Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
7, N
o
. 3
,
Ju
n
e
201
7, p
p
. 1
133
~114
4
I
S
SN
: 208
8-8
7
0
8
,
D
O
I
:
10.115
91
/ij
ece.v7
i
3.p
p11
33-
114
4
1
133
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Wind Farm Management using Artificial Intelligent
Techniques
Boualem
Benl
ahbib
1
,
F
a
rid
Bouch
a
faa
2
, Sa
ad
Mekhilef
3
, Noure
ddine
Bouarroudj
4
1,2
L
a
b
o
r
at
o
i
r
e
d'
In
st
r
u
me
n
t
a
t
i
o
n,
Fa
c
u
l
t
é
d'
El
ec
t
r
on
ique
et d'Infor
m
atique USTHB
,
Alger
(Algérie)
1,4
Unité de Rech
erche Appliqu
é
e en
Energ
i
es R
e
n
ouvelab
l
es, URAER, Cen
t
re de
Dé
veloppement des
Energ
i
es
Renouvelables, CDER,
47133, Ghardaïa,
Alg
e
ria
1,3
P
o
wer Ele
c
tro
n
ics
and
Ren
e
wa
ble
Energ
y
Res
e
arch
Labor
at
or
y
(PEARL) Department of
Electrical of
Eng
i
neer
in
g
University
of
Malay
a
50603
Kuala
Lumpur Malay
s
ia
Article Info
A
B
STRAC
T
Article histo
r
y:
Received
July 10, 2016
Rev
i
sed
May
5, 201
7
Accepted
May 26, 2017
This
paper pres
e
n
ts
a com
p
arativ
e s
t
ud
y
betwe
e
n
the geneti
c algo
rithm
and
particle swarm optimization methods
to determin
e the optimal pr
oportional–
integr
al (PI) controller parameter
s
for
wind farm
supervision algo
rithm. The
main objective o
f
this study
is to
obtain a rapid and stable s
y
stem
b
y
tuning
of the PI controller
,
ther
eb
y
pro
v
idi
ng an excellent monitor for our wind
farm b
y
sending
separate set points to
all wind
generators
. A superviso
r
y
s
y
s
t
em
controls
the ac
tiv
e and
reac
tive power
of the ent
i
re wi
nd farm
b
y
sending out s
e
t points to
all
wind turbin
es.
A machine
con
t
rol s
y
s
t
em
ens
u
res
that
the
s
e
t points
a
t
th
e
wind turbine
le
vel ar
e re
ach
ed.
The
entir
e
control is add
e
d
to the normal o
p
erating power r
e
feren
ce of th
e
wind farm
establish
e
d b
y
a supervisor
y
contro
l
.
F
i
na
ll
y the p
e
rform
an
ce of
the
proposed algor
ithm is verified
through MATLAB/Simulink
simulation
results b
y
consid
ering
a wind
far
m
of
three doubly
-
f
e
d indu
ction
generators.
Keyword:
DFI
G
GA
MPPT a
n
d PC
C
PI con
t
ro
ller
PSO
W
i
nd
f
a
r
m
su
per
v
ision
Copyright ©
201
7 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
B
oual
e
m
B
e
nl
ahbi
b,
Lab
o
rat
o
i
r
e d'
Inst
r
u
m
e
nt
at
i
on,
Faculté d'
Electronique et
d'
Inform
atiq
u
e
USTHB, Alg
e
r (Alg
érie).
Em
ai
l
:
boual
l
a
m
30@gm
ai
l
.
com
1.
INTRODUCTION
The
h
o
m
e
of t
h
e e
nvi
ro
nm
ent
,
t
h
e st
rat
e
gy
o
f
whi
c
h i
s
e
nvi
ro
nm
ent
a
l
pr
ot
ect
i
o
n
,
has
spa
r
ed
n
o
eff
o
rt
si
nce
i
t
ope
ne
d
i
t
s
d
o
o
r
s fo
r
t
h
e pr
ot
ect
i
on fo
rm
al
nui
sa
nces.
R
e
c
e
nt
l
y
,
t
h
i
s
gre
e
nh
o
u
se ha
ndl
ed
a
n
i
m
p
o
r
tan
t
top
i
c related
to
ren
e
wab
l
e en
erg
y
and
env
i
ron
m
en
tal p
r
o
t
ectio
n
.
In
th
e t
h
ird
millen
n
i
um
,
th
e
im
port
a
nce
of
rene
wa
bl
e ene
r
gy
i
s
of c
ont
i
nui
ng c
o
n
cer
n
t
o
researc
h
ers
and e
nvi
r
o
nm
ent
a
l
i
s
t
s
worl
d
w
i
d
e.
Exp
e
r
t
s
h
a
v
e
r
e
po
r
t
ed
t
h
at
man
y
cli
m
ate
ch
ang
e
n
u
i
san
ces, su
ch as f
l
ood
s, cyclon
es, gr
een
house g
a
s
em
i
ssi
ons, acc
el
erat
ed s
o
i
l
e
r
osi
o
n
,
a
n
d
l
o
sses i
n
ge
net
i
c
di
ve
rsi
t
y
,
ha
ve a
ppea
r
e
d
i
n
n
u
m
e
rous
co
unt
ri
es.
Ex
pert
s
have
al
so ex
pl
ai
ne
d
t
h
at
a
ll these nuisa
nces
pos
e
an unprece
dent
ed ec
ologic
al thr
eat on a
global
scale. Thu
s
, the q
u
e
stion
th
at
cu
rren
tly arises is h
o
w to
ad
dress th
is situ
atio
n
and
ho
w t
o
co
n
t
ro
l en
erg
y
. Th
e
onl
y
sol
u
t
i
on t
h
at
co
ul
d sav
e
t
h
e Eart
h i
s
o
r
i
e
nt
i
n
g t
o
wa
r
d
re
newa
bl
e e
n
er
gy
fr
om
t
h
e sun
,
t
h
e wi
n
d
, an
d
tid
es [1
].
The m
a
jor difficulty associated with
dece
nt
ralized ene
r
gy
sources (e
.g
.,
win
d
farm
and
solar plant
)
is th
at th
ese so
urces
do
n
o
t
p
a
rticip
ate in th
e
g
e
n
e
ra
l services system
(i.e.,
vo
l
t
a
ge
ad
ju
st
m
e
nt
, freq
u
enc
y
,
p
o
s
sib
ility to
o
p
e
rate on
islan
d
i
n
g
). Th
is
case is p
a
r
ticu
l
arly tru
e
with
ren
e
wa
b
l
e
en
erg
y
sources, th
e
production
of which is
unpre
dictable
a
n
d c
o
nsidera
b
l
y
fluctuating.
The
integrat
ion of decent
r
alized
pr
o
duct
i
o
n
uni
t
s
i
n
a
net
w
or
k
p
o
ses se
veral
pr
o
b
l
e
m
s
, i
n
cludi
ng
ra
nd
om
and
u
n
p
re
di
ct
abl
e
p
r
o
d
u
ci
bl
e
(e.
g
.
