The Journal of Engineering

The 7th International Conference on Renewable Power Generation

(RPG 2018)

Impact of EV fast charging station on eISSN 2051-3305

Received on 30th October 2018
Accepted on 10th January 2019
distribution system embedded with wind E-First on 20th June 2019
doi: 10.1049/joe.2018.9322
generation www.ietdl.org

Akanksha Shukla1 , Kusum Verma1, Rajesh Kumar1

1Department of Electrical Engineering, Malaviya National Institute of Technology Jaipur, India
E-mail: shukla.akanksha19@gmail.com

Abstract: This study presents the synergy between the electric vehicle (EV) fast charging station (FCS) and wind generation. A
queuing analysis based 24-h FCS charging demand is presented in the study. The uncertainty in wind speed variation is
represented with a Weibull distribution function. The impact on the distribution system is analysed for EV penetration up to 50%
and wind generation penetration up to 30%. The excess wind energy, reverse power flow, reduction in network losses and
voltage deviation are investigated for the IEEE 123 bus radial distribution system to quantify the interaction between FCS and
wind generation. The results show that 30% of wind generation is sufficient to supply 50% EV penetration and reduction in
active and reactive energy loss is observed by 10.08% and 10.94%, respectively.

1 Introduction depends not only on uncertain factors but also on the charging
characteristics of the battery and the supply voltage [10]. The
Rapid depletion of fossil fuels, environmental concerns, volatility queuing model was developed in [11] to model EAECP of FCS and
in fuel prices and advancement in the battery technologies led to the power demand by EV depends on the SOC of the battery.
the fast growth of electric vehicles (EVs). The EV charging Furthermore, it has been demonstrated in [12] that the exponential
infrastructure is categorised into three categories based on the load model is the better representation of FCS charging demand
charging rates. Slow charging levels I and II are used at home or compared to the constant power load (CPL) model in power flow
office and fast charging level III is used during emergency analysis. The literature mentioned has studied the impact of FCS
conditions and long trips [1]. To meet the anxiety of the EV and wind generation on the distribution system individually. Sherif
owners, it is expected that the penetration of level III infrastructure et al. [13] studied the synergy between slow charging of EVs and
known as a fast charging station (FCS) will increase in the future wind generation. However, the analysis was done for slow charging
[2]. With increased penetration of FCS, the charging demand will of EVs and they were modelled as CPL model.
become an essential part of the overall load demand of the From the literature, it is clear that very few authors have studied
distribution system. the synergy between fast charging of EVs and wind generation in
Furthermore, the introduction of distributed generation (DG) at the distribution system considering the effect of charging
the local end reduces stress on the centralised generation. It helps characteristics and variation in supply voltage on EAECP. On the
in reducing line losses and improving the voltage profile of the view of the above, the main objectives of the study are as follows:
system. Currently, wind generation is the fastest growing clean
source of renewable energy source [3]. Despite its numerous • Estimation of the number of EVs in the distribution system
economic and environmental benefits, its intermittent nature based on the number of residential houses and maximum
introduces technical challenges. The amount of wind power a diversified demand. Also, develop a queuing model considering
distribution system can absorb compared to the power supplied by non-homogeneous arrival rates derived from Waterloo Region
the distribution system known as wind penetration will depend on Transportation Tomorrow Survey (TTS) and service times using
the intermittency associated with it [4]. Therefore, in this study, the daily driven distance to determine the 24-h EAECP at FCS.
impact of FCS energy demand and wind energy generation on the
• Modelling of the exponential load model of the FCS for power
distribution system is analysed rather than investigating them
flow analysis to state the relationship between active power
independently.
demand by EV, SoC, and supply voltage.
Many studies show that the increased penetration of EVs has
