Professional Documents
Culture Documents
A new smart charging method for EVs for frequency control of smart grid
Saber Falahati a, Seyed Abbas Taher a,⇑, Mohammad Shahidehpour b
a
Department of Electrical Engineering, University of Kashan, Kashan, Iran
b
Department of Electrical and Computer Engineering, Illinois Institute of Technology, United States
a r t i c l e i n f o a b s t r a c t
Article history: Nowadays, due to the shortage of fossil fuels on the one hand and their high prices on the other hand,
Received 29 June 2015 using electric vehicles (EVs) has been increased. Charging of EVs has imposed new loads on power sys-
Received in revised form 15 March 2016 tems. These new and major loads have faced the frequency control and stability of power systems with
Accepted 4 April 2016
new challenges. One way to deal with this new challenge is smart charging of EVs. In this method, grid
condition is a key parameter that affects the charging of EV. In other words, in smart charging method,
charging is performed with respect to power system parameters such as frequency. In this paper, a smart
Keywords:
charging method based on fuzzy controller is proposed, in which charging process is performed with
Smart grid
Smart charging
respect to the frequency deviation of grid and state of charge (SOC) of EV battery. To evaluate the perfor-
Fuzzy control mance of the proposed controller in control of grid frequency, IEEE 39-bus system in the presence of
Frequency control renewable energy sources is considered as test system. In order to the frequency analysis, this system
Electric vehicle is converted into a three-area system and, for each area, several EV categories with different numbers
Renewable energy resources of EVs, battery capacity, start time of charging, and initial SOC are supposed. Moreover performance of
proposed method is compared with an optimized PI controller in terms of frequency control. To investi-
gate performance of proposed method in charging of EVs, a two area system is assumed and charging of
EVs is verified by applying step loads to both areas. Simulations are carried out in MATLAB/SIMULINK
environment. Results of the simulations reveal the good performance of the proposed controller in terms
of frequency control of grid and charging of EVs.
Ó 2016 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2016.04.039
0142-0615/Ó 2016 Elsevier Ltd. All rights reserved.
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 459
is used for this purpose, as in most of the conventional power sys- area i and synchronous torque factor between areas i and j,
tems. The parameters of PI are obtained by imperialist competitive respectively. DPTie;i and Tij are obtained from the following equa-
algorithm (ICA) [20]. ICA has been used in various papers owing to tions [19]:
its high speed and accuracy in finding the solutions of optimization 2 3
problems; theory of this method has been described in detail in the XN X N XN
2p 6 7
related literatures [21–23]. In order for simplicity, conventional DPTie;i ¼ DPtie;ij ¼ 4 T ij Df i T ij Df j 5 ð3Þ
generation resources are assumed as non-reheat steam units. j¼1
s j¼1 j¼1
j–i j–i j–i
Governor and turbine units are modeled as follows:
1 jV i jjV j j
GGov ðsÞ ¼ ð1Þ T ij ¼ cosðd0i d0j Þ ð4Þ
1 þ sT g X ij
1 In Eq. (4), Vi, di , and Xij show terminal voltage and phase of area i
GTur ðsÞ ¼ ð2Þ and reactance between area i and j, respectively. In Fig. 6, R, D, H,
1 þ sT t
and b are droop characteristic, equivalent damping coefficient,
where Tg and Tt are governor and turbine time constants, respec- inertia constant, and frequency bias, respectively. Frequency bias
tively. In Fig. 6, DP Tie;i and Tij represent tie line power deviation of is obtained using Eq. (5):
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 461
G10 G8 `
PV 23
<30> <37> Area 2 <26>
<29>
<28>
<2> <25>
PV 21 <27> <38>
PV 22
<17>
G9
<1> <18>
EVs 23
EVs 21 <24>
<3>
G6
EVs 22 <16> EVs 33 <35>
<4> EVs 32
Area 1 <15>
<5> PV 14 <22>
PV 32
G1
<12> EVs 31
<6>
<39> <14> <19>
<7> <11> <23>
<13>
<8> PV 13 <20>
<36>
<10> <33>
EVs 12
<9>
EVs 13 PV 31
<31> G7
PV 12 <34> G4
<32>
EVs 11 G2 G5
Area 3
PV 11 G3
Fig. 5. Modified IEEE 39 bus system in the presence of photovoltaic systems and EVs charging stations.