,
wind
p
o
wer, so
lar), lack
o
f
frequ
e
n
c
y–p
ower an
d vo
lta
g
e
adj
u
stm
e
n
t
s, an
d sen
s
itiv
ity to
vo
ltag
e
d
i
ps. Th
e
fai
l
u
re t
o
part
i
c
i
p
at
e i
n
servi
ces sy
st
em
bri
ngs t
h
i
s
t
y
pe of e
n
er
gy
so
ur
ce t
o
beha
ve
sim
i
l
a
rly
t
o
passi
ve
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E
V
o
l
.
7,
No
. 3,
J
u
ne 2
0
1
7
:
11
3
3
– 11
44
1
134
gene
rat
o
rs o
f
el
ect
ri
ci
t
y
. Penet
r
at
i
on o
f
di
st
ri
b
u
t
e
d
ge
nera
t
i
on m
u
st
be l
i
m
i
t
e
d (20% o
r
30% o
f
t
h
e p
o
we
r
consum
ed after a fe
w fee
dbac
k
s) to e
n
sure
ne
twork stability in
acceptabl
e c
o
nditions; thus, powe
r
su
perv
ision
o
f
th
ese farm
s is n
ecessary [1
].
Current research in t
h
e fiel
d of
wind
farm
s
is o
r
ien
t
ed to
ward
t
h
e dev
e
lop
m
en
t o
f
sup
e
rv
ision
al
go
ri
t
h
m
s
t
o
di
st
ri
but
e t
h
e t
o
t
a
l
powe
r
refe
re
nce bet
w
ee
n w
i
nd ge
ner
a
t
o
r
s
.
In t
h
i
s
cont
e
x
t
,
several
al
g
o
ri
t
h
m
s
have
bee
n
pr
op
ose
d
.
Pr
op
o
r
t
i
onal
di
st
ri
b
u
t
i
on al
go
ri
t
h
m
[2]
,
[
3
]
,
w
a
s de
vel
o
ped
t
o
di
st
ri
b
u
t
e
po
we
r
refe
rences
in a
proportional
manner. From
a safety conc
e
r
n, t
h
is algorithm
ensure
s that
each
wind
ge
nerat
o
r
con
s
t
a
nt
l
y
fu
n
c
t
i
ons
far
fr
o
m
i
t
s
l
i
m
i
t
s
, as de
fi
ne
d by
t
h
e (
P
,
Q)
di
a
g
ram
[2]
.
T
h
e
al
go
ri
t
h
m
s
based
on
o
p
tim
ized
o
b
j
e
ctiv
e fu
n
c
tion
[4
-6
] p
e
rm
its an
op
ti
m
a
l d
i
strib
u
tio
n of th
e activ
e and
reactiv
e p
o
we
r re
fe
re
nces
on t
h
e
wi
n
d
gene
rat
o
rs. It
need
s o
p
t
i
m
i
zat
i
on m
e
t
hods
, suc
h
as gen
e
t
i
c
al
gori
t
h
m
(GA
)
[
7
]
,
ne
ur
o
n
s
net
w
or
ks
[8]
,
part
i
c
l
e
s s
w
ar
m
opt
im
i
zat
i
on (
P
S
O
)
[
5
]
,
and
m
e
t
hods t
h
at
com
b
i
n
e t
h
e l
a
t
t
e
r wi
t
h
fuzzy
lo
g
i
c[09
], [4
].
Th
e last sup
e
rv
ision
algorithm
s
are b
a
se
d o
n
pro
p
o
r
ti
o
n
a
l–
in
tegral
(PI) regu
lato
rs. Th
i
s
class
of
al
g
o
ri
t
h
m
s
regul
at
es t
h
e
pr
obl
em
of s
u
per
v
i
s
i
o
n
by
usi
n
g
a si
m
p
l
e
PI re
gul
at
o
r
[1
0]
,
[1
1]
, [
2
9]
.
The cu
rre
nt
re
search
wo
r
k
pr
esent
s
a com
p
arat
i
v
e st
u
d
y
o
f
t
h
e G
A
an
d PSO m
e
t
hods t
o
det
e
rm
i
n
e
th
e o
p
tim
al
PI co
n
t
ro
ller p
a
ra
m
e
ters fo
r the win
d
farm
s
u
p
e
rv
isio
n
algo
rith
m
,
an
d
com
p
ared
with
th
e no
n-
o
p
tim
ized
PI co
n
t
ro
ller, in wh
ich
t
h
e
p
a
ram
e
ters are adju
st
ed
m
a
n
u
a
lly.
2.
POWER SY
STEM
CONFIGUR
A
T
ION
The sy
st
em
studi
e
d
i
n
t
h
i
s
s
t
udy
as
prese
n
t
e
d i
n
t
h
e
f
o
l
l
o
wi
ng
di
ag
ra
m
m
a
i
n
l
y
co
m
p
o
u
nd
fr
om
d
i
f
f
e
r
e
n
t
electr
i
c Ele
m
en
ts, t
h
e w
i
nd
f
a
r
m
co
nn
ect th
rou
g
h
a tr
an
sf
or
m
e
r
(
20K
V
/ 690V
)
to
th
e elect
r
i
cal
net
w
or
k,
ad
di
t
i
onal
di
f
f
ere
n
t
va
ri
abl
e
l
o
ads
al
so c
o
nnect
t
o
t
h
e
net
w
o
r
k
b
u
t
wi
t
h
a
not
her
t
r
a
n
sf
orm
e
r.
w
e
have
f
o
cu
sed
m
o
re on ce
nt
r
a
l
supe
r
v
i
s
i
o
n
uni
t
t
h
at
ca
n c
ont
rol
t
h
e
wi
n
d
fa
rm
i
n
act
ive a
nd
react
i
v
e p
o
we
r
(
P
W
E
, QW
F)
fo
llo
w
i
n
g
th
e netw
or
k
system
op
er
at
o
r
TSO
r
e
qu
ir
ed
p
l
an
.
W
i
nd f
a
r
m
T
r
a
n
sm
i
ssi
o
n
O
p
er
at
or
Sy
s
t
e
m
(
T
SO
)
Ce
n
t
r
a
l
S
u
pe
r
v
i
s
or
y
Un
i
t
Lo
a
d
s
A
v
ai
lab
l
e p
o
wer
(
Q
W
G
_m
ax_
i
)
TS
O
p
l
a
n
(
P
F
_
ref
, Q
F
_
ref
)
B
U
S H
T
A
20k
V
Tra
n
sm
is
sio
n
Net
w
o
r
k
T
r
ansf
or
m
e
r
20k
V
/
690
V
Lin
e
1.5k
m
P
F
,Q
F
P
L
,Q
L
U
gd
(
P
F_
m
a
x
,
Q
F_
m
a
x
)
ev
ery
cy
cl
e
P
o
w
e
r re
fer
e
n
c
e sig
n
a
l
s fo
r
ea
ch
w
i
nd
g
e
ne
rat
o
r
(
P
WG
_
re
f_i
Q
WG
_
re
f_i
)
Fi
gu
re
1.
