Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
8,
No.
6,
Decem
ber
20
18, p
p.
5303
~
5310
IS
S
N: 20
88
-
8708, DO
I:
10
.11
591/ijece
.v8i6
.
pp
5403
-
53
10
5303
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Des
i
gn
o
f
a
M
onitorin
g
-
c
ombin
ed Sitin
g Scheme
f
or El
ectric
Vehicle
Charger
s
Ju
n
gho
on
Lee
,
G
yu
n
g
-
Le
en
Park
Depa
rtment
o
f
C
om
pute
r
Scie
n
ce a
nd
St
a
ti
st
ic
s,
Jeju
Na
ti
ona
l
Uni
ver
sit
y
,
Republi
c
of
Kor
ea
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r 23
, 201
8
Re
vised
Jun
14
, 201
8
Accepte
d
J
ul
22
, 2
01
8
Thi
s
pap
er
d
esig
ns
a
sit
ing
sch
e
m
e
for
public
elec
tr
ic
v
ehicle
ch
arg
ers
bas
ed
on
a
geneti
c
al
gorit
hm
work
ing
on
cha
rg
er
m
onit
oring
strea
m
s
.
The
m
onit
oring
-
combined
allocation
sche
m
e
runs
on
a
long
-
te
rm
basis,
it
er
at
in
g
the
proc
ess
of
c
oll
e
ct
ing
d
ata,
a
naly
z
ing
deman
d,
and
se
lecti
ng
ca
ndid
ates.
The
ana
l
y
sis
of
spati
o
-
t
emporal
arc
h
ive
s
,
ac
qu
i
red
from
the
f
a
st
cha
rg
ers
cur
ren
t
l
y
in
ope
rat
ion
,
foc
uses
on
the
per
-
ch
ar
ger
hot
hour
an
d
proximit
y
eff
ect
to
just
if
y
demand
bal
an
cing
in
geogr
aph
i
c
cl
ust
er
le
v
el.
I
t
le
ads
to
the
def
ini
t
ion
of
a
f
it
ness
func
t
ion
r
epr
ese
nt
ing
the
standa
rd
dev
ia
t
i
on
of
per
-
cha
rge
r
loa
d
an
d
cl
uster
-
by
-
cl
u
ster
distri
but
ion
.
In
a
chr
om
osom
e,
ea
ch
bina
r
y
int
eg
er
is
associate
d
with
a
ca
nd
ida
t
e
and
it
s
st
at
i
c
f
ie
lds
inc
lud
e
th
e
inde
x
to
the
cl
us
te
r
to
which
it
is
bel
onging
.
The
per
form
anc
e
res
ult
obta
in
ed
from
a
prototy
pe
implementa
t
ion
rev
eal
s
tha
t
the
proposed
sche
m
e
ca
n
st
ab
l
y
distri
bute
th
e
ch
arg
ing
l
oad
wi
th
an
addi
t
ion
of
a
new
ch
arg
er
,
achie
ving
the
red
uction
of
standa
rd
dev
ia
t
ion
from
8.
7
%
to
4.
7
%
in
th
e
rea
l
-
world
sce
n
a
rio.
Ke
yw
or
d:
Cl
us
te
r
-
le
vel
de
m
and
s
har
e
EV
c
hargin
g
i
nfrastr
uctu
re
Gen
et
ic
al
gorithm
Publi
c cha
rg
e
r si
ti
ng
Re
al
-
tim
e
m
on
it
or
in
g data
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Gyu
ng
-
Lee
n
P
ark
,
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nce a
nd
Stat
ist
ic
s
,
Jej
u Nati
on
al
Un
i
ver
sit
y
,
Jej
ud
ae
ha
kno 102, R
ep
. of K
or
ea
.
Em
a
il
: jh
le
e@je
j
un
u.
ac.
kr
1.
INTROD
U
CTION
Fo
r
the
pe
netr
at
ion
of
EVs
(
Ele
ct
ric
Veh
ic
le
s),
it
is
nece
ssary
to
buil
d
a
well
-
organ
iz
ed
c
hargi
ng
infr
a
struct
ur
e
ov
e
r
the
ta
r
get
area
[
1].
Be
s
ides
slo
w
AC
charger
s
usua
ll
y
instal
le
d
in
dr
i
ver
s
’
hom
es
for
ov
e
r
night
chargin
g,
hi
gh
-
vo
lt
age
DC
cha
rg
e
rs
are
w
orkin
g
in
public
places
under
the
c
on
t
ro
l
of
res
pons
ibl
e
auth
or
it
ie
s
[2
]
.
In
s
uffici
ent
chargin
g
capaci
ty
br
ing
s
un
ac
ceptable
wait
in
g
tim
e
to
EV
dri
vers,
m
ai
nly
du
e
t
o
long
cha
rg
i
ng
tim
e
wh
ic
h
la
sts
te
ns
of
m
inu
te
s
f
or
a
sin
gl
e
transacti
on
even
with
fast
DC
cha
rg
e
rs.
Hen
ce
,
m
any
countries
are
try
ing
to
instal
l
DC
char
ge
rs
in
a
ppropr
ia
te
places.
At
first,
they
s
el
ect
tho
se
pla
ces
easy
to
supp
ly
p
ower
an
d
guara
nt
ee
el
ect
rical
s
afety
[3
]
.
The
n,
the
pen
et
rati
on
of
E
Vs
w
il
l
create
a
sp
eci
fic
dem
and
patte
r
n,
w
hic
h
m
us
t
be
co
ns
ide
re
d
for
the
ne
xt
ste
p
cha
rg
e
r
insta
ll
at
ion
.
As
s
uc
h,
c
harger
e
xpansi
on
and
dem
and
pa
tt
ern
will
interact
with
each
oth
e
r
re
pea
te
dly,
m
aking
it
essenti
al
to
keep
analy
zi
ng
the
dem
and
b
e
ha
vi
or of a ta
rg
et
c
hargin
g
i
nfrastru
ct
ur
e.
