Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
1
,
Jan
uar
y
201
9
,
pp.
2
72
~
2
78
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
1
.pp
272
-
278
272
Jou
rn
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Improve
ment of
CH el
ec
t
ion in thre
e
-
l
evel
hetero
gene
ous WS
N
Jong
-
Y
ong Le
e
1
, Daesu
ng Le
e
2
1
Inge
nium
Coll
e
ge
of lib
era
l
ar
ts,
Kw
angWoon Unive
rsit
y
,
Seoul
01897,
Kore
a
2
Depa
rtment of
Com
pute
r
Engi
n
ee
ring
,
C
at
hol
ic
Univer
sit
y
of
Pus
an,
Busan
4625
2,
Kore
a
A
rticl
e In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
2
, 2
018
Re
vised N
ov 3
, 2018
Accepte
d Nov
1
9
, 201
8
A
W
ire
le
ss
Sen
sor
Network
is
a
wire
le
ss
l
y
co
nfigure
d
Sensor
Node
with
li
m
it
ed
power
such
as
a
bat
t
er
y.
The
r
e
are
m
an
y
W
ire
l
ess
Sens
or
Network
Protocol
s
to
inc
r
ea
se
ene
rg
y
eff
i
ci
en
c
y
,
among
which
LE
ACH
Protocol
and
SEP
are
t
y
pi
ca
l
.
The
LE
ACH
Protocol
is
m
ai
nl
y
used
for
hom
ogene
ous
sensor
net
work
s
with
the
sa
m
e
ini
tial
ene
r
g
y
,
and
SEP
i
s
used
for
het
ero
g
ene
ous
sensor
net
works
with
diffe
ren
t
in
it
ial
ene
rg
i
es.
In
the
ca
se
of
SEP
-
E,
anot
her
het
ero
g
ene
ous
sensor
with
different
ini
t
ia
l
en
er
g
y
is
adde
d
.
SEP
and
SEP
-
E
provide
a
highe
r
proba
bilit
y
o
f
Cluste
r
Hea
d
el
e
ct
ion
fo
r
node
t
y
pes
with
m
ore
ene
rg
y
than
Norm
al
Nodes
.
Since
th
e
cur
re
nt
residual
ene
rg
y
of
the
n
ode
is
not
conf
i
rm
ed,
ev
en
if
th
e
en
erg
y
is
low,
the
Cluste
r
Hea
d
m
a
y
be
el
e
ct
ed
b
ecause
of
t
he
node
t
y
pe.
In
thi
s
paper,
conside
ring
th
e
residua
l
en
erg
y
of
a
node
when
a
Cluste
r
Hea
d
is
el
ec
t
ed,
we
i
ncr
ea
se
th
e
proba
bil
i
t
y
of
elec
t
ing
a
C
luste
r
He
ad
w
it
h
m
ore
r
esid
ual
ene
r
g
y
.
Cluste
r
Hea
d
co
nsum
es
a
lot
of
ene
rg
y
.
A
node
with
a
lot
of
r
esi
dual
en
er
g
y
is
el
e
cted
as
a
Cluste
r
Hea
d
,
s
o
the
n
et
work
l
ife
ti
m
e
ca
n
be
used
for
a
long
ti
m
e
.
Ke
yw
or
ds:
Cl
us
te
r
Ele
ct
Heter
og
e
ne
ou
s
Level
Pr
ot
oc
ol
WSN
Copyright
©
201
9
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
:
Daes
ung
Lee
,
Dep
a
rtm
ent o
f C
om
pu
te
r
E
ng
i
neer
i
ng,
Ca
tho
li
c Unive
rsity
o
f
P
us
an
, B
us
an
46
252,
Korea
.
Em
a
il
:
ds
le
e@cup.ac.
kr
1.
INTROD
U
CTION
A
W
irel
ess
S
ens
or
Net
wor
k
ref
e
rs
to
a
network
in
wh
ic
h
Senso
r
Nodes
w
it
h
colle
ct
ion
an
d
transm
issi
on
/re
cepti
on
f
unct
ion
s
are
wirel
essly
config
ured.
Be
ca
us
e
t
he
Se
nsor
N
odes
a
re
c
onfigure
d
wirelessl
y,
eac
h
no
de
ha
s
a
lim
it
ed
energy,
su
c
h
as
a
ba
tt
ery.
Sensor
Nodes
ca
n
be
in
sta
ll
ed
in
a
ho
m
e,
a
natur
al
e
nv
i
r
on
m
ent,
a
ro
a
d,
et
c.
to
m
easur
e
or
obser
ve
changes
in
t
he
env
ir
onm
ent.
These
net
wor
ks
are
us
e
d
in
m
any
a
reas
of
ever
y
da
y
li
fe
and
m
ake
our
li
ves
m
or
e
conven
ie
nt.
I
n
ad
diti
on
,
si
nc
e
the
Sensor
Nodes
are
wirelessl
y
config
ur
e
d,
t
he
y
can
be
instal
le
d
wh
e
re
pe
ople
can
not
act
ually
go
.
W
i
rel
ess
Sensor
Networks
hav
e
t
hese
ad
va
ntages
,
but
th
ey
al
so
hav
e
draw
bac
ks
.
