TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 16, No. 3, Dece
mbe
r
2
015, pp. 439
~ 453
DOI: 10.115
9
1
/telkomni
ka.
v
16i3.938
2
439
Re
cei
v
ed Au
gust 1, 201
5; Re
vised Novem
ber 8, 201
5; Acce
pted
No
vem
ber 2
8
,
2015
A Novel Protective Logic for Optimal Coordination of
Overcurrent Relays
Sajad Samadinasab*, Fa
rhad Namda
r
i, Nader Sh
ojaei
Dep
a
rtement o
f
Electrical Eng
i
ne
erin
g, Lores
tan Univ
ersit
y
,
Dan
e
shg
ah Str
eet, 712
34-9
8
6
53, Khorram
a
b
ad, Loresta
n, Iran
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: sajad.sam
adi
nasa
b
@gm
a
il.
com
A
b
st
r
a
ct
Usual
ly coor
di
natio
n of overc
u
rrent
rel
a
ys is
done by tak
i
n
g
into a
cco
unt the
specific
str
u
cture of
the syste
m
w
h
ich d
oes n
o
t show
the rea
l
s
t
ate of
the sys
tem. On the
other
h
and, dyn
a
mic
ch
ang
es i
n
netw
o
rk can
o
ccur du
e to sh
ort circuit co
nd
itions,
the
malf
unctio
n
in
g re
la
ys, devel
op
me
nt, operati
on
a
n
d
repairs on any
part of the power
system
. Also the
most of the new
protective schemes
are based on
a
communic
a
tio
n
chann
el, w
h
ich cann
ot be g
uara
n
tee
d
in
p
r
actice. T
herefore, solv
in
g the probl
e
m
of rela
y
coord
i
nati
o
n
is
extremely
diffic
u
lt i
n
c
a
se
of d
y
na
mi
c
cha
n
g
e
s
in
the
n
e
tw
ork structure
an
d
the
abs
enc
e o
f
communic
a
tio
n
links
b
e
tw
ee
n so
me re
la
ys
.
In
this articl
e,
a nov
el pr
otective log
i
c base
d
o
n
ph
a
s
or
me
asur
e
m
ent
units (PMUs)
d
a
ta is pr
op
ose
d
for o
p
tima
l
c
oord
i
nati
on of overcurr
ent
relays. In this
m
e
t
hod,
by usi
ng the P
M
U measur
e
m
ents, phas
or i
n
formatio
n
can
be o
b
tain
ed c
ontin
uo
usly at
any n
ode w
h
e
r
e
PMUs ar
e i
n
s
t
alle
d i
n
th
e
p
o
w
e
r gri
d
. F
o
r
this
purp
o
se,
in
the
first the Opti
mal
P
M
U pl
ace
m
ent
is
deter
mi
ned for
full netw
o
rk o
b
serva
b
il
ity. Then, the dyn
a
m
i
c
chang
es of n
e
tw
ork w
ill be observ
e
by usi
n
g
w
i
de are
a
me
asure
m
ents ba
sed o
n
PM
Us
data. Fina
lly th
is infor
m
ati
on
i
s
sent via c
o
mmu
n
ic
ation
link
s
PMUs for the
opti
m
al c
oor
din
a
tion
of
ov
ercurre
nt
rel
a
y
s
. T
he us
e of
PMU for th
e
coor
din
a
tio
n
of
overcurr
ent rel
a
ys improv
e the d
e
cisi
on
makin
g
cap
abi
lit
y and p
e
rfor
mance
of protect
i
ve rel
a
ys an
d
hel
p
them t
o
for
m
a
relia
bl
e an
d ro
bust protecti
on
syst
em. The
p
r
opos
ed
meth
o
d
is tested
on I
EEE 8-bus
an
d
14-b
u
s stand
ar
d netw
o
rks.
Ke
y
w
ords
:
overcurr
ent re
l
a
y, opti
m
al c
o
ordi
natio
n of
relays, phas
or me
asur
e
m
ent units,
o
b
serva
b
ility
,
communic
a
tio
n
chann
el, pow
e
r
system prote
c
tion
Copy
right
©
2015 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Today’s p
o
wer
system
s a
r
e ve
ry
comp
lex,
large an
d interconn
e
c
ted. Beca
use of the
increa
sing
de
pend
en
ce o
n
ele
c
tricity, e
n
su
ring
it
s d
e
livery
in a secu
re and
rel
i
able
m
ann
er
is
very importan
c
e to both cu
stome
r
s a
nd
sup
p
liers.
On
the other ha
nd, sho
r
t circuit conditio
n
s
can
occur
unexp
e
ctedly in a
n
y
part of a p
o
we
r sy
stem
.
The in
cide
n
c
e of the fau
l
t is harmful
and
must be isol
a
t
ed by a set
of protective
device
s
. The
prote
c
tion sy
stem
s must b
e
reconn
ect the
affected e
qui
pment a
s
so
on a
s
the
co
ndition
s return to no
rmal.
To solve the
s
e p
r
o
b
lem
s
, the
system
s have
to be monito
red, c
ontroll
e
d
, and protect
ed [1]. Indeed
, relays a
r
e th
e co
re an
d th
e
brain
of po
wer sy
stem p
r
otection. Protective re
l
a
ys
are in
stalle
d
as a "fault
sensor" in
po
wer
system a
nd t
o
isolate
a faulty part fro
m
the other
parts
of the
netwo
rk if th
ere exi
s
ts a
fault
event. The
r
ef
ore, m
ode
rn
relays
are
op
e
r
ating
as
sen
s
ors
and
p
r
ot
ectors [2].
Overcurrent
rel
a
ys
are u
s
e
d
a
s
both pri
m
ary
and b
a
cku
p
prote
c
tion fo
r heavily me
she
d
an
d mu
lti-sou
r
ce p
o
w
er
netwo
rk. L
o
w co
st and si
mplicity to impleme
n
t
are the advant
age
s of overcurre
n
t relay
s
for
Powe
r sy
ste
m
prote
c
tion.
The i
s
sue
of coo
r
di
natio
n
of overcu
rre
nt relay
s
in
cl
ude
s time
set
t
ing
multipliers (T
SM)
a
nd plu
g
setting
m
u
ltipliers (PSM) with appl
ying
related
co
nst
r
aint
s on
operating time differen
c
e b
e
twee
n backup and pri
m
a
r
y relays. The
protective sy
stem mu
st have
ability to the
sen
s
itivity, selectivity and reliability
[3].
Over the pa
st
five decade
s, several stu
d
ies
have be
en
carri
ed o
u
t on
optimal
coo
r
dinatio
n of
overcurrent relays.