Most
m
od
ern
facil
it
ie
s,
not
restrict
ed
to
chargin
g
sta
ti
on
s
,
are
c
onnected
to
a
m
anag
em
ent
coor
din
at
or
vi
a
ub
i
qu
it
ous
a
nd
c
hea
p
com
m
un
ic
at
ion
ch
ann
el
s
[4
]
.
I
n
our
ci
ty
,
nam
e
ly
,
Jeju
,
Re
public
of
Korea,
w
hich
is
m
aking
a
n
extensi
ve
e
ffort
to
prom
pt
the
de
plo
ym
ent
of
EV
s,
m
any
cha
rg
e
rs
are
unde
r
const
ru
ct
io
n
a
nd
repor
t
their
real
-
ti
m
e
wo
r
ki
ng
sta
tus
to
a
central
ser
ver
[
5].
Currentl
y,
245
DC
c
harg
ers
are
e
m
br
aced
in
t
hi
s
m
anag
e
m
ent
dom
ai
n
and
t
he
ir
sta
tus
rec
ords
a
re
acc
um
ulate
d
eve
ry
5
m
inu
te
s.
O
ur
res
earch
te
a
m
acqu
ires
t
he
per
m
issi
on
to
us
e
this
arch
i
ve,
sto
res
in
our
local
databas
e
ta
bles,
and
c
om
bin
es
geograph
ic
inf
or
m
at
ion
for
sop
histi
cat
ed
stream
analy
s
i
s
with
rele
van
t
too
ls
s
uch
as
R
and
Te
nsor
flow
[6
]
.
In
t
his
wor
k,
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
I
nt J
Ele
c
& C
om
p
En
g,
V
ol. 8, N
o.
6, Dece
m
ber
2
01
8
:
5303
–
5310
5304
we
are
to
e
xplo
it
the m
on
it
or
ing data st
ream
for
the
locati
on
selec
ti
on
of n
e
w publi
c cha
rgers to
overc
ome
th
e
weakness
of E
Vs
in
c
hargin
g, com
par
ed
w
it
h gas
oline
-
pow
ered ve
hicle
s.
It
m
us
t
be
m
en
ti
on
ed
that
the
sit
e
sel
ect
ion
ha
s
to
ta
ke
into
account
a
var
ie
ty
of
f
act
ors,
m
ai
nly
du
e
to
the
fact
that
chargers
a
re
hi
gh
-
volt
age
de
vices.
A
s
f
or
e
xisti
ng
relat
ed
work
f
or
this
a
sp
ect
,
Flo
rida
Power
&
Ligh
t
Com
pan
y
li
sts
the
crit
ic
al
factor
s
to
con
sider
i
n
decidin
g
ch
ar
ger
in
sta
ll
at
ion
places
[7
]
.
T
he
li
st
include
s
visibi
li
ty
and
li
gh
ti
ng,
pro
xim
i
ty
to
power
s
ources
,
pa
rk
i
ng
sp
ace
siz
e,
weathe
r
an
d
cl
i
m
a
te
,
el
ect
rical
safety
,
ven
ti
la
ti
on,
and
t
he
li
ke.
I
n
ad
diti
on,
T
r
ans
portat
ion
a
nd
Cl
im
at
e
In
it
ia
t
ive
al
so
ch
oo
s
es
sel
ect
ion
crit
e
ria
f
or
c
hargi
ng
sta
ti
on
s
[
8]
.
This
re
port
ad
dr
esse
s
co
nn
ect
io
ns
t
o
powe
r,
netw
orks
a
nd
com
m
un
ic
at
ion
s,
inter
de
penden
cy
with
exi
sti
ng
i
nfrastr
uc
ture,
an
d
EV
i
nterf
ace
s.
Be
si
des,
the
wei
ght
from
above
-
m
entioned
fact
or
s
will
b
e
dif
fer
e
nt
r
egio
n
by
re
gion
a
nd
de
pe
nd
on
the
forecas
t
on
the
popul
at
io
n
grow
t
h, cust
om
er n
ee
d
c
hange,
a
nd
new te
chnolo
gy a
pp
e
aran
ce
.
In
a
dd
it
io
n,
as
for
relat
ed
w
or
k
f
or
pu
blic
ch
arg
i
ng,
[9
]
trie
s
to
m
axi
m
iz
e
the
el
ect
rificat
i
on
rate,
or
the
tra
vel
dista
nce
c
ov
e
re
d
by
EVs
cha
r
ged
at
new
ch
ar
ging
sta
ti
ons.
Thi
s
resea
rch
is
buil
t
on
the
anal
ysi
s
of
m
illi
on
s
of
tri
ps
ta
ken
by
a
bout
11,
000
ta
xis
in
Be
ijin
g.
The
aut
hors
inv
e
sti
gate
the
unde
rly
ing
c
ha
rg
i
ng
dem
and
over
t
he
ci
ty
area
as
well
as
to
locat
e
ho
t
spot
s
f
or
the
dem
and.
The
a
naly
sis
sel
ect
s
tho
se
areas
wh
e
re
m
any
ta
xi
dri
ve
rs
a
re
li
kely
to
sta
y
for
a
rest
a
s
new
cha
r
ging
sta
ti
on
s,
unde
r
the
ass
um
ption
that
dr
i
ver
s
will
possibly
m
ov
e
1
m
il
e
to
ta
ke
pr
efer
red
cha
r
ge
rs
wh
e
n
nece
s
sary.
In
a
dd
it
ion,
[10]
form
ulate
s
a
m
at
he
m
at
ic
a
l
m
od
el
to
find
op
ti
m
al
so
luti
on
s for
sit
ing
c
ha
rg
i
ng
stat
io
ns
,
achievin
g
a
bout 50
% im
pr
ovem
ent
in
the
el
ect
rific
at
ion
rate.
This
w
ork
de
fines
the
obj
ect
ive
f
unct
ion
based
on
t
he
tra
vel
dis
ta
nce
that
can
not
be
cov
e
re
d
by
an
y
EV
chargin
g
fo
r
t
he
giv
e
n
sta
ti
on
placem
ent.