1 1
b¼Dþ ð5Þ
R 0.9
Since the objective of this study is to investigate the perfor- 0.8
mance of the proposed fuzzy controller for smart charging of
0.2
V\0 E\d
I¼ ð7Þ 0.1
jX
0
In (6) and (7), SEV and I are the power and current received by 0 2 4 6 8 10 12 14 16 18 20
the battery, respectively. V and E are the voltages at the sending Time (s)
and receiving ends, respectively. d is the angle between E and V. B
X is the line reactance between the converter and the utility node.
It can be demonstrated that active and reactive powers are 1
obtained as [24]:
0.9
VE sinðdÞ 0.8
PEV ¼ ð8Þ
X
PV output power (pu)
0.7
0.6
E2 VE cosðdÞ
Q EV ¼ ð9Þ 0.5
X
0.4
In this paper, it is assumed that unidirectional converters are
controllers so that QEV = 0. It is supposed that large dynamics exist 0.3
in system and so power electronics converters are not modeled. It 0.2
has been assumed that there are high speed communication links
0.1
between charging stations and smart grid and the delay can be
neglected due to the high speed communication links. 0
0 2 4 6 8 10 12 14 16 18 20
Time (s)
Design of controller
C
Fundamentals of fuzzy control Fig. 7. Aggregated output power of photovoltaic systems of each area; (A) area 1,
(B) area 2, (C) area 3.
Unlike binary logic, fuzzy logic is a many-valued logic, in which
the fuzzy logic variables may have truth values ranging in different
degrees between 0 and 1, known as their membership grade. Fuzzy logic, and a defuzzification interface. The fuzzification interface
logic can deal with the uncertainties in the system through a sim- converts the binary logic inputs into fuzzy variables, while the
ple IF-THEN rule-based approach, thereby eliminating the need for defuzzification interface converts the fuzzy variables into binary
a mathematical model of the system. This issue is especially useful logic outputs. This conversion is achieved by means of a member-
in complex systems, for which a complete mathematical model ship function. The rule base is a collection of IF-THEN rules that
representation may not be possible. However, the fuzzy logic- describe the control strategy. The output from each rule in the rule
based system complexity increases rapidly with more number of base is deduced by the inference logic to arrive at a value for each
inputs and outputs. A fuzzy logic control (FLC) consists of four prin- output membership function. In Fig. 9, a typical fuzzy controller is
cipal components: a fuzzification interface, a rule base, inference shown [24].
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 463
Design of the proposed controller six fuzzy regions represented by linguistic variables; very high
(VH), high (H), medium high (MH), medium low (ML), low (L),
As mentioned before, in this paper, a controller based on fuzzy and very low (VL). It is assumed that EV users do not use their cars
logic is used for the control of EV charging in order for frequency for long distances so 50% SOC is minimum value that required for
control of grid. In this method, each EV is connected to the grid them. This assumption does not question the proposed method.
by a unidirectional converter. The scheme of the proposed method Membership functions for the output of fuzzy controller have been
is illustrated in Fig. 10. In this method, the frequency deviation of obtained by both analysis and trial and error. Fuzzy rules of fuzzy
grid and current battery SOC of EV are considered the inputs to controller are given in Table 1. Fuzzy rule i in this controller is
fuzzy controller. In the next step, with respect to these inputs expressed as follows [25]:
and the corresponding membership functions and fuzzy rules,
Rule i : IF Df is Lx and SOC is M y ; THEN E is Z l ð11Þ
the input power to each EV is determined. Membership functions
of frequency deviation, SOC and power are represented in Fig. 11
x ¼ 1; 2; . . . ; 5; y ¼ 1; 2; . . . ; 5; l ¼ 1; 2; 3; . . . ; 6
(A)–(C), respectively. It is noteworthy to say that membership
functions of frequency deviation and SOC have been optimized where Lx and My show the antecedents and Zl is consequent part.