U
nl
ike
a
wire
d
ne
twork
,
a
W
irel
ess
Sensor
Net
work
op
e
rates
with
a
lim
it
ed
po
w
er
source
s
uch
as
a
batte
ry
fo
r
eac
h
Sens
or
Node.
If
the
batte
ry
is
exh
austed
,
the
Senso
r
N
ode
will
no
l
onge
r
in
f
un
ct
i
on.
Ther
e
f
or
e,
i
n
orde
r
to
m
ai
ntain
a
netw
ork
for
a
long
ti
m
e,
e
nergy
consum
ption
s
hould be
m
ini
m
iz
ed
by opti
m
iz
ing
the e
ne
rg
y c
onsu
m
ed.
In
t
he
W
i
rele
ss
Se
ns
or
Net
work,
the
re
a
r
e
a
ho
m
og
e
ne
ous
W
i
reless
Senso
r
Netw
ork
a
nd
a
heter
og
e
ne
ou
s
W
irel
ess
Sensor
Net
work.
In
t
he
ho
m
ogeneous
Wirele
ss
Se
ns
or
Network,
al
l
the
Sens
or
Nodes
are
t
he
sam
e.
The
he
te
rogen
e
ous
W
i
reless
Sensor
Net
work
is
a
m
ixtur
e
of
Sensor
N
odes
ha
ving
diff
e
re
nt
init
ia
l
energy.
The
re
are
m
any
W
ir
el
ess
Senso
r
N
et
work
Protoc
ols
to
i
ncr
ease
the
ene
r
gy
eff
i
ci
ency
of
the
netw
ork
.
[1
-
3]
Ty
pical
Pr
ot
oco
ls
are
LEAC
H
Proto
col
[4
]
an
d
SE
P
[5
]
.
T
he
LE
ACH
P
ro
t
oco
l
is
a
cl
us
te
r
-
base
d
P
ro
t
oco
l
that
el
e
ct
s
a
Cl
us
te
r
H
ead
no
de
cy
cl
ic
al
ly
us
ing
a
C
luster
He
ad
el
e
ct
ion
f
orm
ula.
SEP
is
a
Protoc
ol
f
or
heter
ogene
ous
W
i
reless
Se
ns
or
Netw
orks
base
d
on
L
E
ACH
Protoc
ol.
The
S
EP
distr
ibu
te
s
the
Cl
us
te
r
He
ads
acc
ordin
g
to
the
node
ty
pe
us
in
g
the
di
ff
ere
nt
Cl
us
te
r
Hea
d
el
ect
io
n
e
qu
at
io
ns
f
or
the
N
orm
al
No
de
an
d
A
dvan
ced
N
ode,
w
hich
has
m
or
e
init
ia
l
energy
tha
n
th
e
N
or
m
al
Nod
e.
SE
P
-
E
[
7]
a
dd
e
d
an
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ov
eme
nt
of
CH
el
ect
ion i
n
th
ree
-
le
vel
he
te
ro
ge
ne
ou
s
WSN
(
Jo
ng
-
Y
ong Lee
)
273
interm
ediat
e
no
de
with
m
or
e
energy
to
t
he
SEP,
a
nd
al
so
gav
e
dif
fer
e
nt
el
ect
ion
f
or
m
ulas
dep
e
ndin
g
on
t
he
node
ty
pe.
H
oweve
r,
in
t
he
case
of
SEP
or
SE
P
-
E
,
t
he
pro
ba
bili
ty
of
e
le
ct
ion
is
i
ncr
e
ased
for
node
s
with
m
or
e
init
ia
l
ener
gy
c
on
si
der
i
ng
only
the
no
de
ty
pe
withou
t
the
resid
ual
e
nergy.
T
her
e
f
ore,
the
Cl
us
te
r
Head
node
m
ay
b
e el
ect
ed
a
node
wi
th low ene
r
gy.
In this
pa
per,
we pr
opos
e
a
m
et
ho
d
t
o
s
olve
thes
e
pro
blem
s.
2.
RESEA
R
CH MET
HO
D
2
.
1.
He
ter
og
e
neou
s
N
e
twor
k
In
a
hete
roge
ne
ous
wi
reless
sens
or
netw
or
k,
t
he
e
nergy
of
al
l
the
no
de
s
is
not
the
s
a
m
e
as
the
ho
m
og
e
neous
wireless
se
nso
r
net
work
w
he
re
the
e
nergy
of
al
l
the
node
s
is
the
sam
e.
m
%
of
al
l
no
de
s
ar
e
Adva
nced
Nod
es
that
has
m
or
e
e
nergy
tha
n
a
Norm
al
Node.
That
is,
w
hen
the
total
num
ber
of
node
s
is
,
×
(
1
−
)
N
or
m
al
No
de
s
ha
ving
init
ia
l
energy
val
ue
0
exist,
a
nd
t
her
e
a
re
×
A
dv
anced
N
odes
hav
i
ng m
or
e in
it
ia
l energ
y t
ha
n Norm
al
N
odes.
2
.
2
.
Thre
e
-
le
vel He
terogen
eous
Ne
twork
A
th
ree
-
le
vel
heter
og
e
ne
ou
s
wireless
sen
s
or
net
work
is
a
netw
ork
of
three
ty
pe
s
of
node
s
with
diff
e
re
nt
ene
r
gi
es.