Th
ese
studie
s
can
be
divided i
n
to t
h
ree
catego
ri
es: 1
)
T
r
ial
a
nd e
r
ror meth
od 2
)
Structu
r
al
analysi
s
method
ba
se
d on
graph theory
3) Optimizat
i
on method [4-7]. In re
cent years, artificial
intelligence m
e
thods
and
nature
-
in
spi
r
e
d
algo
rithm
s
su
ch a
s
Evol
ution Pro
g
ra
mming [8], G
enetic Al
gorit
hm (GA
)
[9-1
1],
Particle Swarm Algorithm (PSO)
[12, 13] among othe
rs a
r
e used to
solve the issue of optim
al
coo
r
din
a
tion
of overcu
rren
t relays.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 439 – 453
440
The traditio
n
a
l relays
whi
c
h respond t
o
pre
s
et tripp
i
ng thre
shol
d
s
value
s
, con
t
inuou
sly
obtain voltag
e and
cu
rren
t values from
these
l
o
cal
measurement
s an
d de
ci
si
ons to
SCADA.
These thre
sh
olds mig
h
t not be valid whe
n
the
sta
t
e of the power
system
cha
nge
s, du
e to
equipm
ent fa
ilure
or
other distu
r
ba
nce
s
. Also, th
e tradition
al rel
a
ys p
r
op
erly
are
not abl
e
to
disting
u
ish be
tween fa
ult a
nd no
rmal
co
ndition
s [
14]. Malfunctio
n
in
g of relay
s
is
among th
e m
o
st
comm
on m
o
des of failu
re
that ma
ke
s
the casca
de
of faults. Every fou
r
mo
n
t
hs, the
Unit
ed
States expe
riences
a bla
c
kout la
rge
enou
gh to
l
eave half a
million h
o
m
e
s in
da
rk [
15].
Acco
rdi
ng to
the histo
r
i
c
al
data, rel
a
y malfunctio
n
ing
is on
e of the
major
co
ntrib
u
ting facto
r
s
to
70% of the
major di
sturb
ances in th
e
Unite
d
States [1
6, 17].
Usually i
s
n
o
t simulta
n
e
ous
measurement
s of the SCA
D
A syste
m
a
nd the sampli
ng rate i
s
very slow. Also, the pro
g
ra
m
of
its state e
s
timation is n
o
n
-line
a
r a
nd
time-con
su
mi
ng. Therefore, the ac
ce
ss to the vari
ous
para
m
eters o
f
the power
system at any time, is one
o
f
the effective steps in
o
r
de
r
to provide t
h
e
approp
riate q
uality and su
stainable
ele
c
t
r
ical
e
nergy
. One
of the
a
ppro
p
ri
ate to
ols
and
u
s
efu
l
in
this field, is the use of pha
sor mea
s
u
r
em
ent uni
ts (PM
U
s) that have been
used in
many different
countries. The use of
sy
nchron
ous m
easurement
s, will increase
si
gnificantly accuracy
of
diagn
osi
s
an
d fault location in transmi
ssi
on line
s
. For this pu
rp
ose, algo
rith
ms and met
hod
s
based on p
h
a
s
or m
e
a
s
ure
m
ents a
r
e provided in
ord
e
r to dete
c
t and locate the fault [18].
In this
articl
e, a n
o
vel p
r
ote
c
tive logi
c
ba
sed
on
ph
aso
r
me
asure
m
e
n
t units (PM
U
s) data
is propo
se
d for optimal
co
ordin
a
tion of
overcurr
ent relays. Co
mp
ared to
conv
entional
relay
s
, in
coo
r
din
a
tion
of overcu
rre
n
t relays ba
sed o
n
PMU are used from the real
time wide a
r
ea
measurement
syste
m
(WAMS). The
r
e
f
ore, they
ca
n a
c
curately
dete
c
t an
d
locate
the i
n
i
t
ial
disturban
ce i
n
the syste
m
, as we
ll a
s
the system
situation or
st
ate after the
isolatio
n of this
disturban
ce.
Namely, PM
Us m
a
kes o
b
s
erva
bility the
amount of cu
rre
nt and faul
t location. In this
method,
by u
s
ing
the PM
U mea
s
u
r
eme
n
ts, ph
asor
in
formation
can
be
obtain
ed
contin
uou
sly
at
any node
wh
ere PMUs a
r
e installe
d in
the powe
r
g
r
id. For thi
s
purp
o
se, initially the Optimal
PMU pla
c
e
m
ent is
determined fo
r ful
l
netwo
rk
ob
serva
b
ility. Then, the
dynamic
ch
ange
s of
network
will be observable
by
using wi
de area
measur
ements
based on PM
Us
data. Finally this
informatio
n is sent via co
mmuni
cation
links PMUs for the o
p
tima
l coo
r
din
a
tion
of overcurre
n
t
relays a
nd th
e relays can
deci
de
wh
eth
e
r
and
when
t
o
trip
a
tran
smissi
on
line.
This can
stop
the
prop
agatio
n (or ca
scadi
ng)
of
failu
re
s a
nd/or
conf
ine
it
to
a limite
d
small area.
PMU will have
this
a
d
vantag
e
to system
s that
relay
s
se
tting
cl
o
s
e
to
PMU, is do
n
e
onli
ne fo
r e
v
ery fault. Th
e
use of PMU for the Coo
r
di
nation of
overcu
rrent relay
s
improve
s
th
e deci
s
ion m
a
kin
g
cap
abili
ty
and pe
rform
a
nce of protective relays and help
them
to form a reliable and ro
bust protecti
on
sy
st
em.
2. A Rev
i
e
w
of the
Non
-
d
o
minated So
rting Gen
e
ti
c Algorithm
-II (NSGA-II)
The most im
portant pa
rt of all optimization
algo
rith
ms is the sel
e
ction pa
rt. A suitable
sele
ction
crit
erion
can b
e
brou
ght abo
u
t
to
obtain a
good
conve
r
gen
ce beh
avior for algo
rith
m.
Non
-
do
minat
ed so
rting a
ppro
a
ch is
a suitabl
e method for
multi-obj
ectiv
e
probl
em
s. This
approa
ch pro
v
ides a
suita
b
le sele
ction crite
r
ion
for
al
gorithm
to di
stinguish bet
ween
solutio
n
s in
multi-obj
ectiv
e
p
r
oble
m
s.