by ICA algorithm. Four parameters of frequency deviation mem- Fuzzy controller’s output E is obtained by:
bership functions as a, b, c and two parameters of SOC membership ,
X
25 X
25
function as e and f have been optimized with respect to following EI ¼ xi Z l xi ð12Þ
fitness function: i¼1 i¼1
Z
where xi denotes the grade for the antecedent and is calculated by:
J¼ ðjDf 1 j þ jDf 2 j þ jDf 3 jÞtdt ð10Þ
xi ¼ xDfi xSi ð13Þ
After optimization, following results were obtained:
where xDfi and xSi are the grade of antecedents for each rule.
a ¼ 0:729985; b ¼ 0; c ¼ 0:001923; d ¼ 0:760524; To define rules of Table 1 analysis has been used. For example
when frequency deviation of grid is negative low (L) and SOC of
e ¼ 27:853731; f ¼ 80;
EV battery is high (H) it is obvious that EV charging power should
Input frequency deviation is fuzzified into the corresponding be very low (VL). But when frequency deviation of grid is very high
fuzzy signals with five linguistic variables; very low (VL), low (L), (VH) and SOC of EV battery is low (L) it can be understood that
medium (M), high (H), and very high (VH). Input SOC is fuzzified charging power of battery should be very high (VH) to lower fre-
into the fuzzy regions with the same membership function names quency deviation of grid and increase SOC of EV battery. Other
as the frequency deviation. The output of controller is fuzzified into rules of Table 1 have been obtain by same process.
464 S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469
C
Fig. 11. Membership functions; (A) frequency deviation, (B) SOC, (C) power.
Table 1
Fuzzy rules of controller.
SOC/Df VL L M H VH
VL VL L ML H VH
L VL L L MH VH
M VL VL L MH H
H VL VL L MH MH
VH VL VL VL VL VL
Fig. 12. Lumped load of areas; (A) area 1, (B) area 2, (C) area 3.
Simulation results and discussion
begin to charge simultaneously. 3 kW power limit has been sup-
Performance of proposed method on frequency control posed for charging of EVs. It is assumed that each area has three
types of EVs with 10, 12, and 16 kW h battery capacities. In Table 2,
There are several charging strategies for charging EVs, namely the number of EVs, initial SOC, and start time of charging for all
dumb charging, dual-tariff, smart charging, V2G charging, etc. In types of EVs in each area is given. It is noteworthy to say that
dumb charging, typically after the last trip of the day or when a the results of simulations are shown from the time of connection
charging point is available, EVs connect to the gird and, without of the first EVs to grid and 0 in the simulations refers to this fact.
any consideration, begin to charge. In this paper, to investigate Frequency and tie line power deviations of each area with pro-
the proposed fuzzy controller for smart charging of EVs in a smart posed controller and with dumb charging have been shown for
grid in order for frequency control of grid, modified IEEE 39 bus 0 6 t 6 50 s in Figs. 13 and 14, respectively. It can be seen in these
system in the presence of solar energy and EVs is considered and figures that by using of proposed controller for smart charging of
EVs charging is carried out with the proposed smart charging EVs maximum frequency and tie line power deviations at areas
method and dumb charging. For each area, a lumped variable load have been reduced. Results of the simulations reveal that by pro-
is assumed. These loads are represented in Fig. 12. For simplicity in posed method maximum frequency deviations of areas 1 to 3 are
this paper, parameters such as daily travel distance, EV connectiv- decreased by 36%, 45% and 22%, respectively, compared with dumb
ity, and charging station technologies are not considered [3]. It is charging. To more investigate this issue RMS value has been
also supposed that EVs with the same battery capacity in each area defined as below
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 465
Table 2
EVs in each area.
