [28]. Specifically, we are investigating the use of thegravitational search algorithm (GSA) method for developingreal-time and large-scale optimizations for allocating power.
The remainder of this paper is organized as follows: nextsection will describe the specific problem that we are tryingto solve. We will provide the optimization objective andconstraints and mathematical formulation of our algorithm,review the GSA method, and describe how the algorithmworks for our optimization problems. The simulation resultsand analysis are then presented. Finally, conclusions andfuture directions are drawn.
2. Problem Formulation
The idea behind smart charging is to charge the vehicle whenit ismost beneficial such aswhen the electricity price and totalpower demand remain lowest or there is excess capacity ofgenerated power [24].
Suppose there is a charging station with the capacityof total power π. Total π numbers of PHEVs need to becharged within 24 hours of time interval. The proposedsystem should allow PHEVs to leave the charging stationbefore their expected leaving time for making the systemmore effective. It is worth to mention that each PHEV isregarded to be plugged in to the charging station once. Themain aim is to allocate power intelligently for each PHEVcoming to the charging station. The state of charge is themain parameter which needs to be maximized in order toallocate power effectively. For this, the objective functionconsidered in this paper is the maximization of average SoCand thus allocates energy for PHEVs at the next time step.
Mathematical Problems in Engineering 3
The constraints considered are charging time, present SoC,and price of the energy.
The objective function is defined as
max π½ (π) = βπ
π€π(π) SoC
π(π + 1) ,
π€π(π) = π (πΆ
π,π(π) , π
π,π(π) , π·
π(π)) ,
πΆπ,π(π) = (1 β SoC
π(π)) β πΆ
π,
(1)
whereπΆπ,π(π) is the battery capacity (remaining) needed to be
filled for π number of PHEV at time step π; πΆπis the battery
capacity (rated) of the π number of PHEV; remaining timefor charging a particular PHEV at time step π is expressedas ππ,π(π); the price difference between the real-time energy
price and the price that a specific customer at the π number ofPHEV charger is willing to pay at time step π is presented byπ·π(π); π€
π(π) is the charging weighting term of the π number
of PHEV at time step π (a function of charging time, presentSoC, and price of the energy); SoC
π(π+1) is the state of charge
of the π number of PHEV at time step π + 1.Here, the weighting term indicates a bonus proportional
to the attributes of a specific PHEV. For example, if a PHEVhas a lower initial SoC and less charging time (remaining),but the driver is eager to pay a higher price, the system willprovide more power to this particular PHEV battery charger:
π€π(π) πΌ [Cap
π,π(π) + π·
π(π) +
1
ππ,π
(π)] . (2)
The charging current is also assumed to be constant over Ξπ‘:
[SoCπ(π + 1) β SoC
π(π)] β Cap
π= ππ= πΌπ(π) Ξπ‘,
SoCπ(π + 1) = SoC
π(π) +
πΌπ(π) Ξπ‘
Capπ
,
(3)
where the sample time Ξπ‘ is defined by the charging stationoperators and πΌ
π(π) is the charging current over Ξπ‘.
The batterymodel is regarded as a capacitor circuit, whereπΆπis the capacitance of battery (Farad). The model is defined
as
πΆπβ πππ
ππ‘= πΌπ. (4)
Therefore, over a small time interval, one can assume thechange of voltage to be linear:
πΆπβ [ππ(π + 1) β π
π(π)]
Ξπ‘= πΌπ,
ππ(π + 1) β π
π(π) =
πΌπΞπ‘
πΆπ
.
(5)
As the decision variable used here is the allocated powerto the PHEVs, by replacing πΌ
π(π) with π
π(π) the objective
function finally becomes
π½ (π) = βπ€πβ [
[
SoCπ(π) + (2π
π(π) Ξπ‘)
Γ (0.5 β πΆπβ [β
2ππ(π) Ξπ‘
πΆπ
+ π2
π(π)
+ππ(π) ])
β1
]]
]
.
