Int
ern
at
i
onal
Journ
al of
P
ower E
le
ctr
on
i
cs a
n
d
Drive
S
ystem
s
(
IJ
PEDS
)
Vo
l.
12
,
No.
1
,
M
a
r 202
1
, p
p.
612
~
626
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v12.i
1
.
pp612
-
626
612
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
A robust
non
li
ne
ar co
ntr
ol strat
egy of
a PV Syste
m conn
ected
to
the thre
e
-
phase g
rid base
d on ba
ck
steppin
g and PS
O tech
nique
Sa
lm
a
Zo
uga
1
,
M
ohame
d B
enchagr
a
2
,
Abdall
ah Ail
ane
3
1,3
Nati
ona
l
Scho
ol
of
Appl
ie
d
Sc
ie
nc
es,
ISERT
L
abor
at
ory
,
ENSA
,
Sulta
n
Moula
y
Slim
an
e
Univ
e
rsiy
,
Khouribg
a,
Morocc
o
2
Superior
Schoo
l
of Te
chnol
ogy
,
EST
-
Ben
im
e
ll
a
l
,
Sult
an
Moul
ay S
li
ma
ne
Univer
s
iy
,
Khouribg
a,
Morocc
o
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
M
a
y
1
9
, 20
20
Re
vised
Jan
2
8
, 20
21
Accepte
d
Fe
b 1
3
, 20
21
Thi
s
ar
ti
c
le
pre
s
ent
s
a
robust
no
n
-
li
ne
ar
cont
rol
te
chn
ique
of
th
e
three
-
phase
photovol
taic
sys
te
m
.
Th
e
struc
tu
re
chose
n
for
this
PV
sys
te
m
is
tha
t
of
two
power
conve
r
ters
and
DC
vo
lt
ag
e
in
te
rm
ediate
bus.
Th
e
t
wo
power
conve
rt
ers
ar
e:
the
DC
-
DC
con
ver
te
r
a
nd
the
t
hre
e
-
phase
inv
e
rte
r,
which
req
uire
s
two
main
cont
ro
ll
ers
.
T
hese
cont
ro
llers
have
thr
ee
main
obje
c
ti
ves
.
The
first
obj
ec
t
i
ve
is
to
i
mpose
t
he
PV
vo
lt
ag
e
g
ene
ra
te
d
by
the
photovol
taic
pane
l
,
in
ord
er
to
fol
low
a
m
a
xim
um
ref
e
ren
c
e
vol
ta
ge
provi
ded
by
the
MP
PT
bloc
k.
Th
e
sec
ond
on
e
is
t
o
ma
in
ta
in
the
DC
li
nk
vol
ta
g
e
to
a
const
an
t
val
ue
,
in
orde
r
to
op
ti
m
iz
e
the
tra
nsfer
of
en
er
gy
be
twee
n
th
e
two
power
conve
rt
ers.
The
l
ast
obj
ec
t
ive
is
t
o
inj
e
ct
a
three
-
p
hase
sinusoida
l
c
urre
nt
int
o
the
grid
,
whil
e
r
espe
cting
a
un
it
power
factor.
W
it
h
th
e
intenti
on
to
ac
h
ie
v
e
the
se
three
obj
e
ct
iv
es,
we
desig
ned
c
asc
ad
ing
n
onli
ne
ar
cont
rol
l
ers
by
using
the
technique
of
non
-
li
ne
ar
b
ac
ks
te
pping
con
trol
i
n
th
e
synth
esis
o
f
th
ese
two
cont
rollers,
base
d
on
th
e
Lya
pu
nov
func
ti
on
,
w
it
h
reg
a
rd
to
m
axi
mi
se
the
PV
G
output
voltage,
in
ord
er
to
have
a
uni
ta
ry
p
ower
factor
a
t
th
e
grid
sid
e.
In
orde
r
to
r
egulate
DC
-
l
ink
voltage,
w
e
dev
el
op
ed
an
in
te
gr
al
pr
oporti
onal
cont
roller
(PI)
with
par
a
meter
s
tha
t
are
optimi
ze
d
by
the
Par
t
ic
l
e
Sw
arm
Optim
izati
o
n
(PS
O)
me
thod.
T
he
robustness
of
the
cont
rol
ler
designe
d
appr
oac
h
is
te
st
ed
by
a
simul
a
ti
on
in
MA
TLA
B/Sim
uli
nk
software
,
that
im
prove
s
the pe
r
forma
nc
es
of ea
ch
con
trol
l
er
wh
at
ev
er condi
t
ion
s of
clim
at
e
.
Ke
yw
or
d
s
:
Ba
ckstep
ping c
on
t
ro
l
Ca
scade
regula
ti
on
lo
ops
Lya
punov f
un
c
ti
on
c
on
t
ro
l
Op
ti
mize
d al
gorith
m
Thr
ee
phase
P
V
S
ys
te
m
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
:
Salma Z
ouga
Nati
on
al
Scho
ol of
A
pp
li
ed
S
ci
ences
Su
lt
an M
oula
y Sl
imane
U
nive
rsiy
927, A
L Fat
h,
Kho
ur
i
bg
a
-
M
orocc
o
Emai
l:
zouga.
s
el
ma@gmai
l.com
1.
INTROD
U
CTION
Durin
g
the
la
st
deca
de,
we
ca
nnot
ima
gi
ne
a
sin
gle
da
y
without
el
ect
rici
ty
.
We
c
ons
um
e
m
or
e
a
nd
more
e
nerg
y,
a
nd
we
wa
nt
to
pr
oduce
it
co
r
rectl
y.
To
mee
t
this
gro
wing
dema
nd,
we
ha
ve
bee
n
t
urni
ng
for
the
la
st
fe
w
ye
ars
t
owar
ds
cl
ean
an
d
re
ne
wa
ble
e
nergies,
t
o
pro
duce
m
ore
a
nd
m
ore,
wi
thout
destr
oy
in
g
t
he
env
i
ronme
nt.
Nowa
day
s
,
m
or
e
i
nterest
in
photov
oltai
c
energ
y
(PV)
ha
s
fo
c
us
e
d
on
photov
oltai
c
sy
ste
ms
connecte
d
t
o
th
e
gr
id
[
1]
.
H
oweve
r,
t
he
ope
rati
on
of
t
hese
pho
to
volt
ai
c
sy
ste
ms
is
in
flu
enced
by
the
va
riat
ion
of
weat
her
co
nd
it
io
ns
[2]
,
the
non
-
li
near
it
y
of
the
switc
hing
functi
ons
of
c
onve
rters
and
in
ver
te
rs,
and
the
load
var
ia
ti
on
[3]
.
T
he
impa
ct
of
these
dif
fer
e
nt
facto
rs
po
s
es
the
c
on
t
ro
l
of
these
s
yst
ems
in
f
ront
of
the
Diff
e
re
nt
c
halle
ng
e
s
t
o
im
pro
ve
t
he
pe
rform
ance
a
nd
qual
it
y
of
th
e
gr
i
d
.
