TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 12, No. 8, August 201
4, pp. 5897 ~ 5904
DOI: 10.115
9
1
/telkomni
ka.
v
12i8.563
5
5897
Re
cei
v
ed
Jan
uary 16, 201
4
;
Revi
sed
Ap
ril 23, 2014; Accepted Ma
y
10, 2014
Comprehensive Evaluation to Distribution Network
Planning Schemes Using Principal Component
Analysis Method
Wang Ruilian
1
, Gao Shengjian
2
1
School of Elec
tric Po
w
e
r, Nor
t
h Chin
a Univ
e
r
sit
y
of W
a
ter Reso
urces an
d
Electric Po
w
e
r
,
Z
hengzho
u,
Hen
an, Chi
n
a
2
School of Civ
il
Engin
eeri
ng a
nd Comm
unic
a
tion,
North Ch
i
na Un
iversit
y
o
f
W
a
ter Resour
ces and El
ectri
c
Po
w
e
r, Z
hen
gz
hou,
Chi
n
a
A
b
st
r
a
ct
T
h
is pap
er pro
poses a n
e
w
compre
hens
ive
eval
uat
io
n method by n
onl
in
e
a
r princi
pa
l comp
on
en
t
ana
lysis in a
llu
sion to the pr
o
b
le
m i
n
distrib
u
tion n
e
tw
ork pla
nni
ng, such
as a large n
u
m
b
e
r of factors,
intens
e fu
zz
i
n
ess an
d the
n
onli
n
e
a
r rel
a
tio
n
shi
p
betw
e
e
n
the factors. F
i
rstly, accordi
n
g to pre-
existi
ng
achi
eve
m
e
n
t
and c
ons
ider
i
ng l
o
n
g
-ran
g
e
deve
l
o
p
m
ent
of distri
butio
n netw
o
rk, th
e co
mpr
e
h
ens
ive
eval
uatio
n i
n
d
e
x syste
m
tak
i
ng
into
accou
n
t the te
ch
nic
a
l, eco
n
o
m
ic,
envir
on
me
ntal
and
ad
apta
b
i
lit
y
factors and
so
on is c
onstruct
ed. Sec
ond
ly, by dis
posi
ng
al
l of the factors
in t
he in
dex
sy
stem usin
g
fu
zzy
consiste
nt mat
r
ix mo
del to g
e
t the relative
me
mbers
h
ip d
egre
e
matrix of ever
y scheme, the initial d
a
t
a
matrix
in f
u
zz
y
non
li
near
pri
n
cipal
co
mp
on
e
n
t an
alysis
is
obtai
ne
d. T
h
ird
l
y, the w
e
ig
ht of
every princ
i
pal
compo
nent
is acqu
ired by means of
the en
tropy conc
epti
on. T
he su
peri
o
rity of the
me
thod i
n
the p
a
p
e
r
introd
uced
is t
he
meth
od c
a
n
disp
ose th
e fu
zz
y
pr
ob
le
m i
n
complic
ated
pl
ann
ing, si
mplif
y the co
mp
utat
ion
process
in c
o
mpr
e
h
ensiv
e
eval
uatio
n, re
duce th
e su
bj
ectivity an
d a
r
bitrarin
ess, a
nd ca
n
mak
e
the
concl
u
sio
n
mor
e
scie
n
tific a
n
d
more re
aso
n
a
b
le. A c
o
mbi
n
a
t
ion of
a flow
dia
g
ra
m b
a
se
d
and
a c
oncret
e
exa
m
p
l
e, the al
gorith
m
is pr
ov
ed to be corr
ec
tly and practic
a
lly.
Ke
y
w
ords
:
dis
t
ributio
n netw
o
rk pla
nni
ng
,
co
mpr
e
h
ensiv
e e
v
alu
a
tion
,
pri
n
c
i
pal
co
mp
on
en
t analys
is
,
f
u
zzy
consiste
nt matr
ix
,
entropy w
e
ight
Copy
right
©
2014 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
Linki
ng po
wer net
work
and u
s
e
r
s,
distrib
u
ti
on n
e
twork reflect directly the use
r
’
s
requi
rem
ents in the
se
cu
rity, reliability, utilit
y and
other
aspe
cts. The
scie
n
t
ific and
rati
onal
distrib
u
tion n
e
tworks pla
n
n
ing pl
ay an
importa
nt role in the
p
o
we
r net
wo
rk con
s
tru
c
tio
n
,
improvem
ent
and o
peratio
n. So the co
mpre
hen
sive
evaluation in
distrib
u
tion n
e
twork pl
anni
ng
taking into
accou
n
t lots of factors
su
ch
a
s
technolo
g
y, eco
nomic, envi
r
onment an
d
th
e
adapta
b
ility to future
dev
elopme
n
t an
d othe
rs,
es
p
e
cially in
the
un
certai
nty environ
ment
and
adapta
b
ility factors, is very difficult.
