Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
Vol
.
5
,
No
. 3,
J
une
2
0
1
5
,
pp
. 37
9~
39
0
I
S
SN
: 208
8-8
7
0
8
3
79
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
A New Hyb
r
id Artifi
ci
al Neu
r
a
l
Ne
twor
k Ba
se
d Co
ntro
l o
f
Doubly Fed Induction Generator
G. V
e
nu
Ma
dhav
*, Y
.
P. Obulesu**
* Departm
e
nt
of
Electri
cal
and
E
l
ectron
i
cs Eng
i
neer
ing, B. V. R
a
j
u
Institut
e
of
T
e
chnolog
y
,
Indi
a
** Departmen
t
o
f
Electr
i
cal
and
Electroni
cs
Eng
i
neering
,
KL
Uni
v
ers
i
t
y
,
Indi
a
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Nov 14, 2014
Rev
i
sed
Feb
19
, 20
15
Accepte
d
Mar 2, 2015
In this paper
,
H
y
brid
Artificial
Neural Network
(ANN) with Proportional
Integral (PI) control techniqu
e has b
een develop
e
d for Doubly
Fed Induction
Generator
(DFIG) based wind
energ
y
ge
n
e
ra
tion
s
y
s
t
em
and
th
e
perform
ance
of the s
y
s
t
em
i
s
com
p
ared wit
h
NN and P
I
control t
echniqu
e
s
. W
ith the
increasing use of wind power
generati
on, it
is required to instigate th
e
d
y
namic perfor
m
ance an
aly
s
is
of Doubly
Fed
Induction G
e
ner
a
tor und
er
various operatin
g conditions. In
this pa
per,
thr
ee control techn
i
ques have
been proposed,
the first one is
using
PI controller, th
e second
one is ANN
control,
and the third one is base
d on combination of ANN a
nd PI.
The
performance of
the proposed con
t
rol tec
hniqu
es is demonstrated through the
results, determined b
y
using MATLab/
Sim
u
li
nk. From
the results it is
observed th
at
th
e d
y
namic perfo
rmance of
the
DFIG is im
proved with
th
e
H
y
brid
contro
l
techniqu
e.
Keyword:
Co
n
t
ro
l strategy
D
oub
ly f
e
d
indu
ctio
n g
e
n
e
r
a
to
r
Neu
r
al
net
w
or
ks
Propo
rtion
a
l integ
r
al
W
i
nd
p
o
w
e
r
gen
e
r
a
tion
Copyright ©
201
5 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
:
G
.
V
e
n
u
Madhav
Depa
rt
m
e
nt
of
El
ect
ri
cal
and
El
ect
roni
cs
E
n
gi
nee
r
i
n
g,
Pad
m
asri Dr.
B. V. Raju
In
stitu
te o
f
Tech
nolo
g
y
,
V
i
shn
upu
r, N
a
r
s
apur
,
Med
a
k D
i
st., 50
231
3,
AP,
I
ndi
a
.
Em
a
il: v
e
n
u
m
a
d
h
a
v
.
go
p
a
la@g
m
a
i
l
.co
m
1.
INTRODUCTION
M
o
st
of t
h
e
cou
n
t
r
i
e
s i
n
t
h
e w
o
rl
d are
no
w co
nce
n
t
r
at
i
ng
on
re
n
e
wabl
e e
n
er
gy
sou
r
ces, i
n
part
i
c
ul
a
r
l
y
wi
nd e
n
er
gy
. T
h
e
i
n
crease i
n
si
z
e
and l
e
vel
of
penet
r
at
i
on
of t
h
e wi
nd
farm
s has m
a
jor i
m
pact
on
th
e
p
o
wer syste
m
in
clu
d
i
n
g
th
e
un
it co
mmit
m
en
t and
lo
ad
di
spat
c
h
. It
has
i
m
pact
on
t
h
e vol
t
a
ge
of
t
h
e f
eede
r
an
d th
e
sub
s
tatio
n
wh
en a lo
cal p
o
wer system
is co
n
s
i
d
ered
,
[1
]-[2
].
Th
e in
crease i
n
th
e
growth
an
d
i
n
stalled
cap
acity o
f
wind farm
s
m
a
k
e
s it n
ecessary to in
v
e
stig
at
e
an
d
d
e
v
e
l
o
p differen
t
co
n
t
rol cap
ab
ilities an
d strateg
i
es
. W
i
t
h
th
e n
ecessary co
n
t
ro
l tech
n
i
q
u
e
s, th
e wind
farm
s can
b
e
mad
e
with
th
e sam
e
featu
r
es as th
at of con
v
e
n
tion
a
l power
p
l
an
ts. These cap
ab
ilitie
s and
strategies includes the c
o
ntrol
of
out
put active powe
r an
d
reactiv
e p
o
wer of th
e to
tal wi
n
d
farm
, th
e co
n
t
ro
l
of
affect
o
f
pa
ram
e
t
e
rs of
wi
n
d
f
a
rm
for any
ki
nd
of a
b
no
rm
al con
d
i
t
i
on i
n
t
h
e gri
d
. Thi
s
m
a
kes t
h
e
wi
n
d
p
o
we
r
plants t
o
act as
active c
o
m
ponents
of t
h
e
power s
u
pply system
.
Th
e pu
rpo
s
e an
d
th
e m
a
in
ai
m o
f
th
ese contro
l cap
ab
ilitie
s are to
op
ti
m
i
ze th
e g
e
n
e
rated
po
wer and
to
m
a
n
a
g
e
th
e
safe lim
i
t
s alo
n
g
with au
to
matic o
p
e
ration
o
f
th
e
wind tu
rb
in
e. Th
e
op
timizatio
n
o
f
g
e
n
e
rated
p
o
wer redu
ces
th
e op
eratin
g co
sts, en
ab
les
un
ifo
r
m
d
y
n
a
m
i
c resp
on
se and b
e
tter
p
r
od
u
c
t
q
u
a
lity, and
h
e
lp
s to
m
i
nim
i
ze t
u
rbi
n
e l
o
a
d
s
w
h
i
c
h m
a
ke safe
o
p
erat
i
o
n
of
wi
nd
t
u
r
b
i
n
e wi
t
h
m
a
xim
i
zi
ng
t
h
e capt
u
re
o
f
wi
n
d
energy.
W
i
t
h
t
h
e
ai
m
of e
x
t
r
act
i
n
g t
h
e m
a
xim
u
m
wi
n
d
po
wer
a
ppl
i
e
d
t
o
fi
xe
d
or
va
ri
abl
e
spee
d wi
n
d
t
u
r
b
i
n
es,
m
a
ny co
nt
r
o
l
st
rat
e
g
i
es were
de
vel
ope
d
by
s
o
m
a
ny
aut
h
o
r
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
37
9 – 3
9
0
38
0
Th
e
p
e
rfo
r
m
a
n
ces lik
e stab
ility an
alysis and
tu
rb
in
e con
t
ro
l to
im
p
r
ov
e t
h
e po
wer
g
e
n
e
rat
e
d
with
t
h
e
l
o
west
im
pact
on t
h
e g
r
i
d
vol
t
a
ge a
nd f
r
e
que
ncy
d
u
ri
n
g
n
o
rm
al
oper
a
t
i
on an
d u
n
d
e
r several
ab
n
o
rm
al
co
nd
itio
ns,
su
ch
as a tran
sm
i
ssio
n
lin
e eart
h
fau
lt are
d
e
scrib
e
d
in th
is
p
a
p
e
r, fo
r wi
nd
g
e
n
e
ration
m
o
d
e
ls
ope
rat
e
d
wi
t
h
i
n
t
h
e
po
we
r sy
st
em
. The
wi
n
d
t
u
r
b
i
n
es
ba
se
d
on
Do
u
b
l
y
F
e
d
In
d
u
ct
i
o
n
G
e
nerat
o
r
(
D
F
I
G)
an
d
a gri
d
-c
o
nnect
ed c
o
n
v
ert
e
r
wi
t
h
c
onst
a
nt
or
vari
a
b
l
e
bl
ade a
ngl
e a
p
p
l
i
e
d t
o
va
ri
abl
e
spee
d wi
nd
ener
gy
sy
st
em
i
s
consi
d
ere
d
i
n
t
h
i
s
p
a
per
.