T
hu
s
,
it
is
esse
nt
ia
l
to
desig
n
r
obus
t
con
t
ro
ll
ers
w
hi
ch
ens
ure
the
main
obje
ct
ives
of
co
ntr
ol
r
epr
ese
nted
by
the
co
rr
ect
io
n
of
t
he
powe
r
f
act
or,
regulat
ion
of
t
he DC
bu
s
volt
age
a
nd
opti
mi
zat
ion
of
the
powe
r ge
ner
at
ed
by th
e
P
V pane
l
[4
,
5]
.
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
A
rob
us
t
nonlinear c
on
tr
ol st
ra
te
gy
o
f
a
P
V
System
con
nected to
the t
hr
ee
-
phas
e
g
ri
d
…
(
Sa
l
ma Z
ouga
)
613
In
the
li
te
ratu
re,
ma
ny
st
udie
s
hav
e
f
oc
use
d
on
the
de
sign
met
hods
of
li
nea
r
co
ntr
ollers
f
or
photov
oltai
c
sy
ste
ms
c
onnec
te
d
to
gr
i
d.
A
mong
these
m
et
hods
we
ca
n
fi
nd
:
t
he
cl
assic
al
PI
c
ontrolle
r
gen
e
rall
y
us
e
d
in
in
du
st
ry,
be
cause
it
is
c
he
ap,
r
obus
t,
a
nd
it
is
easy
to
impleme
nt
[6
],
it
pro
duces
good
resu
lt
s
in
li
nea
r
s
ys
te
ms
.
H
oweve
r,
it
has
c
ertai
n
li
mit
at
io
ns
,
giv
e
n
t
hel
un
ce
rtai
nties,
a
nd
non
-
li
near
l
oads
.
I
n
[7
,
8],
t
he
c
on
t
ro
l
ba
sed
on
a
PR
c
on
t
ro
ll
er
,
has
al
so
s
how
n
the
good
co
ntinu
at
io
n
of
the
ref
e
ren
ce
,
t
ha
nks
to
th
e
in
finite
gai
n.
D
e
sp
it
e
t
he
us
ef
uln
e
ss
of
t
hese
li
near
c
ontrolle
rs
to
achi
eve
t
heir
obje
c
ti
ves
within
a
sta
ble
op
e
rati
ng
rang
e
,
the
se
li
nea
r
methods
becom
e
in
suf
fici
ent
,
a
nd
unreli
abl
e
in
the
prese
nc
e
of
ra
pid
l
y
c
hangin
g
op
e
rati
ng
c
ondi
ti
on
s
,
i
nclu
din
g
mete
orolo
gi
cal
changes.
Fo
r
this
rea
on,
the
desig
n
of
a
c
on
tr
ol
te
c
hniq
ue,
adequate
t
o per
tur
bations, a
nd to
the
no
n
-
li
ne
arit
y beco
me
necessa
ry
.
A
lot
of
resear
ches
ha
ve
pr
opos
e
d
rob
us
t
non
-
li
nea
r
co
ntr
ol
meth
ods
t
o
e
ns
ure
sta
bili
ty
[9
-
1
1]
,
f
rom
PV
sy
ste
ms
c
onnecte
d
t
o
gr
i
d.
Th
e
sli
di
ng
mode
c
ontr
oller
(S
MC
)
[
12
,
13
],
th
e
pre
dicti
ve
co
ntr
oller
model
[14
]
,
an
d
the
f
eedb
ac
k
li
nea
r
iz
at
ion
te
ch
nique
FB
L
[15
]
,
a
re
me
ntion
e
d.
Each
of
it
s
ap
proac
hes
ha
s
it
s
own
adv
a
ntage
s,
a
nd
disa
dv
a
nta
ges.
T
he
FB
L
te
chn
iq
ue
of
f
ers
the
best
pe
rformance
ov
er
a
wi
de
ra
nge
of
op
e
rati
ng
rang
es
but
it
ca
ncel
s
the
no
nlinear
it
ie
s
of
the
sys
te
m
an
d
bec
ome
s
a
li
near
s
ubsyst
em
.
I
n
a
ddit
ion
,
this
ca
ncell
at
ion
ca
n
be
a
vo
i
de
d
by
us
in
g
a
con
t
ro
ll
er
[
16],
base
d
on
the
nonlinea
r
Ba
c
ks
te
ppin
g
te
c
hniq
ue.
This
met
hod
will
be
prese
nt
ed
in
this
pa
pe
r
,
in
order
to
desig
n
a
n
a
ppr
opriat
e
co
ntr
ol
la
w
,
to
ens
ur
e
t
he
ov
e
rall
sta
bili
ty
of
the
s
ys
te
m
.
N
on
-
li
near
con
t
ro
ll
er
de
sign,
i
n
casca
de
base
d
on
the
B
ackstep
ping
te
chn
i
qu
e
,
ta
kes
into
c
onside
rati
on
t
he
fo
ll
owin
g
c
on
t
ro
l
ob
je
ct
ives:
i)
re
gu
la
te
the
vo
lt
age
wh
ic
h
sup
plies
the
photov
oltai
c
ge
ner
at
or
to
ext
ract
the
ma
xim
um
power
.
ii
)
Ph
asi
ng
t
he
vo
lt
ag
es
of
t
he
gr
id
with
the
c
ur
ren
t
by
regulat
ing
t
his
la
st
,
and
to
as
s
ert
a
perfect
co
rr
ect
io
n
to
the
powe
r
facto
r
(
PFC).
In
orde
r
to
keep
t
he
D
C
bu
s
vo
lt
age
at
it
s
desire
d
ref
e
rence
val
ue,
a
P
I
typ
e
co
ntr
oller
base
d
on
a
pe
rformi
ng
met
a
-
he
ur
ist
ic
a
l
gorith
m
"
par
ti
cl
e
s
warm
op
ti
miza
ti
on"
(P
S
O)
has
been
dev
el
op
e
d
t
o
opti
mize
the
K
p
a
nd
Ki
par
a
mete
rs
of
t
he
con
t
ro
ll
er
.
P
S
O
is a
n op
ti
miz
at
ion
alg
ori
thm
u
se
d
in
this
pa
per
t
o determi
ne
the
par
a
mete
rs of
t
he
P
I reg
ulator
in
orde
r
to
regulat
e
the
DC
bus
volt
age
.
[
17]
.
Co
mp
a
re
d
w
it
h
sever
al
methods,
i
nclu
ding
ge
netic
al
gor
it
hms
(GA),
B
-
sp
li
ne
net
wor
ks
,
f
uz
zy
l
og
ic
an
d
ne
ur
al
n
et
works
[
18]
,
P
SO
has
sho
wn
a
bette
r
opti
miza
ti
on
,
it
is
base
d
on
a
sim
ple
al
gorith
m
c
har
act
erize
d
by
ease
of
imple
mentat
io
n
an
d
rob
us
tness
.
T
his
pa
per
is
struc
ture
d
as
f
ollo
ws:
S
ec
ti
on
2
prese
nts
the
de
scriptio
n
of
t
he
s
ys
te
m,
a
nd
t
he
mat
hemati
cal
m
od
el
.
Sect
io
n
3
de
velo
ps
the
nonli
near
c
on
t
ro
l
strat
e
gies
of
t
he
pr
opose
d
s
ys
te
m.
Se
c
ti
on
4
sho
ws
the
sim
ulati
on
r
esults
as
well
as
an
analysis
of the
se r
es
ults
. F
i
na
ll
y
, conclusi
on
is givi
ng in
sec
ti
on
5
.