Also, in
stead
of on
e o
p
timum
solutio
n
, a
set of
several o
p
timum
solutio
n
s for
multi-obj
ectiv
e
optimizatio
n probl
ems
i
s
achieved. Th
ere is no a
b
solute sup
e
ri
ority
betwe
en the
s
e solution
s. In fact, e
a
ch result
of th
is
set is
equivale
nt to a p
a
rti
c
ular
co
mbinat
ion
of weight val
ues fo
r ea
ch
obje
c
tive function. In
ot
her
wo
rd
s, the re
sult of
weig
hting factors
method is o
n
e
of the results in pareto fro
n
t method.
NSGA-II is o
ne of the alg
o
rithm
s
which wo
rks
based on pareto front. NSGA
-II is
firs
t
time introdu
ced by Prof. K. Deb an
d his collea
gue
s
i
n
[19]. This a
l
gorithm i
s
ba
sed o
n
the id
ea
of conve
r
ting
the N
obje
c
tives to
a si
ngle
fitness me
a
s
ured
by the
creation of
a
nu
mber
of front
s.
The main fea
t
ure of NSGA
-II is its fast non-d
o
mi
n
a
ted
sorting p
r
o
c
e
dure fo
r ran
k
i
ng sol
u
tion
s at
its sel
e
ctio
n stage [20]. NS
GA-II is obtai
ned by
a
ddin
g
two imp
o
rta
n
t operators t
o
co
nvention
a
l
geneti
c
al
gori
t
hm: 1) the
n
on-d
o
min
a
ted
so
rting
2) th
e cro
w
din
g
d
i
stan
ce
ope
rator. Th
e
no
n
-
dominate
d
so
rting op
erato
r
assign
s a su
perio
rity
crite
r
ion (ran
k) to
membe
r
s
of populatio
n, whi
l
e
the
cro
w
ding
distan
ce ope
rator kee
p
s di
versity
of sol
u
tions
betwe
en mem
bers
with eq
ual
ra
nk
.
These ope
rat
o
rs a
r
e d
e
scri
bed mo
re in the followi
ng section
s
.
The aim of multi-obje
c
tive optimizatio
n, is fi
nding the set of solutio
n
s Pareto fro
n
t of the
desi
r
ed p
r
o
b
lem. The gen
eral form of
multi-
obj
ectiv
e
optimizatio
n probl
em a
s
follows:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel Protective Logi
c for Optim
a
l Coordi
nati
on of
Ove
r
current
Rela
ys (Saja
d
Sam
adinasab)
441
(1)
R
x
,
0
h(x)
,
0
g(x)
.
t
.
s
,
)
x
(
n
f
...,
),
x
(
i
f
)
x
(
F
min
Whe
r
e
)
x
(
f
i
is
th
i
the obje
c
tive function,
g(x)
is a se
t of inequality con
s
trai
nts a
n
d
h(x)
is a
set of e
q
ual con
s
traint
s which a
r
e
related to m
u
l
t
i-obje
c
tive o
p
timization
problem. In e
a
c
h
cycle of the algorithm, all solutions will be sorted
by assi
gning a rank to
each of them. This rank
will be obtain
ed by non-d
o
m
inated sorti
ng techni
que.
Fo
r the sa
ke
clarity of this appro
a
ch, it
is
necessa
ry to explain some
subje
c
ts.
2.1. Conce
p
t of Dominati
on
In a m
u
lti-o
b
j
ective mi
nim
i
zation
p
r
obl
em, the
co
m
pari
s
on
bet
ween t
w
o
solu
tions i
s
defined:
(2)
Y
X
:
i
Y
X
:
i
Y)
dom
(X
Y
X
0
i
i0
0
i
i
Whe
r
e X
an
d Y a
r
e t
w
o
sol
u
tion
s of
a multi
-
obje
c
tive proble
m
,
i
is the
nu
mber of
obje
c
tive-fun
ction
s
an
d
0
i
is one
of obje
c
tives [19]. In
Figure 1, the
point A d
o
mi
nates
on th
e
point C, be
ca
use in
no fun
c
tion is
not worse than
C, and in b
o
th function i
s
bett
e
r than it. As
can
be
seen
in Fi
gure
1, for th
e point A, the
functi
o
n
s val
ues F1
and
F
2
are bette
r t
han the
poi
nt C,
i.e. the point
A dominate
s
on the point C.
Also, point
s B and C are equal an
d both point
s are
dominate
d
by
the
point A.
Therefore,
th
e poi
nt A
i
s
chosen
as the
be
st p
a
reto
f
r
ont. Th
e
aim
of
this wo
rk i
s
to
minimize the
function
s F1
and F2.
F2
F1
To
w
a
r
d
s
ge
t
t
i
n
g be
t
t
e
r
To
w
a
rds
g
e
t
t
i
n
g
be
t
t
e
r
B
C
A
Figure 1. Con
c
ept of Do
mination fo
r sel
e
ct the be
st pareto fro
n
t
2.2. Effec
t
iv
e
n
ess o
f
Pres
ence o
f
othe
r Av
ailable Solutions (Cr
o
w
d
i
ng Dis
t
ance
)
Some sol
u
tions
can b
e
comp
are
d
with ea
ch o
t
her after in
trodu
cing
co
nce
p
t of
dominatio
n.
But duri
ng t
he
comp
ari
s
on, we
conf
ront with
so
me solution
s that cann
ot be
comp
ared wit
h
each othe
r by concept of dominat
ion
.
Becau
s
e, some solution
s may be bet
ter
according to
one o
b
je
ctive function
whil
e they ar
e
wo
rse
acco
rdi
n
g
to anothe
r o
b
jective fun
c
ti
on.
Therefore, th
e effectivene
ss
of pre
s
e
n
c
e of ot
he
r
solutio
n
s
can
help u
s
to
overcome thi
s
probl
em. Thi
s
co
ncept will
be explai
ned
by an exa
m
pl
e from
Figu
re
2(a
)
. In thi
s
f
i
gure,
pro
b
lem
spa
c
e
ha
s be
en divide
d to
4 re
gion A, B
,
C and
D, by
con
s
id
erin
g
point x as
a g
oal. Also,
so
me
solutio
n
s ‘a,
b, c, d, e an
d x’, have been spe
c
ified
from all po
ssible
solutio
n
s
of an a
r
tificial
minimization probl
em. As sho
w
n in Fig
u
re 2(a), the
s
e sol
u
tion
s have two values by the two
obje
c
tive function
s F1 an
d F2. We wa
nt to compar
e point ‘x’ with other poi
nts in this prob
lem.
