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.
1
2
,
No.
2
,
Jun
202
1
,
pp.
125
2
~
126
4
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v
1
2
.i
2
.
pp
125
2
-
126
4
1252
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
A neuro
-
fuzzy a
ppro
ach f
or tracki
ng maxi
mum po
wer poin
t o
f
photo
vo
l
taic sola
r system
Aouatif I
bnel
ouad
1
, Ab
delj
alil
El
ka
ri
2
, Hass
an
Aya
d
3
, and
Mos
tafa
Mjahe
d
4
1,2
Depa
rtment
of
Applie
d
Phys
ic
s.
L
abor
a
tory
of
El
e
ct
ri
ca
l
Sys
tems a
nd
Telec
o
m
muni
c
at
ions,
Ca
di
Ayyad
Univ
er
sity,
Facul
ty
of
Sci
en
ce
s a
nd
T
ec
hnol
ogie
s,
Mar
rak
e
c
h,
Moroc
co.
3,4
Depa
rtment
of
Mathematics an
d
Sys
te
ms,
Roy
a
l
School
of
Aero
naut
i
cs,
Marr
akech,
Moro
cc
o.
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
N
ov
30
, 20
20
Re
vised
M
a
r
15
, 2
0
21
Accepte
d
Apr
20
, 20
21
Thi
s
work
pre
s
ent
s
a
hybr
id
s
oft
-
com
pu
ti
ng
me
thodol
ogy
ap
proa
ch
for
int
ellige
n
t
ma
x
im
um
power
point
tra
ck
ing
(MP
PT)
tech
nique
s
of
a
photovol
taic
(PV
)
sys
te
m
unde
r
any
expe
c
te
d
oper
a
ti
ng
cond
it
ions
using
art
if
ic
i
al
neu
ral
net
work
-
fuz
zy
(
neur
o
-
fuz
zy
).
The
pr
oposed
te
chn
iqu
e
pre
dicts
the
ca
l
cul
a
ti
on
of
the
duty
cycle
ensuring
optimal
po
wer
tra
nsfe
r
bet
wee
n
the
PV
gene
r
at
or
and
the
lo
ad.
The
n
eur
o
-
fuz
zy
hybr
id
method
com
bin
es
artifi
c
ia
l
neur
al
n
et
w
ork
(AN
N)
to
dire
c
t
th
e
cont
r
oll
er
to
the
reg
ion
wher
e
th
e
MP
P
i
s
loc
ate
d
with
it
s
r
efe
r
e
nce
vol
ta
g
e
est
i
ma
tor
and
i
ts
bloc
k
of
neur
al
orde
r.
Aft
er
that
,
the
fu
zz
y
logi
c
cont
rol
le
r
(FLC
)
with
rule
infe
ren
ce
b
egi
ns
to
est
abl
ish
the
photovol
t
aic
solar
sys
te
m
a
t
th
e
MP
P.
The
obta
in
ed
si
mul
a
ti
on
result
s
usi
ng
MA
TL
AB/
simul
i
nk
softwar
e
for
th
e
proposed
appr
oa
ch
com
p
are
d
to
AN
N
and
the
p
ert
urb
and
obser
ve
(P&O),
prove
d
tha
t
n
eu
ro
-
fuz
zy
appr
o
a
ch
fu
lfi
l
le
d
to
e
xtra
c
t
the
opt
imum
power
with
per
ti
nen
ce, effi
c
ie
ncy
and
pr
ec
ision
.
Ke
yw
or
d
s
:
Ar
ti
fici
al
n
e
ur
a
l netw
orks
Fu
zz
y
lo
gic c
ontr
oller
M
a
ximum
pow
er
po
i
nt trac
king
Neur
o
-
fu
zz
y
Photo
vo
lt
ai
c 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
:
Aouati
f
Ibnelo
uad
Dep
a
rtme
nt of
Applie
d Ph
ys
ic
s,
La
borato
ry of Elect
rical
S
yst
ems and Telec
om
m
unic
at
ions
Ca
di Ay
yad Unive
rsity, Fac
ul
ty of S
ci
e
nces
and Tec
hnolog
ie
s
112
B
oule
vard
Abd
el
kr
im
A
l
Kh
at
ta
bi,
M
ar
r
akech 4
0000,
M
or
occo
Emai
l:
ao
uatif
.
ibn
el
oua
d@
ce
d.uca.ma
1.
INTROD
U
CTION
Energ
y
pro
duc
ti
on
is
a
chall
eng
e
of
gr
e
at
imp
or
ta
nce
f
or
the
years
to
co
me.
Th
e
ene
rgy
nee
ds
of
industrial
iz
ed
so
ci
et
ie
s
as
w
el
l
as
de
velo
pin
g
co
untrie
s
a
re
ste
adily
inc
reasin
g.
This
pro
du
ct
io
n
has
triple
d
since
the
1960
s
to
the
prese
nt
day.
All
gl
ob
a
l
ene
r
gy
pro
duct
ion
co
mes
f
r
om
fo
ssil
s
ourc
es.
The
c
ons
umpti
on
of
the
se
sou
rce
s
giv
es r
ise
to g
ree
nhouse g
a
s
emissi
ons
an
d
there
f
or
e
a
n
increase
i
n
poll
ution.
I
n
ad
diti
on,
the
excessive
c
ons
umpti
on
of
nat
ur
al
res
ource
s
tocks
reduces
t
he
rese
rv
e
s
of
this
ty
pe
of
e
ne
rgy
in
a
da
ng
erous
way
for
f
uture
ge
ne
rati
ons.
Re
new
a
ble
e
ne
rg
ie
s
s
uch
as
wi
nd
powe
r,
so
la
r
e
nergy,
bio
ma
ss
ene
rgy
an
d
hydro
powe
r
a
r
e
pro
misi
ng
s
ol
ution
s
to
co
m
pete
with
ma
ss
ene
rgy
sou
rce
s
su
c
h
as
fossil
an
d
nucl
ear
e
nerg
y.
Re
new
a
ble
e
ne
rgy
mea
ns
e
ner
gy
from
th
e
s
un,
wind,
earth
heat,
wa
te
r
or
bi
om
ass
.
Un
li
ke
fo
s
sil
f
uels
,
ren
e
wa
ble en
er
gies are
e
nergies with unli
mit
ed
res
ources
. Solar r
a
diati
on
i
s d
ist
rib
uted o
ver
the e
ntire s
urface
of
the
ea
rth;
it
s
den
sit
y
is
not
gr
eat
a
nd
c
auses
no
c
onf
li
ct
between
c
ountries
un
li
ke
oil.