The
form
ulati
on
is
fed
to
a
m
ixed
integ
er
non
-
li
near
program
m
ing
s
olv
er
to fin
d
a
n
a
ns
we
r.
As
fa
r
as
we
know,
the
re
is
no
w
ork
dire
ct
ly
ta
kin
g
int
o
acco
unt
the
current
dem
and
patte
r
n
in
determ
ining
th
e
locat
ion
of
a
dd
it
io
nal
cha
r
ge
rs.
E
ve
n
if
t
he
data
is
a
vaila
ble,
it
is
not
ea
sy
to
fin
d
a
n
optim
a
l
go
al
f
or
this
pro
blem
.
This
pap
er
at
te
m
pts
t
o
ide
ntify
a
char
ge
placem
ent
wh
ic
h
ca
n
di
stribu
te
the
c
ha
rg
i
ng
load
over
the
ta
rg
et
are
a.
W
it
h
this
go
al
de
finiti
on,
a
ge
netic
al
gor
it
hm
is
desig
ned
to
fi
nd
a
s
ub
op
ti
m
al
so
luti
on
within
an
acce
pta
ble
tim
e b
ound.
This
pap
e
r
is
orga
nized
as
fo
l
lows
:
Af
te
r
overv
ie
wing
the
m
ai
n
issue
i
n
Sect
ion
1,
Sect
ion
2
sho
ws
the
data
a
naly
sis
resu
lt
s
to
be
consi
der
e
d
i
n
charger
sit
ing.
The
n,
Sect
io
n
3
c
once
ptu
al
ly
design
s
a
cl
us
te
r
-
integrate
d
sit
in
g
sc
hem
e fo
r E
V
c
hargers. Fi
nally
, S
ect
ion
4
s
umm
arizes and co
nclu
des t
his p
a
per with
a brief
descr
i
ption o
n fu
t
ur
e
w
ork.
2.
DA
T
A ANAL
YS
IS
To
be
gi
n
with
,
Figure
1
sho
ws
the
loc
at
io
n
of
cha
r
ger
s
i
n
Jej
u
Ci
ty
,
w
hich
is
s
urrou
nded
by
ab
out
200
km
lo
ng
coastl
ine.
T
he
m
eaning
of
e
ach
sym
bo
l
w
il
l
be
exp
la
in
e
d
la
te
r.
Her
e
,
the
r
oad
net
w
ork
is
dow
nlo
a
ded
f
r
om
the
op
e
n
data
sit
e
in
a
n
ESRI
s
ha
pe
f
il
e
fo
rm
at
and
plo
tt
ed
on
t
he
R
w
orks
pac
e.
The
charger
distri
buti
on
coi
ncide
s
with
t
he
po
pula
ti
on
de
ns
it
y,
dis
play
ing
hi
gh
co
nce
ntrati
on
i
n
the
cente
r
nort
h
reg
i
on.
Ma
ny
charger
s
are
in
sta
ll
ed
at
the
l
ocal
go
vernm
e
nt
bra
nch
es
as
it
is
easy
to
ge
t
endo
rsem
ent
f
or
E
V
charger
est
abli
sh
m
ent.
They
are
m
ai
nly
us
ed
by
local
resi
den
ts
.
In
ad
diti
on,
as
Je
j
u
is
one
of
the
m
os
t
fam
ou
s
tour
places
in
East
Asia,
a
lot
of
to
ur
ist
at
tract
ion
s
pro
vid
e
ch
ar
ging
equ
i
pm
ent
fo
r
EV
-
dr
i
ving
vi
sit
or
s
.
Accor
ding
to
our
obser
vatio
n,
de
m
and
pe
ak
of
DC
cha
rg
i
ng
arises
be
tween
4
PM
and
6
PM,
s
li
gh
tl
y
dev
ia
ti
ng t
he g
rid pea
k hours.
Figure
1
.
Cha
r
ger locat
io
n
a
nd e
xp
ect
e
d
l
oa
d
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n of
a
M
on
it
ori
ng
-
C
ombine
d Sit
ing Sc
hem
e
for
Elec
tri
c Vehicl
e Ch
ar
ge
rs
(
Ju
ngho
on Lee)
5305
Figure
2
plo
ts
the
occupa
nc
y
rate
of
each
charger.
Occ
up
a
ncy
rate,
a
nalo
gous
to
th
e
chargin
g
dem
and
or
loa
d,
de
note
s
how
m
any
reco
r
ds
are
in
dicat
ing
that
a
charger
is
curre
ntly
wo
r
king
out
of
total
record
s.
I
n
the
figure,
eac
h
do
t
c
orrespo
nds
to
a
sing
le
c
harger
.
F
or
a
charger
,
the
hot
hour,
in
whic
h
it
s
occupa
ncy
rate
touc
hes
hi
ghe
st,
is
ob
ta
ine
d
first.
The
n,
the
hot
rate
is
plot
te
d
on
a
virtua
l
ver
ti
cal
li
ne
of
the
corres
pondin
g
hour. Cha
rg
e
rs
h
avin
g
a com
m
on
h
ot hour ap
pea
r
on the sa
m
e li
ne.
W
e ca
n
see m
or
e d
ot
s than
oth
e
rs
bet
ween
16
t
o
18.
By
this
fig
ure,
m
os
t
chargers
a
re
us
e
d
betw
ee
n
8
to
21.
A
bout
20
c
harge
rs
ha
ve
th
e
occupa
ncy
rate
below
0.05
e
ve
n
in
their
hot
hours,
w
hile
so
m
e
oth
ers
exc
eedin
g
0.6.
One
of
the
goal
s
of
th
e
sit
ing
sc
hem
e m
ay
li
e in the
r
edu
ct
io
n o
f
the
o
cc
up
a
ncy
rate ga
p.
Figure
2
.
Hot
hours a
nd
occup
ancy rate
Nex
t,
Fi
gure
3
traces
the
occupan
cy
rate
acc
ordin
g
to
the
di
sta
nce
to
the
cl
os
est
charger
.
Her
e
a
gain
,
a
sing
le
dot
re
pr
ese
nts
a
si
ngle
charger
.