0.08
0.1
0.06
0.02
0
0
-0.05 -0.02
-0.04
Proposed method
-0.1 Proposed method
-0.06
Dumb charging
Dumb charging
-0.08
-0.15 0 10 20 30 40 50
0 10 20 30 40 50
Time (s)
Time (s)
A
A
0.05 0.04
0.03
Tie line power deviation (pu)
0.02
Frequency deviation (Hz)
0.01
-0.05 0
-0.01
-0.1 -0.02
Proposed method Proposed method
-0.03
Dumb charging Dumb charging
-0.15 -0.04
0 10 20 30 40 50
0 10 20 30 40 50
Time (s) Time (s)
B B
0.1 0.06
0.04
Tie line power deviation (pu)
0.05
Frequency deviation (Hz)
0.02
0 0
-0.02
-0.05
-0.04
Table 3
Maximum and RMS values of frequency and tie line power deviations.
0.3
0.2
Frequency deviation (Hz)
0.1
-0.1
-0.2
Proposed method
-0.3
Optimized PI
-0.4
0 10 20 30 40 50
Time (s)
A
0.3
0.2
Frequency deviation (Hz)
0.1
-0.1
-0.2
Proposed method
-0.3
Optimized PI
-0.4
0 10 20 30 40 50
Time (s)
B
0.3
0.2
Frequency deviation (Hz)
0.1
-0.1
-0.2
Proposed method
-0.3
Optimized PI
-0.4
0 10 20 30 40 50
Time (s)
C
Fig. 15. Comparing the proposed method and optimized PI controller in terms of Fig. 16. Comparing the proposed method and optimized PI controller in terms of tie
frequency deviations; (A) area 1, (B) area 2, (C) area 3. line power deviations; (A) area 1, (B) area 2, (C) area 3.
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 467
Table 4
Comparing the proposed method and optimized PI controller in terms of maximum and RMS values of frequency and tie-line power deviations.
-1500
system considered as above but in absence of solar energies. In
Fig. 15(A)–(C) frequency deviations of area 1 to 3 have been repre- -2000
sented by using of proposed method and optimized PI controller.
As can be seen from these figures by using of proposed method fre- -2500
quency deviations of all areas are reduced than optimized PI con- -3000 Area 1
troller. In Fig. 16(A)–(C) tie-line power deviations of all areas by
aforementioned methods have been shown. These figures show -3500
Area 2
better response of proposed method than optimized PI controller
-4000
too. In Table 4 maximum and RMS values of frequency deviation 0 1000 2000 3000 4000 5000 6000 7000
have been given. It can be found that by proposed method maxi- TIme (s)
mum frequency deviations of areas 1 to 3 are reduced by 31%, A
30% and 32% respectively, and RMS values of frequency deviations
-900
105
-950
Total power (kW)
100 -1000
SOC (%)
-1050
95
Area 1
-1100
90 Area 2
EVs of area 1
-1150
EVs of area 2 1800 1800.5 1801 1801.5 1802 1802.5 1803
TIme (s)
85
0 1000 2000 3000 4000 5000 6000 7000
B
TIme (s)
Fig. 18. Total absorbed power; (A) between 0 and 8100 s, (B) a zoomed view of
Fig. 17. SOCs of EVs. absorbed power when loads disconnected.