affected distribution system considerably [5–10]. To analyse the • Modelling of the intermittent power output from wind
impact, expected aggregated EV charging profile (EAECP) needs generation due to variable wind speed.
to be modelled since EV load characteristics are different from the • Integrate the developed EAECP and wind generation load model
conventional loads. It depends on uncertainties such as charging to investigate the synergy between FCS and wind generation on
start time, stop time, number and types of EVs, battery capacity, the IEEE 123 bus distribution system.
mileage, initial state of charge (SoC), charging characteristics and
charging level. Papadopoulos et al. [5] modelled the EAECP by The rest of the paper is organised as follows. Estimation of the
addressing the uncertainties associated with EVs using Monte– number of EVs in the distribution system, the queuing model for
Carlo simulation (MCS). A probabilistic model of arrival and FCS and exponential load model of FCS for power flow analysis is
departure time, daily driven distance is considered to develop the developed in Section 2. Section 3 describes the mathematical
EV load demand in [6]. Li and Zhang [7] incorporated vehicle model of intermittent power output from wind generation. The
daily mileage, initial SoC, arrival time, battery capacity to develop simulations and results are presented in Section 4 followed by the
EAECP. EAECP developed in these studies are for slow charging concluding remarks in Section 5.
of EVs. The authors of [8, 9] used queuing theory to model the
EAECP for the residential community as well as FCS. The 2 FCS charging demand
stochastic EAECP model is derived from the actual measurements
The stochastic EV fast charging demand depends on uncertainties
and survey data in [9]. The power demand by the EV is considered
such as number and type of vehicles in a planning region, the
constant in most of the literature. However, the power demand
J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697 4692
This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)
capacity of the vehicle, driving profiles, charging time, charging distance (dm) in miles, the efficiency of the charger (η), the battery
level and SoC of the EV battery. In this section, the number of EVs capacity (Cb) in kWh and the specific energy consumption (em) in
in the distribution system is estimated followed by the kWh/mile. Initial SoC can be estimated as
determination of the number of EVs charging simultaneously at
FCS through queuing model of FCS. Then the power demand of Econ
each FCS at the tth hour is determined by stating the relationship SoC = 1 − (5)
Cb
between power demand, SoC, and supply voltage.
where Econ is the daily energy consumed in kWh. SoC is truncated
2.1 Number of EVs in the distribution system within a certain range in order to protect the batteries. So, the SoC
The total number of EVs (NEV) in the distribution system depends becomes
on the number of houses (Nh), penetration (σ) of EVs with respect
SoCmin if SoC ≤ SoCmin
to the total vehicles and the average number of vehicles per
household (ω). It is estimated using (1) Econ
SoC = 1 − if SoCmin < SoC < SoCmax . (6)
Cb
NEV = Nh × ω × σ . (1) SoCmax if SoC > SoCmax
The number of houses (Nh) is determined by replacing the spot The daily consumed energy (Econ) is calculated as
loads of distribution system by the residential houses (nh) and
distribution transformer. The distribution transformer sees the Cb if dm ≥ dmax
maximum diversified demand (MDD) given by (2) Econ = . (7)
em*dm if dm < dmax
nh × Shmax
MDD = , (2) The daily driven distance (dm) according to the TTS data [16]
DF
follows a log-normal distribution given by
where Shmax is the maximum individual demand of each house and
dm = e(μm + σmZ ), (8)
DF is the diversity factor. The value of DF is chosen from the DF
table given in [14] such that MDD is close to the corresponding
spot loads. Therefore, Nh is the nh times the total number of spot where Z is a standard normal variate, μm and σm are the mean and
loads replaced. standard deviation of log-normal distribution calculated from the
mean (μdm) and standard deviation (σdm) of dm are given by
2.2 Queuing model of FCS
μdm
The probability of the number of vehicle charging simultaneously μm = 1n
is determined using the queuing model [15]. Since FCS can serve 1 + (σd2 m / μd2 m) . (9)
multiple EVs at a time, the multiple serve M1/M2/Ko queuing 2
σm = 1n 1 + (σdm / μdm) 2

