Int
ern
at
i
onal
Journ
al of
P
ower E
le
ctr
on
i
cs a
n
d
Drive
S
ystem
(I
J
PE
D
S
)
Vo
l.
11
,
No.
3
,
Septem
be
r
2020
, pp.
12
8
7
~
12
9
7
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v
1
1
.i
3
.
pp
12
8
7
-
12
9
7
1287
Journ
al
h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
Full ve
rsus d
ecou
pled c
onstant ma
trices to
sp
eed up
power
system stat
e esti
mation
Merie
m
Majd
ou
b
1
,
Bo
uchr
a Che
dd
ad
i
2
, Om
ar Sab
ri
3
, Abdelaziz
Bel
fqi
h
4
, Jam
al Bo
ukher
ouaa
5
1,4,5
La
bora
tory
o
f
Elec
tr
ic Sys
tems a
nd
Ene
rgy
,
ENSEM,
Univer
sity
Hass
an
II
of
Casablanc
a, Moroc
co
2,3
La
bora
tory
RI
TM,
EST
,
Univ
e
rsity
Hass
an
II
o
f
Casablanc
a, Moroc
co
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
15, 2
019
Re
vised Ma
r 1
8,
2020
Accepte
d Apr
13, 202
0
Thi
s
pap
er
pr
ese
nts
a
per
for
mance
evalua
t
ion
of
two
solut
ion
s
to
red
u
ce
com
put
at
ion
al
b
urde
n
o
f
the
traditi
ona
l
W
ei
ght
e
d
L
ea
st
Squar
es
Algori
thm
for
power
sys
tem
st
at
e
esti
m
at
i
on:
Simp
li
fi
ed
me
thods
SWLS
1
/
SWLS2
base
d
on
ful
l
consta
nt
m
at
ri
c
es
and
Fast
de
coupl
ed
FD
WLS
base
d
on
dec
oupl
ed
const
ant
m
at
ri
ce
s.
Fir
st,
the
al
gor
it
h
ms
were
t
este
d
on
IEE
E
14
and
118
bus
trans
mi
ss
ion
sys
te
ms.
Second
,
the
soluti
ons
were
te
sted
on
a
rura
l
d
istri
but
io
n
fee
d
er
to
ev
aluate
the
response
of
th
e
al
gori
th
ms
to
high
R/X
ra
ti
o.
R
esul
ts
show
th
at
fo
r
tra
nsmiss
ion
sys
te
ms,
FD
WL
S
is
th
e
f
aste
st
me
thod
but
m
ore
sensit
ive
t
o
err
oneous
m
ea
surem
ent
s.
Si
mpl
ifica
ti
ons
conside
red
in
FD
WL
S,
are
not
val
id
in
distri
bu
ti
on
sys
tems
wit
h
high
R/X
rat
io
thi
s
resul
ts
in
slowing
down
the
al
go
rit
hm
conve
rge
nce
spee
d
co
nsidera
b
ly
co
mpa
red
to
SWL
S2
which
p
erf
or
ms
well.
SW
LS2
al
gor
it
hm
pre
sents
a
pro
mis
ing
soluti
on
to
red
uce
co
mputat
ion
t
im
e
for
app
li
c
at
ion
in
future
sm
art gri
d
.
Ke
yw
or
d
s
:
Decou
pled
c
on
sta
nt matrice
s
Fu
ll
consta
nt
matri
ces
Power syste
m
sta
te
esti
mati
on
Wei
ghte
d
le
ast
squar
e
s
al
gorithm
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
BY
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
M
eriem
M
aj
do
ub,
Lab
or
at
or
y of
Ele
ct
ric Sy
ste
ms a
nd En
e
r
gy, E
NS
E
M
ENS
E
M
,
Uni
ve
rsity
Hassa
n I
I of
Casa
blanc
a,
ENS
E
M
,
El Ja
did
a R
oa
d,
km
7,
BP:
81
18, O
asi
s
–
Ca
sa
bla
nca, M
orocc
o.
Emai
l:
majdou
b.
me
riem
@gm
ai
l.com
1.
INTROD
U
CTION
Re
centl
y,
the
powe
r
sy
ste
m
is
underg
oing
big
c
hanges
due
to
the
massi
ve
integ
rati
on
of
ren
e
wa
ble
energ
y
re
sourc
es
to
tra
nsmi
ssion
a
nd d
ist
rib
ution
syst
ems w
hic
h
le
ad
s
to a
com
plex
bi
directi
on
al
pow
e
r
fl
ow
.
This
c
onjuctu
r
e
incit
es
re
sea
rch
e
rs
t
o
im
prov
e
the
s
peed
an
d
reli
abili
ty
of
sta
te
est
imat
ion
al
gorit
hm
s
to
ens
ur
e a
n
e
ff
ic
ie
nt r
eal
ti
me
monit
or
i
ng of
f
utur S
mat G
rid
.
The
weig
hte
d
le
ast
-
squares
(
WLS
)
meth
od
is
the
most
Sta
te
Esti
mati
on
Algorith
m
us
e
d
i
n
c
ontrol
centers
al
l
ove
r
the
w
orl
d.
In
fact,
WL
S
st
a
te
est
imat
ion
al
gorithm
prov
i
des
the
best
es
ti
mati
on
qual
it
y
an
d
good
c
onve
r
ge
nce
rate.
H
owe
ver,
t
he
gain
a
nd
Jac
ob
ia
n
m
at
rices
nee
d
to
be
recalc
ulate
d
each
it
erati
on
wh
ic
h
needs
a
la
r
ge
amo
un
t
of
c
al
culat
ion
,
a
bi
g
mem
ory
re
qu
i
reme
nt
an
d
long
c
ompu
ti
ng
t
ime
[1].
On
e
of
the
majo
r
so
l
ution
s p
r
opos
e
d
i
n
li
te
ratur
e
to
c
ircum
ve
nt
the co
m
pu
ta
ti
onal
b
ur
den
is
the F
ast
-
dec
oupled
WLS
(F
D
WL
S)
te
chn
i
qu
e
based
on
dec
ouple
d
c
onsta
nt
m
at
rices
[
1
-
5]
as
use
d
to
s
peed
up
loa
d
flo
w
cal
culat
ion
[6
-
8].
T
he
fa
st
-
de
coupled
f
ormu
la
ti
on
ha
s
pr
oven
it
s
ef
fici
enc
y
to
re
du
ce
co
mputat
ion
ti
m
e
an
d
data
sto
rag
e
a
nd
has
f
ound
wide
acce
ptan
ce
in
the
i
ndust
ry
,
va
rio
us
ve
rsions
hav
e
be
en
im
plement
ed
in
con
t
ro
l
ce
nter
s
al
l
over
the
w
or
l
d
[4].