T
h
e m
e
t
hod
ol
o
g
y
i
s
i
n
v
e
st
i
g
at
ed a
n
d
i
m
pl
em
ent
e
d b
y
usi
n
g t
h
e t
o
o
l
bo
xe
s
and
bl
ock
set
s
of
M
A
TLa
b/
Si
m
u
li
nk
[3]
.
The e
ngi
neeri
n
g an
d
desi
g
n
o
f
a D
o
ubl
y
Fe
d I
n
duct
i
o
n
Ge
nerat
o
r
(D
FI
G
)
usi
n
g
AC
/
D
C
/
AC
P
W
M
vol
t
a
ge s
o
u
r
ce
con
v
ert
e
rs i
n
t
h
e rot
o
r ci
rcui
t
and vect
o
r
cont
rol
ap
pl
i
e
d t
o
bot
h Gri
d
Si
de C
o
n
v
ert
e
r (GS
C
)
an
d Ro
tor Si
de Co
nv
erter
(RSC) withou
t con
s
id
ering
an
y abno
rm
al c
o
nd
itio
n is d
e
scrib
e
d
in [4
]. [5
] is
restricted t
o
t
h
e reactive
power c
ont
rol
of a
wind
farm
with
th
e im
p
l
e
m
e
n
tatio
n of
P
I
c
ont
rol
a
n
d
pr
o
p
o
r
t
i
o
nal
di
st
ri
b
u
t
i
o
n
,
c
oncl
udi
ng
t
h
at
PI i
s
m
o
re r
o
bust
c
o
m
p
ar
ed to
pr
opo
r
tional d
i
str
i
bu
tio
n. Th
e
v
ector
con
t
ro
l is
ap
p
lied to
GSC [6
], in th
is su
p
e
rv
isory syste
m
s co
n
t
rols
active and rea
c
tive powe
r
of the whole wi
nd
farm
and a m
achine control system ensures
that
set points at the wind turbine
level are reached. The a
n
alysis of
st
at
or fl
ux
ori
e
nt
ed
vect
o
r
co
nt
r
o
l
t
h
eo
ry
o
f
Do
u
b
l
y
Fed I
n
duct
i
o
n
Gene
ra
t
o
r (
D
F
I
G
)
wa
s i
n
t
r
o
d
u
ced i
n
[7]
.
A
M
a
xi
m
u
m
Power P
o
i
n
t
Trac
ki
n
g
(M
PPT
) and c
ont
rol
st
r
a
t
e
gy
for t
h
e
u
n
i
t
y
powe
r
fac
t
or o
f
st
at
or o
u
t
p
ut
i
s
descri
bed
.
I
n
[
8
]
,
agai
n t
h
e s
t
at
or fl
u
x
ori
e
nt
ed
fram
e is
introduced
for easy
decom
posi
t
i
on
of act
i
v
e an
d
reactive
powe
rs on t
h
e stat
or side
along wi
th introduc
ti
o
n
of lin
ear
PI co
n
t
ro
l aro
und
th
e stator cu
rren
t
s
en
su
ring
g
ood
stab
ility p
r
o
v
i
ded
th
e
g
a
in
s are su
itab
l
y selected
. A
robu
st reg
u
l
ation
of the p
r
im
ary sid
e
activ
e
and reacti
v
e powe
r by the com
ponents of
the stator current
s
is
introduced in [9] and is
synthesized by
using
PI c
ontroller.
In all the m
e
ntione
d lite
rature the
traditional m
e
thod of vect
or
c
o
ntrol is
us
ed with PI c
o
nt
roller
an
d th
e
p
e
rfo
r
man
ce is restri
cted
to
stead
y state an
d
no
t analyzed
fo
r t
r
ansien
t or
g
r
i
d
fau
lt con
d
ition
s
.
The c
o
nt
r
o
l
o
f
r
o
t
o
r
c
u
r
r
ent
s
by
v
o
l
t
a
ge m
ode
c
ont
r
o
l
l
e
r
by
u
s
i
n
g t
h
e
v
ect
or c
o
nt
r
o
l
s
c
hem
e
[10]
,
wh
ich
allows
fo
r adju
stab
le sp
eed
op
eratio
n and
reactiv
e
po
wer con
t
ro
l alo
n
g
with
co
rrectin
g
the
p
r
ob
lem
s
cause
d by
un
b
a
l
a
nced st
at
or
vol
t
a
ge
s. The
r
eact
i
v
e po
wer
gene
rat
i
o
n of
of
fsh
o
r
e wi
n
d
par
k
s usi
ng
D
F
IG
s i
s
expl
ai
ne
d i
n
[1
1]
and i
t
desc
ri
bes t
h
e st
eady
st
at
e operat
i
o
n
wi
t
h
m
i
nim
u
m po
wer l
o
ss o
f
wi
n
d
ene
r
gy
b
y
ne
w
ad
ap
tiv
e Particle Swarm
Op
timizatio
n
(PSO) an
d also
d
e
scrib
e
s th
e con
d
i
tio
n
o
f
fau
lt. Th
e
fau
lt con
d
itio
n i
s
av
o
i
d
e
d
b
y
th
e fau
lt rid
e
throu
g
h
capab
ility
b
y
u
s
ing
cr
owb
a
r
p
r
o
t
ectio
n, wh
ich
b
l
ock
s
t
h
e RSC and
the stato
r
reactive curre
nt cannot be controlled. T
h
e cont
rol of
reactive powe
r inde
pende
n
tly
of the active powe
r flow
usi
n
g b
o
t
h
R
S
C
and GSC
wi
t
h
ap
pl
i
cat
i
on o
f
l
i
n
ear co
nt
r
o
l tech
n
i
qu
es and
it is n
o
t
an
alyzed for the grid fault
co
nd
itio
n [12
]
.
Tw
o di
ffe
re
nt
cont
rol
st
rat
e
gi
es i
.
e., PI an
d Art
i
f
icial Neural Netwo
r
k
(ANN) and
th
e op
eration
of
the DFIG is analyzed in [13], during
the
steady state and tran
si
e
n
t
g
r
i
d
f
a
ul
t
o
p
erat
i
o
n.
The
pape
r
des
c
ri
be
s
the replacem
ent of traditional
m
e
thod of
vect
or control by ANN, the
ANN co
ntrollers’
architecture us
ed is of
conve
n
tional t
y
pe, which ca
n be re
placed by
m
o
re
sophisticated and highly accurate Graphical User
Interfaces
(GUIs)
for
better i
m
provem
e
nt in the
pe
rfor
m
a
nce.
In [13], by using
ANN, there a
r
e still som
e
o
v
e
rsh
o
o
t
s, faster respon
se
co
m
p
ar
ed
t
o
PI con
t
ro
ller
an
d ex
isten
c
e of
o
s
cillato
ry
b
e
h
a
v
i
o
r
d
u
rin
g
the
tran
sien
t or abn
o
rm
al co
n
d
itio
n. Th
is can
be av
o
i
d
e
d
b
y
usin
g
th
e
Hyb
r
i
d
con
t
ro
ller and
th
e p
e
rfo
r
m
a
n
ce of
the DFIG ca
n
be m
u
ch im
proved.