As
the
distan
ce
is
an
a
nalo
g
valu
e,
dots
are
scat
te
red
ov
e
r
t
he
gr
a
ph
sp
ace.
The
m
os
t
isolat
ed
c
harger
is
apa
rt
f
r
om
it
s
cl
os
est
ne
ighbor
by
5.8
km
.
T
ho
se
cha
rg
e
rs
i
ns
ta
ll
ed
in
the
sam
e
bu
il
ding
or
office
app
ea
r
on
the
ve
rtic
al
li
ne
of
0
km
.
As
can
be
se
en
in
the
fi
gur
e,
the
distance
to
the
cl
os
est
neig
hb
or
has
li
tt
le
dep
en
de
ncy
on
th
e
oc
cu
pa
ncy
ra
te
.
Even
t
houg
h
there
a
re
not
so
m
any
cases,
wh
e
n
a
charge
r
is
ne
wly
instal
le
d,
t
he
oc
cu
pa
ncy
rate
ar
ound
the
charge
r
inc
rea
ses
tog
et
her.
D
rivers
seem
to
wan
t
to ch
a
r
ge
at
a
vi
ci
nity
o
f hig
he
r
c
harger
d
e
ns
i
ty
.
Figure
3
.
Pro
xim
ity and
occ
upancy
rate
3.
SITING
S
CHE
ME
3.1.
M
ain ide
a
Figure
4
outl
in
es
the
sequ
e
nc
e
of
the
sit
ing
p
r
ocess
.
For
a
current
cha
r
ger
distribu
ti
on,
t
he
proce
dure
colle
ct
s
the
m
on
it
or
i
ng
d
at
a
a
nd
co
nducts n
e
cessary
analy
si
s.
He
re,
a
ddit
ion
al
inf
orm
ation
s
uch
as p
op
ul
at
ion
grow
t
h
an
d
E
V
pe
netrati
on
is
integrated
i
n
to
the
m
on
it
ori
ng
series.
T
he
n,
pe
rfo
rm
ance
m
et
rics
are
def
i
ned
for
the
sel
ect
ion
of
ne
w
charger
locat
i
on
s
.
The
m
etr
ic
includes
wait
ing
ti
m
e
reducti
on,
pro
xim
it
y
i
m
pr
ovem
ent,
dem
and
balan
ci
ng
,
a
nd
t
he
li
ke.
He
re,
the
annual
budget
of
the
cha
r
ge
r
operati
on
aut
hority
decides
t
he
num
ber
of
c
harg
ers
to
buil
d.
T
hen,
a
hu
m
an
decisi
on
m
aker
rec
omm
end
s
al
l
po
ssi
ble
ca
nd
i
date
locat
ion
s
acco
rd
i
ng
t
o
the
a
bove
-
m
entioned
crit
eria.
Tw
o
or
m
or
e
c
ha
rg
e
rs
ca
n
be
li
ste
d
in
a
sam
e
place
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
I
nt J
Ele
c
& C
om
p
En
g,
V
ol. 8, N
o.
6, Dece
m
ber
2
01
8
:
5303
–
5310
5306
Her
e
,
eac
h
sel
ect
ion
of
a
c
ha
rg
e
r
ca
n
a
ff
e
ct
the
rem
ai
ni
ng
ot
hers.
The
est
i
m
at
ion
of
this
e
ff
ect
is
a
ve
ry
com
plex
pro
ble
m
,
so
we
ta
ke
a
s
uboptim
al
appro
ac
h
[
11
]
.
M
or
e
over
,
the
ef
fect
of
a
ne
w
al
loca
ti
on
is
diff
e
re
nt
accor
ding
to
wh
et
he
r
it
is
inside
a
cl
us
te
r
or
not,
as
charger
s
bel
onging
to
a
sa
m
e
geo
gra
ph
ic
cl
us
te
r
te
nd
t
o
e
ven
ly
sh
are
the l
oad.
Figure
4
.
Cha
r
ger sit
ing
proc
ess
A
pai
r
of
c
hargers
a
pa
rt
fro
m
each
oth
e
r
l
ess
than
1.6
km
bel
ong
t
o
a
com
m
on
cl
us
te
r.
A
dri
ve
r
will
m
ov
e to the
ot
her
c
harge
r
w
he
n
on
e c
harger
o
ccu
pied
i
f
th
e d
ist
ance
between
t
hem
is l
ess than
1.6
km
[9
]
. Out
of
245
c
ha
rg
e
r
s,
the
a
naly
sis
fin
ds
68
cl
us
te
rs
as
s
how
n
in
the
m
ap
of
Fi
gure
1.
The
re
ar
e
3
big
cl
ust
ers
,
one
hav
i
ng
61
c
ha
r
ger
s
in
ce
nter
north
,
the
ot
he
r
tw
o
eac
h
ha
vi
ng
15
a
nd
10
in
cente
r
s
ou
t
h,
res
pecti
vely
.
Figure
5
sho
ws
the
st
and
a
r
d
de
viati
on
i
n
occ
up
a
nc
y
rates
of
cha
rg
e
rs
withi
n
ea
ch
cl
us
te
r
.
Th
ose
cl
us
te
rs
co
nt
ai
nin
g
j
ust
on
e
c
harg
er
ha
ve
no
de
vi
at
ion
.
Th
e
de
viati
on
is
at
m
os
t
0.1
8
an
d
the
cha
rg
i
ng
de
m
and
is
quit
e
even
ly
sh
are
d
withi
n
a
cl
us
te
r.
The
r
e
are
excep
ti
on
al
cases
wh
en
so
m
e
char
ger
s
beco
m
e
ou
t
-
of
-
ser
vice
from
t
i
m
e
to
tim
e.
Af
te
r
al
l,
the
new
c
harg
er
sel
ect
ion
ca
n
be
diff
e
re
ntiat
ed
into
tw
o
c
ases,
one
ad
ding
to
a
cl
us
te
r
a
nd
t
he
oth
e
r
to
an i
nd
epende
nt p
la
ce
.