Amo
ng
these
resou
rces,
so
l
ar
e
nergy
is
c
on
si
der
e
d
t
od
a
y
as
one
of
the
m
os
t
reli
able
re
ne
wab
le
ene
rg
ie
s,
dail
y
a
nd
resp
ect
fu
l
of
t
he
en
vir
onme
nt
the
s
ource
[
1],
[2].
P
ho
t
ovoltai
c
ene
rgy
has
nowa
da
ys
a
n
in
creased
im
port
ance
in
el
ect
ric
al
power
a
ppli
cat
ion
s,
si
nce
it
is
consi
der
e
d
as
an
esse
ntial
ly
inex
ha
us
ti
ble
a
nd
broa
dly
a
va
il
able
energ
y
re
sourc
e [
3]
.
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 n
e
uro
-
f
uz
zy
approac
h f
or
t
ra
cki
ng
m
axim
um power
poin
t of ph
oto
v
oltai
c so
l
ar
syste
m
(
Ao
uati
f Ibnelo
uad
)
1253
Photo
vo
lt
ai
c
(
PV
)
cel
ls
are
usual
ly
ma
nufac
ture
d
of
se
micond
ucto
r
mate
r
ia
ls
capab
le
of
co
nv
e
rtin
g
the
e
nergy of
s
un
li
ght
at
certa
in w
a
vele
ng
t
h
to D
C
el
ect
rici
ty.
W
hen
s
un
li
gh
t
hits
t
he
s
ur
face o
f
a PV
ce
ll
,
the
semic
onduct
or
mate
rial
al
lo
wing
el
ect
r
ons
to
m
ove
from
the
va
le
nce
ba
nd
to
t
he
c
on
du
ct
io
n
ba
nd
a
bs
or
bs
so
me
of
the
photons’
ene
rgy.
The
el
ect
r
on
s
are
the
n
rea
dy
to
flo
w
in
a
cl
os
e
d
-
path
ci
rc
uit
carr
ying
el
e
ct
rical
energ
y
t
o
t
he
l
oad.
Ce
ll
s
a
re
us
ua
ll
y
c
onnec
te
d
in
se
ries
t
o
f
orm
a
P
V
m
odule.
T
he
m
odules
a
re
c
onne
ct
ed
in
diff
e
re
nt
se
ries
an
d
pa
rall
el
topolo
gies
to
r
each
t
he
desi
r
ed
volt
age
an
d
cu
rr
e
nt
le
vel
[4].
T
he
phot
ovoltai
c
sy
s
te
m
c
on
sist
s
of
a
ph
otov
oltai
c
pa
nel
w
it
h
a
po
wer
in
te
rf
ace
an
d
a
load
.
A
sim
ple
DC
/
DC
c
onver
te
r
ci
rcu
it
(Boost)
is
us
ed
as
inter
face
bet
ween
phot
ovoltai
c
pa
nel
PV
an
d
loa
d.
T
his
DC/D
C
conver
te
r
int
end
e
d
to
tra
ns
fe
r
max
imum
e
ne
rgy
f
rom
photov
oltai
c
panel
P
V
t
o
loa
d
a
nd
t
o
e
ns
ure
our
loa
d
cl
os
er
t
o
t
he
MPP.
In
order
to
i
mpr
ov
e
the
ef
fici
ency
of
t
he
photovo
lt
ai
c
generator
(
PV),
i
n
oth
e
r
words
maximize
the
powe
r
delivere
d
to
th
e
loa
d
c
onnect
ed
t
o
t
he
te
r
minals
of
the
generator
,
s
eve
ral
crit
eria
f
or
op
ti
mizi
n
g
the
ef
fici
enc
y
of
t
he
phot
ovoltai
c
sy
ste
m
wer
e
ap
plied
and
te
ch
nique
s
we
re
f
ollo
w
ed
f
or
good
adap
ta
ti
on
a
nd
high
eff
ic
ie
nc
y
[5],
[6].
A
mon
g
t
hese
te
c
hn
i
qu
e
s
is
the
te
c
hn
i
qu
e
of
P
ursuit
of
the
Powe
r
Po
i
nt
M
a
xim
al
or
"Max
i
mum
P
ower
P
oin
t
T
rac
ker, M
PP
T" [7,
8], sev
e
ral
me
thods
a
re m
ent
ion
e
d
i
n
the
b
ibli
ogra
phy:
t
he
P&
O
method
is
a
wi
dely
us
ed
i
n
pract
ic
e
du
e
t
o
it
s
simpli
ci
ty
and
requires
on
ly
measu
res
volt
age
an
d
c
urren
t
of
the
P
V
m
odul
e
[
9]
.
H
oweve
r,
t
his
al
gorith
m
can
osc
il
la
t
e
ar
ound
the
M
a
ximum
P
ower
Po
i
nt
(
MP
P)
unde
r
su
dde
n
s
unli
ght changes
[
10]
-
[12]. Rece
ntl
y, seve
ral resea
r
cher
s
for
phot
ovoltai
c s
ys
te
ms
track
t
he
ma
xi
mu
m
powe
r
by
i
ntell
igent
M
P
PT
te
chn
i
qu
e
s
s
uc
h
a
s
a
rtific
ia
l
neural
netw
ork
(
ANN
)
a
nd
F
uzzy
lo
gic
co
nt
ro
ll
er.
The
a
rtific
ia
l
neu
ral
netw
o
r
k
(ANN)
te
ch
ni
ques
a
re
bei
ng
util
iz
ed
for
photovo
lt
ai
c
ap
pl
ic
at
ion
s,
pr
i
ncipall
y
because
o
f
thei
r
s
ymb
olic
reas
on
i
ng, f
le
xi
bili
ty
a
nd
e
xp
la
nation
capa
bili
ti
es
that
a
re
us
e
f
ul
to
deal
with
st
ron
g
nonlinea
riti
es
and
c
omplex
s
ys
te
ms
[13].
T
he
us
e
of
arti
fi
ci
al
neu
ral
network
(
ANN)
i
n
ph
otovo
lt
ai
c
sy
ste
ms
has
bee
n
c
on
s
idere
d
by
sev
eral
researc
he
r
s
[
14]
-
[
16].
F
uzzy
l
og
ic
co
ntr
oller
has
be
en
co
ns
ide
re
d
a
s
a
n
eff
ic
ie
nt
a
nd
eff
ect
ive
t
oo
l
in
ma
nag
i
ng
un
ce
rtai
nties
a
nd
no
nlinearit
ie
s
of
s
ys
te
ms
[17
].