In t
h
i
s
pa
per
,
t
h
ree
p
o
we
r c
o
nt
r
o
l
t
echni
que
s are i
n
ve
stigat
ed and t
h
ey are
:
Pr
opo
rtion
a
l In
tegral (PI)
co
n
t
ro
ller
b
a
sed
tech
n
i
qu
e; Artificial Neu
r
al
Networ
k (A
N
N
)
based t
e
c
h
n
i
que an
d c
o
m
b
i
n
at
i
on
of A
N
N
an
d
PI c
o
nt
r
o
l
t
ech
ni
q
u
e
fo
r
bet
t
e
r
dy
nam
i
c cont
rol
of
DFI
G
.
I
n
t
h
i
s
w
o
rk
as
m
e
nt
i
oned
,
a
n
e
w
hy
b
r
i
d
co
nt
rol
l
e
r
i
s
i
n
t
r
od
uce
d
,
whi
c
h c
o
nsi
s
t
s
o
f
bot
h
AN
N
and
P
I
c
ont
rol
l
ers t
h
at
m
eans bot
h t
h
e a
dva
n
t
ages
of
A
N
N
and
P
I
cont
rol
l
e
rs a
r
e
em
bedde
d i
n
t
o
o
n
e si
n
g
l
e
cont
rol
l
e
r. C
o
m
p
ared t
o
t
h
e
per
f
o
r
m
a
nce of t
h
e
AN
N a
nd P
I
co
n
t
ro
ller, t
h
e h
y
b
r
i
d
con
t
roller syste
m
est
i
m
ates
th
e co
ntro
l p
a
ram
e
ters o
f
th
e DFIG
m
o
re p
r
o
m
in
en
tly,
whi
c
h can be
easi
l
y
under
s
t
o
od
by
t
h
e sat
i
s
fact
ory
pe
rf
o
r
m
a
nce charact
eri
s
t
i
c
s of DF
I
G
pre
s
ent
e
d i
n
resul
t
s
and
di
scu
ssi
o
n
sect
i
on an
d a
l
so d
u
ri
ng t
h
e
t
r
ansi
ent
c
o
n
d
i
t
i
on t
h
e
hy
b
r
i
d
co
nt
r
o
l
l
e
r
prese
n
t
s
ab
sen
ce or
sm
a
ller o
v
e
rsho
o
t
s, th
e system
m
a
k
e
s to
sta
b
le op
eration
in
lesser tim
e, a
n
d
sm
aller o
s
cillato
ry b
e
h
a
v
i
or and
th
is can
b
e
ob
serv
ed
ev
en
du
rin
g
th
e
fau
lt con
d
ition
.
To
en
su
re
b
e
tter stab
i
lity an
d
p
o
wer
regu
latio
n
g
e
nerated
b
y
th
e
wind
t
u
rb
in
e t
h
e
Hybrid
techn
i
qu
e
wi
ll allo
w th
e
in
crease
o
f
ro
bu
st
n
e
ss, p
e
rform
a
n
ce, and
flex
ibilit
y.
2.
MODELING OF DFIG
In
th
is research
wo
rk
, th
e
o
v
erall
control struct
ure of the
m
odel that
is
used includes a
e
rodynam
i
c,
m
echani
cal
an
d el
ect
ri
c sy
st
e
m
co
m
pone
nt
s,
wi
t
h
an
o
v
e
r
al
l
sy
st
em
of c
o
n
t
rol
m
odel
.
T
h
e el
ect
ri
c m
odel
an
d
the aerodynamic
m
odel interface w
ith t
h
e
grid m
odel and t
h
e
wind
m
odel res
p
ectively, [1]-[2].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A New Hyb
r
id
Artificia
l Neu
r
a
l
Netwo
r
k Based
C
o
n
t
ro
l
o
f
Doub
ly Fed
Ind
u
c
tion
Gen
e
ra
to
r
(
G
. V
.
M
a
dh
av)
38
1
2.
1.
Doubl
y F
e
d Ind
u
cti
o
n
Generator
Th
e
DFIG m
o
d
e
l ado
p
t
ed
in
th
is research
pap
e
r is
the
q
d
0
stationary
r
o
ta
ting re
fere
nce
fram
e
. It is
because the model sim
u
latio
n of
Doubly
Fed Induction Machine (DFIM) is
quite suitable with this fram
e
of
refe
rence
duri
ng tra
n
sie
n
ts. The transient solution
of the DFIM
m
odel is possible because
of the
t
r
ans
f
o
r
m
a
ti
on
fr
om
abc t
o
qd
0
by
w
h
i
c
h t
h
e
di
f
f
ere
n
t
i
a
l
equat
i
ons
wi
t
h
t
i
m
e-vary
i
ng i
n
d
u
ct
an
ces i
s
co
nv
erted
i
n
to
d
i
fferen
tial eq
uatio
n
s
with
constan
t
ind
u
c
tan
c
es, [5
]-[6
]
.
Ψ
Ψ
(1
)
Ψ
Ψ
(2
)
′
Ψ
Ψ
(3
)
′
Ψ
Ψ
(4
)
3
2
2
(5
)
In
t
h
is
p
a
p
e
r, t
h
e
p
e
rform
a
n
ce and
th
e stab
il
ity o
f
th
e
DFIG system
is v
e
rified
fo
r t
r
an
si
en
t cond
itio
n
state whe
n
active and
reactive
powe
rs are
va
ried re
spectiv
e
l
y and also
ve
rified whe
n
a phase to ea
rth
fault is
o
ccurred
,
wh
en
DFIG is con
t
in
uo
usly co
nn
ected
to
th
e g
r
i
d
syste
m
an
d
is co
n
tinuo
u
s
ly
fed
with
m
ech
an
ical
in
pu
t torqu
e
.
2.
2. A
C
/
D
C/
A
C
Co
nver
ter
B
y
usi
ng t
h
e AC
/
D
C
/
AC
co
nve
rt
er,
bi
di
re
ct
i
onal
po
we
r fl
o
w
can be ac
hi
eve
d
fr
om
t
h
e gri
d
t
o
t
h
e
DFIM
and from
the DFIM t
o
t
h
e
gri
d
that
means from
the ro
tor
o
f
th
e
DFIM
, th
e cu
rren
t can b
e
injected
o
r
receive
d
with t
h
e c
u
rrent
flowing in bo
th
the directions [5].
Nom
i
nally
30
to
40%
of t
h
e
gene
rated powe
r is
f
e
d
b
ack
t
o
the g
r
id
t
h
ro
ugh th
e PW
M b
a
ck
-
t
o
-
b
ack
con
v
e
r
t
er
an
d
al
so
b
y
em
p
l
o
y
in
g
t
h
e b
a
ck
-
t
o-
b
a
ck
con
v
e
r
t
e
r t
h
e f
l
uct
u
at
i
o
ns i
n
vol
t
a
ge
, f
r
eq
ue
ncy
,
an
d
wi
n
d
t
u
r
b
i
n
e
out
put
po
we
r can
be e
a
si
l
y
cont
rol
l
e
d. T
h
e
out
put
p
o
we
r of
D
F
I
G
, w
h
i
c
h has fl
uct
u
at
i
n
g
n
o
n
-
g
ri
d fr
eque
ncy
ca
n b
e
co
nt
r
o
l
l
e
d b
y
P
W
M
c
o
nve
rt
er
t
o
a
DC Vo
ltag
e
lev
e
l and
it will b
e
in
j
ected
th
ro
ugh
AC/DC/AC co
nv
ert
e
r in
to
three-ph
ase grid
frequ
en
cy
po
we
r, [4]
.