468 S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469
0.2 Appendix A
System parameters
0
Area 1
Frequency deviation (Hz)
-0.2 R1 = 3 Hz/s
Tg1 = 0.08 s
Tt1 = 0.4 s
-0.4
D1 = 0.0084 pu/Hz
H1 = 0.1677/2 pu s
Area 1 T12 = 0.20 pu/Hz
-0.6
T13 = 0.25 pu/Hz
Area 2
-0.8
0 5 10 15 20 Area 2
Time (s)
R2 = 2.73 Hz/s
Fig. 19. Frequency deviations of area 1 and area 2 in first 20 s. Tg2 = 0.06 s
Tt2 = 0.44 s
D2 = 0.016 pu/Hz
100% faster than EVs in area 2. In Fig. 17 SOCs of EVs in area 1 and
H2 = 0.2017/2 pu s
area 2 have been shown. Battery SOC is calculated by:
T21 = 0.20 pu/Hz
Q1 Q2 T23 = 0.12 pu/Hz
SOC ¼ ð15Þ
Q1
Area 3
where Q1 represents battery capacity and Q2 is output energy. It can
be seen that SOCs of EVs of area 1 which have 90% initial SOC, reach R3 = 2.82 Hz/s
to 100% after elapsing time of approximately 1.5 h and SOCs of area Tg3 = 0.07 s
2 with 85% initial SOC reach to 100% after approximately 2 h. Tt3 = 0.30 s
The total energy that EVs in area 1 absorb from grid for full D3 = 0.015 pu/Hz
charging is 1520 kW h and for EVs in area 2 is 2160 kW h. In H3 = 0.1247/2 pu s
Fig. 18(A) power absorbed by EVs in area 1 and 2 have been shown T31 = 0.25 pu/Hz
and in Fig. 18(B) a zoomed view of total absorbed power changes T32 = 0.12 pu/Hz
has been presented when loads of area 1 and 2 at 1800 s are dis- Sbase = 100 MW
connected. By computing areas of figures in Fig. 18(A) it can be
found that 1520 kW h energy is sent to area 1 and area 2 receives Parameters of PI controllers
2160 kW h energy from grid. It can be seen that changes of power
absorbed in two areas is different from each other. This is because K1(s) = 0.33–0.37/s
frequency deviations and initial SOCs of EVs’ batteries in areas 1 K2(s) = 0.36–0.34/s
and 2 are different and so according to fuzzy rules of Table 1, K3(s) = 0.25–1/s
absorbed powers become different. In order to comparison fre-
quency deviations of areas 1 and 2 have been displayed in Fig. 19
for first 20 s. References
[1] Pillai JR. Electric vehicle based battery storages for large scale wind power
integration in Denmark [Ph.D. thesis]. Aalborg, Denmark, December 2010.
Conclusion
[2] Dán A, Farkas C, Prikler L. V2G effects on frequency regulation and under-
frequency load shedding in a quasi-islanded grid. In: PowerTech, IEEE
In this paper, a new approach based on fuzzy control was pro- Grenoble, 2013. p. 1–6.
posed for the smart charging of EVs. In this approach, unidirec- [3] Garcia-Valle R, Pecas Lopes JA. Electric vehicle integration into modern power
networks. Springer; 2013.
tional converters were used instead of bidirectional converters in [4] Jiang H, Lin J, Song Y, Gao W, Xu Y, Shu B, et al. Demand side frequency control
the vehicle to grid concept. To investigate the proposed method, scheme in an isolated wind power system for industrial aluminum smelting
IEEE 39 bus system in the presence of solar energy resources was production. IEEE Trans Power Syst 2014;29(2):844–53.
[5] Molina-Garciá A, Bouffard F, Kirschen DS. Decentralized demand-side
considered in case study 1. Results of the simulations revealed that, contribution to primary frequency control. IEEE Trans Power Syst 2011;26
by the proposed method, maximum frequency deviations of areas (1):411–9.
1 to 3 was decreased by 36%, 45%, and 22%, respectively, in com- [6] Molina-Garciá A, Bouffard F, Kirschen DS. Demand-side contribution to
primary frequency control with wind farm auxiliary control. IEEE Trans
parison to dumb charging. Moreover, it was illustrated that, using Power Syst 2014;29(5):2391–9.
the proposed method, RMS values of frequency deviations in areas [7] Kempton W, Letendre SE. Electric vehicles as anew power source for electric
1 to 3 were reduced by 60%, 59%, and 63%, respectively, compared utilities. Transp Res D 1997;2(3):157–75.