A
f
uzz
y
lo
gic
con
t
ro
ll
er is
g
e
ner
al
ly
d
esi
gne
d
in
the lig
ht
of e
xp
e
rience
and e
xp
e
rt
knowle
dg
e
[1
7]
-
[
20].
This
pa
pe
r
pre
sents
a
no
vel
M
PP
T
meth
od
ology
base
d
on
a
hy
br
i
d
m
odel
betwee
n
t
wo
im
porta
nt
intel
li
gen
t
M
P
PT
meth
ods.
This
hy
br
i
d
model:
Ne
uro
-
f
uzzy
ap
proa
ch
de
fine
s
of
mu
lt
i
-
la
ye
red
feed
forw
a
r
ded
a
rtific
ia
l
neuron
netw
ork
a
nd
the
infe
re
nce
-
base
d
ta
ble
of
the
f
uzz
y
l
og
ic
c
on
t
ro
ll
e
r.
T
he
arch
it
ect
ure
of
the
arti
fici
al
ne
ur
al
netw
ork
com
posed
of
three
la
yer
s:
i
nputs
,
hi
dd
e
n
a
nd
outp
ut
la
ye
rs.
T
he
pro
po
se
d
i
ntell
igent
M
PP
T
m
et
hod
arti
fici
al
ne
ur
al
netw
ork
is
to
direct
t
he
co
ntro
ll
er
to
the
re
gion
wh
e
re
the
M
PP
is
l
ocated
with
it
s
ref
e
r
ence
volt
age
e
sti
mator
a
nd
it
s
bl
ock
of
ne
ural
or
der
.
A
fte
r
that,
the
f
uzz
y
lo
gic
with rule i
nf
e
r
ence b
e
gins t
o est
ablish the
phot
ovoltai
c so
l
ar s
ys
te
m at th
e max
im
um
po
wer
point (
M
P
P)
. T
he
hybri
d
model:
Neur
o
-
fu
zz
y
a
ppr
oach
ai
ms
t
o
decr
ea
se
the
com
plexity
of
the
photov
oltai
c
so
la
r
s
ys
te
m
and
to
extract
the
m
aximum
po
we
r
at
the
mini
mu
m
ti
me
with
per
ti
nen
ce
and
e
ff
ic
ie
nc
y
under
a
ny
w
eat
he
r
c
onditi
ons
c
ompa
red
to
the
sing
le
A
N
N
a
nd
c
onve
ntio
nal
MPPT
met
hod
P&
O.
T
o
de
velo
p
t
he
Ne
uro
-
F
uzz
y
appr
oach
met
hod,
t
his
wor
k
is
struct
ur
e
d
a
s
f
ollo
ws:
Sec
ti
on
2
over
vie
w
of
photov
oltai
c
so
la
r
s
ys
te
m
by
making
a
fo
c
us
on
the
m
odel
,
the
c
har
act
eri
sti
cs
of
a
P
V
modu
le
a
nd
pr
esents
the
goal
of
DC/DC
c
onve
rter.
Sect
ion
3
desc
ribes
a
cl
assic
al
M
PP
T
met
hod
is
wi
dely
use
d
at
the
li
te
ratur
e
P
&O.
Se
ct
ion
4
pr
ese
nt
s
the
pro
po
se
d
ap
proach
Neuro
-
F
uzzy.
Sect
ion
5
desc
ribes
th
e
detai
l
sim
ulati
on
res
ults
c
ompari
ng
the
nove
l
appr
oach
with
the
si
ng
le
A
NN
an
d
t
he
P
&O
MPPT
me
thod
a
fter
t
hat
it
pr
e
sents
th
e
co
mp
a
rison
of
no
vel
M
PP
T met
hodolo
gy
ne
uro
-
f
uz
zy
in
sta
te
of t
he
a
rt, follo
we
d by the c
oncl
usi
on in Sec
ti
on 6
.
2.
OVERVIEW
OF PH
OTO
V
OLT
AIC SO
LAR SY
STE
M
The
gl
obal
of
the
stud
i
e
d
s
ys
te
m
s
hows
i
n
Fig
ur
e
1,
c
ompose
d
of
a
305W
ph
otovo
lt
ai
c
so
la
r
gen
e
rato
r
c
onne
ct
ed
to
a
po
w
er
el
ect
ronic
el
ement.
This
el
ement
c
onsist
s
of
a
DC
–
DC
c
onve
rter
that
a
ssu
re
s
impeda
nce
a
da
ptati
on
betw
een
the
ph
otovo
lt
ai
c
so
la
r
gen
e
rato
r
an
d
the
load
r
esi
sti
ve
by
tra
ck
ing
the
maxim
um
power
by
t
he
ne
uro
-
f
uzzy
a
ppr
oach
netw
ork
.
In
the
f
ollow
i
ng
parag
raphs
,
al
l
the
blo
ck
s
of
the
photov
oltai
c sol
ar PV s
ys
te
m
are
descr
i
bed in
detai
l.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
12
, N
o.
2
,
J
une
202
1
:
125
2
–
126
4
1254
Figure
1. O
verview
of the
ne
uro
-
f
uzzy ap
proach net
wor
k
M
PP
T
photov
ol
ta
ic
so
la
r
P
V
s
ys
te
m
2.1.
Photov
olt
aic s
ola
r
mo
dule
The
PV
s
olar
modu
le
use
d
i
n
t
his
st
udy
co
ns
ist
s
of
po
l
yc
ry
sta
ll
ine
sil
ic
on
s
olar
cel
ls
e
le
ct
rical
ly.
Its
main elec
tric
al
sp
eci
ficat
io
ns
are s
how
n
in
T
able 1.
Table
1.
T
ech
ni
cal
d
at
a of the
model ma
nufa
ct
ur
er
s
unpowe
r
SPR
-
30
5E.
Maximu
m
Powe
r
(
W
)
305
Op
en
cir
cu
it vo
lta
g
e Voc (
V
)
6
4
.2
Sh
o
rt
-
circuit cur
re
n
t I
sc (
A)
5
.96
Cu
rr
en
t at
m
ax
im
u
m
po
wer
po
in
t I
m
p
(A)
5
.58
Vo
ltag
e at
m
ax
im
u
m
po
wer
po
in
t Vmp
(
V)
5
4
.7
2.2.
Simul
at
i
on
m
od
el
of a PV
gener
ator
The
mathemat
ic
al
models
of
the
P
V
ge
nerat
or
a
re
def
i
ne
d
in
the
f
ollo
wing
e
qu
at
i
ons.