Thi
s
P
W
M
ba
ck-t
o-
back c
o
n
v
ert
e
r
has a co
m
b
i
n
at
i
on of t
w
o c
o
n
v
e
r
t
e
rs,
one i
s
cal
l
e
d as Gri
d
Si
d
e
Converte
r (GSC) and the
other one is called as Rotor Si
de
Con
v
erte
r (RS
C
), as Fi
gure 1. The m
a
in objective
of t
h
e
GSC
i
s
t
o
kee
p
t
h
e
DC
vol
t
a
ge
of t
h
e DC
-l
i
n
k capaci
t
o
r t
o
be co
nst
a
nt
, i
n
depe
n
d
e
n
t
of t
h
e di
rect
i
on
of
t
h
e p
o
we
r
fl
o
w
o
f
r
o
t
o
r, i
n
or
der t
o
gua
ra
nt
ee t
h
e e
x
cha
nge
o
f
react
i
v
e
po
we
r o
n
l
y
t
h
ro
u
gh t
h
e st
at
o
r
as by
main
tain
in
g
t
h
e po
wer
factor
to
b
e
equ
a
l t
o
u
n
ity.
Lik
e
wise, th
e
main
o
b
j
ectiv
e o
f
t
h
e RSC is
to
ach
i
e
ve i
n
di
rect
co
nt
r
o
l
o
f
active and reac
tive powe
r
i
nde
pen
d
e
n
t
l
y
by
co
nt
r
o
l
ove
r
t
h
e
rot
o
r c
u
rre
nt
,
[4]
.
3.
PI CONT
ROLLER BASE
D
DFIG
In
o
r
de
r t
o
opt
im
i
ze t
h
e ope
r
a
t
i
on
of
t
h
e
wi
nd
t
u
rbi
n
e l
i
k
e
w
i
s
e
num
ber
o
f
wi
n
d
t
u
rbi
n
e
s
i
n
a
wi
nd
p
a
rk
, it is n
ecessary to
con
t
rol th
e activ
e an
d
reactiv
e
p
o
wer flows
with
a co
n
s
i
d
eration
th
at th
e wind
tu
rb
ine
will g
e
n
e
rate
max
i
m
u
m
o
u
t
pu
t power to
t
h
e grid
. Thu
s
, an
ad
equ
a
te con
t
ro
l sh
ou
ld h
a
v
e
[4
]:
1
.
Ach
i
ev
ab
le
op
t
i
m
i
zed
o
p
e
ratin
g po
in
t
b
y
con
t
ro
l of activ
e
p
o
wer su
pp
lied b
y
th
e wi
n
d
turb
i
n
e;
2.
Restricted ge
neration of active po
we
r w
h
en
wi
n
d
s
p
ee
d
i
s
very
hi
g
h
;
3.
C
ont
r
o
l
l
e
d i
n
t
e
rcha
n
g
e o
f
r
eact
i
v
e po
we
r
bet
w
ee
n t
h
e
gene
rat
o
r an
d
t
h
e gri
d
w
h
e
r
e i
n
ge
ne
ral
t
h
e
flu
c
tu
ation
i
n
vo
ltag
e
o
c
cu
rs esp
ecially
in
weak
g
r
i
d
s;
4.
C
onst
a
nt
g
r
i
d
vol
t
a
ge
i
n
case
o
f
a
wi
n
d
par
k
o
r
wi
n
d
t
u
rbi
n
e;
5.
M
i
nim
i
zed cos
t
s of
m
a
i
n
t
e
nance a
n
d
t
h
e
ex
pl
o
r
at
i
o
n
i
n
ca
se o
f
wi
n
d
par
k
or
wi
nd
t
u
r
b
i
n
e.
Seve
ral
co
nt
r
o
l
st
rat
e
gi
es
we
r
e
p
r
o
p
o
sed
i
n
or
der
t
o
ach
ieve th
e
po
in
ts m
e
n
tio
n
e
d abov
e. Th
e
v
ector
co
n
t
ro
l strategy is ap
p
lied
to
DFIG in
o
r
d
e
r to
ach
ie
ve t
h
e
i
ndepe
n
d
e
n
t
cont
rol
o
f
act
i
v
e and react
i
v
e
po
we
r
[4]
-
[
5]
, by
dec
o
u
p
l
i
n
g t
h
e rot
o
r cu
rre
nt
i
n
t
o
qua
drat
u
r
e ‘
q
’
and di
rect
‘d
’ axi
s
com
pone
n
t
s. In t
h
i
s
pa
pe
r, t
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
37
9 – 3
9
0
38
2
vect
o
r
co
nt
r
o
l
st
rat
e
gy
i
s
des
c
ri
be
d wi
t
h
PI
co
nt
rol
l
e
r
,
A
NN c
o
nt
r
o
l
an
d Hy
bri
d
c
o
m
b
i
n
at
i
o
n i
.
e.
,
AN
N+
P
I
cont
rol a
n
d re
s
u
lts are c
o
m
p
ared.
A
typ
i
cal
config
uration
o
f
p
r
op
o
s
ed
Hybrid ANN with
PI co
n
t
ro
l
b
a
sed
DFIG is illu
strat
e
d
in Figu
re
1.
In
pl
ace
o
f
Hy
bri
d
AN
N
wi
t
h
P
I
c
o
nt
ro
l
,
an i
ndi
vi
d
u
a
l
con
f
i
g
urat
i
o
n
of
Ne
u
r
al
Net
w
o
r
k
s
(
N
N)
a
n
d
P
I
co
n
t
ro
llers are e
m
p
l
o
y
ed
and
th
e ob
tain
ed si
m
u
latio
n
resu
lts for th
ese co
nfigu
r
ations are com
p
ared and
tabulated.
As s
h
o
w
n i
n
Fi
gu
re
1, t
h
e o
p
t
i
m
al
referen
ce
val
u
e
of
act
i
v
e
po
we
r i
s
gi
ve
n by
Pre
f
,
whi
c
h i
n
ge
neral
obtaine
d
from
the cha
r
acterist
i
cs of
the
wind turbine
and t
h
e op
tim
a
l value of reacti
v
e
powe
r
Qre
f
is c
hos
en
su
ch
th
at at t
h
e grid
con
n
ection
p
o
i
n
t
th
e
vo
ltag
e
is m
a
in
tain
ed
stab
le.
Fi
gu
re
1.
Pr
o
p
o
se
d c
ont
r
o
l
st
rat
e
gy
3.
1.
Gri
d
-Si
d
e
Converter
Control
The vect
or c
o
nt
r
o
l
st
rat
e
gy
i
s
appl
i
e
d t
o
t
h
e GSC
t
o
achi
e
ve t
h
e i
n
de
pe
nde
nt
co
nt
r
o
l
of act
i
v
e a
n
d
reactive powe
r between the c
o
nv
erter and
th
e g
r
i
d
. The strateg
y
is u
s
ed
with
resp
ect to
th
e p
o
s
itio
n
of th
e
st
at
or or
g
r
i
d
r
e
fere
nce vol
t
a
g
e
axi
s
. Th
e typical co
nf
igu
r
ati
o
n of
G
S
C contr
o
l str
a
teg
y
is
sh
own
in Fi
g
u
re 2
.