P
o
we
r Sy
st
em
C
onfi
g
u
r
at
i
o
n [
1
]
Th
e m
a
in
co
mp
on
en
ts of
wind
g
e
n
e
rators
used
in th
is
wind
farm
are tu
rb
in
e, g
e
arb
o
x
,
d
oub
ly-fed
in
du
ctio
n g
e
n
e
rato
r (DFIG)
wh
en
its st
ator is d
i
rectly con
n
ected
to th
e g
r
i
d
and
to
t
w
o inv
e
rters
on
e si
d
e
DFI
G
r
o
t
o
r (R
SC
) pe
rm
it
s t
o
cont
r
o
l
act
i
v
e
and
react
i
v
e
po
we
rs o
f
D
F
I
G
, t
h
e
ot
he
r o
n
e si
de
gri
d
(
G
SC
)
al
l
o
ws t
o
m
a
nage t
r
a
n
si
ent
bal
a
nced
p
o
w
er
a
n
d
cu
rre
nt
t
o
g
r
i
d
,
as s
h
ow
n i
n
Fi
gu
re
2
[1]
,
[1
3]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Win
d
Farm Ma
nag
emen
t
u
s
i
n
g Artificia
l In
tellig
en
t Techniq
u
e
s (Bou
a
l
em Ben
l
ahb
ib
)
1
135
a.
Turbine Model
:
The am
ount
of aer
o
d
y
n
am
ic po
wer
aer
P
capt
u
red f
r
om
wi
nd
t
u
rbi
n
e Fi
gu
re
2(a) can
b
e
ex
pr
essed
by th
e fo
llow
i
ng Equ
a
tion
[3
]:
2
)
,
(
3
SV
C
P
C
P
p
v
p
aer
(
1
)
whe
r
e:
aer
P
is th
e ob
tain
ed
wi
n
d
po
wer(w),
is th
e air den
s
ity(kg
/
m
3
),
V
is th
e wind
speed
(m
/s),
S
is the swep
t
area
of t
h
e t
u
rbine, a
n
d
p
C
is
the Power coe
ffici
ent.
Doub
l
y
f
e
d ind
u
cti
o
n
Ge
n
e
r
a
t
o
r
v
Sli
p
ri
n
g
AC
50
H
z
A
C
v
a
r
i
a
b
l
e
f
r
eq
ue
nc
y
G
r
id
G
e
arbox
Tu
rb
in
e
P
s.
P
Pg
Ro
to
r
s
i
d
e
con
v
ert
er
Gr
i
d
s
i
d
e
co
n
v
ert
er
Grid
si
de co
n
v
erter co
n
t
ro
ll
er
(GS
C
)
Ro
to
r
s
i
d
e
c
o
n
v
ert
e
r
co
n
t
ro
l
l
er
(
RS
C
)
Re
activ
e
p
ow
e
r
dis
p
atch
in
g
Ω
V
dc
V
s
I
t
I
r
I
s
L
o
c
a
l
c
o
nt
ro
l
uni
t
ce
nt
ra
l
su
p
erv
i
s
o
r
y un
it
Q
ref-g
Q
re
f
-
r
P
ref-r
V
dc-re
f
S
r
P
ref
Q
w
g
-r
ef-i
Q
wg
-
m
a
x
DF
IG
(a)
(c)
(b
)
(d
)
Fi
gu
re 2.
D
F
I
G
base
d W
i
n
d
Ener
gy
C
o
n
v
er
si
on
Sy
st
em
with
:
spee
d rat
i
o
defi
ned
as fol
l
ow:
wind
t
V
R
(
2
)
whe
r
e :
t
: is tu
rb
in
e sp
eed
,
wind
V
is th
e
wind
sp
eed
,
and
: is b
l
ad
e p
itch
ang
l
e;
The ae
ro
dy
na
m
i
c t
o
rq
ue i
s
g
i
ven
by
:
turbine
p
turbine
aer
aer
SV
C
P
C
1
*
2
*
3
(
3
)
b.
G
earbo
x m
o
d
e
l
As s
h
ow
n i
n
Fi
gu
re
2(c
)
,
t
h
e
Gear
b
o
x
t
o
r
q
u
e
can
be
p
r
esen
t
e
d by
f
o
l
l
o
wi
n
g
E
q
uat
i
on:
G
C
C
aer
g
g
C
: g
earbo
x to
rque ;
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E
V
o
l
.
7,
No
. 3,
J
u
ne 2
0
1
7
:
11
3
3
– 11
44
1
136
aer
C
:
aer
ody
nam
i
c t
o
r
q
ue ;
a
n
d
G
:
Gearbo
x mu
ltip
lyin
g
factor for sp
eed,
we h
a
v
e
:
G
mec
turbine
whe
r
e:
mec
is the m
echanical spee
d
c.
Gene
ral
DFI
G
M
odel
As shown
i
n
Fi
g
u
re 2(b
)
, t
h
e
DFIG is m
o
d
e
led
in
d, q
Park
m
o
d
e
l, th
e stat
o
r
an
d
ro
t
o
r
voltag
e
s can
b
e
written
as:
dr
s
dr
qr
r
qr
qr
s
dr
dr
r
dr
ds
s
qs
qs
s
qs
qs
s
ds
ds
s
ds
dt
d
I
R
V
dt
d
I
R
V
dt
d
I
R
V
dt
d
I
R
V
)
(
)
(
(
4
)
The stato
r
a
n
d
rot
o
r
flu
x
a
r
e
g
i
ven a
s
f
o
llo
ws
:
qs
sr
qr
r
qr
ds
sr
dr
r
dr
qr
sr
qs
s
qs
dr
sr
ds
s
ds
I
M
I
L
I
M
I
L
I
M
I
L
I
M
I
L
(
5
)
whe
r
e:
s
R
,
r
R
s
L
and
r
L
are the
resistances and
indu
ctances
, re
spectively,
of t
h
e st
at
o
r
an
d
r
o
t
o
r wi
ndi
ng
s,
a
n
d
sr
M
is the m
u
tual inductance
.
ds
V
,
dr
V
,
qs
V
,
qr
V
,
ds
I
,
dr
I
,
qs
I
,
qr
I
,
ds
,
dr
,
qs
and
qr
are the
d
and
q
com
pone
nt
s o
f
t
h
e st
at
or a
n
d
rot
o
r v
o
ltages ,
res
p
ectively
,
c
u
rrents
and
flux
whe
r
e as
is
the rotor s
p
ee
d in electrical de
gree
.
The active
and
reactive
powe
rs at the stat
or s
i
de an
d
r
o
t
o
r
si
de
of
DF
I
G
a
r
e
de
fi
ne
d as:
qs
ds
ds
qs
s
qs
qs
ds
ds
s
I
V
I
V
Q
I
V
I
V
P
(
6
)
qr
dr
dr
qr
r
qr
qr
dr
dr
r
I
V
I
V
Q
I
V
I
V
P
(
7
)
The electrom
a
gnetic torque
is expre
ssed as:
qr
ds
s
em
I
L
M
P
C
(
8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Wind Farm M
a
nagement
usi
n
g Artificial In
telligent Techniques (B
oualem Benlahbib)
1
137
Whe
r
e
P
is th
e nu
m
b
er of
p
o
l
e p
a
irs.
d.