model is considered, where M1 is the non-homogeneous inter

arrival time with mean T λt(T λt > 0) following exponential
2.3 Expected total charging demand of FCS
distribution at the tth hour, M2 is the exponentially distributed
service time with mean T μt(T μt > 0) at the tth hour and Ko is the The expected charging demand of the FCS is the summation of all
maximum number of EVs that FCS can serve at a time. T λt and T μt the individual demands of EV charging simultaneously at FCS. It
depends on the charging level, battery capacity, SoC, and charging
will depend on the density of the vehicles on the road derived from characteristics of EV. The power required by the EV can be
Waterloo TTS survey [16] and SoC of the EV battery at the starting calculated if the terminal voltage of the battery and current injected
time, respectively, at the tth hour, respectively. The exponential in the battery are known. The dynamic model [17] of the battery is
distribution is considered here as the arrival and charging of EVs considered here to describe the relationship between SoC and
are independent of each other. The number of EVs waiting at the terminal voltage given by
FCS is assumed to be infinity for simplicity. The probability of k
number of vehicles charging at the tth hour is modelled as discrete Q 1 − SoC
distribution as follows: V b SoC = V o − rbIb − K + Ae− BQ 1 − SoC , (10)
SoC

where Vo is the battery constant voltage in volts (V), rb is the

internal resistance of battery in Ω, K is the polarisation constant (V/
Pt(k) = Ah), Q is the rated battery capacity (Ah), A is the exponential zone
amplitude (V), B is the exponential zone time constant (Ah−1), Ib is
−1 (3) the battery charging current (A) and Vb is the battery voltage (V). It
Ko − 1
(Ko ρt)i (Ko ρt)Ko 1 is assumed that they will be charged with constant current Ib until it
∑ i!
+
Ko ! (1 − ρt)
, k = 0,
reaches a certain threshold limit in fast charging. In (10), all the
i=0
parameters are constant except the SoC of the battery which
(Ko ρt)k increases with time. Hence, the voltage Vb of the battery increases
ρt(0), k = 1, 2, …, Ko,
k! non-linearly with the increase in SoC of the battery. Once the level
of charging is known, i.e. the current injected in the battery to
where ρt is the occupation rate of the FCS per charging slot at the charge it, the individual charging power demand at rated supply
tth hour. It is defined as voltage can be calculated as

T μt Po SoC = V b SoC *Ib . (11)

ρt = . (4)
KoT λt
The charging power demand increases non-linearly with the SoC of
The occupation rate of the FCS should be ρt < 1 so that the system the battery. The total charging demand for all the k EVs charging
simultaneously at FCS is given by
is stable. The SoC at the time of arrival depends on daily driven

J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697 4693

This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)
 k
PoFCS = ∑ Po SoCi . (12)
i=1

The power demand in (12) is determined at rated voltage (Vo). To

incorporate the effect of variation in supply voltage, an exponential
load model [18] is considered to represent the FCS charging load in
power flow analysis given by

FCS V
Pch = PoFCS* ap + bp ,
Vo (13)
bp = 1 − ap,

where ap, bp, and np are the parameters of the exponential load
model

2.4 Monte–Carlo simulation (MCS)

MCS provides probabilistic solutions to the problem that involves
uncertainties. It uses a random number generator and sampling
techniques. In this study, the cumulative distribution function
(CDF) of daily driven distance and CDF of EVs charging
simultaneously at the tth hour is randomly sampled through the
inversion method of random number generation. Fig. 1 shows the
outline of the MCS process to generate FCS charging demand.
MCS is run for 1000 iterations and it is found that 1000 runs are
sufficient for the convergence of MCS.

3 Output power modelling of wind generation

The output power of wind generation is modelled by modelling the
behaviour of wind speed. The Weibull probability distribution
function is often used to model the wind speed. At the tth hour, it is
given by [3]
l−1
lt vt lt
ft v = e−(vt / ct) , (14)
ct ct

where vt is the wind speed in (m/s) at the tth hour, lt and ct are the
Weibull shape and scale parameters, respectively, at the tth hour.
The value of lt and ct is estimated by (15) and (16) as