Figure
5
.
Per
-
c
luster sta
nd
a
rd
dev
ia
ti
on i
n oc
cup
a
ncy
rate
The
load
is
re
cal
culat
ed
so
that
the
charge
r
in
a
sing
le
cl
us
te
r
has
the
equ
al
occ
upan
cy
rate
by
aver
a
ging
al
l
of
cl
ust
er
m
e
m
ber
s.
W
e
can
est
i
m
at
e
the
increase
in
th
e
dem
and
cha
ng
e
ste
m
m
ing
from
the
pen
et
rati
on
of
Evs
just
i
n
cl
ust
er
le
vel,
not
in
c
harger
-
le
ve
l.
Acc
ordin
g
to
t
he
e
nter
pr
is
e
sche
dule
,
Je
ju
ci
t
y
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n of
a
M
on
it
ori
ng
-
C
ombine
d Sit
ing Sc
hem
e
for
Elec
tri
c Vehicl
e Ch
ar
ge
rs
(
Ju
ngho
on Lee)
5307
will
increase
the
num
ber
of
EVs
by
1.5
ti
m
es
nex
t
ye
ar,
and
s
o
will
the
ci
ty
wide
load.
Assum
ing
the
cl
us
te
r
-
le
vel
dem
and
increases
by
that
rati
o,
the
r
ecal
culat
ed
pe
r
-
cl
ust
er
loa
d
is
disp
la
ye
d
al
so
in
Fi
gure
1.
The
cl
us
te
rs
f
or
eca
ste
d
to
have
la
rg
e
r
than
0.5
in
the
occ
up
a
nc
y
rate
app
ear
a
ll
ov
er
the
ta
rget
area.
It
m
ea
ns
that
EV dr
i
ver
s
w
il
l wait
stat
ist
ic
a
ll
y
m
or
e than o
ne ou
t
of tw
o
ti
m
es.
3.2.
Gene
tic oper
at
i
on
Our
sc
hem
e
rep
rese
nts
a
sit
e sele
ct
ion
as
a bi
nar
y
inte
ger
ve
ct
or
a
s
s
how
n
in Figure
6. H
ere,
m
is
the
nu
m
ber
of
al
l
cand
i
dates
an
d
equ
al
s
t
o
the
l
eng
t
h
of
the
ve
ct
or
,
w
hile
n
is
that
of
sel
ect
e
d
locat
io
ns
de
no
te
d
by
1’s.
Each
lo
cat
ion
ca
nd
i
dat
e
is
associat
ed
with
sta
ti
c
inf
orm
ation
s
uch
a
s
la
ti
tu
de,
lo
ng
it
ud
e,
a
nd
cl
us
t
er
id
if
it
is
inclu
de
d
in
a
cl
us
te
r.
The
cl
us
te
r
i
d
m
akes
it
possi
ble
to
ref
e
r
t
o
the
cl
us
te
r
rec
ord
c
onsist
ing
of
t
he
nu
m
ber
of
m
e
m
ber
charger
s,
ave
rag
e
loa
d,
and
cl
us
te
r
ce
nt
ro
id
.
Fi
gure
6
al
so
s
hows
the
cr
os
s
ov
e
r
ope
rati
on
desig
ne
d
for
th
e g
eneti
c algo
r
it
h
m
. I
n
this exa
m
ple,
m
and
n
are 1
3
an
d
6,
re
sp
ect
ively
. The first p
art con
t
ai
ns
the
vect
or
f
or
t
ho
s
e
can
did
at
e
s
inclu
ded
in
a
cl
us
te
r
wh
il
e
t
he
la
tt
er
not.
A
s
two
pa
rts
are
assessed
dif
fere
ntly
,
cro
ss
over
oper
at
ion
s
ta
ke
place
t
wice,
na
m
el
y,
at
(
C
1
,
C
2
)
an
d
(
C
3
,
C
4
).
A
fter
swi
tc
hin
g
s
ub
st
rin
gs
,
t
he
nu
m
ber
of
1’s
is
highly
li
kely
dif
fer
e
nt
from
n
.
Th
en,
so
m
e
of
them
will
be
cha
nged
to
m
ake
the all
ocati
on
valid.
Figure
6
.
G
e
ne
ti
c o
pe
rati
on
As
for
t
he
fitn
ess
eval
uatio
n
of
a
c
hrom
os
om
e,
or
a
si
ng
le
intege
r
vecto
r,
the
e
ff
ect
of
e
ach
gene
is
interp
reted
as
s
how
n
in
Fig
ure
7.
I
f
a
can
did
a
te
is
inside
an
ex
ist
ing
cl
us
te
r
,
it
first draws
t
he
loa
d
in
a
cl
ust
er.
Fo
r
ex
am
ple, if
the num
ber
of
m
e
m
ber
charg
ers
is 2
a
nd the
av
era
ge
loa
d
i
s 1
.
2,
t
he
ad
diti
on
of
a
ne
w
node
i
n
that
cl
us
te
r
wil
l
cut
dow
n
the
aver
a
ge
loa
d
t
o
0.8
.
The
n,
no
t
j
us
t
reli
evi
ng
the
load
i
n
a
cl
us
te
r,
a
ne
w
ch
arg
e
r
can
abs
orb
the
char
gi
ng
dem
and
f
r
om
adj
acent
charger
s.
I
n
this
exam
p
le
,
the
cl
us
te
r
can
afford
to
sha
re
the
load fr
om
it
s n
ei
ghbors ou
tsi
de
o
f
it
s clu
ste
r by 0.2
x
3. It’
s call
ed
le
fto
ve
r.
Wh
il
e it
is im
po
ssible h
ow
m
uch
load
will
m
i
gr
at
e
to
nei
ghbo
rin
g
no
de
s,
it
can
be
ex
pected
t
ha
t
the
cl
os
e
r
to
a
c
harg
er,
the
m
or
e load
ca
n m
ov
e.
Figure
7
.
Ev
al
uation p
r
ocess
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
I
nt J
Ele
c
& C
om
p
En
g,
V
ol. 8, N
o.