Co
nv
erters M
o
d
e
l
For
t
h
e
DF
IG
m
odel
prese
n
t
e
d i
n
t
h
e
Par
k
m
odel
,
we
f
o
l
l
o
w
a co
nt
i
n
uo
us e
q
ui
val
e
nt
m
odel
o
f
con
v
e
r
ters i
n
t
h
e Pa
r
k
re
fere
n
ce [
1
]
,
[
2
9]
to s
i
m
p
lif
y th
e an
alysis o
f
t
h
e com
p
le
te p
o
wer
g
e
n
e
ration
syst
e
m
.
Th
e cu
rr
en
ts an
d vo
ltag
e
s of
RSC and
G
S
C
sh
own
in Fi
g
u
re 2(
d)
ar
e
d
e
f
i
ned
b
y
th
e
fo
llow
i
ng
Equ
a
tio
ns:
reg
rq
reg
rd
rmq
rmd
V
V
U
V
V
2
(
9
)
rq
rd
reg
rq
reg
rd
mac
m
i
i
V
V
I
2
1
(
1
0
)
whe
r
e:
rmd
V
,
rmq
V
rd
i
,
rq
i
: express
voltag
e
s and
cu
rren
t
s in Park
m
o
d
e
l; an
d
reg
rd
V
,
reg
rq
V
: Ex
press adju
sted
v
o
ltag
e
s in
Park
m
o
d
e
l.
3.
PI Algorithm for Wind Farm Supervision
Th
e m
a
in
ob
j
e
ctiv
e of th
e PI
regu
lato
r-b
ased
al
gorith
m
is t
o
satisfy t
h
e sy
ste
m
o
p
e
rat
o
r referen
ce
o
p
e
rating
set-po
in
t
(
ref
WF
Q
,
ref
WF
P
). T
h
ese
values
are c
o
m
p
are
d
with
t
h
e
active and
reac
tive powe
rs at
t
h
e p
o
i
n
t
o
f
c
o
m
m
on co
upl
i
n
g
(PC
C
)
[
1
7]
, a
n
d
t
h
e
di
f
f
ere
n
ce i
n
po
we
r i
s
det
e
rm
i
n
ed,
w
h
i
c
h i
s
di
st
ri
bu
t
e
d i
n
an ide
n
tical m
a
nne
r(
i
ref
WG
P
,
i
ref
WG
Q
) bet
w
ee
n t
h
e
wi
nd
ge
n
e
rat
o
r
s
of t
h
e
wi
n
d
farm
. Th
e f
o
l
l
o
wi
ng
Fig
u
re
3
p
r
esen
ts th
e prin
ci
p
l
e of th
is al
g
o
rith
m
.
n
ref
WG
P
_
_
WF
P
~
ref
WF
P
_
1
_
_
ref
WG
P
PI
C
o
n
t
r
o
lle
r
1/
n
2
_
_
ref
WG
P
WF
Q
~
ref
WF
Q
_
1
_
_
ref
WG
Q
PI
C
o
n
t
ro
l
l
e
r
1/
n
2
_
_
ref
WG
Q
n
ref
WG
Q
_
_
(a) A
c
t
i
v
e
po
w
e
r con
t
rol
(b) react
i
v
e
pow
e
r
con
t
rol
Figu
re.
3
.
P
I
c
ont
roller
f
o
r
wi
nd
fa
rm
super
v
ision
whe
r
e:
WF
P
~
:
act
i
v
e p
o
w
e
r
gene
rat
e
d
by
w
i
nd
farm
WF
Q
~
:
react
i
v
e po
we
r gene
rat
e
d o
r
abs
o
r
b
ed
by
A
gene
ral
bl
oc
k
di
ag
ram
for
a PI
co
nt
r
o
l
s
y
st
em
i
s
sho
w
n i
n
Fi
gu
re
4.
The c
o
nt
r
o
l
si
gnal
U
(
t
)
i
s
gene
rat
e
d
f
r
om
t
h
e e
r
r
o
r, E(t), as in (1
1):
]
)
(
1
)
(
[
)
(
0
T
i
p
dt
t
E
T
t
E
K
t
U
(
1
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEC
E
1
138
The t
r
whe
r
e
For t
h
t
une
d
suc
h
a
whe
r
e
in
(1
3
4.
P
p
rob
l
e
4.
1.
P
whic
h
algor
i
anim
a
relati
v
t
uni
n
g
com
p
effec
t
swar
m
tu
rb
i
n
i
ndi
v
i
opt
i
m
have
i
opt
i
m
appro
(CS
A
(PS
O
)
syste
m
E
Vo
l.
7
,
N
o
.
r
an
sf
e
r
fun
c
ti
o
K
s
G
p
c
)
(
e
p
K
is th
e
p
r
h
e op
timu
m
p
by
m
i
ni
m
i
zi
n
a
s [
22]
:
(
)
(
0
0
W
t
f
e
(13b
)
is th
e
w
3
a) is
added t
o
P
erfor
man
c
The P
S
O
e
ms
.
A
b
r
i
e
f
r
P
SO Al
g
ori
t
h
The parti
c
h
uses
a pop
u
i
th
m
was
p
r
o
p
a
ls liv
in
g
in
s
w
v
ely complex
The use o
g
PI
r
e
gu
lato
r
ensator (ST
A
t
iv
en
ess of
th
e
m
opt
im
izat
i
o
n
e go
ver
n
or
s
y
i
dual
best
po
s
m
ize a new P
I
i
nvestigat
ed
t
m
izat
io
n
pr
ob
l
o
ach they
hav
e
A
).
H. Bev
r
ani
)
techniques
m
. The perf
o
r
3
,
Jun
e
201
7
o
n of
t
h
e PI
c
o
s
K
i
p
r
op
o
r
t
i
onal
g
a
F
p
erform
ance
o
n
g a perf
o
r
m
a
n
)
)
(
(
)
(
1
2
1
dt
t
e
W
e
W
t
e
W
y
w
ei
ght
e
d
i
n
t
e
g
o
wei
ght
e
d
IA
E
c
e of PSO
a
n
and G
A
m
e
t
r
ev
iew of
thes
h
m
c
le swarm
o
p
u
latio
n of
c
a
p
ose
d
by
Eb
e
w
ar
m,
s
u
c
h
a
s
m
o
t
i
o
n
dy
na
m
f th
is
m
e
th
o
d
r
usi
ng P
S
O
h
A
TC
OM).E
x
p
e
e
p
r
op
ose
d
c
o
n i
s
pr
op
ose
d
y
stem
. Aut
h
o
r
s
itio
n
an
d th
e
I
D-
typ
e
fuzz
y
t
he
case of a
n
em
of dua
l
a
e
use
d
PSO a
l
et al[21] pre
s
fo
r op
timal
r
m
a
nce of pr
o
:
11
3
3
– 11
4
o
n
t
ro
ller is de
a
in and
i
K
is
t
F
ig
ure 4
.