1.86
σwt
lt = , (15)
v̄t
Fig. 1  Flowchart of the proposed random simulation of FCS charging
v̄t demand
ct = , (16)
Γ 1 + (1/lt)
Table 1 Wind speed states of Weibull distribution function
where (σwt, v̄t) are the standard deviation and mean of wind speed at Weibull state, m Wind speed limits, m/s
the tth hour and Γ is the gamma function. The wind power output is 1 0–1
determined by dividing the continuous distribution function into m 2 1–2
states. The wind speed in each state in bounded by the limits. The ⋮ ⋮
number of states m is chosen such that it is neither small nor large. last state vmax − 1 to vmax
Discrete function with a small number of states will lose
information about the continuous distribution function and a large
number of states will increase the problem complexity. In this
0 if 0 ≤ vam ≤ vci,
work, the step is adjusted to 1 m/s as shown in Table 1.
The probability of wind speed being in state (S) m at the tth vam − vci
Prated × if vci < vam ≤ vr,
hour is calculated using (17) for wind speed limits vm1 and vm2 of w
Pm v = vr − vci (18)
state m Prated if vr < vam ≤ vco,
vm2 0 if vco < vam,
P Smt = ∫ vm1
f t v dv . (17)
where vam is the average wind speed of state m, vci, vr, and vco are
the cut in, rated, and cut off speed of wind turbine. The average
The parameters of the power curve of a wind turbine are used to
power output from the wind turbine at the tth hour is
calculate the output power of wind corresponding to each state m is
given by w
Pavg t
= ∑ Pmw v *P Smt . (19)
m

4694 J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697

This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)
Fig. 2  Modified IEEE 123 bus radial distribution system

Table 2 EVs and wind generation penetration scenario

Scenario EV, % DG, % Scenario EV, % DG, %
1 0 0 9 30 0
2 0 10 10 30 10
3 0 20 11 30 20
4 0 30 12 30 30
5 10 0 13 50 0
6 10 10 14 50 10
7 10 20 15 50 20
8 10 30 16 50 30

Table 3 Parameters for EAECP

Parameters Value Parameter Value
w 1.86 μdm (miles) 40
Shmax, kVA 12 σdm (miles) 20
T μt, min 29 Vo, Volts 234.01
Ko 10/15 rb, Ω 0.019047
SoCmin 0.15 K, V Ah 0.010752
SoCmax 0.85 Q, Ah 113.4
Cb, kWh 24 A, V 18.3501
em, kWh/mile 0.2857 B, Ah−1 0.5387
ap 0.0246 np −1.839

Table 4 Parameters for wind power modelling • Spot loads of 44.72 kVA are replaced by ten residential houses
Parameters Value Parameter Value supplied by a 50 kVA transformer.
vmax, m/s 13.6 vci, m/s 3 • Spot load of 22.36 kVA is replaced by four residential houses
supplied by a 25 kVA transformer.
v̄, m/s 4.85 vr, m/s 7
Prated, kW 250 vco, m/s 11 A total of 594 residential houses are obtained and 1.86 vehicles
per household are considered as per the National Household Travel
Survey (NHTS) 2009 data [22]. A total of 16 scenarios are
4 Simulation and results considered in this study listed in Table 2. The parameters required
for the development of EAECP and wind power output modelling
The synergy between FCS and wind generation is investigated on are listed in Tables 3 and 4, respectively.
the IEEE 123 bus distribution system [18, 19] modelled in The impact of FCS and wind generation for the 16 scenarios
OpenDSS [20]. The standard IEEE 123 bus distribution system is listed in Table 2 on the modified IEEE 123 bus distribution system
modified to integrate six FCS and four wind generations as shown is shown. Fig. 3 shows the 24-h EAECP for 10 and 15 slot FCS
Fig. 2. The FCSs are located at 8, 28, 55, 65, 80 and 99 buses [21]. with different penetration levels. The peak is observed at the 8th
The location and size of wind generation are set according to [13]. and 16th hour of the day. The daily wind power output is
The spot loads of the distribution system are replaced by residential represented by Fig. 4 and the maximum power is generated during
houses and distribution transformers. In this study, the power factor the 15th hour of the day.
of the residential houses is kept the same as that of the spot loads. The daily amount of excess wind energy with 50% EV for
different penetration of wind generation is shown in Fig. 5. It

J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697 4695

This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)
Fig. 3  24-h EAECP of 10 and 15 slot FCS Fig. 7  Substation daily active power