At pre
s
ent, t
he
comp
re
he
nsive
evaluat
ion meth
od
s for
distri
buti
on n
e
two
r
k
planni
ng
inclu
de A
H
P, entro
py de
ci
sion, g
r
o
up d
e
ci
sion
-ma
k
in
g and
data
e
n
velopme
n
t a
nalysi
s
an
d
so
on, as
well a
s
the integ
r
at
ions of two o
r
more
meth
ods [1
-5]. Fo
r example, a
combi
nation
of
expert su
bje
c
tive experie
nce a
nd obj
ective entro
p
y
to get compre
hen
sive
weight u
s
e
d
for
scheme
opti
m
ization
[1-2
], prefere
n
ce
ord
e
r
and
t
r
aditional A
H
P to establi
s
h co
mprehe
n
s
ive
evaluation
m
odel [3], g
r
o
up de
ci
sion
-makin
g
the
o
ry and A
H
P to evaluate
comprehe
nsiv
ely
distrib
u
te
net
work [4], dat
a envel
opme
n
t analy
s
is
method
and
AHP to
ev
aluate
plan
ni
ng
scheme of di
stributio
n net
work [5], all of that
methods insepa
rabl
e
from the experts expe
rie
n
c
e
are
difficult t
o
de
al
with t
he mi
scellan
eou
s in
dex a
nd for the to
o hig
h
expe
rt
s exp
e
rie
n
ce
and
techni
cal req
u
irem
ents h
a
v
e large of co
mputing in di
stributio
n net
work pla
nnin
g
.
Princi
pal
co
mpone
nts
an
alysis
metho
d
, by weig
hting an
d synt
hesi
z
in
g several l
e
ss
uncorrelate
d
variable
s
th
at is line
a
r
combinatio
n o
f
more
o
r
igin
al
indexe
s
to
obtain the fi
nal
con
c
lu
sio
n
, can simplify the analysi
s
sy
stem
st
ructu
r
e,
redu
ce the
calculated a
m
ount and
keep
the informatio
n of original
v
a
riabl
es a
s
m
u
ch a
s
po
ssi
b
le[6].
But th
e method n
o
t con
s
ide
r
ing t
he
nonlin
ear
rela
tionshi
p between the ori
g
in
al indexe
s
is
not accord wi
th the pra
c
tical engin
eeri
n
g.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 589
7 –
5904
5898
On the
ba
si
s of traditio
n
a
l
prin
cip
a
l
co
mpone
nt an
a
l
ysis m
e
thod,
nonli
nea
r p
r
incip
a
l
comp
one
nt with
go
od d
e
velopme
n
t and
a
ppli
c
ati
on,
can refl
ect
the nonli
near
rel
a
tion
ship
betwe
en the
origin
al varia
b
les, extra
c
t t
he hig
h
-o
rde
r
s relation
s fro
m
origi
nal variable
s
, also can
redu
ce
the
concl
u
si
on d
e
pend
en
cies
on o
r
igin
al
v
a
riabl
e. However, the
pri
n
cip
a
l
comp
o
nent
analysi
s
is h
a
rd to qualita
t
ive indexes with
vague la
ngua
ge in the compl
e
x distributio
n net
work
evaluation
system, and th
e
weight
s
of ev
ery prin
cip
a
l
comp
one
nt u
s
ing th
e varia
n
ce
co
ntributi
o
n
rate ca
n only reflect the inf
o
rmatio
n ratio
but the amou
nt carried in t
he origi
nal va
riable
s
.
To qualitative indexes
with
vague language, on the
basi
s
of
fuzzy consistent matrix,
fuzzy
de
cisi
o
n
metho
d
by
rigo
ro
us mat
hematical
cal
c
ulatio
n can
get
the relati
ve
memb
ership
degree
can b
e
pre
s
e
n
ted [9-11] a
nd ent
ropy can q
u
a
n
tify information co
ntaine
d
in indexes
o
r
variable
s
[1
2
-
14]. By su
mmari
zing
th
e re
se
arch
result
s,
this pape
r con
s
tructs
di
stribut
ion
netwo
rk plan
ning in
dexe
s
system, int
r
o
duces th
e
rel
a
tive membe
r
ship
deg
ree
a
nd ent
ropy uti
lity
values to n
online
a
r p
r
in
cipal
com
p
o
nent analy
s
i
s
, and q
u
ickly and a
ccurately evalu
a
te
comp
re
hen
si
vely distributi
on network pl
annin
g
.