Blo
c
o
f the
co
nt
ro
l
n
ce i
ndex;
T
h
)
)
(
ot
h
if
dt
t
y
g
ral o
f
a
b
s
o
l
u
E
in
ord
e
r to
a
n
d GA
for
P
t
hods
are em
p
e algorithm
s
i
p
tim
iz
atio
n
(
P
a
ndi
dat
e
s
o
l
u
t
e
r
h
a
r
t
and
R
u
s
sch
ool
s of f
i
m
ics, m
o
re
de
d
i
n
tunin
g
PI
h
as been
a
p
pli
e
e
rim
e
nt resul
t
o
n
t
ro
l appr
oa
c
d
fo
r o
p
tim
al
r
s in
t
h
is wor
k
gl
ob
al
best
p
o
y
l
ogi
c co
nt
ro
electrical D
C
a
x
i
s so
lar tra
c
l
gori
t
h
m
com
p
s
ente
d a com
b
t
uni
ng of PI
-
o
p
o
s
ed
in
telli
4
4
e
fi
ne
d by
(
s
G
c
t
h
e
in
teg
r
al
g
a
c
k di
ag
ram
o
f
l
system
, the
h
erea
fter, this
0
)
(
t
herwise
t
e
y
u
te er
ror
(IA
E
)
a
vo
id
ov
er
sh
o
P
I C
o
ntrolle
r
p
lo
yed b
eca
u
i
s pre
s
ent
e
d i
n
P
SO Particle
t
i
ons t
o
deve
u
ssel Jam
e
s
K
i
sh an
d fl
ock
e
tails ab
ou
t P
S
regu
lator ap
p
ed by Chien-
H
t
s un
der
di
f
f
e
c
h. I
n
a
not
he
r
l
Propo
rt
i
o
na
l
k
i
n
tr
odu
ced
t
o
sition
in
IP
S
o
ller [FLC]
t
u
n
C
dri
v
e
benc
h
c
ker system
w
p
ared to firef
l
b
in
atio
n
o
f
th
e
-
re
gul
at
o
r
-
b
a
s
i
gent
co
nt
rol
)
as fo
llo
ws:
a
in.
f
P
I
c
ont
r
o
l
sy
s
PI con
t
ro
lle
r
is selected to
)
13
(
)
13
(
b
a
)
value, and t
h
o
ot
.
r
Op
timi
za
t
u
se
of t
h
eir e
f
n
th
e
fo
llowin
Swarm
Opti
m
l
op
an
opt
i
m
K
en
ne
dy
[1
4]
.
of
b
i
rd
s.
In
d
e
S
O
algor
ith
m
w
p
ear in s
o
me
a
H
un
g et
al
. [
1
e
ren
t
lo
ad
ing
r
st
udy
H
o
ng
q
l
–I
n
t
eg
r
a
l
–
De
r
t
he n
o
m
i
nal
a
S
O. S
.
Bo
uall
e
n
ing
strateg
y
h
mar
k
.
M. M
u
w
ith
DC
m
o
t
o
l
y
al
gori
t
h
m
(
e
f
u
zz
y lo
g
i
c
s
ed fre
q
u
e
n
c
y
i
s com
p
ared
w
s
tem
.
r
gai
n
s,
p
K
a
sa
t
isfy sev
e
r
a
h
e weighted a
b
i
on
f
fecti
v
ene
ss
i
g
s
ect
i
on.
m
izat
io
n
)
is
a
m
al
so
l
u
tio
n
fo
It b
u
ild
s on
e
d
,
one ca
n
o
b
w
as desc
ri
be
d
a
p
p
licatio
ns a
m
7]
t
o
per
f
o
r
m
co
nd
itio
ns
w
q
i
ng et al. [1
8
r
iv
ativ
e [PID
a
verage p
o
si
t
i
o
e
gue et al. [1
9
,
and for c
h
e
u
ha
mm
ad and
o
r
dr
iv
e,
an
d
FFA) a
nd C
u
and t
h
e pa
rt
i
c
y
cont
r
o
l
l
e
rs
w
ith
th
e
p
u
r
e
ISS
N
:
2
(
a
nd
i
K
, are
a
l co
n
t
ro
l sp
e
c
b
so
lu
te term
i
n
i
n
so
lv
ing
o
p
a
n e
v
ol
ut
i
o
n
f
or
suc
h
pr
o
b
t
h
e social b
e
o
bse
r
ve i
n
t
h
e
s
d
i
n
[
15]
,
[
16]
.
a
m
ong t
h
em
,
a
m
the static sy
n
w
e
r
e used to
8
]
,
an i
m
prov
e
D
]
for m
onito
r
on of
s
w
a
r
m
9
] used
PSO
i
e
c
k
i
ng t
h
ei
r r
e
T. Ali [20
]
s
to tack
le o
p
u
c
k
o
o
S
e
ar
ch
A
c
le swarm
o
p
in th
e AC
m
e
fuzzy PI an
2
088
-87
08
(
12
)
o
p
tim
al
ly
c
i
f
icatio
n
s
,
n
)
(
t
e
y
tim
izat
io
n
algor
ith
m
b
lem
.
Th
is
e
ha
vi
o
r
of
s
e a
n
i
m
als
.
a
n
op
tim
a
l
n
ch
ro
n
ous
pr
o
o
f t
h
e
e
d
particle
r
i
ng wa
t
er
b
eside the
i
n o
r
der t
o
e
su
lt
s,th
ey
s
tud
i
ed
the
timi
zatio
n
A
lgo
r
ith
m
ti
mizatio
n
m
icr
o
g
r
i
d
d Zi
egl
e
r
-
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEC
E
N
ich
o
have
t
gl
o
b
a
to
a
n
4.
2.
G
1
960
-
suc
h
a
devel
o
p
ro
ce
adapt
i
PID
c
t
uni
n
g
Mo
y
e
syste
m
in
teg
r
al [3]
(P&
O
opt
i
m
in
o
r
d
vehi
c
l
han
d
l
5.
R
rep
r
o
d
Matl
a
po
we
r
fu
nct
i
use
d
c
Whe
n
er
ro
r
o
E
o
ls PI con
t
ro
l
t
ak
en
a
hyb
r
i
d
l b
e
st p
a
rticl
e
ew be
st posit
i
G
enetic Al
go
Gen
e
tic a
l
-
1
9
7
0
pe
r
i
od
a
s i
n
heri
t
a
nc
e
o
pm
ent
o
f
a
n
ss
co
ul
d
be u
s
i
ng
PI
con
t
ro
l
T.jim
e
ne
e
c
o
n
t
ro
ller au
t
o
g
tim
e u
p
to
e
n [2] treated
m
, variable s
p
r
al
[
P
I]
t
u
ned
use
d
P
I
D co
n
O
)
ma
x
i
m
u
m
m
ized PID co
n
d
er t
o
contro
l
l
es. F.da
nesh
f
e th
e lo
a
d
-f
re
q
R
ESULTS
A
The wi
nd
d
u
ced
in th
e
a
b/
Sim
u
l
i
nk t
o
r
di
st
ri
b
u
t
i
on
To
conv
e
r
i
on
. Hen
c
e,
it
c
ost
f
u
nct
i
on
(
F
N
i
1
Fig
u
n
1
e
(
i
) is th
e
t
o
f
i
th
sam
p
le
f
Wind F
a
desi
g
n
m
e
t
h
o
d
jum
p
PSO
a
e
s which are
n
i
on
.
o
r
i
thms (G
A)
l
g
o
rith
m
s
th
e
o
a
nd has be
en
e
, m
u
tatio
n, s
e
o
p
tim
iza
t
i
o
n
s
efu
l
with
i
n
a
l
ler s
u
ch as
:
e
t al
[23
]
in
o
o
ma
t
i
c
t
u
n
i
n
g
64
% c
o
m
p
ar
e
th
e p
r
ob
lem
o
p
eed drive
s
a
r
by
[
GA]
, a
n
d
n
tro
ller b
a
sed
power p
o
i
n
t
n
tro
ller p
a
ra
m
e
l
the angul
a
r
f
or
and
H
.