2.
DESCRIPTI
ON A
ND M
O
DELL
ING O
F
THE
ST
UDIED S
YS
TE
M
The
f
ollow
i
ng
sect
ion
present
s
the
mathe
mati
cal
modeli
ng
of
each
c
omponent
in
t
he
PV
conve
rsion
chain.
As
il
lus
trat
ed
i
n
Fi
gur
e
1,
the
ge
ner
al
str
ucture
of
th
e
syst
em
c
onsist
s
of
t
wo
pow
er
sta
ges
t
o
m
anag
e
the
P
V
power
deli
ver
e
d
t
o
the
util
it
y
gr
i
d.
T
he
first
st
ag
e
incl
ud
i
ng
a
chop
per
that
a
ll
ow
s
to
ext
ra
ct
the
maxim
um
po
w
er
ge
ner
at
ed
by
t
he
PV
pan
el
,
by
re
gula
ti
ng
the
ou
t
pu
t
cap
aci
tor’
s
vo
lt
ag
e
to
a
c
onsta
nt
value.
Be
sides
it
al
lows
to
ac
hieve
the
M
P
PT
c
ontrol
by
f
or
ci
ng
the
phot
ovoltai
c
pan
el
to
operate
at
the
op
ti
mu
m
op
e
rati
ng
point
des
pite
the
va
r
ia
ti
on
of
e
nv
ir
onmental
c
ondi
ti
on
s.
T
he
sec
ond
sta
ge
is
re
pr
ese
nted
by
a
three
-
ph
a
se
in
ver
te
r
whose
main
obje
ct
ive
is
to
en
su
re
t
he
c
onve
rsion
of
the
D
C
power
into
a
n
AC
powe
r
in
je
ct
ed
into the
grid
, a
nd contr
ols the
outp
ut curre
nt
.
Figure
1.
S
tr
uc
ture of
the
gri
d
-
co
nnect
ed p
hotovolt
ai
c sy
ste
m
In
or
der
to
ha
ve
a
ze
r
o
phase
sh
ift
with
the
volt
age
of
t
he
gri
d
i
n
orde
r
t
o
ob
ta
in
a
unit
y
FP.
The
use
of
t
he
in
duct
iv
e
filt
er
L
betw
een
the
in
ver
te
r
an
d
t
he
th
ree
-
phase
gr
i
d
is
necessa
ry,
in
orde
r
to
minimi
ze
the
harmo
nic c
ompone
nts
in
the i
nv
e
rter
outp
ut
curre
nt
ge
ner
at
ed by t
he PW
M
c
ontrol.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
D
ri
S
ys
t,
V
ol
.
12
, N
o.
1
,
Ma
rch
20
21
:
612
–
626
614
A
phot
ovoltai
c
pan
el
co
ns
i
sts
of
se
ver
al
cel
ls
as
so
ci
at
ed
in
series
a
nd
i
n
pa
rall
el
,
w
hic
h
a
re
el
ect
ro
nic
c
omp
on
e
nts
t
hat
al
low
t
he
tr
ansfo
rmati
on
of
li
gh
t
i
nto
el
ect
rical
en
ergy.
The
el
ect
rical
char
act
e
risti
cs
of
a
si
ng
le
m
odule
us
ed
i
n
this
stu
dy
a
re
li
ste
d
in
Table
1.
Wh
il
e
the
a
dopted
mathe
mati
cal
model
of
a
PV
cel
l
is
sh
own
in
Fi
gure
2,
thi
s
model
co
n
sis
ts
of
a
cu
rr
e
nt
gen
e
rato
r
0
con
t
ro
ll
ed
by
volt
age,
and
i
nf
l
uen
ce
d
by
te
mp
e
ratu
r
e
and
so
la
r
ra
diati
on
,
an
a
nt
i
-
par
al
le
l
di
ode
D
1,
a
s
hunt
el
ect
rical
resist
ance
ℎ
,
an
d
a
ser
ie
s
re
sist
ance
[19].
Figure
2
.
P
V
c
el
l mod
el
The
relat
io
nsh
ip
betwe
en
P
V
cu
rr
e
nt
a
nd
PV
ou
t
pu
t
volt
age
is
re
presented
by
th
e
fo
ll
owin
g
mathemat
ic
al
e
qu
at
io
n:
I
=
I
0
−
I
D
[
+
−
1
]
−
+
ℎ
(1
)
With
I
represe
nt
the
li
gh
t
ge
ner
at
e
d
c
urre
nt,
I
0
the
di
od
e
sat
ur
at
io
n
c
urre
nt
,
T
K
the
te
m
perat
ur
e
of
cel
l,
In K,
n
th
e ideal
f
act
or,
q
the
ch
a
r
ge of
an
elec
tr
on,
i
n C
and
K
t
he
B
oltzma
n’s c
ons
ta
nt,
in
J/ K
Table
1
P
V mo
du
le
pa
rameter
s
Para
m
eter
Valu
e
Maximu
m
po
wer
260W
Cu
rr
en
t at
th
e
m
ax
im
u
m
p
o
we
r
p
o
in
t
8
,37
A
Vo
ltag
e at
th
e
m
ax
im
u
m
po
we
r
p
o
in
t
3
1
,1V
Nu
m
b
er
o
f
series
m
o
d
u
les
6
Nu
m
b
er
o
f
parallel
m
o
d
u
les
1
The
sta
bili
zat
ion
of
t
he
DC
volt
age
and
the
converge
nce
of
this
volt
age
to
it
s
ref
e
ren
ce
value
is
the
obje
ct
ive
of
the
dev
el
op
e
d
co
ntr
ol
strat
e
gy.
F
or
t
his,
obta
inin
g
a
dyna
mic
model
of
the
DC/DC
c
onve
rter
[20]
is
esse
ntial
by
us
i
ng
the
kirc
hhoff
la
w
wh
e
re
t
he
DC
-
li
nk
ca
pacit
or
and
the
capaci
tor
a
re
c
onnect
ed
and this ca
n be
w
ritt
en
as:
dv
pv
dt
=
1
C
pv
(
i
pv
−
I
L
)
(2
)
di
L
dt
=
1
L
1
[
−
R
1
I
L
+
v
pv
−
(
1
−
μ
)
V
dc
]
(3)
dV
dc
dt
=
1
C
dc
[
(
1
−
μ
)
−
I
dc
]
(4
)
wh
e
re
and
represe
nt
the
gen
e
rated
P
V
ou
t
pu
t
vo
lt
age
and
c
urren
t,
desig
nates
th
e
DC
li
nk
vo
lt
age
,
and
are
res
pecti
vely
the inp
ut and
outp
ut
cu
rr
e
nt c
hopper
, a
nd
is t
he
duty
cy
cl
e
[
21]
.
The
sec
ond
co
nv
e
rsion
sta
ge
represente
d
by
the
th
ree
-
ph
ase
inv
e
rter
is
us
e
d
to
ge
nera
te
the
three
-
ph
a
se
cu
rr
e
nts
injec
te
d
into
the
gr
i
d.
T
he
inv
e
rter
is
bas
e
d
on
P
W
M
co
ntr
ol
and
is
c
onnecte
d
to
the
gr
id
thr
ough
a
ba
x
pa
ss
filt
er.