1.2. The Basi
c Ideas of Fu
zzy
Nonlinea
r Principal Componen
t
Analy
s
is
Nonli
nea
r p
r
i
n
cip
a
l comp
onent a
nalysis (NLP
CA) based o
n
prin
cipal
co
mpone
nt
analysi
s
met
hod is
a nonli
near
analy
s
is method, incl
uding fou
r
proce
s
se
s: firstl
y, the initial data
matrix from the evaluatio
n
s
of every in
dex to
sch
e
m
es i
s
gotte
n usin
g fu
zz
y c
o
ns
is
tent matr
ix
model,
se
co
ndly, nonlin
e
a
r p
r
in
cipal
comp
one
nts
by han
dling
the initial d
a
ta matrix u
s
i
ng
Aitchiso
n.J lo
garithmi
c
tra
n
sformation
to get
th
e
co
varian
ce
mat
r
ix and
obtai
n the
prin
cip
a
l
comp
one
nts
from the eig
envalue
s an
d eigenve
c
to
rs
of the co
varian
ce mat
r
ix, thirdly, the
weig
hts
of every p
r
in
cipal
comp
one
nt u
s
ing
con
c
ept
of entro
py, At last, by a
co
mbination
of t
he
weig
ht and princip
a
l com
p
o
nent, deci
s
io
ns can be ma
de.
Nonli
nea
r pri
n
cip
a
l co
mpo
nent analy
s
is, taking into
accou
n
t the nonlin
ear
rel
a
tionship
betwe
en eve
r
y index in
e
v
aluation
system, redu
cin
g
dimen
s
io
n
a
lity, can settle the probl
em
about
comp
rehen
sive eva
l
uation results in di
re
ct
proportio
n
to th
e co
rrelation
degree b
e
tween
the origin
al variabl
es, but t
o
part of the quantit
ative d
e
scriptio
n ind
e
xes in sy
ste
m
, the Acheson
logarith
m
ic transfo
rmatio
n is po
werl
ess l
eadin
g
to
the initial data matrix hard to
be esta
blishe
d.
The fu
zzy
co
nsi
s
tent matri
x
base
d
an
al
ysis
c
an
qua
ntify qualitative descriptio
n
indexe
s
u
s
i
ng
member
ship,
but it exis
ts
defec
t
s
in fuzzy c
o
ns
is
te
nt matr
ix-
b
as
ed
s
u
c
h
as
difficult to inc
a
r
nat
e
the differe
nce between t
w
o
schem
e
s
by the ele
m
ents i
n
the
establi
s
h
ed
fuzzy p
r
e
c
e
d
ence
relation
matri
x
, poor
simil
a
rity deg
ree
betwe
en o
b
tained fu
zzy
con
s
i
s
tent m
a
trix and fu
zzy
pre
c
ed
en
ce
relation
matrix. In view
of
advantag
es
and
disa
dvantage
s in
the p
r
in
cip
a
l
comp
one
nt a
nd fu
zzy
co
n
s
iste
nt matrix
analy
s
is,
no
nlinea
r p
r
in
ci
pal
comp
one
nt analy
s
is-b
ase
d
total pro
c
e
ss
is co
mpleted
after all of the indexe
s
is
disp
osed u
s
i
ng the fuzzy con
s
i
s
tent m
a
trix
model to g
e
t the ne
w initial
data matrix, a
nd the
info
rm
ation entropy
is intro
d
u
c
ed
in the analy
s
i
s
to get the wei
ght of every princi
pal comp
onent.
2. Calculatio
n Process
2.1. Calculati
on Proces
s Char
t
The nonli
nea
r princi
pal co
mpone
nt anal
ysis calculati
on pro
c
e
s
s is sho
w
n in Fig
u
re 1:
Figure 1. Fuzzy Nonli
nea
r Princi
pal Co
mpone
nt
Ana
l
ysis Based Calcul
ation Flo
w
Ch
art
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
p
re
hen
si
ve Eval
uation
to Distributio
n Ne
twork Plannin
g
Sche
m
e
s Usin
g…
(Wang Ruilia
n)
5899
2.2. Initial Data Matrix
The initial da
ta matrix is constituted by
t
he evaluation of every index to sche
mes. To
compli
cate
d evaluation system,
the
init
ial
data
matrix from the relative mem
bership
deg
ree
matrix by fuzzy co
nsi
s
tent
matrix model
to di
spo
s
e
every evaluatio
n is p
r
e
s
ente
d
. Assu
ming t
hat
n
sch
eme
s
expresse
d
in
n
i
x
i
,
,
2
,
1
and
m
evaluation
indexes
expre
s
sed i
n
m
k
u
k
,
,
2
,
1
c
o
ns
titute a c
e
rtain s
y
s
t
em, the init
ial
data matrix is gotten as foll
ows:
(1) Fu
zzy
prefer
ence r
e
lation matrix
The orde
r of a ce
rtain ind
e
x
k
u
to sch
eme
s
is exp
r
e
s
se
d with the na
tural num
be
r, the
greate
r
num
b
e
rs the
worse sch
eme
s
, the equal
n
u
m
bers the same impo
rta
n
ce of sche
mes.