Be
v
q
u
e
n
c
y contr
o
A
ND
DI
SC
U
S
farm
m
odel
w
e
A
p
pend
i
x
)
o
ols. The
blo
c
is show
n in F
r
g
e
to
ward
th
e
sh
o
u
l
d
be pr
o
(
F
) is
defi
n
e
d
e
i
e
2
1
u
re 5.
B
l
oc
k
d
t
raject
ory e
r
r
o
f
or reactive
p
o
I
S
a
rm Ma
nag
e
m
o
ds. M
.
say
e
d
a
lgo
r
ith
m
for
n
ot
i
m
prove
d
i
o
ry
(G
As Ge
n
fu
lly elab
o
r
a
e
lectio
n
an
d
c
r
algo
rith
m
,
b
u
com
put
i
n
g s
y
rd
er to
m
a
n
a
g
g
proce
ss wit
h
e
d to
trad
itio
n
o
f usi
n
g p
o
w
e
r
ea, they intr
o
d
c
o
m
p
ared
w
i
on ge
net
i
c
al
g
tracking alg
o
e
ters attached
p
osi
t
i
on of t
h
v
r
a
n
i
[5
]
p
ro
p
o
l(LFC) issu
e
,
SI
O
N
w
ith
three wi
n
is u
s
ed
to
c
k di
ag
ram
o
f
i
gu
r
e
5.
e
op
timal so
l
u
o
perl
y
de
fi
n
e
d
b
y
th
e
fo
llo
w
i
d
iagram
o
f
Tu
n
fo
r
w
o
r of
i
th
sam
p
l
e
o
we
r
(Q
re
f
-
WF
Op
t
al
g
o
Kp
SSN
:
208
8-8
7
m
e
n
t usi
n
g
Art
if
d
et al [22] i
n
t
u
ni
n
g
gai
n
s
i
n a pre
d
efi
n
e
n
etic Al
g
o
rit
h
a
t
e
d i
n
hi
s
bo
ross
o
v
er "
p
u
b
u
t rath
er
t
h
e
m
y
ste
m
. A
nu
m
g
e net
w
or
ks
p
h
GA;
ob
tai
n
n
al
m
e
t
hods
a
er c
o
nversio
n
o
duce
d
a cas
c
i
th
Tagu
ch
i a
p
g
o
rith
m
to
a
d
o
rith
m
with
a
d
to electro-h
y
h
e rota
ry act
u
p
s
e
d t
h
e t
u
ni
n
g
,
wh
ich is th
e
n
d ge
ne
r
a
t
o
rs
s
obse
r
ve t
h
e
f
the cont
rol
s
u
tio
n
,
th
e PS
O
d
bef
o
re
t
h
e
P
S
w
ing
Equ
a
tio
n
n
in
g
PI
p
a
r
a
m
w
in
d farm
su
p
e
fo
r activ
e
p
o
)
Q
~
-
WF
f
. T
h
t
im
is
at
ion
ori
t
hm
es
PI
Ki
7
08
if
icia
l Intellig
e
n
or
de
r t
o
co
n
s
of PI reg
u
l
a
t
e
d n
u
m
b
er of
h
ms
)
w
a
s
o
r
ig
ok "
A
dapt
at
i
o
b
lish
e
d
in
19
7
m
odel
i
ng p
r
oc
e
m
ber of
resear
c
p
a
r
am
eter
s in
p
n
ed
resu
lts
sh
o
a
s Ziegler Ni
c
n
syste
m
unit
c
aded c
o
ntro
l
p
p
r
oach un
de
r
d
ju
st a
n
e
w m
o
a
dapt
i
v
e dut
y
y
d
r
au
lic serv
o
u
at
or w
h
i
c
h c
o
g
of PI re
gul
a
t
maj
o
r subj
ec
t
s
ituated i
n
di
f
beha
vi
or o
f
s
yste
m with
a
O
, and
GA al
g
S
O al
g
o
rith
m
[2
8]
:
m
eters with P
S
p
ervision.
o
we
r
(P
ref
-
WF
h
e re
gi
o
n
of
t
h
Process
to
re
g
ulat
e
e
nt
Tech
n
i
qu
e
n
tro
l
th
e
b
oi
l
t
o
r
, based o
n
o
i
t
e
rat
i
ons a
n
d
i
nal
l
y
de
ve
l
o
p
o
n
in
Natu
ral
7
5
[1
9]
. It
s m
a
e
s
s
of
ad
ap
ta
ti
c
h
e
s ha
ve
use
d
p
assi
ve
opti
m
o
w that th
e c
c
ho
ls (ZN)
.
h
in re
newa
bl
e
schem
e
bas
e
r
th
e g
r
i
d
fau
l
o
di
fied Pe
rtu
r
y
cycle. K.
M
actuator syst
e
o
n
t
ro
l th
e
m
o
t
or pa
ram
e
t
e
r
s
t
in
a
p
o
w
e
r
s
y
f
fere
nt win
d
p
f
th
is co
n
t
ro
l
a
PSO–PI an
d
g
ori
t
hm
s m
u
s
is exec
uted.
I
S
O a
n
d
G
A
al
g
)
P
~
-
WF
and
e
h
e para
m
e
ters
e
e
s
(
B
o
ual
e
m
B
l
e
r
-
t
ur
bi
ne u
n
o
b
s
ervi
ng t
h
e
d
m
ovi
ng t
h
es
p
ed
b
y
Jo
hn
H
an
d
Artifici
a
a
in
o
b
j
ectiv
e
w
i
on
, a
n
d
s
h
o
w
d a
ppl
i
cat
i
o
n
m
al n
e
tw
orks
c
o
n
t
rol strate
g
h
.
M
.hasa
n
ie
n
energy ,ene
r
g
e
d on f
o
u
r
p
r
l
t co
nd
itio
n.
A
r
b
a
t
i
on a
nd
O
b
M
. Elbzyom
y
em
(EHSAS
)
o
va
bl
e sur
f
ac
e
s usi
n
g
GA
a
p
y
st
e
m
.
p
ro
files (sy
s
te
m
l alg
o
r
ith
m
d
GA
–P
I
co
n
t
s
t
be gui
de
d
b
I
n the
prese
n
t
(
g
orith
m
s
2
e
(
i
) is th
e
tr
a
to
be
op
ti
m
i
z
B
en
lahb
ib
)
1
139
i
t, au
tho
r
s
e
lo
cal and
e
p
a
rticles
H
o
lland
i
n
a
l Syste
m
s
w
asn’t th
e
w
s how
t
h
is
of
G
A
fo
r
ha
ve
us
e
d
g
y
re
duce
s
n
and
S.