Fi
gure
2
sh
ows
the
eq
ui
valent
ci
rcu
it
of
a
so
la
r
cel
l
usi
ng
a
si
ngle
di
od
e
m
od
el
du
e
to
accu
rac
y
f
or
ph
otovo
lt
ai
c
(PV
)
stud
ie
s.
A
s
ola
r
pa
nel
is
co
m
po
s
ed
of
sev
e
r
al
photovo
lt
ai
c
cel
ls
employin
g
series or
p
a
ra
ll
el
or
series
–
pa
rall
e
l
exter
nal con
ne
ct
ion
s.
Figure
2. Eq
ui
valent circ
uit o
f
s
olar
cel
l
The follo
wing
equ
at
io
ns desc
ribe
t
he
I
–
V
c
ha
racteri
sti
c of
a so
la
r
cell
[21]:
ℎ
=
+
+
(1)
=
[
(
+
)
−
1
]
(2)
=
(3)
=
(
+
(
)
)
(4)
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 n
e
uro
-
f
uz
zy
approac
h f
or
t
ra
cki
ng
m
axim
um power
poin
t of ph
oto
v
oltai
c so
l
ar
syste
m
(
Ao
uati
f Ibnelo
uad
)
1255
ℎ
=
(
)
(
+
(
)
)
(5)
Af
te
r
co
mb
i
nation
of
the
e
qua
ti
on
s
a
bove
,
th
e
ge
ner
al
iz
e
c
urre
nt
volt
age
e
qu
at
io
n
of
a
phot
ovoltai
c
(P
V
) mo
del is:
=
ℎ
−
(
(
+
)
−
1
)
−
(
+
)
(6)
Wh
e
re:
I
pv
:
The
P
V
c
urre
nt;
I
ph
:
has
a
li
near
relat
io
ns
hi
p
with
li
gh
t
intensit
y
a
nd
var
ie
s
with
te
mp
e
ratur
e
va
ri
at
ion
s;
I
d
:
The
Shoc
kley
diode
eq
uat
ion
(A);
I
o
:
T
he
sat
ur
at
e
d
re
ve
rse
c
urren
t;
“
a
”:
the
c
onsta
nt
known
a
s
the
diode
ideal
it
y
facto
r;
V
T
:
The
ther
ma
l
vo
lt
age
ass
oc
ia
te
d
with
the cel
ls;
N
s
:
The
numb
e
r
of
cel
ls
connecte
d
in ser
ie
s;
“
q
”:
T
he
c
harg
e
of
the
el
ect
r
on;
K:
The
Bolt
zman
n
c
onsta
nt
;
T
:
The
abs
ol
ute
te
m
per
at
ur
e
of
the
p
–
n
j
unct
ion
;
I
sc
:
The
s
hort
ci
rcu
it
cu
rr
e
nt;
K
i
:
The
coe
ff
i
ci
ent
of
short
-
ci
rcu
it
cu
rr
e
nt
var
ia
ti
on
with
te
mp
erat
u
re;
G
:
The
li
gh
t
intensit
y.
R
s
and
R
p
:
ar
e
the
series
an
d
pa
rall
el
equ
i
valent
resist
a
nc
es
of
t
he
so
la
r
pa
nel
res
pect
ively;
∆T=T
-
T
n
: T
he deviat
io
n from
stan
dard tem
pe
ratur
e
.
2.3.
Inf
luence
of
te
mpera
tu
re
and ir
radi
at
i
on
on
PV op
era
ting
Fo
r
va
rio
us
va
lues
of
t
he
ir
rad
ia
ti
on
G,
a
nd
cel
ls’
te
m
pe
ratur
e
T,
the
I
-
V
c
ha
racter
ist
ic
s
of
t
he
analyse
d PV
panel are
sho
wn r
es
pecti
vely
in
Fig
ur
e
3 an
d F
igure
4.
Figure
3. I
nf
l
ue
nce
of tempe
r
at
ur
e
with
const
ant ir
rad
i
at
ion
Figure
4. I
nf
l
ue
nce
of irra
diati
on
with
c
onsta
nt
te
mp
erat
ur
e
Dep
e
ndin
g
on
weathe
r
co
ndit
ion
s
,
a
P
V
generator
co
nnect
ed
to
a
loa
d
ca
n
op
e
rate
in
a
l
arg
e
mar
gi
n
of
c
urre
nt
an
d
vo
lt
age
[
22].
F
igure
3
a
nd
Fi
gure
4
sho
w
th
at
the
open
ci
r
cuit
vo
lt
a
ge
V
co
is
incre
asi
ng
with
the
ir
rad
ia
ti
on
and
decr
e
asi
ng
sli
ghtl
y
a
s
t
he
cel
l
te
m
per
at
ur
e
inc
reases
.
On
the
on
e
ha
nd,
t
he
short
c
ircuit
current
Isc
is l
inearl
y
de
pe
nd
i
ng
on the am
bi
ent irr
a
diati
on
in d
irect
propo
rtion, w
hile t
he
o
pe
n
ci
rc
uit v
oltage
decr
ease sl
i
gh
t
ly as th
e cel
l t
empe
ratur
e i
ncrea
ses. Th
e
r
ef
or
e, the
ma
ximum p
ow
e
r
that c
ou
l
d
be ge
ner
a
te
d
by
a
P
V
s
ys
te
m
is
sli
gh
tl
y
depend
i
ng
on
the
te
mp
e
ratu
re
and
ir
rad
ia
ti
on
va
riat
ion
s:
th
e
ma
ximum
powe
r
increases
as
t
he
irrad
ia
ti
on
in
creases
a
nd
v
ic
e
ver
sa
,
on
the o
the
r
ha
nd
a
P
V
ge
nerat
or
p
e
rforms
bette
r
f
or
lo
w
te
mp
erat
ur
e
th
an rai
sed o
ne [
12].
2.4.
The ch
anges
on tem
pera
tu
re
a
n
d irra
dia
tion
It
is
known
tha
t
te
mp
erat
ur
e
may
be
hi
gh
de
sp
it
e
the
ve
ry
li
ttle
pr
ese
nce
of
an
y
i
rr
a
diati
on
cl
ou
ds
.
It
is
al
so
know
n
that
the
te
mp
e
ratur
e
c
ha
nge
and
t
he
irr
adia
ti
on
dis
posed
r
el
at
ively,
the
i
rr
a
diati
on
i
ncrea
ses,
more
heat tra
diti
on
al
ly i
ncr
eas
ed,
a
nd
vic
e ve
rsa.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
12
, N
o.
2
,
J
une
202
1
:
125
2
–
126
4
1256
2.5.