As s
h
ow
n i
n
F
i
gu
re
2, t
h
e P
I
co
nt
r
o
l
l
e
rs ar
e ad
opt
e
d
fo
r t
h
e i
n
t
e
rnal
c
o
nt
r
o
l
l
o
op
o
b
t
a
i
n
ed
by
t
h
e
Laplace tra
n
sform
of (6) a
nd (7) that
re
pres
ents the
GSC
voltage
s in its
dq
c
o
m
pone
nts
,
[4]. In this re
search
work, the
gene
ralized PI c
o
ntrollers
a
r
e re
pl
aced by
norm
a
lized cont
rolle
rs like ANN a
n
d Hy
brid c
o
ntrollers
and the
pe
rformances are
obs
e
rve
d
.
(6
)
(7
)
Applying the
L
a
place tra
n
sform
s
to the above two equations,
(8
)
(9
)
C
o
n
s
id
e
r
in
g
,
(1
0)
(1
1)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISS
N
:
2088-8708
A New Hyb
r
id
Artificia
l Neu
r
a
l
Netwo
r
k Based
C
o
n
t
ro
l
o
f
Doub
ly Fed
Ind
u
c
tion
Gen
e
ra
to
r
(
G
. V
.
M
a
dh
av)
38
3
In
o
r
de
r t
o
desi
gn
t
h
e c
u
rre
nt
cont
rol
l
o
o
p
s t
h
e
fol
l
o
wi
n
g
e
quat
i
o
ns
can
b
e
use
d
[4]
:
1
(1
2)
Sub
s
titu
tin
g
(10
)
, (11) in
(8), (9
)
resp
ecti
v
ely, an
d
b
e
i
n
g
v
q
(s) = 0 the r
e
fere
nce f
o
r th
e voltage
s
val
u
es
v
q_ref
and
v
d_ref
ca
n be obt
ai
ne
d by
:
_
(1
3)
_
(1
4)
The val
u
es of
v
q_ref
and
v
d_ref
are conside
r
ed to be the refe
ren
ce in
pu
t v
a
l
u
es to
the PWM converte
r
by
whi
c
h l
e
vel
of DC
v
o
l
t
a
ge
and re
qui
red
p
o
we
r fact
o
r
i
s
achi
e
ve
d. I
n
t
h
e sim
i
l
a
r way
,
t
h
e cont
rol
des
i
gn o
f
the RSC ca
n
be obtained as t
h
e sam
e
way c
onsi
d
ere
d
for t
h
e
GSC.
Fig
u
re
2
.
Grid-sid
e co
nv
erter
co
n
t
ro
l st
rategy
3.
2.
Rotor
-
Si
de Con
v
er
ter
Contr
o
l
Th
e sy
n
c
hr
onou
s
r
o
tating
r
e
fer
e
n
c
e dq
f
r
a
me is u
s
ed
i
n
th
e RSC co
n
t
ro
l of
D
F
IG
w
i
th
th
e
d
-
ax
is
al
i
gned
wi
t
h
t
h
e st
at
or fl
u
x
p
o
s
i
t
i
on. The t
y
pi
cal
confi
g
u
r
at
i
on
of t
h
e R
S
C
cont
rol
st
rat
e
g
y
i
s
show
n i
n
Fi
gu
re
3
.
By setting up
th
e referen
ce
fram
e
, deco
up
ling
b
e
t
w
een th
e electro
m
ag
n
e
tic to
rqu
e
an
d ro
to
r t
h
e
m
a
gnet
i
z
i
ng c
u
r
r
ent
i
s
obt
ai
ned
.
B
y
t
h
e c
a
l
c
ul
at
ed val
u
es of t
h
e rotor and the stat
or c
u
rre
n
ts, the stator
v
o
ltag
e
and
t
h
e ro
to
r po
sition
,
th
e co
n
t
ro
l
of
th
e PW
M co
nverter
fed to
t
h
e
g
e
n
e
rator
ro
t
o
r is ob
tain
ed
.
As s
h
o
w
n i
n
Fi
gu
re
3, t
h
e
PI c
ont
r
o
l
l
e
rs
are em
pl
oy
ed fo
r t
h
e c
o
nt
rol
of t
h
e R
S
C
o
f
t
h
e
DF
I
G
,
likewise, in
place of PI cont
rollers
the
ANN and the
Hybri
d
controllers
are em
ploye
d
and the re
sults are
descri
bed
.
T
h
e
dq
re
fere
nt
i
a
l
r
o
t
o
r
vol
t
a
ge
ca
n
be
obt
ai
ne
d
f
r
om
(1
5)
an
d
(
1
6
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
37
9 – 3
9
0
38
4
(1
5)
(1
6)
A
s
sh
own
in
Fig
u
r
e
3, th
e
vo
l
t
ag
es
v
r
d
and
v
r
q
resp
ectiv
ely are ob
tain
ed
b
y
ap
p
l
ying
th
e err
o
r
sign
als
i
r
d
and i
r
q
to the PI con
t
ro
ller. To
ob
tain
t
h
e
referen
c
e vo
ltag
e
s
v
r
d_re
f
and
v
r
q_ref
,
[4], t
h
e
decoupling term
s are
adde
d t
o
(15),
(16) res
p
ective
l
y in
orde
r to c
o
m
p
ensate the
cont
rol, th
e e
q
uations in accorda
n
ce a
r
e
give
n
by
_
(1
7)
_
(1
8)
Fi
gu
re
3.
R
o
t
o
r
-
si
de c
o
nve
rt
er
co
nt
r
o
l
st
rat
e
g
y
4.
HYBRID AND
ANN
CONTROLLER B
A
SED DFIG
Th
e n
e
w con
t
ro
l syste
m
wil
l
u
s
e a Hyb
r
id
and
NN to
su
bstitu
te so
me b
l
o
c
k
s
o
f
the trad
itio
n
a
l
sy
st
em
of
vect
or
co
nt
r
o
l
,
[1
4]
.
Th
us, i
t
i
s
i
n
t
e
nde
d t
o
p
r
ese
n
t
a cont
rol
sy
st
em
based on
h
y
b
ri
d ne
u
r
al
n
e
t
w
o
r
ks t
o
be
use
d
i
n
t
h
e
co
n
t
ro
l system altern
ativ
ely to
th
e
o
n
e
b
a
sed
on
PI c
o
n
t
rollers, with
t
h
e
in
ten
tio
n of efficien
tly ex
tract th
e
wind
en
erg
y
,
i.e. to
b
e
ab
l
e
to
ex
tract th
e m
a
x
i
m
u
m
p
o
wer o
f
th
e
tu
rb
in
e du
ri
n
g
so
m
e
situ
atio
n
s
of
fun
c
tion
i
ng
thro
ugh
th
e estimatio
n
o
f
th
e con
t
ro
l p
a
ram
e
ter
s
fo
r t
h
e
g
r
i
d
-si
d
e an
d ro
tor-si
d
e co
nv
erters.
4.1. Archi
t
ecture
of Neur
al
Netw
o
r
k (N
N)
C
o
ntr
o
ller
In M
A
TLa
b/
Si
m
u
li
nk
Neu
r
al
Net
w
or
k T
o
o
l
bo
x s
o
ft
wa
re
can
be use
d
i
n
fo
ur
di
f
f
ere
n
t
way
s
vi
z.