The point ‘x’ dominate
s
all
points in re
gi
on A. It m
eans that values
of F1 and F2
for the point ‘x’
are
le
ss than
value
s
of F1
and
F2
for t
he all
poi
nts i
n
this regio
n
A. For exam
ple, F1
(x)<F1
(c)
and F2
(x)<F
2
(c), therefore x dominat
e
s
‘c’. Also, all
points of r
e
g
i
on C do
mina
te ‘x’. It means
that F1 and
F
2
of the all p
o
ints of regio
n
C h
a
ve
le
ss value th
an
F1 and
F2 of
the point ‘x’. On
the other ha
n
d
, compa
r
in
g points of re
gi
on B and D with point ‘x’ is difficult.
The valu
es
of F1 fro
m
all
p
o
ints of th
e region
B
are l
e
ss than
valu
e of F1 f
r
om
point ‘x’,
but the value of F2 from point ‘x’ is less tha
n
valu
es of F2 fro
m
all points
of the region
B.
Furthe
rmo
r
e,
the value
s
of
F2 from
all p
o
i
nts of
the
re
g
i
on D are le
ss than valu
e of
F2 fro
m
poi
nt
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 439 – 453
442
‘x’, but the v
a
lue
of F1
from p
o
int ‘x’ i
s
le
ss th
an v
a
lue
s
of
F1 f
r
om
all p
o
int
s
of
the
regi
o
n
D.
Therefore, th
e be
st point
cannot
b
e
spe
c
ified by
con
c
ept of
domi
n
a
t
ion between
‘x’ and poi
nts
of
regio
n
B an
d
D. In this
situ
ation, we
use
effectiv
ene
ss of pre
s
e
n
ce
of other
avail
able
solutio
n
s to
c
o
mpare. At firs
t, we as
s
u
me that there is
no
poi
nt i
n
re
gion
C.
We
want to
compa
r
e p
o
int
‘x’
with p
o
int ‘b’
in region
B.
As
see
n
, we
have
F
1
(x
)
>
F
1
(b
) a
n
d
F2
(
x
)<F2
(b
) fo
r t
hes
e tw
o
poi
nts.
Therefore, it is not possibl
e to
compa
r
e
points ‘x’ with ‘b’ at
first.
On the other hand, there
is
point ‘a’ that
F2(a
)<F2
(b)
and F1
(a
)<F1
(b), it mea
n
s
that point ‘b
’ i
s
domi
nated
by point ‘a’. But
there i
s
no p
o
int that domi
nates ‘x
’. The
r
efore, effectiveness of pr
e
s
en
ce of p
o
in
t ‘a’ could h
e
l
p
to compare
point ‘x’
with ‘b’.
Thus,
point ‘x’ is bet
ter than po
int ‘b’. Also,
we have the
same
difficulty to compare point ‘x’ with point ‘d’. It
means that, F2(x)>F2
(d) an
d F1(x
)<F1
(d) a
nd th
en
there i
s
th
e p
o
int ‘e’ that F
2
(e
)<F
2
(d)
a
nd F1
(e
)<F1
(d), but th
ere
i
s
n
o
poi
nt th
at domin
ates
‘x’.
Therefore, eff
e
ctivene
ss of pre
s
en
ce of
point ‘e’ co
ul
d help to co
mpare point ‘
x
’ with point ‘d’.
Thus,
point ‘
x
’ is bette
r th
an p
o
int ‘d’.
Finally,
there
is th
e same
difficulty to compa
r
e
between
point
‘x’
with point
‘e’
a
nd point
‘a’ with each
othe
r. In this
situatio
n, we
can
not
com
pare the
s
e
points with concept
of
do
minati
on a
n
d
effectivene
ss of presen
ce
of other
available
soluti
ons.
Therefore, th
ere i
s
n
o
poi
nt that domin
ates th
e
s
e
p
o
ints
com
p
let
e
ly. Thus, th
ese
point
s
kn
own
as a pa
reto front of this cycle of algorith
m
[19].
The
cro
w
di
ng
distan
ce
ope
rator i
s
expl
ai
ned
b
e
fore
a
nd its m
a
the
m
atical
expre
ssi
on fo
r
a probl
em wit
h
two obje
c
tives, at point i according to
Figure 2(b
)
, is as follo
win
g
equation
s
:
(3)
f
f
f
f
d
min
1
max
1
1
i
1
1
i
1
1
i
(4)
f
f
f
f
d
min
2
max
2
1
i
2
1
i
2
2
i
(5)
2
i
d
1
i
d
D
1
i
d
and
2
i
d
are proportio
n
of the area in
whi
c
h poi
nt
i
is incl
uded,
to entire area
corre
s
p
ondin
g
to each O
F
(F1an
d F2
), resp
ectivel
y
. The sum of these two
paramete
r
s is D
whi
c
h indi
cat
e
s ove
r
all a
r
ea of point
i
and is
calle
d crowding
dista
n
ce. Th
us, e
a
ch
point whi
c
h
has
high
er
crowdi
ng di
sta
n
ce,
covers l
a
rge
r
a
r
ea
an
d its re
moval
resulted in l
o
ss
of diversity of
membe
r
s i
n
a
large
rang
e
of results. It should b
e
men
t
ioned that be
ginnin
g
and e
nding p
o
ints
of
pareto front (rank
= N) are i
m
porta
nt poin
t
s and t
hey h
a
ve the lowe
st priority for removing.
F2
F1
C
BA
D
a
b
x
c
d
e
(a)
F2
F1
E
n
d
i
ng
po
i
n
t
i
i+
1
max
1
f
i-
1
B
e
gi
nni
ng
po
i
n
t
mi
n
1
f
mi
n
2
f
ma
x
2
f
(b)
Figure 2. (a)
Some sol
u
tio
n
s from all p
o
ssi
ble solutio
n
s of an a
r
tificial minimi
zat
i
on pro
b
lem. (b)
Cro
w
di
ng di
stance for point
i
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel Protective Logi
c for Optim
a
l Coordi
nati
on of
Ove
r
current
Rela
ys (Saja
d
Sam
adinasab)
443
The sele
cted
individual
s from set of pa
rents
an
d offspring
are
use
d
for gen
erating ne
w
offsprin
g. It can be perf
o
rmed by tourn
a
ment se
l
e
cti
on, beca
u
se there a
r
e two
paramete
r
s for
each individu
al: 1) nond
o
m
ination ran
k
and 2) cro
w
ding di
stan
ce
. Between two individual
s with
different n
o
n
dominatio
n ranks, the i
n
d
i
vidual with
t
he lo
we
r ran
k
is better. If
both in
divid
uals
exist in the same front, th
e individual
with the
highe
r cro
w
di
ng di
stance i
s
bette
r. Therefore, the
fathers and mothers
for gene
rating
of
fsprin
g
c
an
be sel
e
cte
d
by thes
e two
operators. The
cro
s
sove
r an
d the mutation pro
c
e
ss in
NSGA-II are l
i
ke conventio
nal GA.