%
of
al
l
node
s
ar
e
Adva
nced
N
odes
wi
th
m
or
e
e
nergy
tha
n
Norm
al
Nodes
an
d
0
%
of
the
A
dvanc
ed
Nodes
are
sup
erno
des.
I
n
othe
r
words,
if
the
total
num
ber
of
nodes
is
,
the
re
are
×
(
1
−
)
Norm
al
No
des
with
a
n
init
ia
l
energy
val
ue
0
and
×
(
1
−
0
)
A
dvance
d
Nodes
se
ve
ral
tim
es
m
or
e
than
the
i
niti
al
en
erg
y.
And
th
ere
a
re
×
×
(
1
−
0
)
supe
r
nodes
t
hat
are
ti
m
es
as
m
uch
as
t
he
init
ia
l energ
y.
2
.
3
.
LE
A
CH
Protoc
ol
The
L
EAC
H
P
ro
t
oco
l
is
a
Cl
us
te
r
-
base
d
routing
Proto
col.
The
Senso
r
Fi
el
d
is
di
vid
e
d
i
nto
Cl
ust
ers
,
and
t
her
e
is
on
e
Cl
us
te
r
Hea
d
node
pe
r
Cl
ust
er.
The
L
EAC
H
Protoc
ol
sto
chasti
cal
ly
el
e
ct
s
the
Cl
us
te
r
Hea
d
and
g
ives
al
l
th
e
nodes
in
the Cl
us
te
r
the
op
port
un
it
y
to
be
e
le
ct
ed
as
the
Cl
us
te
r
Hea
d.
T
he
LE
ACH
P
r
oto
c
ol
has
a
set
-
up
phase
in
wh
ic
h
C
luster
Hea
d
el
ect
ion
s
are
m
ade
and
a
ste
ady
-
sta
te
ph
ase
in
wh
ic
h
tran
sm
is
sion
is
act
ually
eff
e
ct
ed.
The
Cl
ust
er
Head
co
nsum
es
a
lot
of
energy
beca
use
it
colle
c
ts
th
e
data
of
the
m
e
m
ber
nodes
in
the
C
luster
an
d
tran
sm
it
s
it
to
the
Ba
se
Stat
ion
.
Wh
e
n
a
node
i
s
con
ti
nu
ou
sly
el
ect
ed
as
a
C
luster
Head,
th
e
e
nergy
of
t
he
node
is
co
ns
um
ed
i
m
m
ediat
ely.
Ther
e
fore,
al
l
no
des
ca
n
be
el
ect
ed
as
a
Cl
us
te
r
Head
by u
si
ng the t
hresh
old eq
uatio
n
in
the
LEAC
H
P
ro
t
oco
l
. In t
he
set
-
up phas
e, the
Cl
us
te
r Head
is ele
ct
ed
u
si
ng
the foll
owin
g
e
qu
at
io
n.
(
)
=
{
1
−
(
mod
1
)
if
∈
0
ot
h
erwise
(1)
As
s
how
n
i
n
(
1),
is
the
curr
ent
r
ound.
If
t
he
set
is
e
m
pty
and
t
her
e
a
re
no
m
or
e
node
s
that
can
beco
m
e
Cl
us
te
r
Hea
ds,
pu
t
al
l
rem
ai
nin
g
en
erg
y
nodes
in
t
he
set
so
t
hat
they
bec
om
e
Clu
ste
r
Hea
ds
.
S
et
is
the
set
to
wh
ic
h
nodes
tha
t
are
no
t
el
ect
ed
to
the
Cl
us
te
r
Hea
d
bel
ong.
If
the
ra
ndom
nu
m
ber
is
le
ss
tha
n
the
th
res
ho
l
d
(
)
,
the
node
is
el
e
ct
ed
as
the
Cl
ust
er
Head
in
th
e
cu
rr
e
nt
r
ound.
O
nce
al
l
the
Cl
us
te
r
Hea
ds
are
el
ect
ed,
th
e
m
e
m
ber
nodes
in
t
he
Cl
us
te
r
tra
ns
m
it
the
data
to
t
he
Cl
us
te
r
He
ad.
T
he
Cl
us
t
er
He
a
d
transm
it
s
the
data
of
the
re
cei
ved
m
e
m
be
r
nodes
a
nd
it
s
own
data
to
the
Ba
se
Station
.
Fi
gure
1
sh
ow
s
flo
wch
a
rt of L
EACH
.
Figure
1. Flo
w
char
t
of LE
AC
H
protoc
ol
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
2
7
2
–
2
7
8
274
2
.
4
.
SEP
SEP
is
s
uitable
for
heter
og
e
ne
ous
net
wor
ks
and
gi
ves
a
dif
fer
e
nt
Cl
us
te
r
Head
pro
ba
bili
ty
equ
at
io
n
dep
e
ndin
g
on
t
he
ty
pe
of
no
de
s.