H
owe
ver,
the
decou
pled
meth
od
m
ay
fail
to
pro
v
i
de
a
s
olu
ti
on
on
il
l
-
conditi
on
e
d
syst
ems
as
high
R/
X
rati
o
of
di
stribu
ti
on
br
a
nch
e
s
or
in
pr
esence
of
er
r
oneo
us
meas
ur
e
ments.
Ther
e
f
or
e,
it
w
il
l
be
inte
resti
ng
t
o
de
velo
p
te
chn
i
qu
e
s
wh
ic
h
e
nsure
a
co
mpro
mise
between
the
reli
a
bili
ty
of
fu
ll
W
ei
ghte
d
Least
Squa
res
Algorith
m
a
nd Converge
nce
s
peed
of f
ast
de
coupled
W
LS an
d
able
t
o
ov
e
rcome
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
1
1
, N
o.
3
,
Se
ptembe
r
2020
:
12
8
7
–
12
9
7
1288
li
mit
at
ion
s
dis
cusse
d
a
bove.
A
uthors
of
[
5],
pr
opos
e
d
t
wo
va
riants
of
WLS
meth
od
with
f
ull
c
onsta
nt
matri
ces
(SW
LS1
a
nd
S
WL
S2
)
ba
sed
on
Dishone
st
Ga
uss
Ne
wton
M
e
thod
us
ed
f
o
r
Loa
d
Flo
w
[2,
9
-
11]
t
o
reduce
cal
cula
ti
on
ti
me
with
ou
t
decou
pled
simpli
ficat
ions.
Th
ough
the
pr
i
nciple
of
t
hese
met
hods
is
ver
y
popula
r, fe
w
st
ud
ie
s a
nal
ys
ed
it
s co
nver
ge
nc
e p
er
f
ormances
[
5,
12
-
15] on
Transmi
ssio
n
s
ys
te
ms a
nd
no
study
app
l
y
it
be
fore
to
a
distri
bu
ti
on
s
ys
te
m.
T
hi
s
arti
cl
e
is
the
first
one
e
xplo
rin
g
t
he
pe
rformance
of
f
ull
ver
s
us
decou
pled
c
onsta
nt
matri
ces
on
distrib
ution
s
ys
te
m
sta
te
est
imat
ion
with
W
LS
al
gorithm
.
The
rest
o
f
the
pa
per
is
or
gan
iz
e
d
as
fo
ll
ow
:
S
ect
ion
2
pr
ese
nts
a
des
cri
ptio
n
of
th
e
tradit
iona
l
we
igh
te
d
le
ast
squares
al
gorithm
an
d
the
al
te
r
native
s
olu
ti
ons
to
r
edu
ce
c
omp
utati
on
al
bur
den.
T
he
al
gorith
ms
st
ud
ie
d
are
te
ste
d
bo
t
h
on
tra
ns
m
issi
on
a
nd
distrib
ution
syst
ems,
e
mp
l
oy
i
ng
accu
rate
an
d
e
rron
e
ous
meas
ur
e
ments.
Sim
ulati
on
resu
lt
s a
re
pr
es
ented
a
nd
disc
us
se
d
in
Secti
on3. T
he mai
n f
ind
in
gs o
f
the
pap
e
r
a
re as
foll
ow
s:
-
Fo
r
Tra
ns
missi
on
Sy
ste
ms,
F
DWLS
based
on
dec
oupled
c
on
sta
nt
matri
c
es
is
the
best
method,
it
is
2
to
5
ti
mes
faster
than ba
sic
W
L
S
and
requires
hal
f
sto
rag
e
c
ap
aci
ty.
-
SWL
S1
meth
od
c
on
si
der
i
ng
const
ant
Jac
obia
n
matri
x
on
l
y
is
no
t
reli
abl
e
an
d
do
es
not
represe
nts
a
ny
adv
a
ntage
on r
edu
ci
ng com
puta
ti
on
ti
me.
-
Fo
r
distri
bu
ti
on
s
ys
te
ms,
F
D
WLS
base
d
on
dec
oupled
co
nst
ant
matri
ces
conve
rg
es
slo
wly
a
nd
re
quir
es
high
it
er
at
ion
s
numb
e
r
w
hile
the
SWL
S2
al
gorithm
e
valua
ti
ng
the
Jac
obia
n
an
d
gain
m
at
rices
on
ce
a
t
the
flat
sta
rt
is
2
ti
mes
faster
tha
n
basic
W
LS
with
t
he
sa
me
reli
a
bili
ty
wh
ic
h
make
it
le
ss
se
ns
it
ive
to
erron
e
ous
mea
su
re
ments
a
nd
high
R/
X
rati
o
com
par
e
d
to
FDWLS.
The
r
efore,
S
WLS
2
pr
ese
nts
a
go
od
al
te
rn
at
ive to
re
du
ce
comp
uta
ti
on
ti
me i
n
f
ut
ur
e
po
wer
sy
st
ems.
2.
WEIGHT
ED
LE
AS
T SQ
U
AR
ES
A
L
GO
RITH
MS
2.1.
Basic al
go
ri
thm
The
Net
wor
k
model
empl
oyed
is
the
sin
gl
e
-
phase
m
od
e
l
with
N
bus
es
and
m
me
asur
e
ments
gathe
red
f
rom
rem
ote
mete
r
s.
M
ost
c
om
m
on
l
y
us
e
d
me
asur
e
ments
ar
e
the
li
ne
power
fl
ow
s
,
bu
s
po
wer
injec
ti
on
s a
nd
bu
s
volt
age
ma
gn
it
udes
.
The
ai
m
of
sta
te
est
imat
or
is
to
pro
vid
e
t
he
best
possi
ble
va
lues
of
the
bu
s
volt
age
ma
gnit
ud
e
s
an
d
ang
le
s
by pr
oc
essing t
he
a
vai
la
ble
netw
ork
data rec
ognizin
g
that t
her
e
are
erro
rs
in
the
m
easur
e
d q
uan
ti
t
ie
s.
The
sta
rting eq
uation f
or the
WLS
stat
e esti
mati
on
al
gorith
m is [4,
16
-
18]
:
=
ℎ
(
)
+
(1)
Wh
e
re: z is t
he
(
m
x1) mea
s
urement
vecto
r;
x
is
an
(nx
1)
sta
te
ve
ct
or
t
o
be
est
ima
te
d:
T
he
num
be
r
of
est
imat
ed
sta
te
s
is
n=
2*N
-
1,
since
the b
al
a
nce
ph
ase’s is al
ready
know
n
ϴ
=0
.
e is an
(
m
x1)
measu
reme
nt e
rror vect
or
.
h
is t
he vect
or
of no
nlinear
f
unct
ions t
hat
rel
at
e the stat
es to
the mea
surem
ents
def
ine
d:
Re
al
an
d R
eact
ive po
wer i
nje
ct
ion
at
bus i:
=
∑
(
+
)
≠
(2)
=
∑
(
+
)
≠
(3)
Re
al
an
d react
i
ve powe
r
fl
ow
from b
us
i t
o b
us
j:
=
2
(
+
)
−
(
+
)
(4)
=
−
2
(
+
)
−
(
−
)
(5)
Wh
e
re:
Vi
is t
he vo
lt
a
ge
magnit
ude at
bu
s
i
θi
is t
he ph
a
se
ang
le
at
bus i
=
−
Gij + jBi
j
is t
he
ij
th element
of the
Y
-
bus
m
at
rix
gij +
j
bij
is t
he
ad
mit
ta
nce
of
the series
bran
ch betwee
n bus i an
d b
us
j
gs
i +
jbsj
is t
he
ad
mit
ta
nce
of the
shu
nt bra
nch at b
us i
.