,
Graph
i
cal User In
terfaces (GUIs),
b
a
sic comman
d
lin
e op
eration
s
,
cu
st
o
m
iza
tio
n
and th
e ab
ility to
m
o
d
i
fy
any
o
f
t
h
e
f
u
n
c
t
i
ons c
o
nt
ai
ne
d i
n
t
h
e t
ool
b
o
x
,
GU
Is i
s
us
ed f
o
r
neu
r
al
net
w
or
k t
r
ai
ni
ng
i
n
t
h
i
s
pa
p
e
r. T
h
e
task
s lik
e
fu
n
c
tio
n
fittin
g
,
p
a
ttern
recog
n
iti
o
n
,
d
a
ta clu
s
tering
an
d ti
m
e
series an
alysis can
b
e
qu
ick
l
y an
d
easily accessed with
GUIs;
of these tasks
mentioned t
h
e
neural net
w
ork function fitting tool is accessed by
GUI to
d
e
fin
e
th
e NNs in
th
is p
a
p
e
r. In
th
e fun
c
tio
n
fittin
g
th
e stand
a
rd
NN th
at is u
s
ed
is a two
-
layer feed
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A New Hyb
r
id
Artificia
l Neu
r
a
l
Netwo
r
k Based
C
o
n
t
ro
l
o
f
Doub
ly Fed
Ind
u
c
tion
Gen
e
ra
to
r
(
G
. V
.
M
a
dh
av)
38
5
fo
rwa
r
d net
w
o
r
k
,
wi
t
h
a si
g
m
oi
d t
r
ansfer
fu
nct
i
o
n i
n
t
h
e
hi
dde
n l
a
y
e
r
and a l
i
n
ear t
r
ansfe
r
f
unct
i
on
i
n
t
h
e
out
put layer. For all the NNs
whic
h are re
pl
aced in place of
PI controllers
in all the three cases described in
the
paper, the
NNs are trai
ne
d
by the
data set which is taken
from
each i
ndi
vidual input and
out
put
of a PI
cont
roller.
The l
ear
ni
n
g
al
go
ri
t
h
m
upd
at
es t
h
e val
u
e
s
of t
h
e wei
g
ht
s an
d bi
as
val
u
es acc
or
di
ng t
o
t
h
e
Leve
nbe
rg
-M
a
r
q
u
ar
dt
B
ack
Pro
p
a
g
at
i
o
n (
B
P) al
g
o
ri
t
h
m
wi
t
h
si
gm
oi
d
hi
d
d
en
ne
ur
o
n
s an
d l
i
n
ea
r
out
put
n
e
uro
n
s
as t
h
e
activ
atio
n
fu
n
c
tio
n
s
u
s
ed
i
n
each
layer.
Th
e
in
pu
t-targ
et
v
e
cto
r
s will b
e
ran
d
o
m
ly d
i
v
i
d
e
d
i
n
t
o
th
ree sets. Th
e first set o
f
70
% will b
e
used
fo
r trai
n
i
ng th
e p
a
ttern
s,
secon
d
set of
1
5
% will b
e
used
to
v
a
lid
ate th
at th
e NN is g
e
n
e
ralizin
g
and
to
sto
p
train
i
ng
before ov
er
fitti
n
g
an
d
the th
ird
set o
f
15
%
will b
e
u
s
ed
as a co
mp
letely in
d
e
p
e
n
d
e
n
t
test of
NN
g
e
n
e
ralizatio
n
.
For all the NNs th
e trai
n
i
ng
is con
tinued
un
til
th
e v
a
lid
ation
ch
ecks are reach
e
d
t
o
six
iteratio
n
s
i.e, th
e
v
a
lid
ation
error failed
to
d
e
crease for six
iteratio
ns.
After
perform
ing all the training step
s the NN which is cre
a
ted is use
d
for the re
placement of c
o
rres
pondi
ng
PI con
t
ro
ller an
d sim
i
lar p
r
o
c
ed
ure is ado
p
t
ed
to ob
tain
o
t
her
NNs.
For t
h
e r
o
t
o
r-s
i
d
e co
nt
r
o
l
l
e
r was u
s
ed a
1-
40
-
1
ne
ural
ne
t
w
o
r
k c
o
n
f
i
g
ur
at
i
on,
whe
r
e t
h
e i
n
put
s a
r
e
the stator and rotor curre
n
ts
, the rotor-side
refere
nce curre
nts and the
rot
o
r a
n
g
u
l
a
r s
p
e
e
d, an
d t
h
e o
u
t
put
s are
the re
fere
nce
v
o
ltages t
o
c
ont
rol t
h
e r
o
to
r-
side c
o
n
v
e
r
ter.
For t
h
e
gri
d
-si
d
e co
nt
rol
l
e
r
w
a
s used a 1-
4
0
-
1
ne
ur
al
net
w
o
r
k co
n
f
i
g
urat
i
o
n,
where the inputs are the
st
at
or
vol
t
a
ges
an
d cu
rre
nt
s,
t
h
e g
r
i
d
-si
d
e
r
e
fere
nce c
u
r
r
e
n
t
s
, a
n
d
t
h
e a
n
gul
a
r
f
r
eq
ue
nc
y
ω
s, an
d t
h
e
out
put
s
are the
re
fere
nce voltages t
o
cont
rol t
h
e
gri
d
-si
d
e c
o
nve
r
ter.
5.
RESULTS
A
N
D
DI
SC
US
S
I
ON
5.
1. Re
acti
ve Pow
er Co
ntr
o
l
The
pr
o
pose
d
cont
rol
st
rat
e
g
i
es have
bee
n
im
pl
em
ent
e
d u
s
i
ng M
A
TLa
b/
Sim
u
l
i
nk a
nd
resul
t
s
a
r
e
p
r
esen
ted. In
th
is sectio
n, th
e p
e
rfor
m
a
nces of the c
o
ntrollers are a
n
alyzed
fo
r t
h
e va
ri
at
i
on
of
react
i
v
e
po
we
r.
The
reactive
powe
r is
va
ried with step c
h
a
nge
in val
u
e
f
r
o
m
-
0
.1
6p
.u
. to
-
0
.04p
.u
ap
plied
at 0.03
secs up
t
o
0.
13s
ecs as
sh
ow
n i
n
Fi
gu
re
6(
b)
. T
h
e
GSC
an
d R
S
C
’
s re
f
e
rence
di
rect
a
n
d
q
u
a
d
rat
u
re
axi
s
v
o
l
t
a
ges
f
o
r t
h
e
PI (
r
ed i
n
col
o
r),
N
N
(
g
ree
n
in
col
o
r)
, an
d
Hy
bri
d
(
b
l
u
e i
n
col
o
r
)
co
nt
r
o
l
l
e
rs,
whi
c
h a
r
e ge
nerat
e
d b
y
t
h
e
cu
rr
en
t
r
e
g
u
l
at
o
r
s ar
e show
n
in
Figur
e
4
(
a)
,
(
b
)
an
d 5(
a),
(b) res
p
ectively. T
h
e
figu
res
s
h
ows
clearly that the
per
f
o
r
m
a
nces of
al
l
t
h
e c
ont
r
o
l
l
e
rs
a
r
e alm
o
st identical for the
v
ds
and when
com
p
are
d
with v
qs
, v
dr
, a
n
d v
qr
,
th
e resp
on
se
of th
e h
y
b
r
i
d
co
n
t
ro
ller
h
a
s l
e
ss am
p
l
i
t
u
d
e
o
f
the transients wh
ile du
ri
ng
th
e starting
o
f
the
perform
a
nce and e
v
en
whe
n
there is an
y step cha
nge a
n
d also it is
m
o
re desirable as it has faster res
p
onse in
reachi
n
g the st
eady-state
value after
0.13sec
s
.