Wh
e
reas
the
dyna
mica
l
m
od
e
l
of
volt
age
so
urce
in
ver
te
r
is
e
xpresse
d
by
the
fo
ll
owin
g
e
qua
ti
on
s
[22
,
23]
:
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
A
rob
us
t
nonlinear c
on
tr
ol st
ra
te
gy
o
f
a
P
V
System
con
nected to
the t
hr
ee
-
phas
e
g
ri
d
…
(
Sa
l
ma Z
ouga
)
615
=
−
−
+
3
(2
−
−
)
(5
)
=
−
−
+
3
(
-
+
2
−
)
(6
)
=
−
−
+
3
(
-
−
+
2
)
(7
)
The
pr
e
vious model
is
tran
sf
ormed
i
nto
a ne
w
re
fer
e
nce f
r
ame
acco
rd
i
ng to
the
a
ngular
f
reque
ncy
of
the n
et
wor
k,
w
hich
al
lo
ws
u
s
to r
e
wr
it
e th
e i
nv
e
rter
model
as foll
ows:
d
I
d
dt
=
−
R
f
L
f
I
d
+
w
I
q
−
1
L
f
E
d
+
V
dc
L
S
d
(8)
d
I
q
dt
=
−
R
f
L
f
I
q
−
w
I
d
−
1
L
f
E
q
+
V
dc
L
S
q
(9)
Id
an
d
Iq
are
are
the
act
ive
and
reacti
ve
c
urren
ts
;
E
d
an
d
E
q
a
re
a
re
t
he
act
ive
an
d
reacti
ve
gr
i
d
vo
lt
age
s; S
d
a
nd
S
q
a
re
t
he
s
w
it
ching
c
ontr
ol
inputs.
3.
CONTR
OLL
ERS
DESIG
N
STEPS
In
t
his
sect
io
n,
we
a
re
i
nt
erested
in
de
sign
i
ng
tw
o
r
obus
t
c
on
t
r
ollers
base
d
on
ad
va
nced
mathemat
ic
al
al
gorithms
de
di
cat
ed
to
non
-
l
inear
s
ys
te
ms
.
The
first
c
ontrolle
r
c
oncer
ns
the
P
V
vo
lt
a
ge,
it
fo
ll
ows
t
he
ma
ximum
powe
r
po
i
nt
(
MPPT
)
by
tw
o
c
ontr
ol
lo
op
s
,
a
nd
r
eg
ulate
s
the
volt
age
ge
ner
at
e
d
by
the
PV
ge
ner
at
or
.
F
or
the
seco
nd
co
ntr
oller
is
ba
sed
on
tw
o
ca
scade
d
lo
ops,
a
n
inte
rn
al
re
gu
l
at
ion
loop
t
o
c
on
t
ro
l
the
act
ive
an
d
reacti
ve
po
wer
injec
te
d
into
t
he
gri
d,
an
d
th
e
exter
nal
lo
op
serv
e
s
ma
ke
the
DC
bus
vo
l
ta
ge
sta
ble acco
rd
i
ng to
the
desire
d
re
fer
e
nce
vo
l
ta
ge
. T
he
f
ull mo
del of the
de
sign
of
t
hese
co
nt
ro
ll
ers
is s
how
n
in
the F
ig
ure
3,
a
nd it
w
il
l be
d
e
ta
il
ed
in this
se
ct
ion
.
Figure
3. N
onli
near co
ntr
ol sc
heme o
f
the
th
r
ee
-
phase
-
gri
d
c
onnected
ph
otovo
lt
ai
c s
ys
te
m
3.1.
P
V outpu
t vo
lt
age c
ont
rolle
r desig
n
The
pr
i
nicipal
go
al
of
this
re
gula
tor
is
to
ens
ur
e
the
e
xtracti
on
of
t
he
ma
xi
mu
m
po
wer
ge
ner
at
e
d
by
the
PV
pan
el
wh
at
e
ver
the
operati
ng
co
ndit
ion
s
.
For
this
r
easo
n,
the
design
of
t
his
c
ontr
oller
ai
ms
t
o
st
abili
ze
the
P
V
volt
ag
e
at
it
s
ref
e
re
nc
e
val
ue
obta
ined
by
th
e
MPPT
al
gorith
m
ba
sed
on
inc
r
emental
c
ondu
ct
ance
te
chn
iq
ue.
T
hi
s
re
gula
ti
on
of
the
P
V
volt
age
is
ens
ure
d
by
the
Ba
c
ks
te
ppin
g
te
c
hniq
ue
a
ppli
ed
to
t
he
dyna
mics
m
od
el
(
4),
a
nd
bas
ed
on
the
sta
bi
li
zat
ion
f
unct
ion
of
L
ya
punov,
t
he
de
sig
n
of
the
c
on
t
ro
l
le
r
is
dev
el
op
e
d
i
n
t
wo steps
[
24]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
D
ri
S
ys
t,
V
ol
.
12
, N
o.
1
,
Ma
rch
20
21
:
612
–
626
616
I
n
t
he
first
des
ign
ste
p:
w
e
de
fine
t
he
er
r
or
betwee
n
t
he
volt
age
gen
e
rat
ed
by
the
PV
pan
el
a
nd
it
s
ref
e
ren
ce
by:
e
1
=
V
pv
−
V
pv
ref
(1
0
)
Wh
e
re
rep
res
e
nts
t
he
volt
ag
e
ge
ne
rated
by
t
he
MPPT
blo
c
k
w
hich
s
erv
es
as
a
re
f
eren
ce
vo
lt
age
.
By r
e
placi
ng
w
it
h
(
4)
t
he deri
vative
of (1
0
),
resu
lt
s t
he
er
r
or
dynamics,
wh
ic
h
is:
e
̇
1
=
1
C
pv
(
i
pv
−
I
L
)
−
V
̇
pv
ref
(1
1
)
To
c
onfir
m the
stabil
it
y,
L
ya
puno
v
ca
nd
i
date
fun
ct
io
n (CLF
)
is cal
culat
e
d as:
V
1
=
1
2
e
1
2
(1
2
)
Its ti
me d
e
rivat
ive is
giv
e
n
as
fo
ll
ows:
V
̇
1
=
e
1
e
̇
1
(1
3
)
If
we
c
onside
r
̇
1
=
−
1
1
the
der
i
vative
of CLF ca
n be
wr
it
te
n
as
:
V
̇
1
=
−
k
1
e
1
2
(1
4
)
M
a
king
̇
ne
gat
ively
def
i
ne
d
def
i
nite
(
̇
1
<
0
)
or
se
mi
-
de
finite
(
̇
1
≤
0
)
al
lows
t
o
sta
bili
ze
̇
1
an
d
to h
a
ve
a
null
c
onve
rg
e
nce e
rror
.
w
hic
h
ma
ke
s it
possible t
o wr
it
e
̇
1
in a si
mp
li
fied
form
:
V
̇
1
=
1
C
pv
(
i
pv
−
I
L
)
−
V
̇
pv
ref
=
−
k
1
e
1
(1
5
)
If
w
e
c
hoos
e
as
virt
ual
co
ntr
ol
input,
we
de
duce
the stabil
iz
ing
f
unct
io
n
(
=
)
,
w
hich
e
nsure
the
asy
mp
t
otic
sta
bili
ty
of
the
su
bsystem
(1
7
).