In
pr
act
ic
e,
it
is
require
d
to
ha
ve
the
numb
e
r
of
meas
ur
e
m
ents
la
rg
e
r
tha
n
num
ber
of
sta
te
s,
this
is
cal
le
d
re
dunda
ncy
[
19].
S
o,
s
ta
te
est
imat
or
can
c
on
si
der
the
var
i
ou
s
ope
rati
on
la
youts
us
e
d
an
d
t
o
c
over
f
or
the
una
vaila
bili
ty
of
tran
smis
sion
a
nd
te
le
m
et
ering
e
quip
ment
fail
ures
[
20].
A
m
easu
r
e
of
t
he
redu
ndanc
y
may be
de
no
te
d by the
re
dundanc
y fact
or
ɳ, w
hich
is
d
e
fin
ed
as
[21]:
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N:
20
88
-
8
694
Full vers
us
dec
oupled
const
ant ma
tri
ces t
o
s
pe
ed up p
ower
s
yst
em
state e
sti
ma
ti
on
(
Meri
e
m
M
ajdo
ub
)
1289
=
=
=
2
−
1
(6)
The me
asu
rem
ent er
rors
ei
a
r
e assume
d t
o s
at
isfy
the
foll
owin
g
sta
ti
sti
cal
propertie
s:
First,
t
he
e
rrors
ha
ve
z
er
o
m
ean:
E
(ei)
=
0,
i
=
1,
.
..,
m.
Sec
ond,
the
error
s
are
ass
ume
d
to
b
e
ind
e
pende
nt,
s
uch that t
he
c
ovaria
nce m
at
rix
is
diag
onal
:
Cov(e)
= E (e,
e
T
)
= R =
d
ia
g {
σ
1
2
,
σ
2
2
, …,
σ
m
2
}
(7)
The
s
olu
ti
on
to
the
sta
te
est
imat
ion
pro
blem
can
be
f
or
m
ulate
d
as
a
mi
nimiza
ti
on
of
t
he
f
ollow
i
ng
ob
je
ct
ive
fun
ct
ion
:
(
)
=
∑
(
−
ℎ
(
)
)
2
=
[
−
ℎ
(
)
]
−
1
[
−
ℎ
(
)
]
=
1
(8)
To
fin
d
the
minimi
zat
ion
of
this
obje
ct
ive
functi
on
t
he
de
rivati
ve
sho
uld
be
s
et
to
z
er
o.
The deri
vative
of the
obje
ct
ive fun
ct
io
n
is
de
no
te
d by g(
x)
:
(
)
=
(
)
=
−
(
)
−
1
[
−
ℎ
(
)
]
=
0
(9)
wh
e
re:
H(x)
=
∂
h(x
)/∂
x, ca
ll
ed
the
meas
ur
e
ment Jac
obia
n mat
rix.
Ign
or
in
g
the
hi
gh
e
r
orde
r
te
r
ms
of
t
he
T
ay
lor
series
e
xp
a
ns
io
n
of
t
he
de
rivati
ve
of
th
e
obje
ct
ive
functi
ons
yield
s an i
te
rati
ve
s
olu
ti
on as
sho
wn b
el
ow
:
+
1
=
+
[
(
)
]
−
1
[
[
(
)
]
[
]
−
1
[
−
ℎ
(
)
]
]
(10)
Whe
re t
he gain
matrix
, G, is
de
fine
d
as:
(
)
=
(
)
=
−
1
(11)
Fo
r
t
he
first
it
erati
on
o
f
t
he
opti
miza
ti
on
pr
ob
le
m
,
an
i
niti
al
gu
ess h
as
to b
e
ma
de
f
or
th
e
sta
te
vecto
r
x0
co
rr
e
spo
nding
t
o
t
he
flat
volt
age
pro
file
,
or
flat
sta
rt
.
A
flat
sta
rt
ref
e
rs
to
a
sta
te
vecto
r
whe
re
al
l
the
vo
lt
age
ma
gn
it
udes
are
1.
0
per
un
it
a
nd
al
l
the
volt
ag
e
an
gles
a
re
0
de
gr
ee
s.
With
res
pect
to
the
sta
te
vecto
r:
the
me
asur
e
ment
f
unct
ion
s,
Jaco
bia
n
a
nd
gai
n
mat
rices
are
cal
c
ul
at
ed
each
it
erati
on
unti
l
the
ab
so
lute
diff
e
re
nce
between
tw
o
s
ucc
essive
values
of
x
is
le
ss
tha
n
a
c
ho
se
n
toler
ance
ɛ
.
WLS
fl
ow
c
ha
rt
prese
nt
ed
in
Figure
1, res
ume
s the
princi
pal steps
of t
he
algorit
hm.
Figure
1. Ba
sic
W
L
S f
l
ow
c
ha
rt
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
1
1
, N
o.
3
,
Se
ptembe
r
2020
:
12
8
7
–
12
9
7
1290
2.2.
Simpl
ifie
d a
l
gori
th
m
s w
i
th
f
ull
co
ns
tant
m
at
ri
ces
M
ost
ly,
it
is
obser
ve
d
t
hat
t
he
el
eme
nts
of
Jaco
bian
an
d
ga
in
m
at
rices
do
not
si
gn
i
f
ic
antly
c
ha
nge
betwee
n
flat
st
art
init
ia
li
zat
io
n
an
d
t
he
co
nver
ge
d
so
l
utio
n
[
5].
S
o,
t
heir
eval
uation
c
ou
l
d
be
restrai
ned
t
o
so
me
fi
rst
it
er
at
ion
s
with
out
al
te
rin
g
t
he
est
imat
ion
qua
li
ty.
Ba
sed
on
this
pri
nciple
tw
o
meth
od
s
ha
ve
emer
ged
:
SWL
S1
a
nd S
WL
S
2.
2.2.1.