(6)
2.1. SystemConstraints. Possible real-world constraints couldinclude the charging rate (i.e., slow, medium, and fast), thetime that the PHEV is connected to the grid, the desireddeparture SOC, the maximum electricity price that a useris willing to pay, and certain battery requirements. Further-more, the available communication bandwidth could limitsampling time, which would have effects on the processingability of the vehicle.
Power obtained from the utility (πutility) and the maxi-mum power (π
π,max) absorbed by a specific PHEV are theprimary energy constraints being considered in this paper.The power demand of a PHEV/PEV cannot exceed the ratedpower output of the battery charger:
β
π
ππ(π) β€ πutility (π) Γ π,
0 β€ ππ(π) β€ π
π,max (π) .
(7)
The overall charging efficiency of a particular charginginfrastructure is described by π. From the system point ofview, charging efficiency is supposed to be constant at anygiven time step. Maximum battery SoC limit for the π numberof PHEV is SoC
π,max. When SoCπreaches the values close to
SoCπ,max, the π number of battery charger shifts to a standby
mode. The state of charge ramp rate is confined within limitsby the constraint ΞSoCmax. To accommodate the systemdynamics, the energy scheduling is updated when when (i)system utility data is updated; (ii) a new PHEV is pluggedin, and (iii) time period Ξπ‘ has periodically passed. Table 1shows all the objective function parameters that were tunedfor performing the optimization.
Energy allocation to PHEV charging station is subjectedto various constraints as mentioned in the problem formu-lation section. Different constraints make the entire searchspace limited to a particular suitable region. So, a powerfuloptimization algorithm should be implemented in order toachieve high quality solutions with a stable convergence rate.
3. Gravitational Search Algorithm
GSA is an optimization method which has been introducedby Rashedi et al. in the year of 2009 [20]. In GSA, the
4 Mathematical Problems in Engineering
Table 1: Parameter settings of the objective function.
Parameter Values
Fixedparameters
Maximum power, ππ,max = 6.7 kWh
Charging station efficiency, π = 0.9Total charging time, Ξπ‘ = 20 minutes(1200 seconds)Power allocation to each PHEV: 30W
Variables
0.2 β€ state of charge (SoC) β€ 0.8Waiting time β€ 30 minutes (1800seconds)16 kWh β€ battery capacity (πΆ
π) β€
40 kWh
Constraints
β
π
ππ(π) β€ πutility (π) Γ π
0 β€ ππ(π) β€ π
π,max (π)
0 β€ SoCπ(π) β€ SoC
π,max
0 β€ SoCπ(π + 1) β SoC
π(π) β€ ΞSoCmax
specifications of each mass (or agent) are four in total, whichare inertial mass, position, active gravitational mass andpassive gravitational mass. The position of the mass presentsa solution of a particular problem and masses (gravitationaland inertial) are obtained by using a fitness function.GSA canbe considered as a collection of agents (candidate solutions),whose masses are proportional to their value of fitnessfunction. During generations, all masses attract each otherby the gravity forces between them. A heavier mass has thebigger attraction force. Therefore the heavier masses whichare probably close to the global optimum attract the othermasses proportional to their distances.
3.1. Law of Gravity. The law states that particles attract eachother and the force of gravitation between two particles isdirectly proportional to the product of their masses andinversely proportional to the distance between them.
3.2. Law of Motion. The law states that the present velocityof any mass is the summation of the fraction of its previousvelocity and the velocity variance. Variation in the velocityor acceleration of any mass is equal to the force divided byinertia mass.
The gravitational force is expressed as follows:
πΉπ
ππ(π‘) = πΊ (π‘)
πππ(π‘) Γ π
ππ(π‘)
π ππ(π‘) + π
(π₯π
π(π‘) β π₯
π
π(π‘)) , (8)
whereπππis the active gravitational mass related to agent π,
πππis the passive gravitational mass related to agent π, πΊ(π‘)
is gravitational constant at time π‘, π is a small constant, andπ ππ(π‘) is the Euclidian distance between two agents π and π.