A
nd
f
rom
(
2
7
)
the
desire
d
value
of
t
he
fir
st
virtu
al
co
ntr
ol
ca
n
be writt
en
as
foll
ows:
α
=
i
pv
−
C
pv
(
V
pv
ref
−
k
1
e
1
)
(1
6
)
As
is
no
t
t
he
f
inal
co
ntr
ol
in
pu
t,
a
nd
not
al
way
s
e
qu
al
t
o
t
he
sta
bili
zi
ng
f
un
ct
io
n,
in
the
fo
ll
owin
g
a
new
er
ror
va
riable
2
betwee
n
the
virtu
al
c
on
t
ro
l
and
it
s
desire
d
va
lue
is
introduc
ed
s
te
p
to
def
in
e
the
final c
on
tr
ol i
nput
sig
nal
.
I
n t
he
sec
ond
de
sign st
ep:
we defi
ne
the
er
ror bet
wee
n
the
c
urren
t a
nd it
s refere
nce
value
as:
e
2
=
I
L
−
α
(1
7
)
Whose
dy
nami
cs is ba
sed
on t
he
s
ubsy
ste
m (
18
)
a
nd
it
can
be
e
xpresse
d b
y
:
e
̇
2
=
1
L
[
v
pv
−
(
1
−
μ
)
V
dc
−
R
1
I
L
]
−
α
̇
(
18
)
In co
ntem
plati
on of sta
bili
zi
ng
the
s
ub
s
ys
te
ms (4
)
a
nd (5),
w
e c
onside
r
th
e seco
nd ca
nd
i
date L
yapu
nov
functi
on:
V
2
=
V
1
+
1
2
e
2
2
(
19
)
Be
fore
obta
ini
ng
t
he
de
rivati
on
of
t
he
Ly
ap
unov
can
did
at
e
Functi
on
2
,
it
is
necessa
ry
t
o
simpli
fy
the
value
of
t
he
der
i
vative
of
̇
1
represe
nte
d
by
(1
5
).
For
this
,
th
e
dyna
mics
of
the
error
1
can
be
r
e
-
expresse
d b
y:
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
A
rob
us
t
nonlinear c
on
tr
ol st
ra
te
gy
o
f
a
P
V
System
con
nected to
the t
hr
ee
-
phas
e
g
ri
d
…
(
Sa
l
ma Z
ouga
)
617
e
̇
1
=
1
C
pv
i
pv
−
e
2
C
pv
−
α
C
pv
−
V
̇
pv
ref
(
20
)
Fr
om
(
1
3
)
,
(
1
6
)
an
d
(2
0
)
the
dynamics
of
tr
ackin
g
er
r
or
1
an
d
L
yapu
nov
functi
on
can
be
wr
it
te
n
as
fo
ll
ows:
e
̇
1
=
−
k
1
e
1
−
e
2
C
pv
(2
1
)
V
̇
1
=
−
k
1
e
1
2
−
e
1
e
2
C
pv
(2
2
)
The
ti
me
de
riv
at
ive of
V
2
is gi
ve
n by usi
ng the
(
19
)
a
nd v
al
ue
of
V
̇
1
,:
V
̇
2
=
−
K
1
e
1
2
+
e
2
(
e
̇
2
−
e
1
C
pv
)
(2
3
)
I
f we s
et
2
<
0
or
2
≤
0
,
w
e can
e
nsure
th
e stabil
it
y
of
t
he
erro
r derivati
ve
2
,
an
d we ca
n
put:
−
2
2
=
̇
2
−
1
(2
4
)
Wh
e
re,
2
is
a
re
al
po
sit
ive
desi
gn
pa
rameter
,
and
t
he
dynam
ic
s
of
t
he
Lya
puno
v
can
dida
te
functi
on
becomes:
V
̇
2
=
−
K
1
e
1
2
−
K
2
e
2
2
≤
0
(2
5
)
The
e
quat
ions
(2
4
)
and
(
18
)
l
ead to
:
−
K
2
e
2
=
1
L
[
v
pv
−
(
1
−
μ
)
V
dc
−
R
1
I
L
]
−
̇
−
1
(
2
6
)
By
re
placi
ng
(2
6
)
in
the
e
xpr
ession
of
t
he
dyna
mics
of
CF
L
(2
5
)
,
we
get the
real
co
ntr
ol
input
sign
al
no
te
d
μ
, th
at
e
nsures
the
sta
bili
ty
of the s
ys
te
m
(
e
1
,
e
2
)
:
μ
=
1
−
1
V
dc
(
−
R
1
I
L
+
v
pv
+
L
K
2
e
2
−
L
α
̇
−
L
e
1
C
pv
)
(2
7
)
3.2.
Ac
tive
an
d rea
c
tive
p
ower c
ontrolle
r
design
The
ob
je
ct
ive
of
t
his
c
on
tr
oll
er
is
to
achie
ve
a
unit
y
pow
er
fact
or
by
re
gu
la
ti
on
of
t
he
direct
a
nd
qu
a
drat
ic
com
pone
nt
of
the
current
,
to
be
sinu
s
oid
al
a
nd
in
phase
with
the
gri
d
s
uppl
y
volt
age
.
Thi
s
internal
c
urre
nt
regulat
ion
lo
op
ai
ms
t
o
ma
ke
the
c
urren
t
and
pr
ese
nt
a
de
qu
at
e
val
ues
with
it
s
gen
e
ra
te
d
ref
e
ren
ce
v
al
ue
s,
in
ord
e
r
t
o
in
je
ct
an
opti
mal
acti
v
e po
wer a
nd zer
o react
iv
e pow
e
r.
The
i
niti
al
ste
p
of
t
his
c
on
t
r
oller
is
to
de
fi
ne
the
trac
king
erro
r
betwee
n
the
act
ive
cu
r
ren
t
a
nd
i
ts
re
f
eren
c
e
value b
y:
e
3
=
I
d
−
I
dref
(
28
)
Wh
e
re
the
re
fe
ren
ce
c
urren
t
is
gen
e
rated
by
the
e
xter
nal
volt
age
re
gu
la
ti
on
l
oop
disc
usse
d
in
the foll
owin
g
s
ect
ion
, a
nd
wh
i
ch
is e
xpresse
d as f
ollo
w:
I
dref
=
β
E
d
(
29
)
With
is
an
y p
osi
ti
ve
real c
on
s
ta
nt g
e
ner
at
e
d by the
oute
r v
ol
ta
ge
loop.
The
ti
me
der
i
va
ti
ve
of
the
t
ra
ckin
g
e
rror
bet
ween
the
act
iv
e
curre
nt
an
d
i
ts
ref
e
ren
ce
va
lue
3
,
is
expresse
d b
y
:
e
̇
3
=
−
R
f
L
f
I
d
+
w
I
q
−
1
L
f
E
d
+
V
dc
L
f
S
d
−
I
̇
dref
(
30
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
D
ri
S
ys
t,
V
ol
.
12
, N
o.