The first
simp
li
fied meth
od
SWLS1
The
first
sim
plifie
d
meth
od
(
SWL
S1)
cal
c
ul
at
es
the
Jaco
bi
an
matri
x
at
e
very
it
erati
on
bu
t
prese
rv
es
the
gai
n
matri
x
c
on
sta
nt
afte
r
an
it
erati
on
k
ch
os
en
as
pr
e
sented
i
n
Fi
gure
2.
T
he
obvi
ou
s
ad
va
ntage
of
t
his
method is
to
re
du
ce
the
num
be
r of
calc
ulati
ons
of the
gain
matri
x.
2.2.2.
The seco
nd
si
mpli
fied met
h
od
SWLS
2
The
sec
ond
me
thod
S
WLS
2
a
dm
it
s
that
t
he
gain
a
nd
the
Ja
cob
ia
n
mat
rices
remai
n
c
on
st
ant
after
an
it
erati
on
k
c
hosen
as
prese
nted
in
Fig
ure
3
[
5,
13
-
15]
.
Figure
2. S
WL
S1
flo
wc
har
t
Figure
3. S
WL
S2
flo
wc
har
t
2.3.
Fast dec
oupl
ed w
ei
gh
t
ed le
ast sq
ua
re
s
algorithm
(
F
DW
LS)
The
pr
i
nciple
of
Fast
D
ecoupled
Stat
e
Esti
mati
on
Tech
niques
consi
sts
on
exp
l
oiti
ng
the
act
ive/r
eac
ti
ve
decou
plin
g
pro
per
t
y
bas
ed
on
fast
dec
oup
le
d
l
oad
fl
ow
met
hods
[
2,
9
-
11].
I
nde
ed,
f
or
la
rg
e
scal
e
po
wer
s
ys
te
ms
th
e
tra
ns
missi
on
li
nes
hav
e
a
ve
ry
high
X/R
r
at
io.
In
s
uc
h
a
case,
the
real
powe
r
changes
a
re
le
ss
sensiti
ve
to
c
hanges
in
volt
age
ma
gn
it
ude
and
ca
n
be
ignore
d.
Simi
la
rl
y,
the
reacti
ve
powe
r
change
is
le
ss
sensiti
ve
to
c
hanges
i
n
a
ng
l
es.
M
a
ki
ng
t
he
se
simpli
ficat
ion
s
,
the
gain
matri
x
in
W
L
S
sta
te
est
imat
ion
alg
ori
thm ca
n be
simpli
fied
[1
0
-
11]
.
In
the
dec
oupl
ed
f
ormulat
io
n,
the
meas
ur
e
ment
vect
or
Z
is
portion
e
d
in
to
two
par
ts:
a
ct
ive
ZA
a
nd
reacti
ve
ZR c
omp
on
e
nts
[2
-
4].
Z=
(
)
, Z
A
=
(
)
, Z
R
=
(
)
(12)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N:
20
88
-
8
694
Full vers
us
dec
oupled
const
ant ma
tri
ces t
o
s
pe
ed up p
ower
s
yst
em
state e
sti
ma
ti
on
(
Meri
e
m
M
ajdo
ub
)
1291
Wh
e
re:
Pinj,
Qinj: are
resp
ect
ivel
y re
al
an
d react
ive
powe
r
injec
ti
on mea
sureme
nt
s.
Pf
lo
w,
Q
flo
w:
are r
e
sp
ect
ivel
y real
and
react
ive po
wer flo
w
meas
ur
eme
nts
.
V: Volt
age
me
asur
e
ments
.
By the
d
e
finiti
on abo
ve
the
J
acob
ia
n
m
at
rix
H
,
and c
ovaria
nce m
at
rix
R
c
an be
wr
it
te
n
a
s:
H=
[
]
, R=
[
0
0
]
(13)
Ba
sed
on
P
-
ϴ
an
d
Q
-
V
dec
ouplin
g,
the
off
-
dia
gonal
bl
ocks
a
nd
in
t
he
meas
ure
ment
ja
cob
ia
n H
are
ignore
d:
H=
[
0
0
]
(14)
The
n
the
g
ai
n mat
rix
is e
xpre
ssed
as
foll
ow:
G=
[
0
0
]
(15)
Wh
e
re:
=
−
1
(16)
=
−
1
(17)
Seve
ral
ve
rsions
of
t
he
fast
-
decou
pled
sta
t
e
est
imat
or
ha
ve
been
pro
po
sed
in
li
te
rature
de
pendin
g
upon
the
ass
umpti
ons
a
nd
a
ppr
ox
imat
io
ns
adopted
.
T
he
ver
si
on
rec
ogni
zed
as
present
ing
best
pe
rformanc
e
has
t
he
f
ollo
wing f
eat
ur
es
[2]
:
-
The
Jac
ob
ia
n
and
gain
matri
ces
are
eval
ua
te
d
on
ce,
at
th
e
flat
vo
lt
age
sta
rt:
al
l
V=1
p.u
an
d
ϴ
=
0
degrees.
-
The
branc
h
ser
ie
s
resist
ance
s ar
e
ig
nore
d
in
f
ormin
g
the
el
ements of
t
he
Ja
cob
ia
n
H, wh
ic
h
is
eq
uiv
al
e
nt
to
re
placi
ng
br
anch
s
us
ce
ptan
ces
bij
by
-
1
/
xij.
Wh
e
re
xij
i
s
the
reacta
nce
of
t
he
se
ri
es
branc
h
betwee
n
bu
s
i a
nd bus
j.
-
A
tra
nsfo
rme
d
meas
ur
e
ment
vect
or
s
Z
A’
an
d
ZR’
are
us
e
d
by
di
vid
in
g
t
he
flo
w
an
d
injec
ti
on
measu
reme
nts
by the c
orres
pondin
g
cal
c
ulate
d vo
lt
age
ma
gnit
ud
e
.
-
Fo
r
bet
te
r
co
nv
erg
e
nce c
ha
rac
te
risti
cs, a b
l
oc
k
se
quentia
l so
luti
on
sche
me i
s u
se
d w
her
ei
n:
∆
=
(18)
TA=
HAA
T
*RA
-
1
*
Δ
ZA
’
(19)
ΔZA’=
(Z
A
-
hA)/V
(20)
hA is the
v
ect
or
of
acti
ve
mea
su
re
ments
fun
c
ti
on
s.
∆
is sol
ved an
d
t
he update
d
′
are
u
se
d
i
n
the
RHS of
:
∆
=
(21)
Wh
ic
h
is t
hen
so
lve
d for
∆
.
TR=
HRR
T
*RR
-
1
*ΔZ
R
’
(22)
ΔZR’
=(ZR
-
hR
)
/V
(23)
hR is the
v
ect
or
of
reacti
ve m
easur
e
ments
fu
nctions.
Fo
r
m
ore
deta
il
s,
ab
ou
t
ho
w
to
cal
culat
e
t
he
el
eme
nts
of
the
fast
-
dec
ouple
d
Ja
co
bian
,
rea
de
r
ca
n
ref
e
r
to
[2
-
4].