Table 5 Active energy loss reduction and voltage deviation

Scenario Active energy Reactive energy Voltage
loss reduction, % loss reduction, % deviation, %
1 0 0 2.26
2 11.98 12.48 2.13
3 22.92 23.87 2.00
4 32.83 34.18 1.88
5 −4.94 −5.05 2.31
6 7.40 7.79 2.18
Fig. 4  Daily expected wind power output 7 18.70 19.56 2.05
8 28.96 30.24 1.93
9 −15.85 −16.22 2.41
10 −2.77 −2.61 2.28
11 9.25 9.89 2.15
12 20.22 21.32 2.02
13 −28.21 −28.88 2.51
14 −14.38 −14.50 2.37
15 −1.62 −1.23 2.25
16 10.08 10.94 2.12

Fig. 5  Daily excess wind generation power output

wind generation. The positive active loss reduction and less voltage
deviation of 2.12% than the base case for the system with 50% EV
and 30% wind generation suggest that the 30% wind generation
penetration is sufficient for the system with 50% EV.

5 Conclusions
In this study, interaction between FCS and wind generations and
their impact on distribution system is investigated for different
penetration of EVs and wind generations. The FCS charging
demand also known as EAECP with different penetration is
modelled using queuing theory which incorporates uncertainty
such as different inter-arrival times and service times. Additionally
Fig. 6  Substation active energy in per unit wind generation power output is modelled by representing the
uncertainties associated with wind speed by Weibull distribution
shows that even 30% penetration of wind generation is not function. It can be inferred that the peak demand of EAECP occurs
sufficient to satisfy the FCS charging demand during the peak during peak vehicle density on the road. While peak power output
hours of demand. The unsatisfied FCS demand will be satisfied by of wind generation output occurs at 3:00 p.m. The FCS charging
importing the power from the substation. demand for 50% penetration is partially satisfied by the wind
Fig. 6 shows the daily active energy drawn from the substation generation even for 30% penetration level due to more charging
in p.u. for different penetration of EVs and wind generations. It is demand than the wind generation power output. At the same time
found that the energy drawn from the substation for 50% of EV and reduction in active and reactive losses of 10.08% and 10.94% and
30% of wind generation penetration is less than that of the system voltage deviation of 2.12% suggest that the integration of wind
without EV and DG. The reverse active power of the system generation has the positive effect on the distribution system with
without EVs for different penetration of wind generation is shown 50% penetration levels of EVs. Also, the daily substation energy
in Fig. 7. The positive value of active power implies that active increases linearly with increase in penetration of EVs and
power flow does not flow back to the main substation. decreases with increase in penetration of wind generations. The
Active energy loss reduction and voltage deviation for all the daily substation active energy demand for 30% wind generation
scenarios listed in Table 2 are given in Table 5. The positive values penetration and 50% EV penetration is less than the base case
of loss reduction implies the active power losses are less than the suggest that the imported energy from substation is less.
base case. Value of voltage deviation closer to 0 shows that the
magnitude of the bus voltages is close to the reference bus voltage.
Scenario 1 is the base case system without EV and wind
6 References
generation. It is observed that there is a maximum reduction in the [1] Shareef, H., Md Mainul Islam, A.M.: ‘A review of the stage-of-the-art
active loss by 32.83% and least voltage deviation of 1.88% for charging technologies, placement methodologies, and impacts of electric
vehicles’, Renew. Sustain. Energy Rev., 2016, 64, pp. 403–420
scenario 4 when no EVs are present in the system with 30% of