DC
/
DC b
oo
s
t c
onverter
A
DC/DC
co
nverter
th
e
tra
nsfer
of
ma
ximum
e
nerg
y
from
ph
otovo
lt
ai
c
pa
nel
P
V
to
loa
d.
A
DC/D
C
conve
rter
is
th
e
interface
tha
t
regulat
es
the
ada
ptati
on
be
tween
t
he
phot
ovoltai
c
PV
pa
nel
an
d
t
he
l
oad
to
ens
ur
e
our
loa
d
cl
os
er
to
the
M
PP
.
Fig
ur
e
5
sho
ws
t
he
el
e
ct
rical
ci
rcu
it
of
the
DC
-
DC
conve
rter
B
oos
t
typ
e
.
The
Bo
os
t
t
ype
co
nverte
r
is
a
volt
age
boost
er.
In
t
his
c
onver
te
r
,
t
he
val
ue
of
t
he
outp
ut
vo
lt
a
ge
is
a
lways
gr
eat
er
tha
n
th
at
of
t
he
in
put. Th
e
in
duct
anc
e
curre
ntly
sto
r
e
s
ene
rgy.
Wh
en
the
switc
h
is
off
(the
i
deal switc
h
is
ope
n)
,
t
he
l
oa
d
receive
s
t
his
e
nergy
i
n
a
dd
it
ion
t
o
t
he
GPV
e
ne
rgy.
I
n
th
is
ty
pe
of
c
onve
rter,
if
we
c
onside
r
that
Vin
is
the v
oltage
of
t
he
GPV,
V
out
is
t
he
volt
age o
f
t
he
load
an
d
D
is
the d
ut
y
c
ycle,
then
the r
el
at
ion
s
hi
p
betwee
n
the
se
vo
lt
age
s a
nd th
e load res
ults i
n
the
(
7
)
:
=
1
(
1
−
)
(7)
Figure
5.
Bo
ost
conve
rter
DC
/DC
3.
MPPT
US
I
N
G (
P
&O)
M
E
THOD
The
pri
nci
ple
of
this
ty
pe
of
c
on
tr
ol
is
bas
ed
on
the
distu
rb
a
nce
of
the
value
of
the
v
oltage
of
the
GPV
an
d
th
e
ob
s
er
vation
of
the
be
ha
viou
r
of
the
res
ulti
ng
powe
r
[23]
.
Fi
gure
6
shows
the
al
go
r
it
hm
associat
ed
with
a
P&
O
co
m
man
d.
We
note
that
we
nee
d
two
sens
ors
to
measu
re
t
he
powe
r
of
the
G
PV
as
a
functi
on
of
ti
me.
To
da
y,
t
he
P&
O
al
go
rithm
is
wide
ly
us
e
d
beca
us
e
of
it
s
simpli
ci
ty
an
d
ease
of
impleme
ntati
on
.
I
n
an
oth
e
r
s
en
se,
it
has
s
ome
disad
va
ntages.
F
or
e
xam
ple,
acco
rd
i
ng
to
the
c
har
act
erist
ic
curve
P
-
V
of
PV
pa
nel
we
c
an
ne
ver
re
ach
ΔP
=
0
.
Eac
h
ti
me
V
i
ncr
ea
ses
or
decr
eas
es
the
power
will
be
change
d
w
hic
h
ma
kes
the
i
mp
le
me
ntati
on
of
t
he
ste
p
P
pv
k+1
=P
pv
k
in
the
al
gorith
m
without
pr
of
it
.
This
instabil
it
y
in
t
he
value
of
P
will
le
ad
to
in
sta
bili
ty
arou
nd
the
opti
mal
value
of
the
powe
r.
H
ow
e
ve
r,
this
instabil
it
y
can
be red
uce
d by
minimi
zi
ng the
incr
e
ment
val
ue of
the sea
rc
h
al
go
rithm.
Figure
6. Flo
w
char
t
of the
al
gorith
m
of
a
P&
O
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 n
e
uro
-
f
uz
zy
approac
h f
or
t
ra
cki
ng
m
axim
um power
poin
t of ph
oto
v
oltai
c so
l
ar
syste
m
(
Ao
uati
f Ibnelo
uad
)
1257
4.
NEU
RO
-
FUZ
ZY
MAXI
MUM P
OWER
P
OIN
T
APP
R
OACH
The
Neur
o
-
F
uz
zy
a
pproach
consi
sts
of
tw
o
sta
ge
s;
the
first
one
is
c
ompo
s
ed
of
m
ulti
-
la
yer
e
d
fee
d
forw
a
r
ded
arti
fici
al
ne
ur
al
ne
twork
.
T
he
a
rch
it
ect
ure
c
omp
os
e
d
of
th
r
ee
la
ye
rs:
in
pu
ts,
hi
dden
a
nd
outp
ut
l
a
y
e
r
s
w
hile t
he
seco
nd
on
e
is a fuzz
y
-
ru
le
-
ba
sed.
Figure
7
s
how
the
pro
posed
structu
re
of
th
e
neuro
-
f
uzz
y
appr
oach.
The
hybri
d
m
odel
is
com
posed
of a
neural m
odel
and a
fuzz
y
l
o
g
i
c
c
ontr
oller.
The
r
ole
of
th
e
ne
ural
m
ode
l
is
to
sea
rch
for
t
he
re
gion
w
he
re
M
P
P
i
s
locat
e
d
a
nd
the
fu
zz
y
con
t
ro
ll
er
hel
ps
to
fin
d
a
nd
e
sta
blish
t
he
MPP
in
that
re
gio
n.
T
his
a
ppr
oa
ch
c
onsist
s
th
e
same
MPPT
Fu
zz
y
log
ic
c
ontr
oller,
but
we
will
decr
ease
the
pa
ce
of
the
duty
cycle
beca
us
e
we
need
a
hi
gh
de
gr
ee
of
pr
eci
sion
,
on
the
one
ha
nd.
On
the
othe
r
ha
nd,
t
he
rol
e
of
t
he
ne
ur
a
l
netw
ork
is
t
o
direct
t
he
c
ontrolle
r
to
the
r
egio
n
wh
e
re
t
he
MP
P
is
l
ocated
.
T
her
e
fore,
we
m
us
t
first
buil
d
t
he
ne
ur
al
net
w
ork
t
hat
is
pre
pa
rin
g
a
le
ar
ning
base
and
le
ar
n
the
netw
ork,
t
he
n
impleme
nt
this
ne
ur
al
netw
ork
in
the
c
ontr
ol
ci
rc
uit,
f
ollow
e
d
by
f
uzz
y
lo
gic
con
t
ro
ll
er.