The πΊ(π‘) is calculated as follows:
πΊ (π‘) = πΊ0Γ exp(βπΌ Γ iter
max iter) , (9)
where πΌ andπΊ0are descending coefficient and primary value,
respectively, and current iteration and maximum number of
iterations are expressed as iter and max iter. In a problemspace with the dimension π, the overall force acting on agentπ is estimated as the following equation:
πΉπ
π(π‘) =
π
β
π=1,π =π
randππΉπ
ππ(π‘) , (10)
where randπis a random number with interval [0, 1]. From
law of motion we know that an agentβs acceleration is directlyproportional to the resultant force and inverse of its mass, sothe acceleration of all agents should be calculated as follows:
acππ(π‘) =
πΉπ
π(π‘)
πππ(π‘), (11)
where π‘ is a specific time andπππis the mass of the object π.
The velocity and position of agents are calculated as follows:
velππ(π‘ + 1) = rand
πΓ velππ(π‘) + acπ
π(π‘) , (12)
π₯π
π(π‘ + 1) = π₯
π
π(π‘) + velπ
π(π‘ + 1) , (13)
where randπis a random number with interval [0, 1].
Gravitational and inertia masses are simply calculatedby the fitness evaluation. A heavier mass means a moreefficient agent. This means that better agents have higherattractions and walk more slowly. Assuming the equalityof the gravitational and inertia mass, the values of massesare calculated using the map of fitness. We update thegravitational and inertial masses by the following equations:
πππ= πππ= πππ= ππ, π = 1, 2, . . . , π. (14)
In gravitational search algorithm, all agents are initializedfirst with random values. Each of the agents is a candidatesolution. After initialization, velocities for all agents aredefined using (12). Moreover, the gravitational constant,overall forces, and accelerations are determined by (9), (10),and (11), respectively. The positions of agents are calculatedusing (13). At the end, GSAwill be terminated bymeeting thestopping criterion of maximum 100 iterations.The parametersettings for GSA are demonstrated in Table 2. Moreover, GSAflowchart is shown in Figure 2.
4. Simulation Results and Analysis
The GSA algorithm was applied to find out global bestfitness of the objective function (Algorithm 1). All the cal-culations were run on an Intel (R) Core i5-3470M [email protected], 4.00GBRAM,Microsoft 32 bitWindows 7OS, andMATLABΒ© R2013a.
Many optimization algorithms involve local search tech-niques which can get stuck on local maxima. Most searchtechniques strive to find a global maximum in the presence oflocal maxima [29]. One of the most important characteristicsof GSA is its significant performance during explorationprocess.The capability of an algorithm to extend the problemin search gap is known as exploration while the ability of analgorithm to recognize optimal solution near a favorable oneis exploitation [30, 31].
Mathematical Problems in Engineering 5
Primary population generation
Global best and worst population update
Mass and acceleration calculation for each agent
Position and velocity update
Return best solution
No
Yes
Meeting the final criteria?
Each agentβs fitness function evaluation
Figure 2: The GSA flowchart.
0 10 20 30 40 500
100
200
300
400
500
600
700
800
Number of runs
Fitn
ess v
alue
J(k)
Figure 3: Fitness value versus number of runs (50 PHEVs).
0 10 20 30 40 500
100
200
300
400
500
600
700
800
900
Number of runs
Fitn
ess v
alue
J(k)
Figure 4: Fitness value versus number of runs (100 PHEVs).
0 10 20 30 40 500
100
200
300
400
500
600
700
800
Number of runs
Fitn
ess v
alue
J(k)
Figure 5: Fitness value versus number of runs (300 PHEVs).
Table 2: GSA parameter settings.
Parameters ValuesPrimary parameter, πΊ
0100
Number of mass agents, π 100Constant parameter, πΌ 20Constant parameter, π .01Power of βπ β 1Maximum iteration 100Number of runs 50
Figures 3, 4, 5, 6, and 7 show the simulation results for50,100, 300,500, and 1000 plug-in hybrid electric vehicles(PHEVs), respectively, for finding themaximum fitness valueof objective function J. In order to evaluate the performanceand show the efficiency and superiority of the proposedalgorithm, we ran each scenario 50 times in total.
For Figure 3 (50 PHEVs), the maximum best fitness andminimum best fitness were 781.1267 and 0.2191, respectively.