4696 J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697

This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)
[2] Zhang, H., et al.: ‘PEV Fast-charging station siting and sizing on coupled [12] Shukla, A., Verma, K., Kumar, R.: ‘Voltage dependent modelling of fast
transportation and power networks’, IEEE Trans. Smart Grid, 2018, 9, (4), pp. charging electric vehicle load considering battery characteristics’, IET Electr.
2595–2605, DOI: 10.1109/TSG.2016.2614939 Syst. Transp., 2018, 12, (4), pp. 221–230, DOI: 10.1049/iet-est.2017.0096
[3] Atwa, Y.M., El-Saadany, E.F.: ‘Probabilistic approach for optimal allocation [13] Sherif, A., Walid, F., Morsi, G., et al.: ‘Impact of wind-based distributed
of wind-based distributed generation in distribution systems’, IET Renew. generation on electric energy in distribution systems embedded with electric
Power Gener., 2011, 5, (1), pp. 79–88 vehicles’, IEEE Trans. Sustain. Energy, 2015, 6, (1), pp. 79–87
[4] Atwa, Y.M., El-Saadany, E.F., Salama, M.M.A., et al.: ‘Optimal renewable [14] Kersting, W.H.: ‘Distribution system modeling and analysis’ (CRC Press,
resources mix for distribution system energy loss minimization’, IEEE Trans. Washington, D.C., 2012, 3rd edn.)
Power Syst., 2010, 25, (1), pp. 360–370 [15] Bhat, U.N.: ‘An introduction to queueing theory: modeling and analysis in
[5] Papadopoulos, P., Skarvelis-Kazakos, S., Grau, I., et al.: ‘Electric vehicles’ applications’ (Birkhäuser, Boston, USA, 2015)
impact on British distribution networks’, IET Electr. Syst. Transp., 2012, 2, [16] ‘The 2011 Transportation Survey’. Available at http://dmg.utoronto.ca/
(3), pp. 91–102 reports/ttsreports.html
[6] Darabi, Z., Ferdowsi, M.: ‘Aggregated impact of plug-in hybrid electric [17] Tremblay, O., Dessaint, L.-A.: ‘Experimental validation of a battery dynamic
vehicles on electricity demand profile’, IEEE Trans. Sustain. Energy, 2011, 2, model for EV applications’, World Electr. Veh. J., 2009, 3, (1), pp. 1–10
(4), pp. 501–508 [18] Price, W., Taylor, C., Rogers, G.: ‘Standard load models for power flow and
[7] Li, G., Zhang, X.-P.: ‘Modeling of plug-in hybrid electric vehicle charging dynamic performance simulation’, IEEE Trans. Power Syst., 1995, 10, (3), pp.
demand in probabilistic power flow calculations’, IEEE Trans. Smart Grid, 1302–1313
2012, 3, (1), p. 492–499 [19] ‘IEEE PES Distribution Test Feeders, 123-Bus Feeder’. Available at http://
[8] Zhang, P., Qian, K., Zhou, C., et al.: ‘A methodology for optimization of ewh.ieee.org/soc/pes/dsacom/testfeeders/ index.html, accessed 2 May 2017
power systems demand due to electric vehicle charging load’, IEEE Trans. [20] OpenDSS Software: Available at http://www.sourceforge.net/projects/
Power Syst., 2012, 27, (3), pp. 1628–1636 electricdss/
[9] Leou, R.C., Teng, J.H., Su, C.L.: ‘Modelling and verifying the load behaviour [21] Shukla, A., Verma, K., Kumar, R.: ‘Consumer perspective based placement of
of electric vehicle charging stations based on field measurements’, IET Gener. electric vehicle charging stations by clustering techniques’. IEEE 2016
Transm. Distrib., 2015, 9, (11), pp. 1112–1119 National Power Systems Conf. (NPSC), Bhubaneswar, India, 2016, pp. 1–6
[10] Dharmakeerthi, C., Mithulananthan, N., Saha, T.: ‘Impact of electric vehicle [22] Santos, A., McGuckin, N., Nakamoto, H.Y., et al.: ‘Summary of travel trends:
fast charging on power system voltage stability’, Int. J. Electr. Power Energy 2009 national household travel survey’. Technical Report, 2011
Syst., 2014, 57, pp. 241–249
[11] Hafez, O., Battacharya, K.: ‘Queuing analysis based PEV load modelling
considering battery charging behavior and their impact on distribution system
operation’, IEEE Trans. Smart Grid, 2016, 9, (1), pp. 261–273

J. Eng., 2019, Vol. 2019 Iss. 18, pp. 4692-4697 4697

This is an open access article published by the IET under the Creative Commons Attribution -NonCommercial License
(http://creativecommons.org/licenses/by-nc/3.0/)