Figure
7. The
pro
po
se
d
st
ru
ct
ur
e
of the
ne
ur
o
-
fuz
z
y
a
ppr
oa
ch
4.1.
The MP
PT
co
nt
r
oller
with
ANN
c
ontroll
er
The
ne
w
te
c
hniqu
e,
w
hic
h
c
hoose
s
the
purs
uit
of
the
ma
xi
mu
m
power
point,
is
the
ne
ural
meth
od.
We
will
app
ly
it
to
appr
ox
im
at
e
the
outp
ut,
wh
ic
h
is
the
volt
age
that
c
orr
esp
onds
t
o
this
powe
r,
as
a
f
unct
ion
of
irra
diati
on
c
hanges
,
a
nd
te
mp
e
ratur
e
,
is
the
tra
cki
ng
of
the
var
ia
ti
on
of
the
ma
xim
um
power
po
i
nt.
Wh
e
re
ou
r
s
ys
te
m
n
ee
ds
t
o
e
vo
l
ve,
quic
kly an
d
e
ff
i
ci
ently.
4.1.1.
Mathem
ati
cal
mo
d
el
li
ng
of
an
ar
tifici
al n
euron
The
mat
hemat
ic
al
mo
del
of
an
arti
fici
al
neur
on
is
il
lu
strat
ed
in
Fi
gure
8.
A
ne
uron
c
onsist
s
essenti
al
ly
of
an
integ
rato
r
that
perf
orms
the
wei
gh
te
d
s
um
of
it
s
inpu
ts.
The
re
su
lt
n
of
this
s
um
is
the
n
trans
forme
d
by
a
tra
nsfer
f
unc
ti
on
f,
w
hic
h
pro
duces
t
h
e
ou
tpu
t
D
of
t
he
ne
uro
n.
The
R
i
nputs
of
t
he
ne
uro
ns
corres
pond
to
the
vecto
r
P
=
[
p
1
p
2
…
…p
R
]
T
,
wh
il
e
W=
[
W
1,1
W
1,2
...
.
W
1,R
]
T
,
rep
re
s
ents
the
vect
or
of
the
weig
hts
of
t
he neu
r
on. T
he out
pu
t
n of t
he
i
nt
egr
at
or
is
giv
e
n by the
f
ollowi
ng
e
quat
io
n
[
15
]
,
[
24]:
=
∑
(
1
,
=
1
)
−
;
=
[
(
1
,
1
1
)
+
(
1
,
2
2
)
+
.
.
.
.
.
+
(
1
,
)
]
−
(8)
This ca
n
al
s
o b
e wri
tt
en
in
ma
trix fo
rm:
=
(
)
−
(9)
=
(
)
=
(
(
)
−
)
(10)
This
outp
ut
c
orres
ponds
to
a
weig
hted
s
um
of
weig
hts
a
nd
inputs
min
us
wh
at
is
cal
le
d
t
he
bias
b
of
the
neur
on.
Th
e
res
ult
n
of
t
he
wei
gh
te
d
sum
is
cal
le
d
t
he
act
ivati
on
le
ve
l
of
t
he
ne
uro
n.
T
he
bias
b
i
s
al
so
cal
le
d
the
act
ivati
on
t
hr
es
hold
of
t
he
ne
uro
n.
W
he
n
the
a
ct
ivati
on
le
vel
reaches
or
e
xc
eeds
the
th
reshold
b
,
then
the
a
rgu
ment
of
bec
ome
s
posit
ive
(
or
ze
ro).
Othe
rw
ise
,
it
is
ne
gative
[
15
]
,
[
24]
.
The
re
is
a
n
ob
vious
analo
gy w
it
h b
iolog
ic
al
n
e
u
r
o
n
s
as s
how
n
i
n
Ta
ble
2.
Unde
r
MATL
AB/
simuli
nk
,
t
he
r
ole
of
th
e
neural
netw
or
k
is
to
direct
the
co
ntr
oller
t
o
t
he
re
gion
wh
e
re
t
he
MP
P
is
l
ocated.
T
hu
s
,
it
is
nec
es
sary
to
b
uild
th
e
ne
ur
al
net
work,
i.e.
to
p
re
pa
re
a
le
ar
ning base
an
d
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
12
, N
o.
2
,
J
une
202
1
:
125
2
–
126
4
1258
to
le
arn
t
he
ne
twork
,
an
d
the
n
impleme
nt
this
ne
ur
al
network
i
n
the
c
ontr
ol
ci
rcu
it
.
T
he
act
i
vatio
n
f
un
ct
io
n
makes
it
possi
ble
to
de
fine
t
he
inter
nal
sta
te
of
t
he
ne
uron
a
ccordin
g
t
o
it
s
total
input.
T
he
re
are
se
ver
al
typ
es
of
act
ivati
on
f
un
ct
io
ns
[25].
The
act
ivati
on
f
unct
ion
us
e
d
in
our
ne
ur
al
netw
ork,
wh
ic
h
is
a
neural
ne
twork
mu
lt
il
ayer
is
th
e
sigm
oid f
un
c
ti
on
for
the
h
i
dden
laye
r
a
nd the linea
r
f
unct
ion f
or the
outp
ut laye
r.
Figure
8. The
a
rtific
ia
l neuron
models
Table
2.
A
nalo
gy b
et
ween t
he
b
iol
og
ic
al
ne
uro
n
a
nd the
forma
l
neur
on
Bio
lo
g
ical
n
eu
ron
Fo
rm
al neu
ron
Sy
n
ap
se
W
eig
h
t of
co
n
n
ectio
n
s
Den
d
rites
Co
n
n
ectio
n
s o
f
o
th
er
n
eu
ron
s to
neu
ron
K
Ax
o
n
Co
n
n
ectio
n
s fr
o
m
neu
ron
k to
oth
er
n
eu
ron
s in
the
n
etwo
rk
Co
re
Activ
atio
n
f
u
n
ctio
n
4.1.2.
Multilaye
r
net
wo
r
k (
multil
ay
er perceptr
on
MLP
)
An
M
LP
is
ma
de
up
of
se
ver
a
l
la
yer
s:
an
in
put
la
ye
r,
on
e
or
m
or
e
hi
dden
or
inte
rme
diate
l
ayer
s
,
a
nd
an
outp
ut
la
ye
r
.
Tw
o
s
uccessi
ve
la
ye
rs
are
f
ully
c
onnected
,
an
d
al
l
co
nn
e
ct
ion
s
a
re
unid
irect
ion
al
.