2.3. Perform non-domin
ated sorting
Non
-
do
minat
ed sorting
di
vides the
sol
u
tions
of
ea
ch cycl
e to di
fferent front
s (level).
After prod
uci
ng con
c
ept o
f
domination
and explai
nin
g
the effectiv
ene
ss
of pre
s
en
ce
of other
available sol
u
tions in com
parin
g betwe
en ea
ch
co
u
p
le of them, non-domi
nate
d
sortin
g will
b
e
perfo
rmed
by algorithm in
each cycl
e. At first,
we compa
r
e ea
ch
coupl
e of so
lutions
with the
con
c
e
p
t of domination, se
p
a
rately.
The
solutio
n
s, whi
c
h
we
re
not d
o
minat
ed by ot
h
e
rs,
are
kept in
the first fro
n
t
or b
e
st
front (calle
d
set F1
). The
n
Among, th
e sol
u
tion
s
which we
re no
t
dominate
d
by
others without
con
s
id
erin
g the effectiven
ess of front F
1
, are kept
in
the second f
r
ont (calle
d set F2). Similarly,
the solutio
n
s
whi
c
h were n
o
t dominate
d
by others
wit
hout co
nsi
d
e
r
ing the effecti
v
eness of fro
n
t
F1
F2, are
kept in the third front (calle
d set F3
). Th
is process is
r
epe
ated u
n
til there is
no
solutio
n
in th
is cy
cle
with
out having it
s o
w
n
front. Subse
que
ntly, thes
e g
ene
rated fronts
are
assign
ed thei
r co
rre
sp
ondi
ng ran
k
s. Thus, F1 is a
ssi
gned ran
k
1, F2 is assig
n
e
d
ran
k
2 and
so
on [19].
3. The propo
sed meth
od
The mo
st of the new p
r
otective sche
mes a
r
e b
a
sed on a
reli
able commu
nicatio
n
cha
nnel, which cann
ot always b
e
gu
arante
ed in
pra
c
tice. As
well a
s
, duri
ng bla
c
kouts or
ca
scadin
g
failure
s in the p
o
w
er
sy
ste
m
, that system
condition
s ch
a
nge rapidly, more info
rma
t
ion
excha
nge
s
be re
quired
by the co
ntrol cent
e
r
s and sub
s
tations. In o
t
her words,
the
comm
uni
cati
on ch
ann
els
are o
perating
with high lo
ad and the
r
e
f
ore be
com
e
more vuln
era
b
le
whe
n
the p
o
w
er sy
stem i
s
in u
n
interru
p
tible
conditi
ons. M
o
re
ove
r
, the ne
w p
r
otective sch
e
m
es
are b
a
sed o
n
the logi
c em
ployed by
tra
d
itional ove
r
current an
d di
st
an
ce
relays. This me
an
s
that
the modern relays are also base
d
on the assum
p
tio
n
s mad
e
for traditional rel
a
ys, which are
clea
rly invalid
sometime
s. Therefore, wi
thout cha
ngin
g
the basi
c
p
r
inci
ple
s
of protective relay
s
,
the malfu
n
cti
oning
of th
e
m
cann
ot be
avoided.
Thu
s
, a
ne
w
and
more
com
p
rehen
sive l
ogi
c i
s
need
ed in pro
t
ective relays.
Figure 3. Phasor m
e
asurement units that
function wit
h
the aid of GPS satellite
While
mo
st o
f
relay
s
still o
n
ly use m
agn
it
udes of volt
age
and
current me
asure
m
ents,
a
new technol
ogy is availa
ble for accu
rately
measu
r
ing voltage and cu
rrent pha
sors. The
s
e
measurement
s offe
r n
e
w in
formation
in
o
r
de
r to
imp
r
o
v
e the fu
nctio
nal lo
gic of
protective
relay
s
.
The ide
a
of p
hasor m
e
a
s
u
r
eme
n
t wa
s i
n
trodu
ce
d a
fter the
bla
cko
ut in No
rth-E
a
st US. Th
e f
i
rst
prototype
ph
aso
r
m
e
a
s
urement u
n
it (PMU) i
s
dev
elope
d by a
Virginia
Te
ch
re
sea
r
ch tea
m
in
1988 [21]. PMU utilizes
powerful
sign
al pro
c
e
s
sing
technol
ogy, have ca
pabl
e of measuri
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 439 – 453
444
voltage and current pha
so
rs with high a
c
cura
cy
(le
ss than 0.1% error) and ve
ry high speed
(60
sampl
e
s p
e
r se
co
nd
). P
M
Us me
asure the
volt
ag
e an
d
cu
rre
nt sig
nal
s at
two
sub
s
tations
sep
a
rate
d by hund
red
s
of miles which are syn
c
h
r
oni
zed pre
c
i
s
ely with the aid o
f
a GPS satellite
system that
i
s
sh
own
in Figure 3.
Th
e time-tag
s are attache
d
with
sample
s, an
d this inform
a
t
ion
is exch
ang
ed
over co
mmu
nicatio
n
ch
an
nels a
nd
co
lle
cted by control cente
r
s an
d/or sub
s
tatio
n
s.
By extracting
the relevant
inform
ation
from
the
s
e
measurement
s, ph
asor i
n
formatio
n
can
be
obtaine
d at a
n
y node
wh
ere PMUs a
r
e i
n
stalle
d in th
e po
wer gri
d
. This i
n
form
ation can b
e
u
s
ed
to do more a
c
curate state e
s
timation, co
ntrol, and p
r
o
t
ection [22].