6, Dece
m
ber
2
01
8
:
5303
–
5310
5308
A
li
nea
r
loa
d
m
igrati
on
is
de
sign
e
d
base
d
on
the
distan
ce
betwee
n
t
w
o
c
harger
s
or
cl
us
te
rs,
as
sh
ow
n
in
E
q.
(
1).
=
×
−
−
(1)
,
w
her
e
P
n
is
th
e
ov
e
rloa
de
d
dem
and
of
a
nei
ghbori
ng
c
harg
er
or
cl
us
te
r
,
a
nd
P
a
is
the
m
i
gr
at
a
ble
am
ou
nt.
[
L
,
U
]
is
the
ra
nge
a
new
c
harger
ad
diti
on
will
be
i
n
e
ff
ect
.
H
ere,
a
cl
us
te
r
is
co
ns
ide
re
d
a
s
ing
le
c
harge
r
with
it
s
centr
oid
act
in
g
as
a
po
sit
io
n
va
lue
of
a
c
ha
rger.
d
is
th
e
dist
ance
to
a
new
charger
a
nd
on
ly
char
ge
rs
res
idin
g
in [
L
,
U
]
w
il
l b
e con
si
der
e
d.
If
d
is
U
,
P
a
is 0,
w
hic
h
ind
ic
at
es that t
he
new charger
can
no
t t
ake lo
ad
fro
m
tha
t
charger
.
If
it
i
s
L
,
al
l
overlo
aded
am
ou
nt
c
an
be
s
har
e
d.
The
s
um
of
al
l
P
n
can
no
t
ex
ceed
the
am
ount
of
le
ftov
e
r.
The
f
it
ness
va
lue
of
a
new
cha
r
ger
ad
diti
on
will
be
th
e
s
um
of
the
loa
d
s
ha
re
d
in
a
cl
us
te
r
(
if
it
is
inside
a
cl
ust
er
)
an
d
the
loa
d
dr
a
w
n
f
ro
m
it
s
neig
hbors.
T
he
genet
ic
op
e
r
at
ion
s
it
erate
to
im
pr
ov
e
t
he
fitness
v
al
ue ge
ner
at
io
n by ge
ne
rati
on
with the
ob
j
e
ct
f
unct
io
n
a
nd en
c
od
i
ng sc
he
m
e.
4.
PERFO
R
MANC
E
ME
ASU
REME
NT
This
sect
io
n
m
easur
e
s
the p
er
form
ance
of
th
e
pro
po
se
d
sc
hem
e
by
a
proto
ty
pe
i
m
ple
m
entat
ion
us
i
ng
the
C
pro
gr
am
m
ing
la
ngua
ge
.
For
sim
plicit
y,
we
assum
e
that
we
can
se
le
ct
any
place
for
a
ne
w
ch
arge
r
instal
la
ti
on
. H
e
nce,
the d
ist
an
ce within a cluster d
oes no
t m
at
te
r.
Fo
r
n
cha
rg
e
rs
to add,
how
to assig
n
th
e
m
to
resp
ect
ive
cl
ust
ers
is
the
m
ain
pro
blem
.
The
def
a
ult
par
a
m
et
ers
are
as
fo
ll
ows:
The
nu
m
ber
of
ne
w
c
harger
s
is
50
a
nd
that
of
cl
us
te
rs
is
78,
der
i
ved
fro
m
the
current
placem
ent
in
Jeju.
T
he
po
pu
l
at
ion
s
iz
e
is
10
0,
wh
il
e
the
ge
netic
loop
it
erates
1,000
tim
es.
In
ad
di
ti
on
,
occ
upan
cy
rate
is
the
pr
oba
bili
ty
that
a
charger
is
us
ed
an
d
denotes t
he
c
ha
rg
i
ng d
em
and on a c
ha
rg
e
r [
12
]
.
The
first
e
xpe
rim
ent
m
easur
es
the
basic
beh
a
vior
of
t
he
ge
netic
it
erati
on.
Fi
g
ure
8
pl
ots
the
i
m
pr
ovem
ent
i
n
the
sta
nd
a
r
d
dev
ia
ti
on
in
the
occ
up
a
ncy
rate
of
each
c
harger
acc
ordi
ng
to
the
pr
ogress
of
gen
et
ic
it
erati
ons.
It
will
sho
w
how
eve
nly
the
c
ha
rg
i
ng
load
is
distrib
uted
over
the
ta
rg
et
a
rea
with
t
he
add
it
io
n
of
ne
w
c
harge
rs.
Cu
r
ren
tl
y,
t
he
st
and
a
r
d
dev
ia
ti
on
of
the
occ
upancy
rate
for
245
cha
rg
e
rs
is
8.7%
,
wh
il
e
the
total
aver
a
ge
is
36.
0
%.
W
it
h
a
n
a
dd
it
io
n
of
a
ne
w
cha
rger
to
a
heav
y
-
loa
ded
cl
us
te
r,
the
c
ha
rg
i
ng
load
is
distrib
ut
ed.
A
t t
he
first
stage of th
e
ge
netic
it
erati
on
, t
he
sta
ndar
d de
viati
on
is
10.56 %,
higher
tha
n
the
current
val
ue,
ind
ic
at
ing
tht
an
inap
pro
pr
i
at
e
assignm
ent
m
akes
wo
rs
e
the
diff
e
rence
in
the
util
i
zat
ion.
Howe
ver,
re
pe
at
ed
exec
utio
n
of
ge
netic
op
erati
on
s
im
pr
ov
es
the
pe
rform
ance,
reac
hi
ng
4.7
%
afte
r
25
0
it
erati
on
s.
Be
y
ond
t
his
po
i
nt, no m
or
e im
pr
ov
em
ent is o
bs
e
rv
e
d.
Figure
8
.
I
te
rat
ive im
pr
ov
em
ent
Nex
t
, F
ig
ur
e 9
sh
ows
t
he
ef
fe
ct
of
the popul
at
ion
siz
e
in g
e
netic
op
e
rati
on
.