[8] Ota Y, Taniguchi H, Nakajima T, Liyanage KM, Baba J, Yokoyama A.
with the case of using dumb charging. In another case study, per- Autonomous distributed V2G (Vehicle-to-Grid) satisfying scheduled
formance of proposed method is compared with an optimized PI charging. IEEE Trans Smart Grid 2012;3(1).
controller. The results of the simulations illustrated good perfor- [9] Vachirasricirikul S, Ngamroo I. Robust LFC in a smart grid with wind power
penetration by coordinated V2G control and frequency controller. IEEE Trans
mance of the proposed method in this case as well so that, by using
Smart Grid 2014;5(1).
the proposed method, maximum frequency deviations of areas 1 to [10] Liu H, Hu ZC, Song YH, Lin J. Decentralized vehicle-to-grid control for primary
3 were reduced by 31%, 30%, and 32%, respectively, than the opti- frequency regulation considering charging demands. IEEE Trans Power Syst
mized PI controller. Performance of the proposed method in charge 2013;28(3):3480–9.
[11] Galus MD, Koch S, Anderson G. Provision of load frequency control by PHEVs,
of EVs in a two area system was verified in another case study, controllable loads, and a cogeneration unit. IEEE Trans Ind Electron 2011;58
which showed that the proposed method properly charged EVs. (10).
S. Falahati et al. / Electrical Power and Energy Systems 83 (2016) 458–469 469
[12] Masuta T, Yokoyama A. Supplementary load frequency control by use of a [19] Bevrani H. Robust power system frequency control. 2nd ed. Springer; 2014.
number of both electric vehicles and heat pump water heaters. IEEE Trans [20] Atashpaz-Gargari E, Lucas C. Imperialist competitive algorithm: An algorithm
Smart Grid 2012;3(3). for optimization inspired by imperialistic competition. In: IEEE congress on
[13] Datta M, Senjyu T. Fuzzy control of distributed PV inverters/energy storage evolutionary computation. p. 4661–7.
systems/electric vehicles for frequency regulation in a large power system. [21] Taher SA, Hajiakbari Fini M, Falahati Aliabadi S. Fractional order PID controller
IEEE Trans Smart Grid 2013;4(1). design for LFC in electric power systems using imperialist competitive
[14] Bahmani MH, Abolfazli M, Afsharnia S, Sadegh Ghazizadeh M. Introducing a algorithm. Ain Shams Eng J 2014;5:121–35.
new concept to utilize plug-in electric vehicles in frequency regulation service. [22] Hadji MM, Vahidi B. A solution to the unit commitment problem using
In: 2nd international conference on control, instrumentation and automation imperialistic competition algorithm. IEEE Trans Power Syst 2012;27
(ICCIA). p. 106–9. (1):117–24.
[15] Sharma I, Canizares C, Bhattacharya K. Smart charging of PEVs penetrating into [23] Nazari-Shirkouhi S, Eivazy H, Ghodsi R, Rezaei K, Atashpaz-Gargari E. Solving
residential distribution systems. IEEE Trans Smart Grid 2014;5(3):1196–209. the integrated product mix-out sourcing problem using the imperialist
[16] Hilshey AD, Hines PDH, Rezaei P, Dowds JR. Estimating the impact of electric competitive algorithm. Expert Syst Appl 2010;37(12):7615–26.
vehicle smart charging on distribution transformer aging. IEEE Trans Smart [24] Singh M, Kumar P, Kar I. Implementation of vehicle to grid infrastructure using
Grid 2013;4(2):905–13. 2014. fuzzy logic controller. IEEE Trans Smart Grid 2012;3(1):565–77.
[17] Gonzalez Vaya M, Andersson G. Centralized and decentralized approaches to [25] Datta M. Fuzzy logic based frequency control by V2G aggregators. In: IEEE 5th
smart charging of plug-in vehicles. In: IEEE power and energy society general international symposium on power electronics for distributed generation
meeting. p. 1–8. systems (PEDG). p. 1–8.
[18] Jiang T, Putrus G, Gao Z, Conti M, McDonald S, Lacey G. Development of a
decentralized smart charge controller for electric vehicles. Int J Electr Power
Energy Syst 2014;61:355–70.