1
,
Ma
rch
20
21
:
612
–
626
618
Alte
rn
at
ivel
y,
we defi
ned
the
trackin
g
e
rror
betwee
n
the
r
e
act
ive curre
nt
and it
s r
e
fer
e
nc
e v
al
ue
er
r
or
as:
e
4
=
I
q
−
I
qr
e
f
(3
1
)
U
sin
g
t
he valu
e of
̇
, w
e
can
e
xpress
t
he deri
vative
of
4
by:
e
̇
4
=
−
R
f
L
f
I
q
−
w
I
d
−
1
L
f
E
q
+
V
dc
L
f
S
q
−
I
̇
qr
e
f
(3
2
)
The
determi
nation
of
t
he
c
ondi
ti
on
s
on
t
he
c
on
t
ro
l
la
w
re
quires
t
he
ch
oic
e
of
t
he
L
ya
punov
sta
bili
ty
functi
on
w
hich
al
lows
t
o
fi
nd
the
bala
ncin
g
po
i
nt
of
the
syst
em
.
T
h
us
,
a
new
CLF
is
be
ing
c
onside
red
wh
ic
h
makes
it
possi
ble
t
o de
du
c
e t
he
c
on
t
ro
l si
gnal
s
an
d
V
3
=
V
2
+
1
2
e
3
2
+
1
2
e
4
2
(3
3
)
The deri
vative
of this CLF
is re
pr
ese
nted
by:
V
̇
3
=
V
̇
2
+
e
3
e
̇
3
+
e
4
e
̇
4
(3
4
)
By
re
placi
ng
t
he values
of
V
̇
2
from (
2
5
),
e
̇
3
f
rom
(3
0
)
, a
nd
e
̇
4
from
(3
2
)
int
o (3
4
)
yields
we’ve
got:
V
̇
3
=
−
K
1
e
1
2
−
K
2
e
2
2
+
e
3
(
R
f
L
f
I
d
+
w
I
q
−
1
L
f
(
E
d
−
V
dc
S
d
)
−
I
̇
dref
)
+
e
4
(
R
f
L
f
I
q
−
w
I
d
−
1
L
f
(
E
d
−
V
dc
S
q
)
−
I
̇
qr
e
f
)
(3
5
)
to
make
th
e
de
rivati
ve
of
V
3
ne
gativel
y
def
i
ne
d
al
lows
us
to
hav
e
a
global
sta
bili
ty
of
th
e
s
ys
te
m,
and this
by
c
hoos
in
g
t
he
f
ollo
wing s
witc
hing
contr
ol laws:
=
(
−
+
̇
−
3
3
+
)
(
3
6
)
=
(
+
+
̇
−
4
4
+
)
(
3
7
)
By
us
in
g
a
de
rivati
ve of
the
C
LF, we ca
n
si
mp
li
f
y
the
equ
at
ion
(
3
5
)
as:
̇
3
=
−
1
1
2
−
2
2
2
−
3
3
2
−
4
4
2
≤
0
(
38
)
This
ens
ures
a
gen
e
ral
sta
bili
zat
ion
relat
ed
to
the
dynamic
operati
on
of
the
V
SI,
a
nd
w
hic
h
ma
kes
it
po
s
sible t
o i
nje
ct
an
op
ti
mal
va
lue of t
he
act
ive
powe
r
a
nd a
zer
o
injec
ti
on
of the
reacti
ve powe
r.
3.3.
Desi
gn o
f
DC
-
li
nk
volt
age c
ontrolle
r
T
he
DC
bus
volt
age
c
ontrol
l
oop
must
ens
ure
tw
o
c
ontr
ol
ob
je
ct
ives
the
first
is
t
o
keep
this
volt
age
sta
ble, a
nd the
seco
nd is to
ge
ner
at
e t
he
act
iv
e ref
e
re
nce c
urren
t.
re
qu
i
red f
or the i
nternal
current l
oop.
Ther
e
f
or
e,
t
he
desig
n
of
this
con
t
ro
ll
er
base
d
on
a
P
I
volt
age
lo
op,
w
hich
has
as
in
pu
ts:
the
square
d
of
the
DC
bus
v
oltage
2
a
nd
it
s
re
fere
nce
c
omp
on
e
nt
2
,
an
d
wh
ic
h
ge
ner
at
e
s
at
it
s
outp
ut
a
c
on
tr
ol
sign
al
mu
lt
ipli
ed
by
gri
d
volt
age
to
deter
mi
ne
the
am
plit
ud
e
of
t
he
c
urre
nt
injec
te
d
i
nto
the
gr
id
.
T
he
blo
c
k
diag
ram of
the
con
t
ro
l l
oop
is
sh
ow
n
in
Fi
gur
e 4
.
In close
d
lo
op,
the tunin
g
la
w
is e
xpresse
d b
y
=
(
2
−
2
)
.
(
)
(39)
(
)
is t
he
tra
nsfer
functi
on
of
c
on
ven
ti
onal
P
I
c
ontr
oller,
e
xpres
sed b
y:
+
∗
(
1
)
.
The
deter
minat
ion
of
t
he
c
on
t
ro
l pa
rameters
an
d
m
us
t be precise
in
or
der to
e
ns
ure
t
he
opti
mal
sta
bili
ty
point
and
a
slo
wer
dynamic
co
m
pa
red
to
the
int
ern
al
cu
rr
e
nt
l
oops
,
for
this
pur
po
se
the
P
arti
cl
e
Sw
ar
m
Op
ti
mi
zat
ion
(
PS
O)
c
on
t
ro
l
te
c
hn
i
que
h
as
been
a
dopted
w
hich
has
become
one
of
t
he
fa
vo
rite
s
in
op
ti
miza
ti
on
al
gorithms
s
olu
ti
on
s
,
t
hanks
to
it
s
simpli
ci
ty
of
im
pleme
ntati
on,
it
s
r
obus
tn
ess
a
nd
it
s
ca
pa
ci
ty
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
A
rob
us
t
nonlinear c
on
tr
ol st
ra
te
gy
o
f
a
P
V
System
con
nected to
the t
hr
ee
-
phas
e
g
ri
d
…
(
Sa
l
ma Z
ouga
)
619
to
imp
rove
propo
rtion
al
inte
gr
al
(
PI)
pe
rfo
rma
nces,
by
s
earchi
ng
for
t
he
best
values
in
orde
r
to
m
ake
the
error as s
mall
as possib
le
or ze
ro.
Fig
ur
e
4.
P
SO al
gorithm
b
a
s
ed DC
-
li
nk
Volt
age contr
ol lo
op
3.3.1.
P
art
ic
le
sw
arm
op
tim
iz
at
io
n
algorit
hm
The
PS
O
al
gorith
m
is
co
nsi
der
e
d
one
of
the
m
os
t
ex
cel
le
nt
op
ti
mi
zat
ion
meth
od
s
that
have
unde
rgo
ne
ma
ny
cha
nges
si
nc
e
it
s
introd
uction
i
n
1995,
by
Ke
nn
e
dy
a
nd
Eberha
rt.
PS
O
has
pro
ven
r
obus
t
i
n
so
lvi
ng
c
on
ti
nuous
nonlinea
r
op
ti
miza
ti
on
pro
blems,
it
s
pri
nci
ple
a
dopts
a
conven
ti
on
a
l
swar
m
ma
na
geme
nt
for
bir
ds
.