A
la
r
ger
popu
la
ti
on
le
ads
to
a
bette
r
ge
ne
div
ersit
y.
Ac
tuall
y,
ou
r
im
ple
m
entat
ion
preven
ts
duplica
te
d
ch
ro
m
os
om
es
from
ta
kin
g
place
at
the
sam
e
t
i
m
e
fo
r
t
he
sa
ke
of
m
aking
the
po
pu
la
ti
on
set
m
or
e
di
ver
se
.
Mo
reov
er,
the
le
ng
t
h
of
a
chrom
os
om
e
is
78
,
na
m
ely,
the
nu
m
ber
of
cl
us
te
rs
.
It
is
qu
it
e
lon
g
an
d
po
ssibly
ho
sts
a
var
ie
ty
of
so
l
ut
ion
s.
Howe
ver,
the
f
igure
s
hows
th
at
the
popula
ti
on
siz
e
has
li
tt
le
eff
ect
on
th
e
sta
ndar
d
de
vi
at
ion
im
pr
ov
e
m
ent.
Eve
n
wit
h
50
chrom
os
om
es,
we
ca
n
ac
hiev
e
the
sam
e
i
m
pro
vem
ent
as
with
500
c
hro
m
os
o
m
es.
This
com
es
from
th
e sit
uation t
hat a
fe
w
c
lusters
dom
inate
the
whole
oc
cup
a
ncy
rate.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Desig
n of
a
M
on
it
ori
ng
-
C
ombine
d Sit
ing Sc
hem
e
for
Elec
tri
c Vehicl
e Ch
ar
ge
rs
(
Ju
ngho
on Lee)
5309
Figure
9
.
P
opul
at
ion
size e
ff
e
ct
Finall
y,
Figu
r
e
10
plo
ts
ho
w
the
propose
d
s
chem
e
can
stab
ly
i
m
pr
ov
e
th
e
load
distrib
ut
ion
with
a
new
cha
r
ger
i
ns
ta
ll
at
ion
.
T
h
e
ex
per
im
ent
changes
the
num
ber
of
new
charge
rs
fro
m
10
to
200.
In
eac
h
par
am
et
er
set
tin
g,
our
sc
hem
e
fin
ds
a
rea
s
on
a
ble
qu
al
it
y
sub
op
ti
m
al
s
olu
ti
on
with
1,0
00
it
erati
ons
.
T
he
sta
nd
a
rd
de
via
ti
on
sta
rts
f
rom
7.
26
%
wit
h
10
cha
r
ger
s
and
en
ds
up
at
just
1.6
%
wi
th
20
0
c
harger
s.
T
he
perform
ance cu
r
ve
is t
otall
y con
ti
nu
ous, ha
ving
no ex
ce
pt
ion
al
os
ci
ll
at
ion
s
or sp
i
kes.
Figure
10
Cha
r
ger ad
diti
on
5.
CONCL
US
I
O
N
In
this
pa
per,
we
ha
ve
pr
e
se
nted
a
co
ncep
t
ual
desig
n
of
a
sit
ing
schem
e
fo
r
pu
blic
EV
charge
rs,
m
ai
nly
based
on
the
analy
si
s
res
ult
of
c
ha
rg
e
r
m
on
it
or
i
ng
stream
s.
A
ge
netic
al
gor
it
h
m
is
ex
plo
i
te
d
to
ov
e
rc
om
e
the
diff
ic
ulty
in
es
tim
a
ti
ng
the
ef
fect
of
a
c
ha
r
ge
r
placem
ent,
wh
il
e
a
fitnes
s
functi
on
is
de
fined
f
or
the
eval
uation
of
a
cha
rger
locat
ion
acc
or
ding
to
w
heth
er
a
c
harger
i
s
inside
of
an
y
cl
us
te
r
or
not.
T
he
accum
ulati
on
of
m
or
e
m
on
it
or
i
ng
se
ries
w
il
l
ref
ine
our
evaluati
on
m
od
el
an
d
re
veal
the
eff
ect
o
f
a
new
charger
instal
la
ti
on
,
m
aking
the
pro
cess
of
charger
sit
ing
a
cy
berp
hysic
al
syst
e
m
app
r
oa
ch
[
13]
.
Mo
r
eov
e
r
,
our
researc
h
te
a
m
is
dev
el
opi
ng
a
var
ie
ty
of
business
m
od
el
s
base
d
on
t
he
m
assive
dat
a
analy
sis,
incl
ud
i
ng
the
ad
ver
ti
sem
ent
an
d
tourist
good
s
trade
duri
ng
the
le
ng
t
hy
chargin
g
ti
m
e.
Partic
ularly
,
we
are
pla
nnin
g
to
integrate
re
ne
wab
le
e
nergy
into
the
c
harger
operat
ion
,
i
nclu
ding
locat
ion
det
erm
inati
on
,
ba
tt
ery
charger/
discha
rg
e
sch
e
duli
ng,
and the
li
ke [1
4].
ACKN
OWLE
DGE
MENTS
This
researc
h wa
s s
upporte
d by the
2018
sci
entifi
c prom
otion p
r
ogram
f
unde
d by Je
ju
N
at
ion
al
Uni
ver
s
it
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
I
nt J
Ele
c
& C
om
p
En
g,
V
ol. 8, N
o.
6, Dece
m
ber
2
01
8
:
5303
–
5310
5310
REFERE
NCE
S
[1]
C.
Deve
ld
er,
N.
Sadeghi
anpour
hamam
i,
M.
Strobbe,
N.
Ref
a,
“
Quanti
f
y
ing
Flexi
bilit
y
in
EV
Chargi
ng
as
DR
Potent
ial:
Anal
ysis
of
Rea
l
-
W
orld
Data
se
ts,
in
I
EE
E
In
te
rnation
al
Confe
ren
ce
o
n
Smar
t
Gr
id
C
omm
unic
ati
ons
,
2016.
[2]
E.
Kara
,
J.
Mac
donal
d,
D.
Bl
ack,
M.
Berge
s,
G
.
Hug,
S.