Heter
ogene
ou
s
netw
orks
a
re
dif
fer
e
nt
f
rom
ho
m
og
ene
ous
net
wor
ks
in
wh
ic
h
al
l nodes
ha
ve t
he
sam
e energ
y. A n
od
e
w
it
h m
or
e ener
gy t
han a
Norm
al
Node
is
call
ed a
n Adva
nced
Node.
In
case
of
Cl
ust
er
Head
el
ect
ion,
it
is
el
ect
e
d
by
pro
ba
bili
ty
equ
at
io
n
li
ke
LEACH
P
r
otoc
ol.
I
n
case
of
SE
P,
wei
ght
is
app
li
ed
to
Adva
nce
d
Node
with
hi
gh
e
r
ene
r
gy
to
increa
se
the
el
ect
ion
pro
ba
bili
ty.
The
wei
gh
te
d
pro
bab
il
it
y
fo
r
m
ula
of
the
N
orm
al
No
de
a
nd
the
weigh
te
d
pro
bab
il
it
y
fo
r
m
ula
of
the
A
dvance
d
Node
a
re sh
own
in
(2).
=
1
+
×
=
(
1
+
)
1
+
×
(2)
The
Cl
us
te
r
H
ead
el
ect
io
n
prob
a
bili
ty
(
)
an
d
(
)
us
in
g
t
he
weig
hted
pro
ba
bili
t
y
equ
at
i
on
of (2)
are sh
ow
n
in
(3).
(
)
=
{
1
−
(
mod
1
)
if
∈
0
ot
h
erwise
(
)
=
{
1
−
(
mod
1
)
if
∈
′
0
ot
h
erwise
(3)
As
sho
wn
i
n
(
3),
is
the
curr
ent
rou
nd
a
nd
is
the
set
of
N
or
m
al
No
des
t
hat
are
not
Cl
ust
er
Hea
d
within
1
r
ound o
f
Ep
oc
h.
And
(
)
is
a
thr
esh
old app
li
ed
to
×
(
1
−
)
N
orm
al
No
des
.
T
hi
s
ens
ur
es
that
N
or
m
al
N
od
e
s
ar
e
Cl
us
t
er
Head
s
e
xactl
y
on
ce
eve
ry
1
p
opt
×
(
1
+
×
)
r
ound
of
E
po
c
h.
T
he
nu
m
ber
of
aver
a
ge
Cl
us
te
r
Hea
ds
of
th
e
Ep
och
is
×
(
1
−
)
×
.
G
is
a
set
of
Adva
nced
N
odes
that
are
not
Cl
us
te
r
Hea
ds
within
the
la
st
1
r
ound
of
E
poch.
A
nd
(
)
is
a
t
hr
es
hold
a
pp
li
ed
to
×
A
dvanc
ed
Nodes.
This
e
ns
ures
th
at
the
Adva
nced
N
odes
are
Cl
us
te
r
Head
s
e
xactl
y
on
ce
e
ve
ry
1
×
(
1
+
×
)
1
+
r
ounds.
This
per
i
od is
def
i
ned as a
Sub
-
e
poch
.
Each
E
poc
h
ha
s
1
+
su
b
-
e
po
c
h.
As
a
re
su
lt
,
A
dv
a
nce
d
Node
s
in
a
hete
roge
neous
ep
oc
h
be
com
e
Cl
us
te
r
Hea
ds
exactl
y
1
+
tim
es.
The
ave
ra
ge
num
ber
of
Cl
ust
er
Hea
ds
per
r
ound
of
hete
rogen
e
ous
ep
oc
h
is
×
×
.
×
(
1
−
)
×
+
×
×
=
×
(4)
As
sho
wn
i
n
(4),
the
s
um
of
the
a
vera
ge
num
ber
of
Cl
us
te
r
Hea
ds
pe
r
N
orm
a
l
Node
pe
r
heter
og
e
ne
ou
s
epo
c
h
a
nd
t
he
aver
a
ge
nu
m
ber
of
Cl
us
te
r
He
ads
pe
r
A
dva
nc
ed
N
ode
pe
r
s
ub
-
e
po
c
h
is
eq
ual
to
the av
e
ra
ge num
ber
o
f
Cl
us
te
r
H
eads
p
e
r
e
poch
p
e
r rou
nd. Fi
gure
2
s
hows
num
erical
ex
am
ple.
Figure
2. A
nu
m
erical
ex
am
p
le
f
or a
heter
og
eneous
n
et
wor
k
2
.
5
.
SEP
-
E
SEP
-
E
is
a
S
EP
-
base
d
Protocol
f
or
th
ree
-
le
vel
het
er
ogeneous
se
ns
or
networks.
Ex
ist
ing
SEP
s
consi
der
only
two
ty
pes
of
Sensor
Nodes
with
dif
fer
e
nt
energy,
but
S
EP
-
E
a
dds
an
oth
e
r
node
ty
pe
with
diff
e
re
nt
ene
r
gy.
It
has
N
orm
al
No
de,
a
nd
A
dv
a
nce
d
Node
that
has
m
or
e
ene
r
gy
than
Norm
al
Node,
interm
ediat
e n
od
e
in
w
hich h
as m
or
e ene
rg
y
than A
dvance
d Node.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ov
eme
nt
of
CH
el
ect
ion i
n
th
ree
-
le
vel
he
te
ro
ge
ne
ou
s
WSN
(
Jo
ng
-
Y
ong Lee
)
275
In
ste
a
d
of
in
(1)
of
LE
AC
H
P
ro
t
oco
l,
is
assig
ned
t
o
Norm
al
No
de,
is
assigne
d
t
o
Adva
nced N
od
e, and
is assi
gned
to I
nterm
e
diate
nod
e
.