In
s
uch
a
netw
ork, t
her
e
are
no con
necti
on
s
bet
wee
n
t
wo n
e
uro
ns
of
the same
laye
r. A
n MLP
h
as
t
h
ere
fore:
1
)
An i
nput
layer
that r
ece
ives t
he
d
a
ta
to
b
e p
rocess
ed;
2
)
One
or mo
re
int
er
m
ediat
e or
hidde
n l
ayers
perform
ing t
h
e sp
ecific
proc
essing
of t
he
n
etwor
k;
3
)
An o
utpu
t lay
er
t
hat pr
es
ent
s th
e netw
or
k respon
ses.
The
pur
pose
of
le
arn
i
ng
is
to
est
imat
e
netw
ork
par
a
mete
rs
by
mi
nimizi
ng
a
n
e
rror
f
un
ct
io
n.
Learn
i
ng
is
sup
er
vised
.
T
he
error
functi
on
thu
s
represe
nts
the
distan
ce
that
exists
between
t
he
cal
c
ulate
d
respo
ns
e
of
t
he
net
work
a
nd
it
s
de
sired
res
pons
e
.
T
he
le
arn
i
ng
co
ns
ist
s
in
a
pp
l
ying
t
o
the
netw
ork
pa
irs
of
inputs
an
d
ou
t
pu
ts (d
e
sired
outp
uts),
a
nd
th
en
ap
ply
i
ng
a
le
arn
i
ng
al
gorithm
to m
od
if
y
t
he
va
rio
us
p
ara
mete
r
s
of
the
netw
ork
.
T
he
le
ar
ning
al
gorithm
use
d
f
or
t
his
ty
pe
of
netw
ork
is
t
he
gra
dient
bac
k
pro
pa
gation
(G
BP
)
[23].
The
struct
ur
e
of
the
ne
ural
netw
ork
us
e
d
in
the
c
on
tr
ol
sy
ste
m.
T
his
ne
twork
ha
s
a
n
in
p
ut
la
yer
con
ta
ini
ng
t
w
o
in
pu
ts
(
Irra
diati
on
a
nd
T
empe
rature),
a
hidden
la
yer
of
9
ne
uro
ns
and
a
n
outp
ut
la
yer
con
ta
ini
ng a si
ng
le
ne
uro
n
(t
he
volt
age
V).
At
the
e
nd
of
t
he
le
a
rn
i
ng
phase,
we
obta
in
the
final
ne
ur
al
net
wor
k
im
ple
mentat
io
n,
w
hi
ch
giv
e
s
us
a
value
very
cl
os
e
to
t
he
exac
t
value
of
the
M
PP
.
It
a
dm
it
s
as
inputs
the
irrad
ia
ti
on
a
nd
te
mp
e
ratu
re
and
a
s
ou
t
pu
t,
the
volt
age clo
se to
th
e M
P
P
[26].
4.2.
Th
e
MPP
T Contr
oller
with
Fuzzy
L
ogic
A
Fu
zz
y
L
og
i
c
Co
ntr
ol
(
FL
C
)
is
use
d
to
w
ork
as
a
n
M
PP
T
c
ontr
oller
that
trac
ks
the
op
ti
ma
l
op
e
rati
ng
poin
t
of
a
P
V
pa
nel.
F
uzz
y
Lo
gic
Co
ntr
ol
is
on
e
of
the
mo
st
us
ed
te
chn
i
qu
e
s
in
dif
fer
e
nt
eng
i
neer
i
ng
c
ha
ll
eng
es
of
it
s
mu
lt
i
-
r
ule
-
base
d
cha
racteri
sti
cs
[27].
F
uzzy
l
og
ic
c
on
t
ro
l
ha
s
a
simple
an
d
cl
ear
proce
dure
beca
us
e exact mat
he
mati
cal
mo
del
li
ng
and tech
ni
cal
q
ua
ntit
ie
s o
f
a sy
ste
m a
re
no
t re
quire
d
f
or this
con
t
ro
ll
er
[28
]
.
The
fu
z
zy
c
ontr
oller
co
ns
ist
s
of
t
hr
ee
blo
c
ks
:
the
fi
rst
bl
ock
f
uzzifica
ti
on
w
hich
num
erical
input
va
riables
(
Vpv
,
P
pv
)
a
r
e
conve
rted
i
nt
o
li
nguisti
c
va
riable
(E
,
DE
)
base
d
on
a
membe
rs
hip
f
un
ct
io
n.
The
sec
ond
bl
ock
is
devoted
to
infe
ren
ce
rul
es,
wh
il
e
the
l
ast
blo
ck
is
the
defuzzifi
cat
io
n
f
or
returnin
g
to
the
real
domai
n
(
D
).
T
his
la
st
ope
rati
on
us
es
the
centre
of
mass
to
deter
mine
t
he
value
of
the
ou
t
pu
t
[29
].
Fi
gure
9
s
hows
the
ba
sic
stru
ct
ur
e
of
the
us
e
d MPPT F
uzzy co
ntr
ol
le
r
[29
]
.
Figure
9.
Bl
oc
k diag
ram of
th
e FLC “S
UP
R
I
M
E”
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 n
e
uro
-
f
uz
zy
approac
h f
or
t
ra
cki
ng
m
axim
um power
poin
t of ph
oto
v
oltai
c so
l
ar
syste
m
(
Ao
uati
f Ibnelo
uad
)
1259
Fo
r
t
he
M
P
PT
con
t
ro
ll
er
with
f
uzz
y
l
og
ic
,
th
e
in
pu
ts
a
re
ta
ken
as
a
cha
ng
e
in
po
wer
a
nd
volt
age
as
well
. T
h
ere is
a b
loc
k f
or
cal
culat
ing
t
he
e
rror (E
)
a
nd the
change
of the
e
rror (
DE) at
sa
mp
li
ng in
sta
nts
k
:
(
)
=
(
)
−
(
−
1
)
(
)
−
(
−
1
)
(11)
=
(
)
−
(
−
1
)
(
12)
Wh
e
re,
Ppv
(k)
is
the
po
wer
de
li
ver
ed
by
P
V
pa
nel
an
d
V
pv
(k)
is
t
he
te
rmi
nal
volt
age
of
t
he
m
odule
at
samp
le
k.
Fu
zzi
ficat
io
n
:
The
r
esulti
ng
l
inguist
ic
var
ia
bles
ha
ve
been
us
e
d
f
or
t
he
M
PP
T
f
uzz
y
c
on
t
ro
ll
er:
PB
(
posit
ive
big
),
PS
(
posit
ive
s
mall
),
Z
E
(
zer
o
)
,
NS
(
negat
ive
s
mall
)
a
nd
NB
(
ne
gative
b
ig
)
for
e
xpres
sing
the
reel
in
puts
a
nd
outp
ut
va
riabl
es.