Algorith
m step
s ar
e
de
picte
d
i
n
Fi
gure
4:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
1
1
, N
o.
3
,
Se
ptembe
r
2020
:
12
8
7
–
12
9
7
1292
Figure
4.
F
DWLS f
l
ow
c
ha
rt
3.
SIMULATI
O
N RESULTS
The
obje
ct
ive
of
this
sect
ion
is
to
c
ompare
the
perf
orma
nce
of
t
he
Weig
hted
Le
ast
S
qu
a
res
Algorith
ms
de
scribe
d
i
n
t
he
pre
vious
Sec
ti
on
:
Ba
sic
W
LS,
S
WLS
1,
SWL
S2
a
nd
FDWLS
in
te
rms
of
com
pu
ta
ti
on ti
me, c
onve
rg
e
nc
e rate, esti
mat
ion
qual
it
y
a
nd num
ber o
f
it
er
at
ion
s.
3.1.
Descripti
on
of
simul
ati
on
s
The
fou
r
al
gor
it
hm
s
hav
e
bee
n
te
ste
d
on
th
r
ee
stu
dy
cases
IEEE
14,
118
bu
s
s
ys
te
ms
a
nd
a
12
-
bus
rural distrib
ution
s
ys
te
m
un
de
r
tw
o
sce
nar
i
os:
I
n
s
ce
na
rio
1,
acc
ur
at
e
mea
su
re
ments
e
qual
to
l
oad flow r
esults
hav
e
been
em
pl
oy
e
d.
I
n
scena
rio
2,
measu
re
ments
we
re
al
t
ered
by
a
noise
of
10%
to
eva
luate
the
res
po
ns
e
of
the stu
died
alg
or
it
hms t
o er
roneous
meas
ur
e
ments.
Fo
r
IE
EE
bus
sy
ste
ms
,
the
ne
twork
data
file
s
can
be
dow
nlo
a
ded
f
rom
Power
Syst
em
s
Test
Ca
se
Ar
c
hiv
e
[22].
To
c
ompare t
he
stat
e esti
mate
accurac
y of
t
he
f
ollo
wing si
mu
la
ti
ons,
mea
n
ab
so
l
ute p
e
rc
entage e
rror
(MAPE
)
is i
ntr
oduce
d
as
foll
ow [2
3]
:
=
1
∑
|
−
|
×
100%
=
1
(24)
Wh
e
re,
X
is
the
tr
ue
value
of
s
ys
te
m
sta
te
ob
ta
ine
d
from
load
flo
w
resu
l
ts
and
Xe
is
th
e
est
imat
ed
sta
te
. A
s
mall
er
value of
M
A
PE in
dicat
es a
more acc
ur
at
e
sta
te
esti
mati
on
resu
lt
.
M
A
PE
V: mea
n
a
bsolute
per
c
entage e
sti
mati
on er
ror of
volt
age m
a
gnit
ude.
M
A
PE
ϴ: m
ea
n
a
bsolute
per
c
entage e
sti
mati
on er
ror of
volt
age a
ng
le
.
The
sim
plifie
d
al
gorithms
ha
ve
bee
n
te
ste
d
for
di
ff
e
ren
t
value
of
t
he
it
erati
on
k
f
r
om
wh
ic
h
just
the
gain
matri
x
is
c
on
si
der
e
d
c
onsta
nt
S
WLS
1
or
t
he
gain
an
d
Jac
obia
n
matri
ces
are
c
onsta
nt
(
SWL
S2
).
Conver
ge
nce t
olera
nce
us
ed
i
n
the
se test
s is:
ɛ=1
0
-
4 for
both
vo
lt
age
ma
gnit
ud
e
and
volt
age a
ng
le
sta
te
s.
3.2.
Te
st
syste
ms
Fo
r
al
l
te
st
cas
es,
meas
ur
eme
nts
we
re
set
to
ensure
a
re
dund
a
nc
y
facto
r
>1.
T
hey
wer
e
cho
se
n
of
diff
e
re
nt
typ
es
and
unifo
rml
y
distri
bute
d
th
r
ough
the
netw
ork
to
e
nsure
obser
va
bili
ty
[16,
17].
Weig
ht
of
al
l
measu
reme
nts
is assu
med
1.
a.
IEEE 1
4 bus s
ys
te
m:
Fo
r
IEEE
14
bus s
ys
te
m
test
case, a set
of
41 mea
sureme
nt
s (
η =
1,5) i
s c
hose
n
as:
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N:
20
88
-
8
694
Full vers
us
dec
oupled
const
ant ma
tri
ces t
o
s
pe
ed up p
ower
s
yst
em
state e
sti
ma
ti
on
(
Meri
e
m
M
ajdo
ub
)
1293
-
1 vo
lt
age
ma
gnit
ud
e
at b
us 1
.
-
8 real
po
wer
i
nj
ect
ion
s a
nd 8 r
eact
ive powe
r
i
nject
ion
s
at b
use
s: 2, 3
, 7, 8,
10, 11,
12, 14.
-
12
real
power
flo
w
a
nd
12
reacti
ve
po
wer
flo
w
on
bran
ches:
1
-
2,
2
-
3,
4
-
2,
4
-
7,
4
-
9,
5
-
2,
5
-
4,
5
-
6,
6
-
13, 7
-
9, 11
-
6, 12
-
13.
b.
IEEE 1
18 bus
sy
ste
m:
A
set
of 72
6
m
easur
e
ments
(η
=
3)
is
chose
n a f
ollow
:
-
118 v
oltage
m
agn
it
udes
for a
ll
b
us
es
.
-
118 real
power i
nject
ion
s
and
118 react
ive
powe
r
i
nject
ion
s
for
all
buses.
-
186 real
power flo
w
a
nd
186 r
eact
ive powe
r flo
w
f
or all
n
et
work bra
nch
e
s.
c.
12 bus dist
ri
buti
on
s
ys
te
m:
A
ph
ys
ic
al
ly
e
xi
sti
ng
rural
di
stribu
ti
on
Fee
der
[24]
is
c
onside
red.
Fig
ure
5
s
hows
t
he
sing
le
-
li
ne
diag
ram
of
t
he
sy
ste
m
,
w
he
re
node
1
is
t
he
s
ub
sta
ti
on.
The
sy
ste
m
data
in
cl
ud
in
g
li
ne
da
ta
and
l
oad
dat
a
can
be fo
und i
n [25
].
Figure
5. Sin
gl
e
-
li
ne diag
ram of
the
12
-
bus
di
stribu
ti
on s
ys
t
em.
A
set
of 25
me
asur
e
ments
(
η
=1,1)
is ch
os
e
n as f
ollo
w:
-
1 vo
lt
age
ma
gnit
ud
e
at b
us 1
.