The average best fitness is 158.8289. Figure 4 depictsthe maximum fitness value for 100 PHEVs. In this case,the maximum best fitness and minimum best fitness were579.3955 and 3.2523.The average best fitness is decreased into139.7536.
For Figure 5 (300 PHEVs), the maximum best fitness andminimum best fitness were 743.1251 and 2.3279, respectively.The average best fitness is 172.4296.
Figure 6 depicts the maximum fitness value for 500PHEVs. In this case, themaximumbest fitness andminimumbest fitness were 836.2707 and 0.9818.The average best fitnessis decreased into 152.36437.
Figure 7 shows the maximum fitness value for 1000PHEVs. In this case, themaximumbest fitness andminimumbest fitness were 968.7652 and 7.2747. The average best fitnessis decreased into 161.52349.
6 Mathematical Problems in Engineering
(1) Initialization of totalπmass agents randomly(2) Computation of πΊ(π‘), Fitness (Best and Worst )(3) For each of the agent πΌ, evaluate:
(3.1) Fitnessπ
(3.2) Massπ
(3.3) Force of Massπ
(3.4) Acceleration of Massπ
(3.5) Massπvelocity update
(3.6) New position of Agentπ
If (Probabilityπ>Thershold){
If (Probabilityπ> Rand
π)
Then return Best Fitnesssolution so far
ElseModification of solution
}
(4) Failed to meet stopping criteria,Go To Step 2, Else Stop
Algorithm 1
0 10 20 30 40 500
200
400
600
800
Number of runs
Fitn
ess v
alue
J(k)
Figure 6: Fitness value versus number of runs (500 PHEVs).
Finally, Table 3 summarizes the result. From that it can beconcluded that average best fitness remains almost in similarpattern for five (05) different scenarios.
4.1. Performance Evaluation of GSA
4.1.1. Convergence Analysis. It can be apparently seen thatalthough the algorithm has been set to run for maximum 100iterations, the convergence happened in about 20 iterations.The result derived in this paper reveals that each object ofthe standard GSA converges to a stable point. Here, theassumption was that the gravitational and inertia masses arethe same. However, for some applications different valuesfor them can be used. A heavier inertia mass provides aslower motion of agents in the search space and hence a more
0 10 20 30 40 500
200
400
600
800
1000
Number of runs
Fitn
ess v
alue
J(k)
Figure 7: Fitness value versus number of runs (for 1000 PHEVs).
precise search [20]. On the contrary, a heavier gravitationalmass causes a higher attraction of agents. This allows a fasterconvergence. The analysis results confirm the convergencecharacteristics of GSA according to the given parametersranges of the algorithm. Figures 8, 9, 10, 11, and 12 showthe convergence behavior of GSA. The best fitness functionshows convergences after the same iterations (35 iterations)for both 50 and 100 numbers of PHEVs while for 500 and1000 numbers of PHEVs, it shows early convergence (before20 iterations).
4.1.2. Robustness. The similar numeric patterns of averagebest fitness show the robustness of GSAmethod.Thismethodresists change without adapting its initial stable configuration
Mathematical Problems in Engineering 7
Table 3: Fitness evaluation of GSA.
Fitness function For 50 PHEVs For 100 PHEVs For 300 PHEVs For 500 PHEVs For 1000 PHEVsπ½(π)Maximum best fitness 781.1267 579.3955 743.1251 836.2707 968.7652Average best fitness 158.8289 139.7536 172.4296 152.36437 161.52349Minimum best fitness 0.2191 3.2523 2.3279 0.9818 7.2747
100 20 30 40 50 60 70 80 90 100Iteration
β108.15
β108.2
β108.25
β108.3
β108.35
β108.4
β108.45
β108.5
β108.55
β108.6
β108.65
Best
fitne
ssJ(k)
Figure 8: Best fitness versus iteration (50 PHEVs).
0 10 20 30 40 50 60 70 80 90 100Iteration
β106.9
β107
β107.1
β107.2
β107.3
β107.4
β107.5
β107.6
β107.7
β107.8
β107.9
Best
fitne
ssJ(k)
Figure 9: Best fitness versus iteration (100 PHEVs).
for different cases (number of PHEVs) which proves GSA asa robust algorithm.