Eac
h
"
bir
d"
is
re
presente
d
by
a
par
ti
cl
e.
a
nd
r
epr
ese
nts
a
ca
nd
i
date
s
olu
ti
on
t
o
t
he
opti
miza
ti
on
pro
blem.
A
s
w
arm
inclu
des
of
S
-
par
ti
cl
es
fl
ow
i
ng
t
hroug
h
a
mu
lt
idimens
ion
al
searc
h
s
pa
ce,
an
d
each
of
it
s
par
ti
cl
es
a
dju
s
t
its
po
sit
io
n
in
searc
h
sp
ac
e
acco
rd
i
ng
to
it
s
own
e
xp
erience
a
nd
th
at
of
nei
ghbo
ur
i
ng
par
ti
cl
es.
Th
us,
a
par
ti
cl
e
us
e
s
the
good
pos
it
ion
e
nc
ounter
ed
by
it
sel
f
a
nd
t
hat
of
it
s
ne
ighbou
rs
pa
rtic
le
s
to
po
sit
io
n
it
sel
f
towa
rd
s
t
he
m
os
t
opti
m
al
sol
ution
.
The
se
cond
esse
ntial
factor
in
this
te
chn
iq
ue
a
nd
wh
ic
h
al
lows
each
pa
rtic
le
to
c
ha
nge
it
s
posit
io
n
is
t
he
velocit
y
i
nformat
io
n,
w
hich
al
lo
ws
each
par
ti
cl
e
durin
g
updatin
g proce
ss at eve
r
y
it
er
at
ion
to
f
i
nd it
s b
est
a
nd
on
l
y solutio
n
.
every
pa
rtic
le
is
re
pr
ese
nted
as
a
po
i
nt
in
a
j
-
dime
ns
i
on
a
l
sp
ace.
T
he
i
th
par
ti
cl
e
is
r
epr
ese
nted
as
=
(
1
,
2
,
…
,
)
.
T
he
best
previ
ou
s
po
sit
ion
of
it
h
pa
rtic
le
swa
rm
is
re
pr
es
ented
a
s
=
(
1
,
2
,
…
,
)
,
this
is
cal
le
d
pb
e
st.
T
hu
s
,
a
par
ti
cl
e
us
es
th
e
good
posit
ion
e
ncou
ntere
d
by
it
sel
f
an
d
th
at
of
it
s
neig
hbou
rs
par
ti
cl
es
to
po
sit
ion
it
sel
f
to
wards
the
m
ost
opti
mal
s
olu
t
ion
.
T
he
sec
ond
e
ssentia
l
fac
tor
i
n
this te
ch
nique
and whic
h al
lo
ws
eac
h partic
l
e to c
hange
.
V
ij
n
+
1
=
w
∗
V
ij
n
+
r
1
c
1
(
P
ij
n
−
X
ij
n
)
+
r
2
c
2
(
P
gj
n
−
X
ij
n
)
(
40
)
Wh
e
re
is
the
inerti
a
wei
gh
t,
1
and
2
are
acce
le
r
at
ion
coe
ff
ic
ie
nts,
1
and
2
are
ra
ndom
num
ber
s
un
i
formly
d
ist
r
ibu
te
d bet
wee
n
0
a
nd
1,
a
nd
r
epr
ese
nts t
he
it
erati
on num
be
r
.
T
he ne
w po
sit
ion o
f
t
he parti
c
le
s is obtai
ne
d on the
basis
of
it
s
velocit
y
a
nd
it
s p
re
vious
po
sit
ion
:
+
1
=
+
+
1
(
41
)
The
perf
or
m
an
ce
of
eac
h
par
t
ic
le
is
measu
re
d
us
in
g
a
pred
efine
d
ob
je
ct
iv
e
(f
it
ne
ss)
f
un
c
ti
on
,
w
hic
h
def
i
ne
the
c
ha
r
a
ct
erist
ic
s
of th
e opti
miza
ti
on
pro
blem [2
5
].
3.3.2.
PI c
ont
rolle
r based
on PS
O t
ec
hni
que
In
this
w
ork,
the
PSO
al
go
rithm
is
us
e
d
to
deter
mine
the
opti
mal
value
s
of
a
nd
of
the
P
I
con
t
ro
ll
er.
T
he
impleme
ntati
on
str
uctu
re
of
the
PS
O
te
ch
nique
for
t
un
i
ng
the
PI
c
on
t
ro
ll
er
in
t
he
D
C
bus
con
t
ro
l
lo
op
is
represe
nted
in
Figure
4
,
it
s
pe
rformance
e
valuati
on
base
d
on
t
he
obje
ct
ive
f
unct
ion
def
in
ed
by
the mini
mum t
ime mult
ipl
y
s
qu
a
re
d
er
r
or
IT
SE,
a
nd
wh
ic
h ca
n be
represe
nted b
y:
ITSE
=
∫
t
.
e
2
∞
0
dt
(
42
)
Wh
e
re
is
the
PI
co
ntr
oller
input
that
e
xp
resse
d
the
er
r
or
bet
ween
th
e
square
of
t
he
meas
ur
e
d
vo
lt
age
and
t
he
squar
e
of its
r
efere
nce
value
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
D
ri
S
ys
t,
V
ol
.
12
, N
o.
1
,
Ma
rch
20
21
:
612
–
626
620
Figure
5.
imple
mentat
io
n
al
go
rithm
of the
PSO
-
P
I
c
ontr
oller meth
od
The
c
ompu
ta
ti
on
al
flo
wch
a
rt
in
Fig
ur
e
5
re
presents
t
he
al
gorith
m
use
d
i
n
the
impleme
nt
at
ion
of
t
he
PSO
f
or
the
pu
rpose
of
op
ti
m
iz
ing
th
e
par
a
mete
rs
of
the
P
I
.
T
he
Para
mete
r
s
init
ia
li
zat
ion
at
th
e
beg
i
nning
of
the
PS
O
al
gori
thm incl
ud
es
the ma
xim
um
num
ber
of it
e
rati
on
s
that i
s equal t
o 100 itera
ti
on
s,
t
he
siz
e
of
t
he
popula
ti
on
th
at
is
c
onside
re
d
by
30
in
div
i
du
al
s
(p
a
rtic
le
s)
i
n
the
popu
la
ti
on
,
as
well
as
the
acce
le
r
at
ion
coeffic
ie
nts
1
a
nd
2
and
the
i
nert
ia
weig
ht
.
Since
th
e
dimens
ion
of
the
pr
oble
m
co
ncerns
two
pa
ramete
rs
an
d
to
be
co
ntr
olled,
t
he
posit
ion
an
d
t
he
velocit
y
are
r
epr
ese
nted
by
matri
ces
with
the
dimensi
on
of
2
×
30
by
us
in
g
the
(
4
0
)
an
d
(4
1
)
.
At
the
en
d
of
t
he
al
gorith
m
t
he
best
posit
ion
of
a
pa
rtic
le
represe
nts
the
op
ti
mal
values
of the c
ontrolle
r gains
an
d
.
3.
RESU
LT
S
AND A
N
ALYSIS
Simulat
io
ns
a
re
done
i
n
M
A
TLAB/
Sim
ulink
s
of
t
ware
in
orde
r
t
o
ve
rif
y
a
nd
evaluate
t
he
performa
nce
of
th
e
nonlinea
r
co
ntr
oller
bas
ed
on
t
he
Ba
c
ks
te
ppin
g
te
ch
nique,
de
sig
ne
d
i
n
t
he
a
b
ov
e
sect
ion
and
il
lustrate
d
in
Fig
ur
e
3.