Figure 4. Flowchart of the
prop
osed
met
hod ba
se
d on
NSGA-II alg
o
rithm
The flo
w
cha
r
t of the
pro
p
o
s
ed
metho
d
i
s
sho
w
n
in Fi
gure
4. In thi
s
meth
od, fro
m
data
measured by
the PMU
are used fo
r co
ordin
a
tion of
overcurrent
relays, so that
is
satisfie
d the
constrai
nts
related to the main and backup
re
lay
s
. On the other hand, PM
U will
have t
h
is
Initializ
e
P
op
ul
ation
max
g
g
Ye
s
No
Start
g
=0
Determine the
o
p
timal location an
d number of
PM
Us
o
p
timal coordination of overcurren
t
rela
y
s
Nondominated S
o
rtin
g
and C
r
o
w
d
i
n
g
Distance
Selection
o
f
c
hromosomes
Cr
ossover
&
Mutation
1
g
g
Form
set
o
f
n
ond
ominated
solutions
Final
Decision
M
a
king
Output
Results
Multiobjective
E
v
aluation
Determining online and offline
rela
y
s
according
to the optimal
location
of
PMUs
Extracting phaso
r
Data
b
y
using the r
eal-
t
ime
m
easu
r
ed
PM
U
da
t
a
sending information via communication links of PM
Us
D
etermin
ing
TS
M
and
Iset
of
the
online
and
offline
rel
a
y
s
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel Protective Logi
c for Optim
a
l Coordi
nati
on of
Ove
r
current
Rela
ys (Saja
d
Sam
adinasab)
445
advantag
e to systems that
relays setting clo
s
e to
PMU, is don
e online for ev
ery fault. In o
t
her
words, PM
Us make
s ob
se
rvability the amount of
cu
rrent and fault
locatio
n
. As a re
sult, whe
n
locatio
n
of th
e fault
wa
s id
entified by th
e PMU,
a
sig
nal i
s
sent to
the Pha
s
o
r
Data Con
c
entrator
(PDC) and the setting value to be
considered for
it fault. With
this task there are ability to
cha
nge
th
e relays
settings and re
du
ce
th
eir ope
rating
tim
e
durin
g
fault occurre
n
ce. But
con
s
id
erin
g to the high co
st of installati
on PMU on a
ll network b
u
s
e
s
, initially the Optimal P
M
U
placement i
s
determi
ned fo
r full net
work
observability.
Therefore, th
e ultimate O
b
jective fun
c
tio
n
of pro
b
lem i
s
a two
-
obj
ect
i
ve function
that in
cl
ude
s
the followi
ng
function
s: 1
)
determi
ne t
he
optimal location and n
u
mb
er of PMUs 2
)
minimizi
ng the ope
rating t
i
me of overcu
rre
nt relays.
Therefore, i
n
itially the
Optimal PM
U pla
c
e
m
ent
is d
e
termi
ned fo
r full
netwo
rk
observability. Then, by usi
ng PMU me
a
s
ureme
n
ts
, the values
of voltage an
d current pha
so
r in
the line
s
a
n
d
buses po
we
r sy
stem a
r
e
obtaine
d
for the de
sired
fault. Now a
c
cordi
ng to t
h
e
locatio
n
of PMU, the relay
s
su
rroundi
ng
bus wh
i
c
h P
M
U ha
s bee
n
installed a
s
online relays
and
the other rel
a
ys as
offline
relays
are co
nsid
ere
d
. Th
e setting
s of
offline relay
s
is fixed an
d the
setting
s of o
n
line relays
is obtai
ned
by usin
g the
real
-time m
easure
d
PM
U data. T
h
e
s
e
informatio
n is sent via co
mmuni
cation
links PMUs for the o
p
tima
l coo
r
din
a
tion
of overcurre
n
t
relays. In
o
r
d
e
r to
obtain
the b
e
st
solut
i
ons, th
ey sh
ould
be
ran
k
ed b
a
sed o
n
a b
e
n
c
hma
r
k.
Therefore, in
the NSGA-II algorit
h
m
is a
ssi
gne
d a ra
n
k
to ea
ch an
swer
whi
c
h th
ey are ba
se
d on
the numbe
r of non-d
o
min
a
ted points,
comp
ared to
other point
s. At
the end of the algorit
hm,
points that h
a
ve the best
ran
k
, namely
have the ra
nk 1, are ch
o
s
en a
s
the set of solution
s or
Pareto fro
n
t. Also, the crowdi
ng di
sta
n
ce p
a
ramet
e
r is
cal
c
ulat
ed for e
a
ch
membe
r
of e
a
ch
grou
p, which
indi
cate
s
the
si
ze
of
the
nearby
samp
le to the
oth
e
r m
e
mbe
r
s
of those g
r
o
ups.
Great val
ue
this pa
ram
e
ter lea
d
s to
diverge
n
ce a
nd bette
r ra
nge
will cre
a
te
in the se
t
of
membe
r
s of t
he
po
pulatio
n.
The
strin
g
s
with
hig
h
e
r value
are
sele
cted
an
d with a
pply
i
ng
cro
s
sove
r an
d mutation
op
erato
r
s are d
e
termin
ed tim
e
setting
mult
ipliers (TSM)
and the
curre
n
t
setting of relay (Iset). Therefore,
the dy
namic
changes of network
will be observe by using
wide
area m
e
a
s
urements
ba
se
d on PMUs
data. Finally,
this informat
ion is sent via comm
uni
ca
tion
links PM
Us for the
optim
al
co
ordi
nation
of over
cu
rre
n
t
relays an
d t
he relays can
de
cide
whet
her
and when to
trip a transmissi
on line.
In the fo
llowi
ng the two
-
o
b
jective opti
m
ization p
r
o
b
lem
functions will
be explained
separately.
3.1. Optimal Placement o
f
Phasor Me
asureme
nt u
n
its
Phaso
r
m
e
a
s
urem
ent u
n
its a
r
e
a n
e
w
techn
o
logy fo
r po
we
r
syst
em state
e
s
timation.
Usi
ng th
e te
chnolo
g
y of P
M
U, the
po
wer
system
i
s
conve
r
ted
fro
m
a
static inf
r
ast
r
u
c
ture
to
a
flexible and live infrastru
c
t
u
re [23]. Phasor
me
asure
m
ent unit (PMU) a
s
one
of the the main
exceptio
n fie
l
d of tran
smi
ssi
on
sm
art
grid, i
s
th
e
fundame
n
tal
solutio
n
s for the
real
ti
me
monitori
ng po
wer
grid
s. Th
at able to co
n
v
ersio
n
of no
nlinea
r state
estimation
eq
uation
s
to lin
ear
equatio
ns, which imp
r
ove
s
the spe
ed
control sy
ste
m
s, safety and mana
gem
ent system
s, that
use the
results of state esti
mation [24].
3.1.1. Obser
v
a
bilit
y
analy
s
is based o
n
PMU
The p
o
wer
n
e
twork i
s
co
u
l
d be
ob
se
rve
d
, wh
en a
r
e
cal
c
ulate
d
th
e all
state va
riable
s
in
orde
r to sy
ste
m
state e
s
timation. That it mean
s,
ca
n to cal
c
ulate
d
the voltage p
h
aso
r
for all b
u
s
and also cu
rrent to all bran
che
s
are co
n
necte
d to its [25]. PMU inst
alled on a certain bus is a
b
l
e
to
measure the
voltage pha
sor of
that
bus and also cu
rrent
pha
sor
of the all b
r
an
ches
con
n
e
c
ted to
it. As a re
sult
, the bus volt
age
size and
pha
se a
ngle
of a co
nne
cte
d
to bu
s ha
s
a
PMU is
cal
c
ulated to u
s
i
ng ki
rchhoff
equatio
ns.