Thus
the fuzzy pr
ec
edenc
e
r
e
lations
h
ip
matr
ix is
s
how
n:
C
n
n
k
ij
c
n
j
i
m
k
,
,
2
,
1
,
;
,
,
2
,
1
In the form
ula:
k
ij
c
are the
pre
c
e
den
ce
deg
ree
coe
fficients
abo
ut schem
e
i
x
and
scheme
j
x
to the index
k
u
. Assumin
g
that the order
of sche
m
e
i
x
to some an
index
m
k
u
k
,
,
2
,
1
is
n
i
a
i
,
,
2
,
1
, s
o
m
fuzzy p
r
eferen
ce
rel
a
tion m
a
trixe
s
of
m
indexes are
sho
w
n:
C
k
n
n
k
ij
c
k
l
n
l
k
l
n
l
k
i
k
j
a
a
a
a
1
1
min
max
5
.
0
5
.
0
)
(
(
1)
(2) Fu
zzy
consiste
nt matrix
Fuzz
y c
o
ns
istent matr
ix
n
n
k
ij
k
r
R
m
k
,
,
2
,
1
converted fro
m
fuzzy pref
eren
ce
relation m
a
tri
x
in the condi
tion of
1
)
min
max
(
1
1
n
r
r
k
p
n
p
k
p
n
p
is co
mputed by the formula:
5
.
0
2
1
1
k
p
n
p
k
p
n
p
k
j
k
i
k
ij
r
r
r
r
r
(
2)
In the formula:
n
j
c
r
n
l
k
il
k
i
,
,
2
,
1
1
and
k
ij
r
reflect
relative sup
e
rio
r
ity degree
about sch
e
m
e
n
i
x
i
,
,
2
,
1
and
n
j
x
j
,
,
2
,
1
to the evaluation in
dex
k
u
.
(3) O
p
timiza
tion
v
a
lue of single index
The optimi
z
a
t
ion value
k
i
s
about every scheme
i
x
to index
k
u
is com
puted by the
formula:
n
j
k
ij
k
i
r
n
n
s
1
1
2
1
1
(3)
(4) Op
timiza
tion member
ship matrix
of ev
er
y
index
Optimizatio
n
membershi
p
matrix of ever
y inde
x in evaluation system
from the
optimizatio
n value of sin
g
le
index is sh
o
w
n:
m
n
ij
s
S
]
[
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 589
7 –
5904
5900
2.3. Nonline
a
r principal componen
t
s
(1) Initial data matrix
Optimizatio
n
membe
r
ship
matrix of every index
is the initial data matrix from the formula
(4).
Thu
s
t
he n
onlin
ear pri
n
ci
pal
compon
ent
s
analysi
s
by Ach
e
son l
ogarith
m
ic cente
r
transfo
rmatio
n method is p
r
esented.
(2) Ache
son logarithmic
transformatio
n
All of the d
a
ta in the initial data matr
ix is co
nverted by Ache
son lo
g
a
rithmi
c
transfo
rmatio
n sho
w
n:
m
l
il
ij
ij
s
p
s
y
1
ln
1
ln
(5)
Thus th
e new data matrix is
Y
=
nm
ij
y
.
(3) Cov
a
riance
m
a
trix
Covari
an
ce
matrix from the new d
a
ta matrix
by logari
t
hmic center
i
s
gotten, expressed
in:
m
m
ij
ij
z
Z
(6)
In the formula:
j
pj
n
p
i
pi
ij
y
y
y
y
n
z
1
1
1
where
n
p
pi
i
y
n
y
1
1
and
n
p
pj
j
y
n
y
1
1
.
(4) Nonlinea
r principal componen
t
s
The Nonli
n
e
a
r
p
r
in
cipal
comp
one
nts from
the
fo
rmula
m
m
ij
ij
z
Z
are
o
b
tained
sho
w
n:
SA
f
F
q
n
ij
)
(
T
=
mq
nm
a
s
(7)
In the formul
a: Matrix
A
=
(
a
mq
)
is
corresp
ondi
ng ei
genve
c
tors o
f
every positive eigenvalu
e
of
matrix
Z
m
m
ij
z
. Usu
a
lly, the po
sitive eigenval
u
e
s i
s
le
ss tha
n
all of th
e ei
genvalu
e
s,
so th
e
corre
s
p
ondin
g
eigenve
c
tors is le
ss too. That is
to say
,
the corresp
ondin
g
prin
ci
pal com
pon
e
n
ts
are le
ss tha
n
the origin
al variabl
es. Th
e
dimen
s
ion
a
lity reductio
n
is succe
s
sful.
2.4. Entrop
y
Weigh
t
Usually, the varian
ce
cont
ri
bution rate of prin
cipal
co
mpone
nts i
s
the
wei
ght dist
ribution.