4.1.3. Diversity. Here, the average best fitness gives differentvalues with the increment of PHEVs population. The rateof convergence of mass agents in GSA is good through the
0 10 20 30 40 50 60 70 80 90 100Iteration
β12.316
β12.318
β12.32
β12.322
β12.324
β12.326
β12.328
β12.33
β12.332
β12.334
Best
fitne
ssJ(k)
Figure 10: Best fitness versus iteration (300 PHEVs).
0 10 20 30 40 50 60 70 80 90 100 Iteration
β149.6
β149.7
β149.8
β149.9
β150
β150.1
β150.2
β150.3
β150.4
β150.5
Best
fitne
ssJ(k)
Figure 11: Best fitness versus iteration (500 PHEVs).
fast information flowing among mass agents, so its diversitydecreases very quickly in the successive iterations and leadsto a suboptimal solution.
4.1.4. Computational Cost. Here, we measured the computa-tional cost of the algorithm in terms of total running time.Table 6 shows the computational time ofGSA for five differentscenarios. Here, average CPU time is measured in seconds.
8 Mathematical Problems in Engineering
0 10 20 30 40 50 60 70 80 90 100Iteration
β160.6
β160.65
β160.7
β160.75
β160.8
β160.85
β160.9
β160.95
Best
fitne
ssJ(k)
Figure 12: Best fitness versus iteration (1000 PHEVs).
As GSA needs a good number of parameter tuning, thecomputational cost increases with the increment of the totalnumber of PHEVs.
4.1.5. Quality of Solution. When an algorithm finds an opti-mal solution to a given problem, one of the important factorsis speed and rate of convergence to the optimal solution.For heuristics, the additional consideration of how close theheuristic solution comes to optimal is generally the primaryconcern of the researcher [32]. InGSA, the faster convergenceand better exploitation rate ensure good quality solution,which is best fitness function.
For this optimization, the initial state of charge wasexpressed as a random number which is continuous anduniform between 0.2 and 0.6.The sample timewas set around1200 seconds (20 minutes). The remaining charge time wasdefined as continuous random number between 0 and 6hours. The price according to customerβs choice for payingthe bill for electricity was expressed as a continuous randomnumber which is in between $1 and $2.
The capacity of the battery was assumed to be identicalfor all vehicles.
4.2. Comparison between GSA and PSO. Particle swarmoptimization (PSO) with the parameter settings stated inTable 4 was also performed for the same objective functionand compared with the performance of gravitational searchalgorithm in terms of average best fitness. The swarm sizeand maximum iterations were set exactly the same as thoseof GSA algorithm for the comparison purpose. The values ofparameters c1, c2, and π€ were set as standard values, 1.4, 1.4,and 0.9, respectively.
From Figure 13 it is clear that gravitational search algo-rithm outperformed particle swarm optimization in terms ofaverage best fitness. Starting from 50 numbers of PHEVs upto 1000 PHEVs, GSA shows better fitness value than PSO.
Table 4: PSO parameter settings.
Parameters ValuesSize of the swarm 100Maximum number of steps 100PSO parameter, π1 1.4PSO parameter, π2 1.4PSO inertia (π€) 0.9Maximum iteration 100Number of runs 50
Table 5: Summarizes the comparisons of GSA with PSO algorithmin terms of average best fitness.
Average best fitness for PSO GSA50 PHEVs 142.839 158.8289100 PHEVs 171.102 182.3097300 PHEVs 169.312 172.4296500 PHEVs 150.869 152.364371000 PHEVs 156.802 161.52349
Table 6: Computational time for PSO and GSA.
Computational Time (sec.) PSO GSA50 PHEVs 1.650 2.721100 PHEVs 1.686 4.439500 PHEVs 1.990 18.1651000 PHEVs 2.398 36.275
Table 7 illustrates the advantages and disadvantages ofboth GSA and PSO for solving different optimization prob-lems.