T
herefo
r
, the
buses th
at in their have b
een
installe
d PM
U, are di
re
ctl
y
obser
ved,
and
bu
se
s th
at are
conn
e
c
ted to
the
b
u
s
with PM
U,
the
y
are in
directly
observed [2
6]. Bus ob
se
rvability i
ndex (BOI) is p
r
op
o
s
ed
as
perfo
rmance indi
ca
tor
on qu
ality of the optimi
z
ati
on. BOI for b
u
s
i
)
i
(
β
is defin
ed
as the
numb
e
r of ph
asor
measuri
ng
units
whi
c
h a
r
e a
b
le to o
b
s
erve
a give
n
bus. Sy
stem
observability
red
unda
ncy
index (SO
R
I) is
defined a
s
th
e total set of BOIs of syste
m
buse
s
.
In the other
wo
rds, if bus i wi
th the numbe
r of
i
β
PMU is ob
se
rvable, SORI
is achieved a
s
(6
) [27].
(6)
n
1
i
i
SORI
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 439 – 453
446
3.1.2. The Fo
rmulation of
Optimal Placement PMU
The fo
rmul
ation of th
e o
p
timal pla
c
e
m
e
n
t of PMU in
a sy
stem
with
n b
u
ses is p
r
ese
n
ted
as Equatio
n (7) [28]:
(7)
n
0
i
b
Ax
y
s.t
i
x
i
w
min
Whi
c
h
w
is the cost fun
c
tion
for the instal
led PMUs, and in normal stage, pla
c
e
m
ent
equali
ng to m
a
trix of unit
n
n
is con
s
ide
r
ed.
A is co
nne
cti
on matrix of
n
n
whi
c
h reveal
s the
way of con
n
e
c
tion of bu
se
s whi
c
h i
s
def
ined a
s
(8
):
(8)
otherwise
0
connected
are
j
and
i
buses
if
1
j
i
1
j)
,
i
(
n
n
A
The discrete
nature of the
optimal pla
c
e
m
ent of
PMU make it nece
s
sary that X vector to
be
d
e
fined
a
s
e
quatio
n (9
) su
ch that the element
s of
that position sho
w
the installatio
n
of this
equipm
ent in each bu
s:
(9)
otherwise
0
i
bus
at
installe
is
pmu
if
1
i
x
i
]
x
[
Also b matrix
for at least one
observabilit
y is shown as (10):
(10
)
T
1]
1
...
1
1
1
[
1
n
b
3.2.
Setting Ov
ercurrent Rela
y
s
The
obje
c
tive functio
n
a
nd
con
s
trai
nts
of
the p
r
o
b
lem,
to obtai
n the
pa
ramete
rs
of TMS
and
set
I
is
defined as
follows
[
29]:
(11)
seti
I
sci
I
log
i
TMS
3
)
i
set
I
,
i
TMS
(
f
opi
t
,
n
1
i
opi
t
:
Minimize
Whe
r
e
n is
the num
ber
of overcurre
n
t rela
y
s
. Constraint o
p
timization
pro
b
lem a
s
follows
:
(12)
i
max
TMS
i
TMS
i
min
TMS
(13)
CTI
)
m
z
(
m
op
t
)
m
z
(
b
op
t
(14)
Min
i
fault
I
i
set
I
Max
i
ioad
I
Whe
r
e
i
op
t
is o
p
e
r
ating time
th
i
rel
a
y
,
m
op
t
and
b
op
t
are
o
peratin
g time
of prim
ary a
n
d
backu
p relay
s
re
spe
c
tively and CTI is th
e Coo
r
din
a
tio
n
Time Interv
al.
Con
s
trai
nt (1
3) i
s
u
s
ed
for each p
a
ir m
a
in and
ba
cku
p
rel
a
y
b)
,
(
m
and fo
r erro
rs rel
a
ting
to zo
ne of
protection
m
z
. With
re
spe
c
t to th
e Figu
re
5, th
e failures
are
identified
by
the F1
and
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
A Novel Protective Logi
c for Optim
a
l Coordi
nati
on of
Ove
r
current
Rela
ys (Saja
d
Sam
adinasab)
447
F2 poi
nts. T
a
king
into
account
Con
s
trai
nt (14
)
, the
pi
ckup val
ue
of an
overcu
rre
nt relay
mu
st
be
set between t
he maximum
load current a
nd the mini
m
u
m fault curre
n
t experien
c
e
d
by the relay.
t
(
B
ack
u
p
)
b
(M
a
i
n
)
F1
F2
>C
I
>C
I
t
b
t
m
m
Figure 5. Coo
r
dinatio
n of overcu
rrent rel
a
ys
In the relays coo
r
din
a
tion
probl
em, the deci
s
io
n variabl
es
are
the TMS an
d
set
I
variable
s
for each relay. NSGA-II
al
go
rithm,
p
r
obl
em'
s
va
r
i
ab
les a
r
e
en
co
d
ed in
to
s
t
r
i
ng
s
,
s
o
each
strin
g
repre
s
e
n
ts an
an
swer to th
e p
r
obl
em
of
co
ordination.
Figu
re
6 sh
ow
s st
ru
ct
ur
e
of
the string
wh
en the network co
nsi
s
ts of
n overcurrent
relays.
1
set
I
1
TMS
2
set
I
2
TMS
…
n
set
I
n
TMS
Figure 6. Structure of the
st
ring in the NS
GA-II method
The main o
b
j
e
ctive functio
n
(OF
)
that is alr
eady u
s
e
d
in most of the
literature is
the total
weig
hted su
m of operatin
g times (O
Ts) of
primary rel
a
ys as follo
ws [30]:
(15
)
t
.
w
OF
mim
m
1
i
i
i
Whe
r
e m i
s
t
he num
be
r of
relay
s
,
i
t
is th
e ope
rating ti
me of the
rel
a
y
th
i
per fault In
front of the relay and
i
w
is
the wei
ght a
ssi
gne
d for t
he op
eratin
g
time of the relay
th
i
and i
s
usu
a
lly set to one
[30]. This obj
ect
i
ve f
unctio
n
ha
s two p
r
oble
m
s.