=
1
+
(
−
0
(
−
)
)
=
(
1
+
)
1
+
(
−
0
(
−
)
)
=
(
1
+
)
1
+
(
−
0
(
−
)
)
(5)
The
rati
o
of
Adva
nced
Node
a
nd
I
nter
m
ediat
e
node
is
%
of
th
e
total
node
.
The
rati
o
of
interm
ediat
e n
od
e
s is
0
%
of
th
e Adv
a
nce
d N
od
e
.
3.
PROP
OSE
D MET
H
OD
Since
the
Cl
ust
er
Hea
d
el
ect
ion
th
reshold
of
the
LE
ACH
Pr
ot
oc
ol
does
no
t
c
onside
r
t
he
resid
ua
l
energy
of
t
he
node,
a
node
with
ins
uffici
e
nt
ene
r
gy
to
operate
as
a
Cl
us
te
r
Hea
d
node
ca
n
be
el
e
ct
ed.
To
s
olv
e
t
hese
pro
blem
s,
M.
J.
Ha
ndy
[6
]
has
im
pr
oved
th
e
Cl
us
te
r
He
ad
el
ect
ion
t
hresh
old
to
ta
ke
into
account t
he
r
e
s
idu
al
e
nergy
of the
nodes
. Ha
nd
y'
s prop
os
al
is sh
own
i
n (6)
.
(
)
×
(6)
As
sho
wn
i
n
(
6)
is
T
hr
es
hold
co
ns
ide
rin
g
t
he
resi
du
al
e
ne
rg
y
that
m
ultip
li
es
the
(
)
valu
e
of
the
LEAC
H
Proto
col
by
the
resi
du
al
e
nergy
rat
io
of
the
node.
(
)
has
a
value
be
tween
0
a
nd
1.
I
n
the
LE
AC
H
Pr
ot
oc
ol,
the
r
andom
nu
m
ber
of
eac
h
no
de
is
com
par
ed
with
the
t
hr
es
ho
l
d
val
ue,
a
nd
if
the
val
ue
of
t
he
thres
ho
l
d
e
qu
a
ti
on
is
gr
eat
er
,
it
is
el
ect
ed
as
the
Cl
us
te
r
Head.
T
her
e
fore,
if
t
he
th
res
ho
l
d
value
e
xc
eeds
1,
the
Cl
us
te
r
He
ad
is
al
ways
e
le
ct
ed
,
s
o
t
he
t
hr
es
hold
val
ue
sho
uld
not
e
xc
eed
1.
I
n
(6),
is
t
he
init
ia
l
energy
of
t
he
node,
a
nd
is
t
he
c
urre
nt
e
nergy
of
the
node
.
T
her
e
f
or
e
,
as
the
resi
dual
e
nergy
of
t
he
node
bec
om
es
sm
a
ll
er,
the
va
lue
of
the
thr
esh
old
e
qu
at
i
on
beco
m
es
cl
os
er
to
0
a
nd
it
is
not
el
ect
ed
as
th
e
Cl
us
te
r Head
.
In
t
he
case
of
e
xisti
ng
he
te
rogen
e
ous
sens
or
netw
or
k
P
ro
t
oco
ls
,
t
he
Cl
us
te
r
H
ead
el
ect
ion
pro
bab
il
it
y
is
i
ncr
ease
d
to
ward
the
init
ia
l
ener
gy
de
pe
nd
i
ng
on
the
node
ty
pe.
Be
cause
it
has
a
lot
of
init
ia
l
energy,
it
is
su
it
able
as
a
Cl
us
te
r
Head.
Howev
e
r,
since
th
e
pr
io
rity
of
the
el
ect
ion
pr
ob
abili
ty
is
determ
ined
accor
ding
to
the
node
ty
pe,
i
neffici
ent
cl
us
te
rs
are
f
or
m
ed
wh
e
n
these
ty
pes
of
nodes
a
re
far
f
ro
m
the
Ba
se
Stat
ion
.
Als
o,
because
the
res
idu
al
ene
rg
y
is
no
t
co
ns
ide
re
d,
the
Cl
us
te
r
Head
ca
n
be
el
ect
ed
first
beca
us
e
of
the
node
ty
pe
even
the
act
ual
residu
al
e
nerg
y
is
insu
ff
ic
ie
nt
.
In
this
pa
per,
we
pro
po
se
a
n
el
ect
ion
prob
a
bili
ty
of form
ula to solve t
his
prob
l
e
m
as it fo
ll
ow
s.
We
first
c
onsi
der
the
resi
du
a
l
energy
i
n
(
)
of
the
existi
ng
L
EACH
P
ro
t
oc
ol.
In
the
case
of
M.J.
Handy,
the
Cl
us
te
r
Hea
d
el
e
ct
ion
pro
ba
bili
ty
was
adjusted
acco
rd
i
ng
to
the
energy
rati
o
by
m
ulti
plyin
g
th
e
el
ect
ion
e
nergy by the
ℎ
ℎ
.