But the va
ria
n
ce
contri
but
ion
rate
refle
c
ts t
he
relati
ve value
of i
n
formatio
n
carri
ed i
n
eve
r
y
origin
al va
ria
b
les but th
e
actual
qu
antity. Inform
atio
n ent
ropy
ca
n qu
antify informatio
n, so
it is
use
d
to determine the wei
ght of t princi
pal com
pon
e
n
ts.
(1) Nonneg
a
tiv
e
processing of princip
a
l compone
nts
In orde
r to
acq
u
ire
entropy, the po
ssi
bl
e ne
gati
v
e quantity existing in
prin
cipal
comp
one
nt
matrix ne
ed
s non
neg
ative pro
c
e
s
sing.
After non
neg
ative pro
c
e
ssing, the
matri
x
is
F
′
:
F
′
=F +
g
(8)
In the form
ul
a: g is the
smallest
natural num
ber le
ading
all of t
he p
r
in
cipal
comp
one
nts
to
positive.
(2) Entr
opy
Entropy ca
n displ
a
y the order de
gree of
info
rmation i
n
variable
s
o
r
indexes. The
amount
of information
is expre
s
sed
in
e
t
(
t
=1
,
2
,
…
,
q
):
q
t
it
it
t
h
h
n
e
1
ln
ln
1
(
9)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
p
re
hen
si
ve Eval
uation
to Distributio
n Ne
twork Plannin
g
Sche
m
e
s Usin
g…
(Wang Ruilia
n)
5901
In the fo
rm
ula:
q
j
ij
ij
ij
f
f
h
1
'
'
where
ij
f
'
is the d
a
ta a
fter non
neg
a
t
ive pro
c
e
ssi
ng of
prin
cipal com
pone
nts.
(3)
Principal componen
t
w
e
i
ght
The di
spe
r
se
degree of the
q
Princi
pal
comp
onent
s is expre
s
se
d in informati
on avail
value
d
t
(
t
=1
,
2
,
…
,
q
)
,t
he formul
a is:
t
t
e
d
1
(
10
)
The d
a
ta in t
he
q
p
r
in
cip
a
l com
pon
ent
more
scattered, the
corre
s
po
ndin
g
dat
a in the
d
t
(
t
=1
,
2
,
…
,
q
)
lag
e
r, indi
cate that the
importan
c
e
degree of th
e index is h
i
gher, o
n
the
contrary, the
lowe
r im
port
ance de
gree
of the in
d
e
x. Thus, th
e
weig
ht dist
rib
u
tion vecto
r
s of
prin
cipal
com
pone
nts by n
o
rmali
z
in
g
d
t
(
t
=1
,
2
,
…
,
q
) are a
c
q
u
ire
d
:
q
,
,
,
2
1
T
,
q
t
t
t
t
d
d
1
(
11
)
2.5. Fuzzy
Nonlinear Principal Compo
n
ent-ba
sed
Compre
hens
iv
e Decision Model
Every princi
p
a
l comp
one
n
t
weighted a
nd sy
nthe
si
zed, the com
p
reh
e
n
s
ive d
e
ci
sion
model is
sho
w
n:
i
O
q
iq
i
i
F
F
F
2
2
1
1
(
i
=1,
2
,
…
,
n
)
(
12
)
3. Planning Scheme
Decisions by
Fuzz
y
Nonlinear Princi
pal Compo
n
e
nt
Analy
s
is to
Distribu
tion Net
w
o
r
k
3.1. Scheme Compre
hens
iv
e Ev
aluation
Index Sy
st
em of Ne
t
w
o
r
k Planning
Network pla
n
n
ing involve
s
many facto
r
s.
The sche
me
compreh
ensive evalu
a
tion
ind
e
x
system
of net
work
plan
nin
g
is summa
ri
zed
from
t
he
[1-4] refere
nces, al
so
kno
w
n a
s
2
4
attrib
ute
comp
re
hen
si
ve evaluation
index
set. T
he ind
e
x system ca
n ove
r
all reflect n
e
t
work pl
annin
g
scheme. Fo
r some repeti
t
ive el
ements in the evaluation index
set, to spe
c
ific issu
es,
the
approp
riate comprehe
nsiv
e eval
uation i
ndexe
s
are
chosen.
3.2. Scheme Decision
-ma
k
ing Proces
s
After sele
ctin
g, evaluation
index syste
m
of distribu
ti
on netwo
rk plannin
g
scheme is
formed. All o
f
the indexe
s
data in the
evaluat
ion sy
stem
a
r
e con
v
erted
to
n
u
m
ber betwee
n
0
and 1 u
s
ing t
he fuzzy con
s
iste
nt matrix deci
s
ion
-
ma
king mo
del. T
hus the d
a
ta betwe
en 0 an
d 1
is the memb
ership de
gre
e
of every index to sch
e
m
es that est
ablishe
s initial data matri
x
of
nonlin
ear p
r
i
n
cip
a
l com
p
o
nent analy
s
is and the non
linear p
r
in
cip
a
l comp
one
n
t
s are obtai
n
ed.