It has been proven that gravitational search algorithmhas good ability to search for the global optimum, but itsuffers from slow searching speed in the last iterations [35].Moreover, the inertia mass is against the motion and slowsthe mass movement. Agents with heavy inertia mass moveslowly and hence search the space more locally. So, it canbe considered as an adaptive learning rate [34]. GSA is amemory-less algorithm. However, it works competently likethe algorithmswithmemory.Our simulation results show thegood convergence rate of the GSA.
5. Conclusion and Recommendations
In this paper, gravitational search algorithm- (GSA-) basedoptimization was performed in order to optimally allocatepower to each of the PHEVs entering into the chargingstation. A sophisticated controller will need to be designedin order to allocate power to PHEVs appropriately. Forthis wake, the applied algorithm in this paper is a steptowards real-life implementation of such controller for PHEVcharging infrastructures.
Here, five (05) different numbers of PHEVs were con-sidered for MATLABΒ© simulation and then obtained resultswere compared with PSO in terms of average best fitness.The
Mathematical Problems in Engineering 9
Table 7: Advantages and disadvantages of PSO and GSA.
Optimizationmethod Advantages Disadvantages
PSO
Less parameters tuningEasy constraint
Good for multiobjectiveoptimization
[33]
Low quality solutionNeeds memory to update velocity
Slow convergence rate
GSAHigh quality solutionGood convergence rate
Local exploitation capability [34]
Needs more computational timeMore parameters tuning
50 100 300 500 10000
20
40
60
80
100
120
140
160
180
200
GSAPSO
Number of PHEVs
Aver
age b
est fi
tnes
s
Figure 13: Average best fitness versus number of PHEVs.
success of the electrification of transportation sector solelydepends on charging infrastructure. Only proper chargingcontrol and infrastructure management can assure the largerpenetration of PHEVs.The researchers should try to developefficient control mechanism for charging infrastructure inorder to facilitate upcoming PHEVs penetration in highways.In future, more vehicles should be considered for intelligentpower allocation strategy and improved versions of GSA andhybrid swarm intelligence based methods should be appliedto ensure low computational cost.
Nomenclature and Acronyms
PHEVs: Plug-in hybrid electric vehiclesEPRI: Electric Power Research InstituteV2G: Vehicle to gridSoC: State of chargeICEVs: Internal combustion engine vehiclesAEVs: All-electric vehiclesHEVs: Hybrid electric vehiclesAER: All-electric rangeπΌπ(π): Charging current over Ξπ‘
πΆπ,π(π): Remaining battery capacity required to be
filled for πth PHEV at time step ππΆπ: Rated battery capacity of the πth PHEV
(Farad)ππ,π(π): Remaining time for charging the πth PHEV
at time step ππ·π(π): Price difference
π€π(π): Charging weighting term of the πth PHEV
at time stepSoCπ(π + 1): State of charge of the πth PHEV at time step
π + 1
SoCπ,max: User-defined maximum battery SoC limit
for the πth PHEVπutility: Power available from the utilityππ,max: Maximum power that can be absorbed by a
specific PHEVπ: Overall charging efficiency of the charging
station.
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper.
Acknowledgments
The authors would like to thank Universiti TeknologiPetronas for supporting the research under Graduate Assis-tance Scheme. This research paper is financially sponsoredby the Centre of Graduate study with the support of theDepartment of Fundamental & Applied Sciences, UniversitiTeknologi Petronas.
References
[1] On certain integrals International Energy Agency (lEA), Trans-port, Energy and C02-Moving Towards Sustainability, 2009.
[2] D. Holtz-Eakin and T. M. Selden, βStoking the fires? CO2
emissions and economic growth,β Journal of Public Economics,vol. 57, no. 1, pp. 85β101, 1995.
[3] S. F. Tie and C. W. Tan, βA review of energy sources andenergy management system in electric vehicles,β Renewable andSustainable Energy Reviews, vol. 20, pp. 82β102, 2013.
[4] βEnvironmental assessment of plug-in hybrid electric vehicles,Volume 1: Nationwide greenhouse gas emissions,β Tech. Rep.