-
1 real
po
wer
i
nj
ect
ion
a
nd 1 r
eact
ive powe
r
i
nject
ion
at
bus 1.
-
11 r
eal
powe
r f
low
a
nd
11 r
ea
ct
ive pow
e
r flo
w
f
or all
n
et
work branc
hes
.
3.3.
Results
of sc
e
na
ri
o 1
(accur
at
e
mea
surem
ents)
In
this
sect
io
n,
WL
S
al
go
r
it
hm
s
hav
e
be
en
te
ste
d
with
accu
rate
mea
su
re
ments
eq
ua
l
to
loa
d
flo
w
re
su
lt
s.
3.3.1.
Simul
at
i
on
re
sults
fo
r
IEEE
14 bus sys
tem
As
see
n
i
n
Ta
bl
e 1
:
FDWLS
is
t
he
fastest
meth
od
even
i
f
num
be
r
of
it
erati
ons
is
increase
d
c
ompare
d
to
t
he
ba
sic
WLS.
FDWLS
so
l
ution cha
nges sli
gh
tl
y b
ut
pr
eci
sion remai
ns
good.
SWL
S1
do
es
not
c
onve
rg
e
at
flat
sta
rt
(
k=1)
,
for
k=
2
num
ber
of
it
erati
on
s
a
nd
c
omp
utati
on
ti
me
is
increase
d
c
ompare
d
t
o
the
o
t
her met
ho
ds
. T
he
s
olu
ti
on is t
he
sa
me as Bas
ic
W
LS
.
SWL
S2
a
pp
li
e
d
at
flat
sta
rt
(
k=1)
co
nver
ge
s
in
half
ti
me
com
par
e
d
to
t
he
basic
WLS
with
a
good
qu
al
it
y o
f
est
i
mati
on
e
ve
n
if
numb
e
r of i
te
r
at
ion
s
require
d i
s imp
or
ta
nt c
ompare
d
t
o
the
o
the
r met
hods.
Table
1.
Per
for
ma
nce e
val
uation o
f WLS
alg
or
it
hms
for IE
EE 14
bus s
ys
t
em un
der scen
ario
1
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ (
%)
Bas
ic WLS
0
.00
4
5
4
1
4
1
.28
2
.41
SW
LS1 (k=1
)
Prog
ram
do
es n
o
t
co
n
v
erge
SW
LS1 (k=2
)
0
.00
5
7
5
1
7
1
.28
2
.41
SW
LS1 (k=3
)
0
.00
4
1
0
2
4
1
.28
2
.41
SW
LS2 (k=1
)
0
.00
2
1
1
3
8
0
.57
0
.87
SW
LS2 (k=2
)
0
.00
2
5
5
0
4
1
.33
2
.5
SW
LS2 (k=3
)
0
.00
3
5
4
5
4
1
.28
2
.41
FDW
LS
0
.00
1
4
1
8
6
0
.34
0
.45
Figure
6
sho
ws
t
hat
F
D
W
LS
a
nd
S
WL
S2
(
k=1)
a
re
cl
os
er
to
t
he
t
ru
e
val
ue
t
han
basic
W
LS
,
the same
f
ig
ur
e ap
pear
a
nce
was ob
ta
i
ned f
or volt
age a
ngle
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
1
1
, N
o.
3
,
Se
ptembe
r
2020
:
12
8
7
–
12
9
7
1294
Figure
6. Esti
mate
d vo
lt
a
ge mag
nitud
e
(P.
U)
by
dif
fer
e
nt W
L
S alg
or
it
hms f
or IE
EE
14
bu
s
sy
ste
m
unde
r
s
ce
nar
i
o 1
3.3.2.
Simul
at
i
on
re
sults
fo
r
IEEE
118 bus
syste
m
Accor
ding
t
o
Table
2:
he
re
al
so
,
S
WLS
1
do
e
s
not
c
onve
rg
e
f
or
k=
1.
It
seem
s
t
hat
const
ant
gain
matri
x
ass
ociat
ed
t
o
va
riable
Jaco
bian
ma
y
l
ead
t
o
c
onve
r
ge
nce
pro
blems.
F
DWLS
co
m
pu
ta
ti
on
ti
me
i
s
the
lowest.
Th
e
all
meth
ods
hav
e
pro
vid
e
d
a
good esti
mati
on
qual
it
y.
Table
2.
Per
for
mance e
val
uation o
f WLS
alg
or
it
hms
for IE
EE 11
8 bus s
yst
em und
e
r
sce
na
rio
1
3.3.3.
Simul
at
i
on
re
sults
fo
r
12 bu
s distribu
tio
n
system
As
noti
ced
i
n
Table
3,
S
W
LS1
has
c
onve
rg
e
d
f
or
k=
1
because
the
final
sta
te
is
too
cl
os
e
to
the init
ia
l st
at
e (Basic
W
L
S c
onve
r
ges on
l
y
i
n 3 iterat
io
ns
)
.
FDWLS
co
nv
erg
e
nce
sp
ee
d
slo
wed
do
w
n
c
ons
ide
rab
l
y
c
ompare
d
to
th
e
oth
e
r
al
gorithms.
The
so
l
utio
n
of
F
DWLS
is
t
he
le
ss
accu
rate,
an
d
th
e p
r
ogra
m r
eq
uire
d
a h
i
gh
it
erati
on
nu
mb
e
r
(60,5
a
ga
inst 3
for
the
basic
f
ormulat
io
n).
S
WLS
2
(k
=
1),
conve
rg
e
d
i
n
le
ss
ti
me
an
d
i
ts
so
luti
on
is
accurate
as
th
e
bas
ic
WLS
ones.
Table
3.
Per
for
mance e
val
uation o
f WLS
alg
or
it
hms
for 12
bu
s
d
ist
rib
utio
n
s
ys
te
m
unde
r
scena
rio
1
Re
su
lt
s
are
de
picte
d
in
Fi
gur
e
7:
the
dif
fe
re
nces
be
twee
n
e
sti
mate
d
volt
ages
by
S
WL
S2(k
=
1),
Ba
sic
WLS
and t
he
tr
ue
values
are s
o
s
mall
that the
y
ca
nnot
be dis
ti
ng
uis
he
d.