After the dat
a in p
r
in
cipal
comp
one
nt m
a
trix is
n
on-n
egative p
r
o
c
e
s
sed, ent
ropy
utility value of
the prin
cipal
compo
nent
is gotten, and by t
he weig
ht distri
bution after
norm
a
lization
and
comp
re
hen
si
ve deci
s
ion
can be mad
e
.
In this
pap
er,
the d
a
ta in
the initial
data
matrix i
s
wei
ghted
nonlin
e
a
r inte
gration
usi
ng
Ache
son
log
a
rithmi
c cent
er tra
n
sfo
r
ma
tion and
then
the pri
n
ci
pal
com
pon
ent
gotten, that i
s
to
say, the prin
cipal com
pon
e
n
t is nonlin
ea
rity.
4. Examples
A distributio
n
network pla
n
n
ing exampl
e
from
the Ref
e
renc
es
[1], verify the c
o
rrec
t
nes
s
of the a
r
ith
m
etic d
e
scri
bed in
this
pape
r.
Lo
ng-term pl
annin
g
sch
e
me
s
of high
-volta
ge
distrib
u
tion n
e
twork in the
plannin
g
are
a
pro
p
o
s
ed i
n
clu
de: sche
me 1 (3
5kV
prog
ram
)
, scheme
2 (11
0
kV
progra
m
) a
nd
scheme
3 (3
5kV an
d 1
1
0
k
V hybri
d
scheme
)
. The
evaluation i
n
dex
system
is
si
mplified, sele
cted and
sho
w
n
i
n
Fi
gure 2.
According
to
the planni
ng schem
es and
comp
re
hen
si
ve evaluation
indexe
s
syst
em, the ev
al
uation
of eve
r
y index to
scheme
s
i
s
sho
w
n
in Table 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 589
7 –
5904
5902
Figure 2. Indice
s set of Simplified Co
m
p
reh
e
n
s
ive Evaluation System to Distrib
u
tion Net
w
ork
Table 1. Evaluation of Every Index to Scheme
s
Attribute
elements
Technical indexe
s
Economic
indexes
Environmental
indexes
Adaptability
indexes
Scalability
indexes
Reliability
Power
supply
quality
Total
investment
(hundr
ed
million
yuan
)
Net
w
ork
l
o
s
s
rate(%)
Number of
high
voltage station
Degree o
f
matching
schemes and
load gro
w
th
Ex
tensible
allow
ance
Scheme
1
Better
Better
13.02
0.73
More
Middling
The
most
Scheme
2
Better
Better
12.73
0.72
Gene
ric
Better
More
Scheme
3
Worse
The best
13.21
0.68
The least
Worse
Middling
Data t
y
pe
Qualitative
Qualitative
quantitative
quantitative Qualitative
Q
ualitative Qualitative
By the formula (1) to (4)
,
t
he initial data
matrix by the obt
ained d
a
ta from the evaluation of
every index to scheme
s
u
s
ing fu
zzy co
nsi
s
tent matri
x
model is sh
own a
s
follo
wing:
08333
.
0
08333
.
0
58333
.
0
58333
.
0
08333
.
0
66667
.
0
08333
.
0
33333
.
0
58333
.
0
33333
.
0
33333
.
0
58333
.
0
16667
.
0
5
.
0
58333
.
0
33333
.
0
08333
.
0
08333
.
0
33333
.
0
16667
.
0
5
.
0
S
By the formula (5), the ne
w matrix is:
0.85305
-
0.85305
-
1.09289
1.09289
0.85305
-
1.22643
0.85305
-
0.1188
-
0.44082
0.1188
-
0.1188
-
0.44082
0.81192
-
0.286675
0.916859
0.35724
1.02909
-
1.02909
-
0.357239
0.33588
-
0.762714
Y
By the formula (6), the covarian
ce m
a
tri
x
of that new matrix is
m
m
ij
ij
z
Z
, shown:
0.7912
0.5014
0.9309
-
0.9309
-
0.5014
0.6289
-
0.6965
0.5014
0.5248
0.6761
-
0.6761
-
0.5248
0.7621
-
0.5633
0.9309
-
0.6761
-
1.1338
1.1338
0.6761
-
0.8912
0.8756
-
0.9309
-
0.6761
-
1.1338
1.1338
0.6761
-
0.8912
0.8756
-
0.5014
0.5248
0.6761
-
0.6761
-
0.5248
0.7621
-
0.5633
0.6289
-
0.7621
-
0.8912
0.8912
0.7621
-
1.1381
0.7676
-
0.6965
0.5633
0.8756
-
0.8756
-
0.5633
0.7676
-
0.6958
Z
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Com
p
re
hen
si
ve Eval
uation
to Distributio
n Ne
twork Plannin
g
Sche
m
e
s Usin
g…
(Wang Ruilia
n)
5903
The po
sitive eigenvalu
e
s
of matrix
m
m
ij
ij
z
Z
are 0.008
、
00.52
33 and 5.4
1
1
1
, that is
to s
a
y
q
=3, th
en by the formula (7
), the prin
cipal
com
pone
nt matrix is:
0.6874
-
0.2764
0.4406
0.0927
0.2481
-
0.1607
0.0238
-
0.0387
-
0.0053
-
F
By the formula (8), the n
a
tural n
u
mbe
r
g=
1, after non-neg
ative data in the prin
cip
a
l
comp
one
nt matrix, and b
y
the formula
(
9), the entropy
e
t
of the three pri
n
ci
p
a
l comp
one
nt is
0.866554
0.984817
0.999911
t
e
. By the formula
(10)
, the entropy utility is
0.1334
0.0152
0.0001
t
d
, and by
the formula
(11), the
weight vector is
0.897108
0.102219
0.000672
t
.