10 Mathematical Problems in Engineering
1015325, Electric Power Research Institute (EPRI), Palo Alto,Calif, USA, 2007.
[5] H. Lund and W. Kempton, βIntegration of renewable energyinto the transport and electricity sectors through V2G,β EnergyPolicy, vol. 36, no. 9, pp. 3578β3587, 2008.
[6] A. R. Hota, M. Juvvanapudi, and P. Bajpai, βIssues and solutionapproaches in PHEV integration to the smart grid,β Renewableand Sustainable Energy Reviews, vol. 30, pp. 217β229, 2014.
[7] J. Soares, T. Sousa, H. Morais, Z. Vale, B. Canizes, and A. Silva,βApplication-Specific Modified Particle Swarm Optimizationfor energy resource scheduling considering vehicle-to-grid,βApplied Soft Computing Journal, 2013.
[8] W. Su and M.-Y. Chow, βComputational intelligence-basedenergy management for a large-scale PHEV/PEV enabledmunicipal parking deck,β Applied Energy, vol. 96, pp. 171β182,2012.
[9] W. Su and M. Y. Chow, βPerformance evaluation of a PHEVparking station using particle swarm optimization,β in Proceed-ings of the IEEE PES General Meeting, pp. 1β6, July 2011.
[10] W. Su and M.-Y. Chow, βPerformance evaluation of an EDA-based large-scale plug-in hybrid electric vehicle charging algo-rithm,β IEEE Transactions on Smart Grid, vol. 3, no. 1, pp. 308β315, 2012.
[11] K. Morrow, D. Karner, and J. Francfort, βPlug-in hybrid electricvehicle charging infrastructure review,β Tech. Rep., The IdahoNational Laboratory, 2008.
[12] βInvestigation into the scope for the transport sector to switchto electric vehicles and plug- in hybrid vehicles,β Tech. Rep.,Department for Transport, London, UK, 2008.
[13] D.Mayfield, βSite Design for Electric Vehicle Charging Stations,ver. 1.0,β Sustainable Transportation Strategies, July 2012.
[14] G. Boyle, Renewable Electricity and the Grid: The Challenge ofVariability, Earthscan Publications, London, UK, 2007.
[15] A. Hess, F. Malandrino, M. B. Reinhardt, C. Casetti, K. A.Hummel, and J. M. Barcelo-Ordinas, βOptimal deployment ofcharging stations for electric vehicular networks,β inProceedingsof the 1st ACM Conference on Urban Networking (UrbaNe β12),pp. 1β6, New York, NY, USA, December 2012.
[16] Z. Li, Z. Sahinoglu, Z. Tao, and K. H. Teo, βElectric vehicles net-work with nomadic portable charging stations,β in Proceeding ofthe IEEE 72nd Vehicular Technology Conference Fall (VTC '10),pp. 1β5, Ottawa, Canada, September 2010.
[17] T. Ghanbarzadeh, S. Goleijani, and M. P. Moghaddam, βReli-ability constrained unit commitment with electric vehicle togrid using hybrid particle swarm optimization and ant colonyoptimization,β in Proceeding of the IEEE PES General Meeting:The Electrification of Transportation and the Grid of the Future,pp. 1β7, San Diego, Calif, USA, July 2011.
[18] A. Shafiei and S. S. Williamson, βPlug-in hybrid electric vehiclecharging: current issues and future challenges,β in Proceedings ofthe IEEE Vehicle Power and Propulsion Conference (VPPC β10),pp. 1β8, September 2010.
[19] J. Yan, C. Li, G. Xu, and Y. Xu, βA novel on-line self-learningstate-of-charge estimation of battery management system forhybrid electric vehicle,β in Proceedings of the IEEE IntelligentVehicles Symposium, pp. 1161β1166, June 2009.
[20] E. Rashedi, H. Nezamabadi-pour, and S. Saryazdi, βGSA: agravitational search algorithm,β Information Sciences, vol. 179,no. 13, pp. 2232β2248, 2009.
[21] D. Martens, B. Baesens, and T. Fawcett, βEditorial survey:swarm intelligence for data mining,βMachine Learning, vol. 82,no. 1, pp. 1β42, 2011.