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ
(%
)
Bas
ic WLS
0
.18
2
3
4
7
4
1
.79
1
.74
SW
LS1 (k=1
)
Prog
ram
do
es n
o
t
co
n
v
erge
SW
LS1 (k=2
)
0
.26
5
6
2
9
7
1
.79
1
.74
SW
LS1 (k=3
)
0
.15
6
1
2
7
4
1
.79
1
.74
SW
LS2 (k=1
)
0
.08
4
7
2
6
10
1
.81
1
.13
SW
LS2 (k=2
)
0
.09
8
5
1
8
5
1
.77
1
.72
SW
LS2 (k=3
)
0
.13
5
4
9
5
4
1
.79
1
.74
FDW
LS
0
.03
7
5
5
8
5
2
.07
1
.1
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ
(%)
Bas
ic WLS
0
.00
2
2
0
1
3
0
.01
1
.02
SW
LS1 (k=1
)
0
.00
3
6
0
1
5
0
.01
1
.02
SW
LS1 (k=2
)
0
.00
2
0
2
1
3
0
.01
1
.02
SW
LS2 (k=1
)
0
.00
1
5
4
5
5
0
.01
1
.02
SW
LS2 (k=2
)
0
.00
1
9
0
6
3
0
.01
1
.02
FDW
LS
0
.00
3
8
7
9
6
0
.5
0
.76
1
9
.87
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N:
20
88
-
8
694
Full vers
us
dec
oupled
const
ant ma
tri
ces t
o
s
pe
ed up p
ower
s
yst
em
state e
sti
ma
ti
on
(
Meri
e
m
M
ajdo
ub
)
1295
Figure
7. Esti
mate
d vo
lt
a
ge mag
nitud
e
(P.
U)
by
dif
fer
e
nt W
L
S alg
or
it
hms f
or
12 bus dist
rib
ution s
yst
em
unde
r
sce
nar
i
o1
3.4.
Results
of sc
e
na
ri
o 2
(
err
oneous
meas
ure
ments
)
In
this
sect
io
n,
measu
reme
nts
wer
e
al
te
re
d
by
a
noise
of
10
%
to
evaluate
t
he
res
pons
e
of
the
stud
ie
d
al
gorithms t
o
e
rron
e
ous
meas
ur
e
ments.
3.4.1.
Simul
at
i
on
re
sults
fo
r
IEEE
14 bus sys
tem
Table
4
sho
ws
that
ba
d
meas
ureme
nts
al
te
r
the
est
imat
io
n
qual
it
y
f
or
al
l
WLS
al
gorith
ms.
S
pecial
ly
,
FDWLS
w
hic
h
pro
vi
ded
the
l
ess
acc
ur
at
e
sol
ution
.
Howe
ve
r,
FDWL
S
m
et
hod
sti
ll
con
s
ta
ntly
the
faste
st
one.
Fo
r
this ca
se,
S
WLS
1 ha
ve
fai
le
d
not
on
l
y
at
flat
start b
ut al
so
for
t
he
sec
ond i
te
rati
on (k
=2).
Accor
ding
to
Figure
8,
All
WLS
al
gorith
ms
s
olu
ti
ons
a
re
fa
r
from
th
e
true
val
ues.
The
FDWLS
curve
is
t
he
f
urt
hest
w
hile
B
asi
c
W
LS
cu
r
ve
is
the
near
e
st
one
to
the
t
ru
e
values
c
urve.
T
ho
s
e
res
ul
ts
are
the in
ver
se
of t
ho
s
e
ob
ta
ine
d
i
n
sce
nar
i
o 1.
Table
4.
Per
for
mance e
val
uation o
f WLS
A
l
gorithms
fo
r
IEEE
14 bus sy
ste
m un
der
sce
nar
i
o2
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ
(
%)
Bas
ic WLS
0
.00
5
8
5
8
5
7
.82
5
.07
SW
LS1 (k=1
)
Prog
ram
do
es n
o
t
co
n
v
erge
SW
LS1 (k=2
)
Prog
ram
do
es n
o
t
co
n
v
erge
SW
LS1 (k=3
)
0
.00
5
2
6
6
5
7
.82
5
.07
SW
LS2 (k=1
)
0
.00
2
3
9
1
8
9
.08
7
.32
SW
LS2 (k=2
)
0
.00
3
1
2
7
5
7
.3
4
.17
SW
LS2 (k=3
)
0
.00
3
5
4
5
4
7
.8
5
.04
FDW
LS
0
.00
1
4
5
9
6
.5
9
.6
8
.15
Figure
8. Esti
mate
d vo
lt
a
ge mag
nitud
e
(p.
u)
by d
if
fer
e
nt
WLS
alg
or
it
hms f
or IE
EE
14
bu
s
s
ys
te
m
unde
r
scenari
o2
3.4.2.
Simul
at
i
on
re
sults
fo
r
IEEE
118 bus
syste
m
Re
su
lt
s
prese
nted
i
n
Ta
ble
5,
are
le
ss
af
fected
by
no
ise
co
mp
a
red
to
the
pr
e
vious
case
I
EEE
14
bu
s
sy
ste
m,
es
peci
al
ly
for
volt
ag
e
ang
le
sta
te
.
In
fact,
the
set
of
meas
urem
ents
ch
os
e
n
f
or
this
stu
died
m
odel
,
pr
ese
nts
a
hi
gh
re
dundanc
y
le
vel
(η
=
3)
wh
i
ch
al
lo
wed
to
t
he
WLS
Al
gor
it
hm
s
to
pro
vi
de
good
est
im
at
ion
.
SWL
S1 has
conv
e
r
ged f
or k
=
2
in
m
or
e
it
erati
on
s
and c
omp
uting t
ime c
ompare
d
t
o
the
o
t
her al
gorith
ms.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
1
1
, N
o.
3
,
Se
ptembe
r
2020
:
12
8
7
–
12
9
7
1296
Table
5.
Per
for
mance
Eval
uation o
f WLS
Al
gori
thms
for I
EEE
118
B
us
s
ys
te
m
unde
r
S
cenari
o2
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ
(
%)
Bas
ic WLS
0
.18
4
7
4
7
4
6
.65
0
.99
SW
LS1 (k=1
)
Prog
ram
do
es n
o
t
co
n
v
erge
SW
LS1 (k=2
)
0
.45
6
2
5
6
14
6
.66
0
.99
SW
LS1 (k=3
)
0
.17
0
3
8
8
4
6
.65
0
.99
SW
LS2 (k=1
)
0
.08
4
1
3
7
10
7
.27
1
.85
SW
LS2 (k=2
)
0
.09
5
1
4
7
5
6
.54
0
.87
SW
LS2 (k=3
)
0
.13
0
6
8
4
4
6
.66
0
.99
FDW
LS
0
.04
1
3
4
3
5
.5
8
.37
2
.93
3.4.3.
Simul
at
i
on
re
sults
fo
r
12 bu
s distribu
tio
n
system
As
noti
ced
in
Table
6,
her
e
al
so
t
he
FDW
LS
has
co
nver
ged
sl
ow
l
y
wi
th
high
it
erati
on
numb
e
rs.
Wh
il
e,
SWL
S
2(k=
1), c
onve
rges in
reduce
d
ti
me compa
re
d
t
o
the
b
a
sic
WL
S, wit
h
the
sa
me r
el
ia
bili
ty.
Table
6.