By the formula (12
)
, the co
mpre
hen
sive
evaluation
co
nclu
sio
n
is:
9747
.
0
0579
.
1
4119
.
0
i
O
The sche
me-orde
r
is
sche
me 2, schem
e 3, sch
eme
1, that is to say, the sche
m
e 2 is
the optimal schem
e. The concl
u
si
on co
nsi
s
tent with
the refe
ren
c
e
[1], proves th
e corre
c
tne
ss of
the method d
e
scrib
ed in th
is pap
er.
5. Conclusio
n
1) Acco
rdin
g
to the rese
arch re
sult
s, i
ndex syste
m
of comp
re
hen
sive eval
uation to
distrib
u
tion n
e
twork pl
ann
ing sche
me
is e
s
tablis
he
d. By prioriti
zation of ev
ery sche
me
to
indexe
s
, fuzzy preferred
relation
matrix
is esta
blish
e
d
and then th
e fuzzy
con
s
i
s
tent matrix. By
the relative membe
r
ship
degree matri
x
of every
in
dex to sche
m
es, the initial data matri
x
in
nonlin
ear p
r
in
cipal
comp
on
ent analysi
s
i
s
gotten.
2) After Ache
son lo
garith
m
ic ce
nter
con
v
ersio
n
to the initial data matrix, the covariance
matrix is obta
i
ned. Accordi
ng to the corresp
ondi
ng ei
genve
c
tors of positive eige
nvalue
s by the
covari
an
ce m
a
trix and
initi
a
l data
matri
x
, princi
pal
compon
ent m
a
trix is
gotte
n. The
prin
ci
pal
comp
one
nt is nonli
nea
r
co
mbination
s
of the initia
l
dat
a matrix
-vect
o
r, it i
s
mo
re
rea
s
on
able
th
an
the comm
onl
y linear pri
n
ci
pal com
pon
e
n
t analysi
s
.
3) The weight
distribution vector of pri
n
ci
pal com
ponent using norm
alized entropy utility
not cumul
a
tive vari
an
ce
contributio
n
ra
te, not o
n
ly
unifies the
selecte
d
stan
dard
of
pri
n
cipal
comp
one
nts
numbe
r, but also
sho
w
s the amou
nt
of variable informatio
n ca
rried in prin
cip
a
l
comp
one
nt from initial data matrix.
4) In
com
p
re
hen
sive eval
uation, the i
n
dexes having
little influen
ce on
co
mpre
hen
sive
evaluation
n
eed b
e
redu
ced, the
stro
ng fuzzin
ess of the com
p
reh
e
n
s
ive e
v
aluation
system
need
s to
be
co
nsi
d
e
r
ed,
and
the
co
mpre
hen
sive
evaluatio
n
con
c
lu
sio
n
s
are
mad
e
m
o
re
convin
cin
g
a
s
mu
ch
as po
ssible, fo
r th
e
many in
dexe
s
d
e
scribed
o
n
ly usi
ng va
g
ue la
ngu
age
i
n
distrib
u
tion
n
e
twork plan
n
i
ng eval
uatio
n sy
stem
. T
he ju
dgme
n
t evaluatio
n
con
c
lu
sio
n
u
s
ing
nonlin
ear prin
cipal com
pon
ent
analy
s
is brou
ght
in
fu
zzy d
e
ci
sio
n
-makin
g an
d
entropy the
o
ry is
easy to obtai
n. The cal
c
ul
ation pro
c
e
ss is simp
le, o
r
derlin
ess, an
d con
c
lu
sio
n
is co
rrect.
Referen
ces
[1]
Nian
L
i
u, L
i
Ma, T
i
eming
Z
hu. S
y
nthetic
al
a
ssessme
n
t on
distri
butio
n n
e
t
w
ork P
l
a
nni
ng sc
he
m
e
consi
deri
n
g
An
ti-Disaster ab
ili
t
y
and
e
g
io
na
l character
i
stics.