M
g
y s
t
o
r
ag
e
o
p
o
r
t
i
onal
A
.ha
r
rag et
b
ser
v
ation
et
a
l
.
[4]
us
i
ng G
A
e
of
space
p
pr
oach
t
o
m
data are
usi
ng t
h
e
tr
ollers fo
r
b
y the cost
st
u
dy, the
(
14
)
a
j
ector
y
e
d
is set
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E
V
o
l
.
7,
No
. 3,
J
u
ne 2
0
1
7
:
11
3
3
– 11
44
1
140
as fo
llo
ws:
0<
k
p
,
k
i
<10
Fo
r the PSO al
g
o
rith
m
,
th
e p
o
p
u
l
atio
n
size
is set to
2
0
p
a
rticles. Th
e p
a
ra
m
e
ters
c
1
,
c
2
and
W
are set
to
2.05
, 2.05
an
d 0.729
8,
r
e
sp
ectiv
el
y
.
T
h
e
m
a
xim
u
m
num
ber
o
f
i
t
e
rat
i
o
n
n
is set to
20
iteratio
n
s
.
For the
GA al
gorithm
param
e
ters a
r
e selected as
below:
a.
Selection:
normalized geom
etric selection
b.
C
r
oss
o
ver:
a
r
i
t
h
m
e
ti
c cross
o
v
e
r
c.
M
u
t
a
t
i
on:
uni
f
o
rm
m
e
t
hod
d.
Po
p
u
l
a
t
i
on
n
u
m
ber:
20
e.
Gen
e
ration
(iteratio
n) n
u
m
b
e
r:
20
f.
The sam
e
search inte
rval a
nd obj
ective func
tion as
PSO al
gorithm
.
First, th
e PI con
t
ro
ller g
a
ins are adju
sted
m
a
n
u
a
lly
. T
h
erea
fter,
we pe
rform an optim
ization
process
usi
n
g t
h
e PS
O and
GA m
e
t
hods. Fi
g
u
r
e 6 sh
ows t
h
e c
o
st
funct
i
o
n ev
ol
ut
i
on
du
ri
n
g
t
h
e opt
i
m
i
zati
on p
r
oces
s
u
s
ing
PSO and GA.
After
2
0
iteratio
n
s
, th
e PSO and
GA con
v
e
rg
e to th
e
op
ti
m
a
l p
a
ram
e
t
e
rs.
Tabl
es 1 a
nd
2
sho
w
t
h
at
al
l
of t
h
e
s
e pa
ram
e
t
e
rs o
b
t
a
i
n
ed
fr
om
bot
h al
go
ri
t
h
m
s
are di
ff
erent
f
r
om
one a
n
othe
r for the active and
reactive power cont
rollers
.
Th
e sim
u
latio
n
resu
lts o
f
t
h
e con
t
ro
lled
system with
t
h
e o
p
t
i
m
i
zed PI are s
h
ow
n i
n
Fi
g
u
r
e
s 7
(
a,
b)
.The
o
p
t
i
m
i
zed PI c
o
nt
r
o
l
l
e
r by
t
h
e P
S
O a
nd
G
A
m
e
t
hod
s are
also c
o
m
p
ared
with the
non
-op
t
i
m
ized
PI
fo
r th
e
active
and reactive powe
rs.
Fi
gu
re
6.
Ev
ol
ut
i
o
n
o
f
t
h
e c
o
st
fu
nct
i
o
n
(
CF
)
Tabl
e 1.
Param
e
t
e
rs P
I
c
ont
r
o
l
l
er o
b
t
a
i
n
e
d
by
di
f
f
ere
n
t
m
e
t
hods
f
o
r
act
i
v
e
po
we
r
Tabl
e 2.
Param
e
ters P
I
c
ontrol
l
er obtaine
d
by
differe
n
t m
e
thods
for reactive powe
r
Fi
gu
res
7(a,
b
)
sho
w
t
h
at
t
h
e
opt
i
m
i
zed PI cont
rol
l
e
r
usi
n
g
t
h
e PSO m
e
t
hod
has
bet
t
e
r p
e
rf
orm
a
nce,
suc
h
as
ra
pi
d
r
e
sp
onse
an
d t
r
aject
o
r
y
t
r
ac
ki
ng
t
a
sk
, c
o
m
p
ared
wi
t
h
t
h
e
G
A
m
e
t
hod a
n
d
t
h
e n
o
n
-
o
pt
im
ized P
I
.
Fi
gu
res 8 a
n
d 9 sh
o
w
s t
h
e e
n
t
i
r
e wi
n
d
fa
r
m
wi
t
h
the active and reactive powe
rs a
n
d distributi
on
order powers
for each wi
nd generat
o
rs
us
i
n
g the PSO–PI c
o
ntroller algorit
h
m
.
0
2
4
6
8
10
12
14
16
18
20
2
4
6
8
10
12
14
16
x 1
0
9
G
ener
at
i
o
n
F
i
tnes
s
func
ti
on (
F
1)
GA
PSO
M
e
thod
Para
m
e
ters
T
r
aditional m
e
thod
GA
PSO
K
p
0.
1 1.
812
0.
5098
K
i
2 6.
649
10
m
e
thod
p
a
ra
m
e
t
e
rs
T
r
aditional m
e
thod
GA
PSO
K
p
0.
1 2.
181
1.
9648
K
i
2 2.
674
4.
1773
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Wind Farm M
a
nagement
usi
n
g Artificial In
telligent Techniques (B
oualem Benlahbib)
1
141
(a) Active Po
w
e
r
(b
) Rective
Po
wer
Fi
gu
re
7.
C
o
m
p
ari
s
on
res
u
l
t
s
obt
ai
ne
d
by
P
I
opt
i
m
i
zed by
P
S
O,
G
A
.
A
n
d
no
n
-
o
p
t
i
m
i
zed PI
(a) Reactive
P
o
we
r
pr
o
duce
d
by
the
wi
nd
fa
rm
(b
) R
eact
i
v
e
P
o
we
r
pr
o
duce
d
by
t
h
e
fi
r
s
t
wi
nd
gene
rat
o
r
(c)
Reactive
power
produce
d
by
the second wind
gene
rat
o
r
(d
)
R
eact
i
v
e
p
o
we
r
pr
o
duce
d
by
t
h
e t
h
i
r
d
w
i
nd
gene
rat
o
r
Fi
gu
re
8.
Si
m
u
l
a
t
i
on R
e
sul
t
s
t
h
e ce
nt
ral
i
zed
su
perv
ision
o
f
th
e
reactiv
e
power
[PSO-PI]
0
5
10
15
20
25
-8
-6
-4
-2
0
2
4
6
8
10
12
x 1
0
5
ti
m
e
(
s
)
re
a
c
t
i
v
e
p
o
w
e
r(v
a
r
)
14.