The
sm
aller
t
he
resid
ual
energy
of
a
node
,
the
le
ss
the
Cl
us
te
r
Head
el
ect
ion
pr
oba
bili
ty
.
Howe
ver,
if
th
e
rem
ai
nin
g
en
erg
y
of
t
he
nodes
is
re
duce
d
as
the
netw
ork
procee
ds
,
t
he
Cl
us
te
r
Hea
d
e
le
ct
ion
pro
bab
il
it
y
is
reduce
d
as
a
whole,
an
d
th
e
Cl
us
te
r
Head
m
a
y
no
t
be
el
ect
ed.
To
ov
erco
m
e
this
pr
ob
le
m
,
resid
ual
ene
rgy
rati
o
of
the
node
is
cha
nged
base
d
on
t
he
m
axi
m
u
m
residu
al
e
nergy
of
the
al
ive
nodes
in
th
e
Sensor
Fiel
d, not ba
sed
on t
he
init
ia
l ener
gy. (
to
(
)
)
The pr
opose
d
Cl
us
te
r Head
e
le
ct
ion
th
res
hold for
m
ula app
ly
ing
im
pr
ov
e
m
ents is sho
w
n
in
(7).
(
)
×
(
)
(7)
4.
SIMULATI
O
N AND
RES
U
LT
S
4
.
1
.
Radio
M
od
el
Wh
e
n
data
is
transm
itted
f
rom
the
Sensor
Node,
it
r
eq
ui
res
tr
ansm
issi
on
e
ne
rg
y
−
(
)
a
nd
a
m
plific
at
ion
energy
−
(
,
)
de
pendin
g
on
the
di
sta
nce.
When
receivin
g
data
from
a
Sens
or
N
od
e
,
it
r
eq
uir
es
rece
iving ene
r
gy
−
(
)
.
The flo
wch
a
rt
of the
rad
i
o
m
od
el
is s
how
n
in
Fig
ur
e
3.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
2
7
2
–
2
7
8
276
Figure
3. Ra
di
o
m
od
el
Energy
co
nsu
m
pt
ion
is
pro
portio
nal
to
the
sq
ua
re
of
th
e
distance
if
the
transm
issi
on
distance
is
within
the
fr
e
e
sp
ace
dista
nce,
a
nd
e
nergy
co
ns
um
ption
is
propo
rtion
al
to
t
he
f
ourth
powe
r
of
th
e
transm
issi
on
di
sta
nce
wh
e
n
the
transm
issi
on
distance
is
outsi
de
the
fr
ee
sp
ace.
The
f
r
ee
sp
ace
range
was
def
i
ned
as
d
0
.
Ther
e
f
or
e,
i
n
a
wireless
net
w
ork,
e
nergy
co
ns
um
ption
i
ncrea
ses
as
the
tr
ansm
issi
on
dis
ta
nce
increases
. T
his
is sho
wn in (
8).
is t
he data
pa
cket siz
e, a
nd
is t
he
tra
ns
m
is
sion dista
nce
.
(
,
)
=
−
(
)
+
−
(
,
)
=
{
+
2
≤
0
+
4
>
0
0
=
(8
)
Wh
e
n
recei
ving
data,
it
requires
recei
ving
ene
rg
y
acc
ordi
ng
to
the
s
iz
e
of
the
dat
a
m
essage.
The
e
nergy e
quat
ion re
quire
d at
this ti
m
e is
sh
ow
n
in
(9
).
(
)
=
−
(
)
=
(9)
The radi
o
m
odel
p
aram
et
ers
use
d
i
n
the
si
m
u
la
ti
on
are
sho
w
n
in
Ta
ble 1.
Table
1.
Ra
dio
Mode
l
Param
et
ers
Para
m
eters
Valu
e
Data Agg
regatio
n
(
E
DA
)
5
0
nJ
/b
it/signal
Energy dis
sip
atio
n
to run
the radio
de
v
ice (
E
elec
)
5
0
nJ
/b
it
Free
sp
ace
m
o
d
el
o
f
T
rans
m
itter
A
m
p
lif
ier
(
ε
fs
)
1
0
pJ
/b
it/
m
2
Multi
p
ath
m
o
d
el
o
f
T
rans
m
itter
A
m
p
lif
ier
(
)
0
.00
1
3
pJ
/
b
it/
m
4
4
.
2
.
Equ
al Pr
obabil
ity
C
lus
ter He
ad El
ected
U
sin
g
the
une
qu
al
pro
bab
il
it
y
cl
us
te
r
hea
d
el
ect
ion
m
et
ho
d,
on
a
verag
e,
the
nu
m
ber
of
cl
us
te
r
heads
that
are
cl
os
e
to
the
(
num
ber
of
node
s)
*
(cl
us
te
r
he
ad
el
ect
ion
pro
bab
il
it
y)
is
el
e
ct
ed.
The
nu
m
ber
of
cl
us
te
r
head
s
i
s
not
co
ns
ta
nt
because
al
l
the
nodes
pe
r
r
ound
el
ect
the
cl
ust
er
hea
d
on
ly
on
ce
a
cco
rd
i
ng
to
the
crit
ic
al
equ
at
io
n.