[22] J. Krause, J. Cordeiro, R. S. Parpinelli, and H. S. Lopes, βAsurvey of swarm algorithms applied to discrete optimizationproblems,β in Swarm Intelligence and Bio-inspired Computation:Theory and Applications, pp. 169β191, Elsevier Science & Tech-nology Books, 2013.
[23] O. Okobiah, S. P. Mohanty, and E. Kougianos, βGeostatisticsinspired fast layout optimization of nanoscale CMOS phaselocked loop,β in Proceedings of the 14th International Symposiumon Quality Electronic Design (ISQED β13), pp. 546β551, SantaClara, Calif, USA, March 2013.
[24] W.-Y. Chang, βThe state of charge estimating methods forbattery: a review,β ISRN Applied Mathematics, vol. 2013, ArticleID 953792, 7 pages, 2013.
[25] T. Ganesan, I. Elamvazuthi, K. Z. Ku Shaari, and P. Vas-ant, βSwarm intelligence and gravitational search algorithmfor multi-objective optimization of synthesis gas production,βApplied Energy, vol. 103, pp. 368β374, 2013.
[26] O. T. Altinoz and A. E. Yilmaz, βCalculation of optimizedparameters of rectangular patch antenna using gravitationalsearch algorithm,β in Proceedings of the International Sympo-sium on INnovations in Intelligent SysTems and Applications(INISTA β11), pp. 349β353, IEEE, June 2011.
[27] O. T. Altinoz, A. E. Yilmaz, and G. W. Weber, βOrthogonalarray based performance improvement in the gravitationalsearch algorithm,β Turkish Journal of Electrical Engineering andComputer Sciences, vol. 21, no. 1, pp. 174β185, 2013.
[28] N. M. Sabri, M. Puteh, and M. R. Mahmood, βA review of grav-itational search algorithm,β International Journal of Advances inSoft Computing and Its Applications, vol. 5, no. 3, pp. 1β39, 2013.
[29] R. E. Haskell, G. Castelino, and B. Mirshab, βEfficient algorithmfor locating the global maximum of an arbitrary univariatefunction,β Journal of Forth Application and Research, vol. 5, no.3, pp. 357β364, 1989.
[30] M. Udgir, H. M. Dubey, and M. Pandit, βGravitational searchalgorithm: a novel optimization approach for economic loaddispatch,β in Proceedings of the Annual International Confer-ence on Emerging Research Areas and International Conferenceon Microelectronics, Communications and Renewable Energy(AICERA/ICMiCR β13), pp. 1β6, 2013.
[31] A. A. Bazmi and G. Zahedi, βSustainable energy systems: roleof optimization modeling techniques in power generation andsupplyβa review,β Renewable and Sustainable Energy Reviews,vol. 15, pp. 3480β3500, 2011.
[32] R. S. Barr, B. L. Golden, J. P. Kelly, M. G. C. Resende, andW. R. Stewart Jr., βDesigning and reporting on computationalexperiments with heuristic methods,β Journal of Heuristics, vol.1, no. 1, pp. 9β32, 1995.
[33] Y.-T. Kao and E. Zahara, βA hybrid genetic algorithm andparticle swarm optimization formultimodal functions,βAppliedSoft Computing, vol. 8, pp. 849β857, 2008.
[34] I. Rahman, P. M. Vasant, B. S. M. Singh, and M. Abdullah-Al-Wadud, βIntelligent energy allocation strategy for PHEVcharging station using gravitational search algorithm,β AIPConference Proceedings, vol. 1621, pp. 52β59, 2014.
[35] S. Mirjalili, S. Z. Mohd Hashim, and H. Moradian Sardroudi,βTraining feedforward neural networks using hybrid parti-cle swarm optimization and gravitational search algorithm,βApplied Mathematics and Computation, vol. 218, no. 22, pp.11125β11137, 2012.
Submit your manuscripts athttp://www.hindawi.com
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttp://www.hindawi.com
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com
Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Stochastic AnalysisInternational Journal of