Per
for
mance e
val
uation o
f WLS
alg
or
it
hms
for 12
bu
s
d
ist
rib
utio
n
s
ys
te
m
unde
r
scena
rio
2
Alg
o
rithm
Co
m
p
u
tatio
n
tim
e
(secon
d
s)
Iter
atio
n
s n
u
m
b
er
MAPE
V (
%)
MAPE
ϴ
(
%)
Bas
ic WLS
0
.00
2
9
3
4
4
1
0
.35
9
.07
SW
LS1 (k=1
)
0
.00
5
5
3
6
9
1
0
.35
9
.06
SW
LS1 (k=2
)
0
.00
2
8
0
1
4
1
0
.35
9
.07
SW
LS2 (k=1
)
0
.00
1
6
0
2
6
1
0
.34
9
.08
SW
LS2 (k=2
)
0
.00
2
2
0
8
4
1
0
.35
9
.07
FDW
LS
0
.00
4
4
0
8
6
8
.5
9
.6
9
.11
4.
DISCU
SSI
ON
Applie
d
on
tra
ns
missi
on
sy
st
ems
with
lo
w
R/
X
rati
o,
FDWLS
is
t
he
fas
te
st
WLS
Algo
rithm.
It
is
2
-
5
ti
mes
faster
than
basic
W
LS
e
ven
if
c
onve
r
gen
ce
re
quires
half
to
t
wo
it
erati
ons
more
tha
n
bas
ic
WLS
(simp
li
ficat
io
ns
ass
ume
d
di
minish
t
he
tr
ue
qua
dr
at
ic
conve
rg
e
nce
pro
per
ti
es
of
ba
sic
WLS
).
F
DWLS
requires
half
s
tora
ge
capaci
t
y
since
off
di
agonal
el
eme
nt
s
of
H
a
nd
G
matri
ces
a
r
e
neg
le
ct
e
d.
D
ia
gonal
Jaco
bian
el
em
ents
a
re
sim
ple
s
an
d
co
ntain
only
reacta
nce
branc
hes.
Howe
ver,
a
pp
li
e
d
on
distri
bu
ti
on
syst
em
with
hi
gh
R/
X
rati
o,
F
D
WL
S
was
t
he
slo
west
al
gorith
m
an
d
re
quire
d
a
hu
ge
it
era
ti
on
s
numb
e
r
because
assumpti
ons c
onside
red are
b
a
sed o
n
tra
ns
mi
ssion sy
ste
m
fe
at
ur
es
(acti
ve/r
eact
ive d
ec
oupl
ing
)
.
SWL
S1
al
gori
thm
do
e
s
no
t
co
nv
e
r
ge
f
or
first
it
erati
on
s.
It
see
ms
t
hat
co
ns
ta
nt
gain
matri
x
associat
ed
to
var
ia
ble
Jaco
bi
an
may
le
a
d
to
co
nv
e
r
gen
c
e
pro
blems.
S
WLS
1
is
not
reli
able
an
d
does
not
pr
ese
nt a
ny advanta
ges o
n
re
du
ci
ng comp
ut
at
ion
ti
me.
SWL
S2
al
go
rithm
ap
plied
a
t
first
it
erati
o
n
(
k=1),
requi
res
the
e
valua
ti
on
of
gai
n
a
nd
Jac
obia
n
matri
ces
just
once
w
hic
h
re
duces
c
ompu
ta
ti
on
ti
me
.
I
ndee
d,
S
WL
S2
(
k=
1)
is
2
ti
mes
f
ast
er
tha
n
basi
c
WLS
even
it
re
quire
s
a
higher
it
er
at
ion
s
num
ber.
S
WLS
2(k=
1)
has
t
he
sa
me
c
har
act
erist
ic
s
of
basic
WLS
wh
ic
h
make
it
le
ss
se
ns
it
ive
to
er
roneous
meas
ur
e
ments
a
nd
high
R/
X
rati
o
c
ompa
red
to
FDWLS
an
d
it
c
ould
be
app
li
ed
in dist
r
ibu
ti
on
netw
ork
sta
te
esti
mati
on.
5.
CONCL
US
I
O
N
This
pa
per
has
prese
nted
a
pe
rformance
c
omparis
on
betwee
n
t
wo
s
olu
ti
on
s
to
re
duce
c
omp
utati
on
al
bur
den
of
tra
di
ti
on
al
Weig
hted
Least
Squa
re
s
(
WLS
)
Algo
rithm:
S
olu
ti
on
1
base
d
on
f
ull
co
ns
ta
nt
m
at
rices
(simp
li
fie
d
m
et
hods
S
WL
S
1
/
S
WLS
2)
and
S
olu
ti
on
2
based
on
de
coupled
co
nst
ant
matri
ces
(F
ast
Decou
pled
W
LS
F
D
WLS
).
Algorith
ms
ha
ve
been
te
ste
d
on
th
ree
stu
dy
cases
IE
EE
14,
118
bus
sy
ste
ms
a
nd
a
12
-
bu
s
r
ur
al
distrib
utio
n
s
ys
te
m.
For
ea
ch
case,
simul
at
ion
s
wer
e
pe
rformed
on
ac
cur
at
e
a
nd
er
r
on
e
ous
measu
reme
nts.
Simulat
io
n
re
su
lt
s
ha
ve
show
n
t
hat
the
pe
rformance
of
F
D
WLS
al
gorithm
de
pe
nds
on
th
e
app
li
cat
io
n
e
nvir
onment:
a
pp
li
ed
on
tra
ns
mi
ssion
s
ys
te
ms
,
FDWLS
is
the
fastest
W
LS
Algorith
m.
H
oweve
r,
on
distri
bu
ti
on
sy
ste
m
with
hi
gh
R/
X
rati
o,
F
DWLS
was
th
e
slow
est
a
l
gor
it
hm
an
d
requi
red
a
hu
ge
it
er
at
ion
s
numb
e
r. The
rfor
e
, FD
WLS
is
not s
uitable
fo
r dist
rib
ution s
ys
te
m stat
e esti
mati
on
.
SWL
S2
(
k=1)
is
2
ti
mes
fa
ste
r
tha
n
basi
c
WL
S
with
t
he
sa
me
reli
abili
ty,
w
hic
h
make
it
le
ss
sensiti
ve
t
o
e
rron
e
ous
meas
urements
an
d
hi
g
h
R/
X
rati
o
c
ompa
red
to
F
D
WLS
.
T
he
refore
,
S
WL
S2
a
pp
l
ie
d
at
the
flat
sta
rt
presents
a
good
al
te
r
nativ
e
to
reduce
com
pu
ta
ti
on
t
ime
in
f
utu
re
distri
bu
ti
on
sy
ste
m
sta
te
esti
mati
on
.
Evaluation Warning : The document was created with Spire.PDF for Python.