Occasio
nally
,
too
m
any
cl
us
te
r
heads
m
a
y
be
el
ect
ed
or
cl
us
te
r
head
s
m
ay
no
t
be
el
e
ct
ed
at
al
l.
This
can
be
no
dif
fer
e
nt
or
le
ss
ef
fici
ent
than
befo
re
cl
us
te
rin
g.
I
n
order
t
o
i
m
prov
e
t
his,
we
use
the
un
i
form
cl
us
ter
hea
d
el
ect
io
n
m
et
ho
d
s
o
tha
t
the
sam
e
nu
m
ber
of
cl
us
te
r
hea
ds
a
re
el
e
ct
ed
eve
ry
r
ound.
T
he
un
i
form
c
lu
ste
r
hea
d
sel
ect
ion
m
e
tho
d
el
ec
ts
the
cl
us
te
r
head
us
in
g
the
crit
ic
al
equ
at
i
on
un
ti
l
it
beco
m
es
(num
ber
of
no
des)
*
(cl
us
te
r
head
el
ect
ion
pro
bab
il
it
y)
ev
ery
r
ound.
Fig
ur
e
4
an
d
Tabl
e
2
sho
w
t
he
num
ber
of
cl
us
te
r
hea
ds
per
r
ound
wh
e
n
t
he
total
num
ber
of
node
s
is
10
0
a
nd
the
nu
m
be
r
of
cl
ust
er
el
ect
ion
pro
bab
il
it
ie
s is 10%.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Impr
ov
eme
nt
of
CH
el
ect
ion i
n
th
ree
-
le
vel
he
te
ro
ge
ne
ou
s
WSN
(
Jo
ng
-
Y
ong Lee
)
277
Table
2.
N
um
ber
of
Cl
us
te
r H
eads
Per
R
ound
Ro
u
n
d
r
Un
eq
u
al elec
t
Equ
al elec
t
1
16
10
2
11
10
3
13
10
4
11
10
5
11
10
6
7
10
7
11
10
8
10
10
9
4
10
10
6
10
Av
g
.
1
0
.00
1
0
.00
Figure
4.
N
umber
of clu
ste
r h
eads
per r
ound
4
.
3
.
Sim
ulat
i
on
We
com
par
ed
the
networ
k
li
fetim
e
of
the
pr
op
os
e
d
m
e
t
hod
an
d
the
e
xisti
ng
LE
AC
H
Protoc
ol.
The Si
m
ulatio
n
Pa
ram
et
ers
are s
how
n
in
Ta
ble 3.
Table
3.
Sim
ul
at
ion
Pa
ram
et
e
rs
Para
m
eters
Valu
e
Nu
m
b
e
r
o
f
Sens
o
r
No
d
es (
N
)
100
Sen
so
r
Field
4
0
0
x 4
0
0
Po
sitio
n
of
Base Statio
n
Cen
ter
(20
0
,
2
0
0
)
Initial Energ
y
(
e
0
)
0
.5J
Ad
v
an
ced No
d
e Add
itio
n
al E
n
ergy
(
)
1
.5
Su
p
er
No
d
e Add
itio
n
al E
n
ergy
(
)
3
Ratio
of
Adv
an
ced
Nod
e (
m
)
0
.2
Ratio
of
I
n
ter
m
ed
i
ate Nod
e (
m
0
)
0
.5
Size of
Packet
1
0
0
0
bits
Assum
ing
that
the
Sensor
Fiel
d
is
400
x
400
a
nd
the
loc
at
ion
of
the
B
ase
Stat
ion
is
locat
ed
at
the
center
(20
0,
2
00)
of the
Se
nso
r
Fiel
d t
he
n
the
Sen
s
or
Nodes a
re r
a
ndom
ly
arr
a
ng
e
d
a
s s
hown in Fi
gure
5.
Figure
5. N
ode
p
la
cem
ent
in s
ens
or
fiel
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
1
,
Ja
nu
a
ry 20
19
:
2
7
2
–
2
7
8
278
4
.
4
.
Sim
ulat
i
on
Resul
t
The
Fig
ure
6
and
Ta
ble
4
sh
ow
the
re
s
ults
of
netw
ork
li
fetim
e
com
par
ison
us
i
ng
the
ab
ove
si
m
ulat
ion
pa
r
a
m
et
ers.
Sim
ul
at
ion
res
ults
s
hows
t
hat
SEP
-
E
ha
s
a
39
3%
of
net
work
li
fetim
e
i
m
pr
ovem
ent
ov
e
r
t
he
LE
A
CH
P
ro
t
oco
l.
Wh
e
n
e
ne
rg
y
i
s
co
ns
ide
re
d
i
n
P
rop
os
e
d
m
et
hod,
it
ca
n
be
co
nf
irm
ed
th
at
it
is
i
m
pr
oved
b
y
22.
6%
c
om
par
ed
to
SE
P
-
E.
Table
4.