Kili
c
c
ote
,
“
Esti
m
ating
the
Bene
f
it
s
of
El
ectri
c
Vehicl
e
Sm
art
Chargi
ng
at
Non
-
Reside
nt
ia
l
Lo
ca
t
ion:
A
Data
-
Drive
n
Ap
proa
ch,
Appl
i
ed
En
ergy
,
vol
.
155
,
no.
1,
pp.
515
-
525,
2015
.
[3]
F.
He,
Y.
Yin,
J.
Zhou,
“
Deplo
y
ing
Publ
ic
Cha
rging
Stat
ions
f
or
El
e
ct
ri
c
Vehi
cl
es
on
Urban
Road
Networks,
Tr
anspo
rt R
ese
a
rch
Part
C
,
pp
.
2
27
-
240,
2015
.
[4]
J.
Yam
az
aki,
D.
Yos
hiro,
H.
Fu
kuhar
a,
T
.
Ha
y
a
shi,
“
A
Com
pre
hen
sive
Da
ta
Pr
oce
ss
ing
Approa
ch
to
the
Futur
e
Sm
art
Grid,
in
4th
Inte
rnationa
l
Confe
ren
ce
on
Re
newab
le
En
e
rgy
Re
search
a
nd
Appl
ic
a
ti
ons
,
2015,
pp.
1033
-
1036.
[5]
J.
Le
e
,
G.
Park,
Y.
Han,
S.
Yoo,
“
Big
Data
Ana
l
y
sis
for
an
El
e
c
tri
c
Vehi
cle
Cha
rging
Infra
struc
t
ure
using
Open
Data
and
Softwa
re,
in
ACM eE
ne
rgy
,
Ma
y
2017,
pp.
252
-
253.
[6]
J.
Le
e
and
G.
Park,
"Inte
gr
ated
Coordination
of
El
e
ct
ri
c
V
ehi
c
le
and
R
en
ewa
ble
Ene
rg
y
Ge
ner
ation
in
a
Microgr
id,
"
Internat
ion
al
Journ
al of
E
lectr
i
ca
l
an
d
Com
pute
r
Eng
ine
er
ing, vol
.
7,
no.
2
,
pp
.
706
-
71
2,
2017
.
[7]
Florida
Pow
er &
Li
gh
t, Com
pan
y
Siti
ng
Plug
-
in E
le
c
tri
c
Vehi
cl
e
C
har
ging,
www.
F
PL.
com/ elect
ri
c
ve
hi
cl
es
.
[8]
Tra
nsporta
ti
on
and
Cli
m
at
e
In
i
ti
ative
,
Siti
ng
a
nd
Design
Guidel
in
es
for
El
e
c
tri
c
Vehi
cle
Suppl
y
Equi
pm
ent,
htt
ps://www.trans
portati
onandcl
i
mate.
org/sit
es/d
ef
ault/fi
le
s/E
V
S
i
ti
ng
and
Design
Guidel
in
es.
pdf
.
[9]
H.
Cai
,
X
.
Jia
,
A.
Chiu,
X
.
Hu,
M.
Xu,
“
Siti
ng
Publ
ic
El
e
ct
ri
c
Ve
hic
l
e
Chargi
ng
S
ta
ti
ons
in
B
ei
j
in
g
using
Big
-
Dat
a
Inform
ed
Tra
v
el Pat
t
ern
s of Ta
xi
Fleet,
Tr
anspor
t
Re
search
Part
D
,
vol
.
33
,
pp
.
39
-
46,
2014
.
[10]
N.
Shahra
ki,
H.
Cai
,
M.
Turkay
,
M.
Xu,
“
Optima
l
Loc
a
ti
on
of
Elec
tr
ic
Public
Ch
arg
ing
Stat
ions
using
Rea
l
W
orld
Vehic
l
e
T
rav
e
l
P
at
t
ern
s,
Tr
anspor
tat
ion
R
ese
arch
Part
D
,
vol
.
41
,
pp.
165
-
176,
201
5.
[11]
A.
Ovall
e
,
A.
H
ably
,
S.
Bac
h
a,
G.
Ramos
,
J.
Hos
sain,
“
Escor
t
E
volut
iona
r
y
G
a
m
e
D
y
namics
Approac
h
for
In
tegra
l
Loa
d
Mana
g
ement
of
El
e
ct
ri
c
Vehic
l
e
Fle
et
s,
IEE
E
Tr
ansactions
on
Industrial
Elec
troni
cs
,
v
ol.
64
,
issue
2,
pp.
1358
-
1369,
201
7.
[12]
J.
Le
e
and
G.
Pa
rk,
"S
erv
ic
e
T
ime
Anal
y
sis
for
E
le
c
tri
c
Vehi
cle
Chargi
ng
Infra
st
ruc
ture,"
Inte
rna
ti
onal
Journal
of
El
e
ct
rica
l
and
C
omputer
Engi
n
e
ering
,
vo
l. 8, no. 2, pp.
818
-
824
,
2018.
[13]
A.
Am
at
o,
R.
Aversa,
B.
Ma
rti
no,
S.
Ven
ticinque,
“
A
C
y
ber
Ph
y
sic
al
S
y
stem
of
Sm
art
Microgr
ids,
”
in
Inte
rnational
Co
nfe
renc
e
on
Networ
k
-
Based
Info
rm
ati
on
Syste
ms
,
2016
,
pp
.
165
-
172.
[14]
B.
Banht
h
asit
,
C.
Jam
roe
n,
S.
D
ec
hanup
apr
ittha
,
Optimal
Gene
r
at
ion
Sch
edul
in
g
of
Pow
er
S
y
st
em
for
Maximum
Rene
wabl
e
En
er
g
y
Harv
esti
ng
a
nd
Pow
er
Losses
Minim
iz
ation,
”
Inte
rnational
Jo
urnal
of
Elec
tri
c
al
and
Compute
r
Engi
ne
ering
,
vol
.
8
,
no
.
4
,
pp
.
195
4
-
1966,
2018
.
Evaluation Warning : The document was created with Spire.PDF for Python.