On
e of them
is
miscoo
rdin
ation an
d an
oth
e
r i
s
in
sen
s
ibi
lity to av
oid h
a
ving large
di
scrimin
a
tion ti
mes i
n
ad
dition
to CTI. T
o
ov
ercome
the
mentione
d dif
f
iculties in [3
1
], a ne
w
OF i
s
p
r
o
p
o
s
ed
fo
r
coo
r
din
a
tion
of
OC relays
, as follows
:
(16
)
n
1
k
).
2
bk
t
).
mbk
t
mbk
t
(
2
mk
t
.
mbk
t
mbk
t
(
2
m
1
i
2
)
i
t
(
1
F
.
O
Whe
r
e n i
s
th
e numb
e
r of
P/B relay pairs,
i
t
is the o
perating time of the rel
a
y
th
i
and
k
rep
r
e
s
ent
s e
a
ch
P/B rela
y pair
and
varie
s
from 1
to
n
.
1
α
an
d
2
α
are u
s
ed
to
co
ntrol the
weig
hting of
N
1
i
2
)
i
t
(
and
P
1
k
2
mk
t
).
mbk
t
mbk
t
(
of the O
F
and
mbk
t
is th
e
discrimi
natio
n time
betwe
en the
main and b
a
ckup ove
r
curre
n
t relays which is obtain
ed
from the equ
ation belo
w
:
(17)
CTI
-
m
t
-
b
t
mbk
t
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 16, No. 3, Dece
mb
er 201
5 : 439 – 453
448
Whe
r
e
m
t
and
b
t
are the operatin
g times of the
primary an
d backu
p relay
s
, re
spe
c
tively.
CTI is th
e
co
ordin
a
tion tim
e
interval th
a
t
is eq
ual to
0.3(sec). To
descri
be th
e
role
of the n
e
w
expre
ssi
on, consi
der
mbk
∆
t
is p
o
s
itive, then th
e third te
rm
of the OF g
a
i
n
s valu
e. Be
cau
s
e
of
multiplying b
y
2
bk
t
, the prog
ra
m tries to further redu
ce
the op
erating
time (O
T)
of the ba
ckup
relay a
nd th
e
r
efore p
r
even
ts the
und
esi
r
able i
n
crea
se
of the
OT
of
the p
r
imary
relay. Ho
weve
r,
the ne
ce
ssity
of a meth
od
by whi
c
h
th
e mentio
ned
probl
em
s co
uld be
solve
d
completely,
is
s
e
ns
ed
.
4. Simulation Res
u
lts a
nd Discu
ssi
on
4.1. IEEE 8-Bus Test S
y
st
em
The pro
p
o
s
e
d
method is a
pplied to an 8
-
bu
s, 9-b
r
an
ch netwo
rk
sh
own in Figu
re
7(a). At
bus
4, there
is a lin
k
to a
nother
netwo
rk
whi
c
h i
s
modele
d
by
a sh
ort ci
rcui
t capa
city of 400
MVA. The p
a
ram
e
ters u
s
ed in th
e ne
twork i
s
p
r
ov
ided in
refe
rence [32]. T
he tra
n
smi
s
si
on
netwo
rk con
s
ists
of 14
rela
ys which thei
r lo
cation
a
r
e
indi
cated
in
Figure 7
(
a
)
.
The T
M
S val
ues
can
ra
nge
co
ntinuou
sly fro
m
0.1 to 1.1,
while
seve
n
available
discrete pi
ckup ta
p setting
s
(0.
5
,
0.6, 0.8, 1.0,
1.5, 2.0
an
d 2.
5) a
r
e
consi
dered. T
he g
ene
ratio
n
si
ze
and
p
opulatio
n si
ze is
dire
ctly rel
a
te
d to the
chro
moso
me l
e
n
g
th; for l
ong
er l
ength
s
, m
o
re
chrom
o
somes
shoul
d
be
prod
uced. T
h
e ge
neration
si
ze
and th
e po
pulation
size a
r
e
co
n
s
ide
r
ed
to b
e
100
0 an
d 1
00,
r
e
spec
tively.
In propo
se
d
method, by usin
g the PMU me
a
s
u
r
e
m
ents, ph
asor informatio
n can b
e
obtaine
d cont
inuou
sly at a
n
y node
wh
ere PMUs a
r
e i
n
stalle
d in th
e po
wer gri
d
. This m
ean
s t
hat,
the dynami
c
cha
nge
s of n
e
twork
will b
e
ob
se
rve by
usin
g wi
de
a
r
ea m
e
a
s
u
r
e
m
ents
ba
sed
on
PMUs data. F
o
r thi
s
p
u
rp
ose initially the
Optimal
PM
U pla
c
eme
n
t fo
r full n
e
two
r
k
observability i
s
determi
ned
b
y
using
NSG
A
-II algorithm
and
with the
aid of e
quatio
n 7, which is
sho
w
n i
n
Fig
u
re
7(b
)
. Then, b
y
using PM
U
measurement
s, the valu
e
s
of voltage an
d cu
rrent ph
a
s
or in the lin
e
s
and
bu
se
s p
o
we
r
system
are o
b
taine
d
for th
e
desi
r
ed fault.
No
w acco
rdi
ng to
the l
o
cation
of
PMU, the rel
a
ys su
rroun
di
ng bu
s whi
c
h
PMU ha
s be
en install
ed a
s
onlin
e rel
a
ys and th
e oth
e
r
relays a
s
offline relays a
r
e
consi
dered. By det
ecting fault location by the PMU
and de
clare it
to
Phaso
r
Data
Con
c
entrato
r (PDC), the
values of o
n
line rel
a
ys
setting is d
e
termin
ed for t
h
e
desi
r
ed fa
ult. Finally this
informatio
n is se
nt via communi
catio
n
links PMUs for the o
p
timal
coo
r
din
a
tion
of overcu
rre
n
t relays an
d the relays can de
cid
e
whethe
r an
d whe
n
to trip a
transmissio
n
line. With thi
s
task there a
r
e ability to ch
ange th
e rela
ys setting
s
a
nd redu
ce th
eir
operating tim
e
durin
g fault occurre
n
ce.
r
1
r
8
1
65
7
23
4
r
13
r
6
r
12
r
5
r
11
r
4
r
10
r
7
r
3
r
9
r
2
r
14
2
1
8
6
7
3
4
5
(a)
(b)
Figure 7. (a)
Single line di
agra
m
of the 8-bu
s sy
stem
. (b) The o
p
timal locatio
n
of PMUs in th
e
IEEE 8-
Bus
Tes
t
Sys
t
em
Evaluation Warning : The document was created with Spire.PDF for Python.