T
he
global
par
a
m
et
ers
are
obta
ined
on
the
bas
is
of
the
the
ore
ti
cal
stud
y
sho
wn
i
n
th
e
desi
gn
pa
rt
of
t
he
c
on
tr
ol
le
r
an
d
that
a
r
e
giv
e
n
in
Tab
le
2
,
T
he
sim
ul
at
ion
res
ults
wer
e
ge
ner
at
e
d
un
der
sta
ble
operati
ng c
onditi
ons
w
hich
a
re :(
G=
1000
w/m²,
T=
25°C)
, a
nd o
t
he
r diffe
ren
t c
on
diti
on
s
.
Table
2
.
param
et
ers
of t
he pr
opose
d PV
s
ys
te
m
Para
m
eter
Sy
m
b
o
l
Valu
e
PV
-
ar
ray p
o
wer
P
pv
1562
DC b
u
s cap
acito
r
C
dc
3
×
10
3
Switch
in
g
f
requ
en
cy
F
s
5
Grid
V
g
3
8
0
V/5
0
Hz
Filter
p
ar
am
e
ters
L
R
1
,8mH
0
,05
Ω
PV vo
ltag
e r
eg
u
lat
o
r
k
1
k
2
3
×
10
−
2
5
×
10
−
3
Cu
rr
en
t r
eg
u
lato
r
k
3
k
4
10
7
10
6
PI
-
PS
O
k
i
k
p
0
.8
0
.05
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
A
rob
us
t
nonlinear c
on
tr
ol st
ra
te
gy
o
f
a
P
V
System
con
nected to
the t
hr
ee
-
phas
e
g
ri
d
…
(
Sa
l
ma Z
ouga
)
621
Ca
se
1:
(S
im
u
lation u
nder c
on
s
tant
ir
r
ad
i
an
ce
and
tem
pera
tu
re)
The
Fi
gure
6
-
10
il
lustrate
th
e
beh
a
vior
of
t
he
main
s
ys
te
m
singa
ls
after
a
simulat
ion
unde
r
sta
ble
conditi
ons
(
=
1000
w/
2
;
=
25°C)
,
in
order
t
o
de
m
on
st
rate
the
m
os
t
s
ign
ific
a
nt
aspe
ct
s
of
t
he
c
on
t
rol
strat
egy pr
opose
d.
Figure
6
s
how
s
the
beh
a
vior
of
the
vo
lt
age
gen
e
rated
by
t
he
PV
a
rr
a
y,
wh
ic
h
f
ollow
s
it
s
ref
e
rence
value
at
near
l
y
93.3
V
t
hank
s
to
the
hybri
d
Mpp
t
c
omma
nd
t
hat
intr
od
uces
bo
t
h
te
ch
ni
ques
(Incco
nd
a
nd
Ba
ckstep
ping)
.
T
he
values
of
the
c
urren
t
a
nd
the
i
nput
volt
age
of
the
ch
op
per
s
in
dicat
e
r
especti
vely
,
16
.74
A
an
d
93.3V,
a
nd
t
he
value
of
the
P
V
powe
r
is
ab
out
1561.
8W.
This
ju
sti
fies
the
reli
abili
ty
of
the
exter
nal
vo
lt
age
co
ntr
oller
base
d
on
t
he
lyap
unov
f
un
ct
ion
w
hic
h
ha
s
ens
ur
e
d
a
global
sta
bili
zat
ion
of
the
sy
ste
m
,
an
d
makes
it
possi
ble to e
xtract t
he
in
sta
ntane
ous
vo
lt
a
ge
a
nd
maxim
um
cu
rrent ge
ne
rated
by the
P
V gen
e
r
at
or
.
Figure
6.
I
nput
Volt
age
Figure
7
repres
ents
the
be
ha
vio
r
of
the outp
ut
vo
lt
age o
f
the
chop
per
,
as
ca
n
be
no
ti
ce
d
in
this
figur
e
the
volt
age
f
ollow
s
perfect
ly
(
on
ave
ra
ge)
it
s
ref
e
re
nce
valu
e
(
360),
w
hich
pro
ves
th
e
perf
ormance
of
t
he
DC
bu
s
vo
lt
age
re
gu
la
ti
on
l
oop,
that
intr
oduces
the
PS
O
base
d
P
I
dc
li
nk
c
on
t
ro
ll
er
,
with
the
a
bili
ty
to
quic
kly
and accu
ratel
y conve
rg
e
to
it
s
ref
e
ren
ce
val
ue
s.
Figure
7.
DC
–
bu
s
volt
age
Figure
8
s
how
s
the
outp
ut
c
urren
t
a
nd
th
e
AC
volt
age
wav
e
f
or
m
s,
from
t
his
fi
gure,
we
note
that
there
is
a
ze
r
o
ph
a
se
di
ff
e
rence
betwee
n
t
he
volt
ag
e
a
nd
t
he
injec
te
d
c
urr
ent.
,
w
hich
ma
kes
it
possi
ble to
ha
ve
the
unit
y
pow
er
facto
r
is
we
ll
achieved
a
nd
the
f
reque
nc
y
remai
ne
d
co
ns
ta
nt
an
d
e
qual
to
the
gr
i
d
-
vo
lt
age
fr
e
qu
e
nc
y
.
M
oreo
ver,
the
c
ur
ren
t
injec
te
d
into
th
e
net
wor
k
co
nver
ges
t
o
it
s
ref
e
re
nce
val
ue
a
fter
a
sh
ort
transient p
ha
se
(0
.
07s)
w
hich
means
t
hat
the n
onli
nea
r
cu
rr
e
nt
co
ntr
oller
de
velo
ped
h
as g
i
ven
u
s
t
he
ex
pe
ct
ed
resu
lt
s
an
d
the
co
ntr
ol
t
h
at
w
e
ha
ve
set
up
a
ll
ow
s
us
to
ha
ve
a
un
it
powe
r
facto
r
.
A
nd
a
ccordin
g
to
fi
gure
9,
we
no
ti
ce
that
the
reacti
ve
po
wer
is
re
gu
la
te
d
to
zer
o
a
nd
the
act
ive
powe
r
injec
te
d
into
the
gri
d
reac
he
s
it
s
maxim
um
val
ue
(156
2W)
,
w
hi
ch
will
ensu
re
a total
transfe
r of PV
po
wer.
The
im
pro
vem
ent
of
t
he
qu
al
it
y
of
the
inje
ct
ed
po
wer
is
ens
ur
e
d
by
the
non
-
li
nea
r
c
ontr
ol
lo
op
s
,
wh
ic
h
is
pro
ve
d
by
Fig
ur
e
10
w
hich
s
how
the
T
HD
in
t
he
curre
nt,
as
it
can
be
no
ti
ced
that
t
he
TH
D
in
th
e
gr
i
d
c
urren
t i
s
3,
34
w
hich
is
unde
r
t
he
sta
nda
rd
val
ue
re
quir
ed no
t t
o
e
xcee
d 5%
.
Evaluation Warning : The document was created with Spire.PDF for Python.