5
15
15.
5
16
16.
5
17
17.
5
9.
2
9.
4
9.
6
9.
8
10
10.
2
10.
4
x 1
0
5
Q
w
f
-ref
Qw
f
-
m
e
s
-
GA
-
P
I
Qw
f
-
m
e
s
-
P
S
O-
P
I
Qw
f
-
m
e
s
-
P
I
0
5
10
15
20
25
30
-2.
5
-2
-1.
5
-1
-0.
5
0
x 1
0
6
tim
e
(
s
)
ac
t
i
v
e
pow
e
r
(
w
at
)
19
.
2
19
.
4
19
.
6
19.
8
20
20
.
2
20
.
4
20.
6
20
.
8
21
21
.
2
-1
.
0
6
-1
.
0
4
-1
.
0
2
-1
-0
.
9
8
-0
.
9
6
x 1
0
6
P
w
f-re
f
Pw
f
-
me
s
-
G
A
-
P
I
Pw
f
-
me
s
-
PSO
-
P
I
Pw
f
-
me
s
-
PI
0
10
20
30
40
50
60
70
80
90
100
-2
-1.
5
-1
-0.
5
0
0.
5
1
x 1
0
6
t(
s
)
Q
w
f[v
a
r
]
Qw
f
-
r
e
f
Qw
f
-
m
é
s
0
10
20
30
40
50
60
70
80
90
10
0
-1
6
-1
4
-1
2
-1
0
-8
-6
-4
-2
0
2
4
x 1
0
5
t(
s
)
Q
w
g-
1[
v
a
r
]
Qw
g
1
-
m
a
x
Qw
g
1
-
m
é
s
Qw
g
1
-ré
f
0
10
20
30
40
50
60
70
80
90
10
0
-1
6
-1
4
-1
2
-1
0
-8
-6
-4
-2
0
2
4
x 1
0
5
t(
s)
Q
w
g-
2[
v
a
r]
Q
w
g2-
m
a
x
Q
w
g2-
m
é
s
Q
w
g2-
r
é
f
0
10
20
30
40
50
60
70
80
90
10
0
-1
6
-1
4
-1
2
-1
0
-8
-6
-4
-2
0
2
4
x 1
0
5
t(
s)
Qw
g-
3
[
v
a
r
]
Q
w
g3-
m
a
x
Q
w
g3
-
m
és
Qw
g
3
-ré
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E
V
o
l
.
7,
No
. 3,
J
u
ne 2
0
1
7
:
11
3
3
– 11
44
1
142
(a) Active Po
w
e
r pr
od
uce
d
by
the win
d
farm
(b
) Act
i
v
e po
w
e
r pr
od
uce
d
by
t
h
e fi
rst
wi
n
d
gene
rat
o
r
(c) Act
i
v
e p
o
w
e
r pr
od
uce
d
by
t
h
e
sec
o
nd
wi
nd
gene
rat
o
r
(d
) Act
i
v
e po
w
e
r pr
od
uce
d
by
t
h
e
t
h
i
r
d
wi
nd
gene
rat
o
r
Fig
u
re.9
.Sim
u
l
atio
n
Resu
lts th
e cen
t
ralized
su
perv
ision
o
f
th
e activ
e
po
wer [PSO-PI]
6.
CO
NCL
USI
O
N
In t
h
is stu
d
y
,
d
i
ffere
nt ap
p
r
oa
ches
fo
r
win
d
f
a
rm
super
v
i
s
i
o
n
w
e
re pre
s
ent
e
d. We foc
u
se
d ou
r
st
u
d
y
on
o
n
e o
f
t
h
e
m
whi
c
h i
s
ba
sed o
n
pr
op
o
r
t
i
onal
i
n
t
e
g
r
al
[PI] algo
rith
m
.
In
o
r
d
e
r t
o
ob
tain
th
e co
n
t
ro
ller
param
e
t
e
rs, t
h
e com
p
arat
i
v
e
st
u
d
y
ha
ve
been
t
a
ke
n
b
e
t
w
een
PS
O
and
G
A
m
e
t
hods
. T
o
ve
ri
f
y
t
h
e
effect
i
v
e
n
ess
o
f
t
h
e
p
r
op
ose
d
m
e
t
hods
, a m
o
del
of t
h
e
wi
n
d
farm
com
poun
d
wi
t
h
t
h
re
e
wi
n
d
ge
nerat
o
rs
was
si
m
u
lated
u
s
i
n
g
Matlab
/
Sim
u
lin
k
.
Th
e sim
u
latio
n
resu
lts sh
ow t
h
at th
e op
ti
m
i
zed
PI
con
t
ro
ller tun
e
d
b
y
th
e
PSO
m
e
t
hod e
xhi
bi
t
s
bet
t
e
r
p
e
rf
orm
a
nce t
h
a
n
t
h
e
o
n
e t
u
n
e
d
by
t
h
e
G
A
m
e
t
h
o
d
a
n
d
t
h
e
n
o
n
-
opt
i
m
i
zed PI.
APPE
NDI
X
Plant Pa
ram
e
ters
1
.
5
M
W
W
i
nd
Tu
rb
in
e
Parameters:
Rotor diam
eter: 35.25 m
. bla
d
es
Num
b
er:
3
In
ertia: 1
000
kg
/m
2
Ai
r de
nsi
t
y
=1
.2
2
kg/
m
3
1.
5 M
W
D
F
I
G
Param
e
t
e
rs :
s
R
=0.012
,
r
R
=0.021
s
L
=2
.0
37
.10
e
-
4
H
:
r
L
=1
.7
5.10
e -4
H
sr
M
=0
.0
35
H
:
s
L
=0.
0
35+
2.
0
3
7
.
1
0
e
-
4 H
r
L
=0
.0
35
+1
.7
5.10
e-
4 H
0
10
20
30
40
50
60
70
80
90
10
0
-2
-1
.
8
-1
.
6
-1
.
4
-1
.
2
-1
-0
.
8
-0
.
6
-0
.
4
-0
.
2
0
x 1
0
6
t(
s)
P
w
f[w
a
t
]
P
w
f
-re
f
Pw
f
-
m
é
s
0
10
20
30
40
50
60
70
80
90
100
-2
-1
.
5
-1
-0
.
5
0
0.
5
1
x 1
0
6
t(
s
)
P
w
g-
1[
w
a
t
]
P
w
g1-
m
a
x
P
w
g1-
m
é
s
P
w
g1-
r
é
f
0
10
20
30
40
50
60
70
80
90
100
-2
-1
.
5
-1
-0
.
5
0
0.
5
1
x 1
0
6
t(
s)
P
w
g-2[
w
a
t
]
Pw
g
2
-
m
a
x
P
w
g2
-
m
és
P
w
g
2
-ré
f
0
10
20
30
40
50
60
70
80
90
10
0
-2
-1
.
5
-1
-0
.
5
0
0.
5
1
x 1
0
6
t(
s
)
P
w
g-3[
w
a
t
]
P
w
g3
-
m
ax
P
w
g3-
m
é
s
Pw
g
3
-
r
é
f
Evaluation Warning : The document was created with Spire.PDF for Python.