Po
we
r System
Te
ch
no
l
o
g
y
.
20
12;
36(5)
:
219-
225.
[2]
F
eng S
h
i, P
e
il
i
Li
u. Dec
i
sio
n
makin
g
meth
o
d
for
urb
an po
w
e
r distrib
u
tion
gri
d
p
l
a
nni
ng
usin
g
e
n
trop
y
w
e
ight.
Engi
ne
erin
g Journ
a
l o
f
W
uhan Univ
e
r
sity
. 2012; 45(
2): 211-2
15, 22
4.
[3]
Gang W
e
i, Ji
a Li
u, Xi
n Z
h
ang, et a
l
.
C
o
mpre
hens
ive
decisi
on-
maki
ng of AHP/P
R
OMET
HEE in
distrib
u
tion n
e
t
w
ork plann
in
g.
Procee
din
g
s of
the CSU-EPS
A. 2009; 21(
3): 36-40.
[4]
Yu
Yi. Compr
e
hens
ive eva
l
ua
tion and
d
e
cisi
on-Maki
ng
for
distrib
u
tion
n
e
tw
o
r
k
pla
nni
ng.
Dissertati
o
n
for Master'
s
Degree. 20
12.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 8, August 2014: 589
7 –
5904
5904
[5]
Luji
e
L
i
u, Ro
ng
Hu, Yang F
u
,
et al. Compr
e
h
ensiv
e eva
l
uati
on of reso
urce
econ
om
y bas
e
d
distrib
u
tio
n
net
w
o
rk p
l
an
ni
ng Schem
e.
Power System
Technology.
200
8; 32(16): 6
6
-7
0.
[6]
Hon
g
zha
n
N
i
e,
Son
g
Ni
e, Yi
Qian, et
al. Co
mpreh
ensiv
e d
e
cisio
n
-mak
i
ng
of alter
nativ
e t
r
ansmissi
o
n
net
w
o
rk pla
nni
ng
b
a
sed on princi
pa
l
comp
one
nt
an
al
ysis
.
Pow
e
r System T
e
ch
no
logy
.
201
0; 34(6
)
:
134-
138.
[7]
Bei Li
u. Appl
i
c
ation of n
o
n
-
Lin
ear pri
n
ci
p
a
l comp
on
ent
anal
ys
is method to rock
mass qua
lit
y
classification.
W
a
ter Resourc
e
s and Pow
e
r.
201
1; 29(1
2
): 78-80.
[8]
Ping
Li.
Class
ificatio
n of
e
x
p
a
n
sive
soi
l
b
a
se
d o
n
n
on-
lin
ear
pri
n
cip
a
l c
o
m
pon
ent
ana
l
y
si
s an
d cl
uste
r
analy
sis.
Ya
ng
tz
e Ri
ve
r
. 20
12;
43(7): 40-4
3
.
[9]
Z
heng
yo
n
g
Z
h
ou, Xia
ohu
i Yu
an, Yon
g
Z
h
o
u
.
T
he pr
edicti
o
n rese
arch a
b
out the stren
g
th of ceme
nt
base
d
o
n
th
e
non-
lin
ear
pri
n
cipal
com
pon
e
n
t ne
ural
n
e
t
w
ork. Mathem
ati
cs in Pr
actice
and
T
heor
y
.
201
3; 43(3): 83
-91.
[10]
Ron
g
Z
h
a
n
, D
e
sha
n
T
ang, Z
i
zeng
Li
u. Opti
mizati
o
n
of r
u
b
ber d
a
m’s st
yl
e
s
bas
ed
on fuz
z
y
c
onsist
ent
matrix
.
Water Resources and Power
. 2009; 27(6): 10
7-1
0
9
.
[11]
Xi
an
g Qiu. Improve
d
fuzzy
cons
istent matrix a
nd it
s applic
at
io
n
in Engin
eer
i
ng Eval
uatio
n
.
Mathe
m
atics i
n
Practice and T
heory.
20
11; 4
1
(17): 44-
47
.
[12]
Jiken
g
L
i
n, T
ongfei
Li, Z
i
min
g
Z
hao, et
a. As
s
e
ssment
on
po
w
e
r s
y
stem
bla
ck-start schem
es b
a
sed
o
n
entrop
y
-
w
e
i
ght
ed fuzz
y
com
p
rehe
nsive ev
al
uatio
n mode
l.
Pow
e
r System T
e
chnol
ogy.
201
2; 36(2):
115-
120.
[13]
Herui C
u
i, Lih
ua Li
ang, Li
ho
ng W
ang.
Reliability evaluati
on index of distribut
ion system
based on
entropy-w
ei
ght
T
O
PSIS meth
od.
T
r
ansactions of the CSAE.
2011; 27(s
u
ppl
ement1): 1
7
2
-17
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.