Sim
ul
at
ion
Re
s
ult
Proto
co
l
FND
I
m
p
rov
ed
Ratio
LE
ACH
Pr
o
to
co
l
45
SEP
-
E
177
3
9
3
%
▲
th
an
L
EACH Pr
o
to
co
l
Propo
sed
m
ethod
217
2
2
.6%
▲
than SE
P
-
E
Figure
6. Com
par
is
on of
net
work li
feti
m
e
5.
CONCL
US
I
O
N
Cl
us
te
rin
g
-
bas
ed
W
i
reless
Sensor
Net
work
Protoc
ols
can
help
t
o
im
pr
ov
e
Netw
ork
li
feti
m
e.
Howe
ver,
sinc
e
a
Cl
us
te
r
H
ead
of
the
cl
ust
er
ag
gregate
s
an
d
t
ran
sm
its
data,
the
en
erg
y
bur
de
n
is
great
.
Heter
og
e
ne
ou
s
sens
or
netw
ork
is
com
po
se
d
of
nodes
that
hav
e
dif
fer
e
nt
init
ia
l
ener
gy.
In
order
t
o
m
a
intai
n
the
Net
work
li
fetim
e
of
the
ne
twork
f
or
a
l
ong
ti
m
e,
it
is
effe
ct
ive
to
el
ect
nodes
ha
ving
hi
gh
i
niti
al
energy
as
Cl
us
te
r Head
.
Howe
ver,
if
th
e
Cl
us
te
r
Hea
d
el
ect
ion
pr
obabili
ty
is
increased
only
for
nodes
with
a
lot
of
init
ia
l
energy,
a
no
de
with
a
low
e
ne
rg
y
m
ay
be
elected
as
a
Cl
us
te
r
Head,
w
hic
h
m
ay
be
ineffi
ci
ent.
In
this
pa
per,
we
c
onsider
th
e resid
ual e
nergy to
so
l
ve
this
problem
, also
the im
pr
ov
em
ent of t
he
act
ual
N
et
w
ork
li
feti
m
e.
REFERE
NCE
S
[1]
I.
F.
Ak
y
il
di
z,
W
.
Su,
Y.
Sankar
a
subram
ani
am,
E
.
Ca
y
ir
ci.
A
Surve
y
on
Sens
or
Networks,
IEEE
Comm
unic
ati
on
s
Magazine
.
2002
.
[2]
J.Y.
Lee,
K.
Jun
g,
H.
Jung,
D.
L
ee
.
Im
proving
th
e
En
erg
y
Eff
ic
i
e
nc
y
of
a
Cluster
Hea
d
E
le
c
ti
on
fo
r
W
ire
le
ss
Senso
r
Networks,
Int
ernati
onal Journal
of
Distribu
te
d
S
ensor Ne
tworks
.
2014;
10(3).
[3]
J.Y.
Lee,
K
.
Jung,
B.
Shresth
a,
J
.
LEE,
S
.
Cho.
Ene
rg
y
Ef
ficien
c
y
Im
prove
m
ent
of
the
of
a
Clus
te
r
He
ad
Select
i
on
for
W
ire
l
ess Sensor Ne
tworks,
In
te
rnational
Jour
nal
of
Smar
t
Hom
e
.
2014;
8(3);
9
-
18.
[4]
W
.
Heinz
e
lman,
A.
Chandr
akas
an,
H.
Ba
la
kr
i
shnan.
En
ergy
-
Ef
fici
ent
Comm
unic
ati
on
Proto
col
for
Wir
el
es
s
Mic
ros
ensor Ne
t
works
,
Proce
edings
of
the 33rd
Hawai
i
In
te
rn
at
i
onal
Conf
ere
n
ce on
S
y
s
te
m
Sci
e
nce
s.
2000
.
[5]
G.
Sm
ara
gdaki
s,
I.
Matt
a
,
A.
Besta
vros.
SEP:
A
Stabl
e
E
le
c
ti
o
n
Protoc
ol
for
cl
ustered
he
te
ro
gene
ous
Wirel
es
s
Sensor Ne
tworks
,
Boston Univ
ers
ity
Com
pute
r
Sc
ie
nc
e
Dep
art
m
en
t.
2004
.
[6]
M.J.
Hand
y
,
M.
Haa
se,
and
D.
T
imm
ermann.
Lo
w
ene
rgy
adaptive
cl
ust
ering
hier
archy
wit
h
det
e
rm
ini
stic
cl
ust
er
-
head
sel
ec
t
ion
,
Mobile
and
W
ir
el
ess Com
m
unic
at
ions Ne
twork,
2002.
4th
Int
ern
at
ion
al
W
orkshop on. IE
E
E, 200
2.
[7]
F.A.
Aderohunm
u,
J.D.
Deng, “
An E
nhan
ce
d
Stab
le
El
e
ct
ion
Proto
col
(SEP) fo
r
Cl
ustere
d
He
te
rog
e
neous
W
SN
”
[8]
W
.
S.
LE
E,
K.D.
Jung,
J.Y.
Le
e.
Im
prove
m
ent
of
Cluste
r
Hea
d
sel
ec
t
ion
of
LE
ACH
Protocol
,
Inter
nati
onal
Journa
l
of
App
li
ed
Eng
in
ee
ring
Re
search
(
IJA
ER
)
.
2017;
1
2(20).
Evaluation Warning : The document was created with Spire.PDF for Python.