Th
e co
m
p
on
ent o
f
th
e qu
adratu
re ax
is is g
e
nerally u
s
ed
to
reg
u
l
ate th
e reactiv
e p
o
wer, here fro
m
th
e
fi
g
u
res
4(
b) a
nd
6(
b
)
, i
t
i
s
cl
earl
y
i
ndi
cat
ed t
h
at
t
h
e
qua
drat
ure c
o
m
ponent
of t
h
e
ref
e
rence st
at
o
r
v
o
l
t
a
g
e
si
gni
fi
ca
nt
l
y
cont
rol
s
t
h
e
rea
c
t
i
v
e p
o
w
er
o
f
DFI
G
.
T
h
e
com
pone
nt
v
qs
h
a
s lesser over
s
hoo
t an
d r
e
ach
es
stability quickly com
p
ared to
ANN a
nd
PI c
ont
rollers
duri
ng t
h
e step
cha
nge i
n
reacti
v
e
powe
r,
whic
h
can be
cl
earl
y
obse
r
ve
d i
n
Fi
g
u
r
e
4(
b
)
.
(a)
(b
)
Figu
re
4.
(a
) R
e
fere
nce
gri
d
-s
ide v
o
ltage
v
ds
,
d
u
e to
reactive
p
o
we
r c
o
ntrol
fo
r P
I
c
o
ntrol a
n
d
N
N
c
o
ntr
o
l; (
b
)
Refere
nce gri
d
-side v
o
ltage
v
qs
, due to reacti
v
e
powe
r c
ontrol for PI c
ontrol, NN c
ont
rol a
n
d Hy
bri
d
c
ont
rol.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
37
9 – 3
9
0
38
6
Si
m
ilarly, th
e co
m
p
on
en
t
of th
e referen
c
e ro
tor-si
d
e
vo
ltag
e
vd
q h
a
s lesser ov
ersh
oo
t, lesser
o
s
cillatio
n
s
and
q
u
i
ck
ly reach
e
s th
e stab
le v
a
lu
e
fo
r
h
ybrid
co
n
t
ro
ller
as sho
w
n
i
n
Fig
u
re 5(a), (b). Th
e
com
pone
nt
v
dr
is alm
o
st zero
and the c
o
m
p
onent
v
qr
is almost anal
ogous t
o
the
active power P.
T
h
e res
p
onse
s
of the active a
nd
reactive powers
deliv
ered to the grid for
the case of reactive powe
r cont
rol are s
hown
i
n
Fi
gu
re
6(a
)
,
(b
). T
h
e
resp
o
n
s
e
s cl
earl
y
sh
o
w
t
h
at
t
h
e
hy
b
r
i
d
c
ont
rol
l
e
r i
s
m
o
re adva
nt
ageo
us a
s
i
t
exhi
bi
t
s
b
e
tter
p
e
rforman
ce in reach
i
ng
th
e stab
le
operatio
n with low tran
sien
ts and
low
p
e
ak
o
v
e
rshoo
t.
(a)
(b
)
Figu
re 5.
(a
)
R
e
fere
nce rot
o
r
-
s
ide voltage
v
dr
, d
u
e t
o
reactiv
e p
o
we
r c
o
ntro
l fo
r P
I
c
o
ntrol
an
d
NN
co
ntr
o
l; (
b
)
Refere
nce r
o
t
o
r-si
d
e
voltage
v
qr
, d
u
e t
o
reac
tive p
o
we
r c
o
n
t
rol f
o
r P
I
c
ont
rol,
N
N
c
o
ntrol
an
d
Hy
b
r
id c
o
ntr
o
l.
(a)
(b
)
Figu
re
6.
(a
)
A
c
tive p
o
we
r,
d
u
e to
reactive
p
o
we
r c
o
nt
r
o
l
fo
r PI con
t
ro
l an
d NN
con
t
ro
l; (b
)
Reactiv
e po
w
e
r
,
due
t
o
react
i
v
e
po
we
r c
ont
r
o
l
fo
r P
I
c
o
nt
rol
,
NN
co
nt
r
o
l
a
n
d
Hy
bri
d
c
o
nt
r
o
l
.
5.2.
Acti
ve P
o
w
er Control
In the case of
active power c
ont
rol, the
vari
ation of
t
h
e act
i
v
e po
we
r i
s
obt
ai
ned
by
t
h
e vari
at
i
o
n o
f
t
h
e q
u
ad
rat
u
re
com
pone
nt
of
t
h
e rot
o
r c
u
r
r
e
nt
v
qr
as ah
o
w
n i
n
Fi
g
u
re
8(
b)
. D
u
e t
o
t
h
e va
ri
at
i
on
of
v
qr
, it
resul
t
s
i
n
a st
e
p
c
h
an
ge
fr
om
0.
8
p.
u t
o
0
.
2
p.
u. i
n
t
h
e
act
i
v
e p
o
w
er as
s
h
o
w
n i
n
Fi
g
u
r
e
9(a
)
.
Fi
g
u
r
e
7(a
)
sho
w
s t
h
e GS
C
refere
nce
di
r
ect
axi
s
com
ponent
of st
at
o
r
vol
t
a
ge
fo
r t
h
e
cont
r
o
l
l
e
r,
w
h
i
c
h cl
earl
y
sho
w
s t
h
e
step c
h
ange
of val
u
e
beca
use
of t
h
e ste
p
c
h
ange
in
active
power. T
h
e
perform
a
n
ces
of th
e co
n
t
ro
llers are
al
m
o
st id
en
tical for
v
ds
. Figu
re
7(
b)
sho
w
s
t
h
e G
S
C r
e
f
e
r
e
n
ce qu
adr
a
ture
axis c
o
m
p
one
n
t of
stator vol
t
age for
t
h
e cont
rol
l
e
r
,
whi
c
h sh
ow
s t
h
at
t
h
e su
dde
n
chan
ge o
f
t
h
e
val
u
e o
f
v
qs
at exactly 0.03se
cs and 0.13sec
s
whe
r
e
th
e step
ch
an
ge is ap
p
lied
an
d
it, im
p
lica
t
es th
e r
eactiv
e p
o
wer
Q, which
is al
m
o
st
zero
v
a
lu
e.
Here th
e
per
f
o
r
m
a
nces of
NN a
n
d h
y
b
ri
d c
ont
rol
l
ers are i
d
e
n
t
i
cal and the
pe
ak val
u
es are
less com
p
ared to PI
cont
roller.
Fig
u
re
8(a
)
,
(b
) s
h
ows t
h
e RSC r
e
fere
nce
di
rect and
quadrature axis
co
m
p
onen
t
s of
ro
tor
vo
ltag
e
fo
r t
h
e c
o
nt
r
o
l
l
e
rs. T
h
e
ri
p
p
l
e
s at
t
h
e st
art
i
n
g a
nd
d
u
ri
ng t
h
e st
ep
cha
n
ge
are
qui
t
e
l
e
ss
fo
r
hy
bri
d
c
ont
rol
l
e
r
com
p
ared t
o
NN a
n
d PI c
ontrollers.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
A New Hyb
r
id
Artificia
l Neu
r
a
l
Netwo
r
k Based
C
o
n
t
ro
l
o
f
Doub
ly Fed
Ind
u
c
tion
Gen
e
ra
to
r
(
G
. V
.
M
a
dh
av)
38
7
The active and reactive powers
resu
lts du
e to
ch
ang
e
in
activ
e po
wer con
t
ro
l is sho
w
n
in
Figure
9(a
)
, (
b
). T
h
e per
f
o
r
m
a
nce of t
h
e p
r
o
p
o
se
d cont
rol
l
e
r i
s
quite co
m
p
arab
le with
th
e pe
rform
a
nce of ANN a
nd
it is far better in com
p
arison
with
PI c
o
ntroller as it has lesser am
plitude
transients a
n
d reaches t
h
e steady-
state co
nd
itio
n
q
u
i
ck
ly. Th
e max
i
m
u
m
d
i
fferen
ce
b
e
tween
t
h
ese con
t
ro
l
typ
e
s is less th
an 0
.
5
%
,
wh
ich
can
b
e
cl
earl
y
obse
r
ve
d
fr
om
t
h
e Fi
g
u
re
9
(
a)
, (
b
).
(a)
(b
)
F
i
g
u
r
e
7
.
(a
)
R
e
fere
nce gri
d
-s
ide
v
o
ltage
v
ds
, du
e to activ
e
po
wer con
t
ro
l
fo
r PI con
t
ro
l an
d NN con
t
ro
l; (b
)
Refere
nce gri
d
-side v
o
ltage
v
qs
, d
u
e to
activ
e p
o
we
r c
o
ntro
l fo
r P
I
c
o
ntrol
,
N
N
c
o
ntr
o
l a
n
d
Hy
bri
d
c
ont
rol.
(a)
(b)
F
i
g
u
r
e
8
.
(a
)
R
e
fere
nce rot
o
r
-
s
ide voltage
v
dr
, du
e t
o
activ
e
p
o
wer co
n
t
ro
l
for PI con
t
ro
l an
d NN con
t
ro
l; (b
)
Refere
nce r
o
t
o
r-si
d
e
voltage
v
qr
, d
u
e t
o
act
i
v
e
po
we
r c
ont
r
o
l
f
o
r
P
I
c
ont
r
o
l
,
N
N
c
ont
rol
a
n
d
Hy
bri
d
c
ont
rol
.
(a)
(b
)
Fi
gure
9.
(a)
Activ
e po
w
e
r
,
du
e to activ
e
pow
er con
t
ro
l fo
r PI
con
t
ro
l and NN co
n
t
ro
l; (b) Reactiv
e
po
wer,
due
t
o
act
i
v
e
p
o
we
r c
o
nt
rol
f
o
r
PI
co
nt
r
o
l
,
NN
co
nt
r
o
l
a
n
d
Hy
bri
d
c
o
nt
r
o
l
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
IJEC
E V
o
l
.
5, No
. 3,
J
u
ne 2
0
1
5
:
37
9 – 3
9
0
38
8
5.
3. Ph
ase-
to
-
E
arth
F
a
ul
t
In
t
h
is section, th
e
respon
se o
f
th
e
DFIG is an
alyzed
for a
phase
-
to-ear
th
fau
lt o
n
th
e
lin
e o
f
electrical network
near to t
h
e
DFIG. T
h
e
fault is consi
d
ere
d
to
be
occurre
d at t =0.02s
for
a duration
of 3m
s.
Si
m
ilar to
th
e
p
r
ev
iou
s
sectio
n
s
, Fi
g
u
res
10
(a),
(b
) a
n
d
1
1
(a
),
(b
) s
h
ow
s the
refe
rence
direct a
n
d
qua
d
r
at
ure
v
o
l
t
a
ge si
gn
al
s fo
r t
h
e control the GSC and RSCs, res
p
ectively. The res
p
ons
e
of
v
ds
as show
n
i
n
Fi
gu
re 1
0
(a
), c
l
earl
y
i
ndi
cat
i
ng t
h
e appl
i
cat
i
on
of fa
ul
t
co
nd
itio
n
at t=0
.
02
s, th
e resu
lt is id
en
tical fo
r
all th
e
descri
bed c
ont
rol
l
e
rs
, ho
we
v
e
r t
h
e hy
b
r
i
d
c
ont
rol
l
e
r p
r
od
u
ces l
o
w o
v
ers
h
oot
d
u
ri
ng t
h
e
t
r
ansi
ent
co
n
d
i
t
i
on.
Th
e r
e
sp
on
se of
v
qs
is al
m
o
st
si
m
ilar fo
r all
t
h
e con
t
ro
llers, still
th
e h
y
b
r
id co
n
t
ro
ller exhib
its lesser tran
sient
am
pl
i
t
ude com
p
are
d
t
o
ot
he
r
t
w
o c
ont
rol
l
e
rs
. I
n
b
o
t
h
t
h
e F
i
gu
res 1
0
(
a),
(
b
)
,
i
t
can be cl
earl
y
obse
r
ved
t
h
at
there is a s
u
dden inc
r
ease
of
value
of
v
ds
an
d s
u
d
d
e
n
dec
r
e
a
se of
val
u
e o
f
v
qs
, ind
i
catin
g
th
at th
e ov
erall v
a
lue
o
f
th
e
referen
c
e stato
r
-sid
e vo
ltag
e
is
falls
d
u
e
to
t
h
e
sudd
en app
lication
o
f
fau
lt conditio
n
.
As illu
st
rated
i
n
Figu
res
1
1
(a
),
(b
), t
h
e c
o
m
p
o
n
ents
o
f
v
dr
and
v
qr
a
r
e al
m
o
st
zero
i
n
di
cat
i
n
g t
h
at
hi
gh
am
ou
nt
s
of
c
u
r
r
en
t
s
pass
t
h
r
o
u
g
h
t
h
e
r
o
t
o
r
wi
ndi
ng
d
u
r
i
ng t
h
e a
b
no
r
m
al
con
d
i
t
i
on,
w
h
i
c
h
i
s
si
m
i
lar ca
use i
n
st
at
or
wi
ndi
ng
al
s
o
.
The
r
e
spon
ses show
s th
e
p
e
r
f
o
r
man
ces of
th
e h
ybr
id
an
d
ANN are id
en
ti
cal an
d
b
e
tter
in
th
e tran
sients wh
en
com
p
ared t
o
PI controller.
(a)
(b
)
Figu
re 1
0
. (a)
Refere
nce gri
d
-side v
o
ltage
v
ds
, d
u
e to
a
pha
se-to
-
earth
fa
ul
t fo
r P
I
c
o
ntrol
an
d
NN
co
ntr
o
l; (
b
)
Refere
nce gri
d
-side v
o
ltage
v
qs
, d
u
e t
o
a
pha
se-t
o
-
eart
h
fa
ul
t
fo
r P
I
c
o
nt
rol
,
N
N
c
o
nt
r
o
l
a
n
d
Hy
bri
d
c
ont
rol
.
(a)
(b
)
Figu
re 1
1
. (a)
Refere
nce
r
o
t
o
r-si
d
e voltage
v
dr
, du
e t
o
a
phase-
t
o-
ear
th f
a
u
lt fo
r
PI
co
nt
r
o
l
an
d
N
N
c
ont
rol
;
(
b
)
Refere
nce r
o
t
o
r-si
d
e
voltage
v
qr
, due t
o
a
phase-to-earth fa
ult for
PI co
nt
r
o
l
,
NN
co
nt
r
o
l
and
Hy
bri
d
c
o
nt
r
o
l
.
Th
e respon
ses
o
f
t
h
e activ
e an
d
reactiv
e
p
o
wers
fo
r th
e ph
ase to
earth
fau
lt con
d
ition
are shown
i
n
Figu
res 1
2
(
a),
(b
). Fr
om
the figu
re it is clear
ly
verified
th
at
tran
sien
ts or th
e d
i
st
u
r
ba
nce
s
whi
c
h are cause
d i
n
P and
Q
d
e
liv
ered
t
o
th
e grid
are qu
ite sm
o
o
th
er
for th
e
cas
e of hy
bri
d
controller whe
n
c
o
m
p
ared t
o
NN a
nd
Evaluation Warning : The document was created with Spire.PDF for Python.