Fi
gure
10a,
Fig
ure
10b
a
nd
Fig
ur
e
11
il
lustrate
t
he
me
mb
e
rsh
i
p
f
un
ct
ion
s
of
five fuzz
y su
bse
ts for t
he
in
pu
t’s v
a
riables
E
and DE
a
nd th
e outp
ut v
a
riab
le
D
.
(a)
(b)
Figure
10. Me
mb
e
rsh
i
p funct
ion
s
(
a
)
the
er
r
or E
(b)
the
ch
ang
e
of t
he
er
r
or D
E
Figure
11. Me
mb
e
rsh
i
p funct
ion
s
of
ou
t
pu
t
var
ia
bles D
Infer
e
nce
ru
le
s
:
Ta
ble 3 s
how
s the
ru
le
s
tabl
e of the
fuzzy
con
t
ro
ll
er
w
here al
l i
nputs in
the mat
rix
are [E,
DE
]
[
30]
.
D
ef
uzzifica
ti
on
:
The
pr
oc
ess of
defuzzifi
cat
ion
c
onve
rts
the infer
re
d
f
uz
zy
c
on
t
ro
l act
ion
into a n
um
e
rical
v
al
ue a
t t
he ou
t
pu
t
(
D)
by
making t
he
c
ombinati
on
of t
he ou
t
pu
ts
r
es
ul
ti
ng
from eac
h r
ule. In
this pa
per the
c
entre
of gra
vity
def
uzzifier
, whic
h
is t
he
m
os
t c
om
m
on
one, is
ado
pted.
In the
Fig
ur
e
12 is
sh
ow
n
th
e s
urf
ace o
utput D=
f
(
E,
DE
) of
t
he
M
P
PT c
ontr
ol
le
r.
Table
3.
T
he
fuzzy
lo
gic c
on
t
ro
ll
er i
nf
e
re
nce
rule
DE
E
NB
NS
ZE
PS
PB
NB
NB
NM
NS
PM
NB
NS
NB
NS
PS
ZE
NM
ZE
NM
ZE
PM
PS
NS
PS
NS
PS
PB
PM
ZE
PB
ZE
ZE
PB
PB
PS
Figure
12. T
he
surf
ace
d = f
(E,
DE) of the
M
PP
T
con
t
ro
ll
er
outp
ut
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8
694
In
t J
P
ow
Ele
c
&
Dr
i
S
ys
t
,
V
ol
.
12
, N
o.
2
,
J
une
202
1
:
125
2
–
126
4
1260
5.
SIMULATI
O
N RESULTS
AND DIS
C
USSION
Figure
1
repres
ent
t
he
ge
ner
al
dia
gr
a
m
of
the
w
hole
s
ys
te
m
,
w
hic
h
c
omp
ose
d
of
the
P
V
ar
ray,
blo
c
k
of
DC/DC
B
oost
c
onve
rter,
blo
c
k
of
the
novel
neuro
-
f
uz
zy
meth
od
a
nd
t
he
resist
ive
loa
d.
In
the
presen
t
study,
a
hybri
d
m
od
el
,
ne
uro
-
fu
zz
y
,
l
ooki
ng
f
or
to
e
xtr
act
the
maxim
um
powe
r
f
or
PV
s
olar
s
ys
t
em
i
n
minimu
m
ti
me
an
d
a
hi
gh
de
gr
ee
of
pr
eci
sion.
T
he
ro
le
of
M
PP
T
f
uz
zy
lo
gic
c
ontr
oller
is
t
o
c
hoos
e
t
he
corres
pondin
g
area
re
gion,
w
hich
fi
nds
the
M
PP
with
decre
asi
ng
th
e
pac
e
of
the
duty
c
ycle.
T
hen
the
neural
ne
tw
ork
is
to
direct
the
c
on
t
ro
ll
er
t
o
the
re
gion
w
he
re
the
M
PP
is
locat
e
d.
Se
ve
ral
pe
rformance
c
rite
r
ia
are
repor
te
d
in
th
e
ANN
li
te
ratur
e
as:
the
r
es
pons
e
ti
me,
le
arn
i
ng
base
a
nd
le
ar
n
the
ne
twork
.
The
re
by,
the
est
imat
ion
pe
rformances
of
t
he
neuro
-
f
uzz
y
ap
pr
oach
a
nd
t
he
sin
gle
A
NN
will
be
e
valua
te
d
only
in
te
r
m
of
est
imat
ion
ti
m
e
for
e
xtract
th
e
maxim
um
po
wer
of
P
V
so
la
r
syst
em.
T
he
same
thin
g
wa
s
com
pa
red
wi
th
the
conve
ntion
al
a
lgorit
hm
(
P&
O)
reg
a
rd
i
ng
t
o
the
ma
xim
um
powe
r
ext
ra
ct
ed
unde
r
MATL
AB/
simuli
nk
.
The
theo
reti
cal
and
simulat
ion
res
ults
acq
uired
with
Ne
uro
-
F
uz
zy,
a
rtific
ia
l
neu
ral
netw
ork
Con
tr
oller
a
nd
P&O,
in
c
heck
i
ng
t
he
M
P
P
of
th
e
anal
ys
ed
P
V
mod
ule,
for
va
rio
us
valu
e
s
of
so
la
r
ir
ra
diati
on
G
a
nd
cel
ls’
te
mp
erat
ur
e
T are
giv
e
n
in
Ta
ble
4.
T
he
refor
e
,
this
ta
ble
c
onfir
ms
t
hat
the
neur
o
-
fu
zz
y
gi
ves
a
quic
k
res
pons
e
with
sta
bili
ty
a
rou
nd
M
PP
th
an
t
he
c
on
ven
t
ion
al
A
NN
a
nd
t
he
P&
O.
It
al
so
e
xtracts
th
e
ma
xim
um
po
wer
in
sh
ort
ti
me
with
ef
fici
enc
y
a
nd
per
ti
ne
nce.
Nev
e
rtheless
,
t
his
ta
ble
e
xpre
sses
the
most
e
ffor
t
meth
od
be
tween
the
c
onve
ntional
A
NN
a
nd
cl
assic
al
P&
O
m
et
hods
,
these
methods
are
li
mit
ed
a
rou
nd
a
small
v
al
ue
es
pecial
ly
of
the
ma
xim
um powe
r
with
long
ti
me
t
o
res
pons
e
the
P
&O
but
t
he
pro
pos
ed
Ne
uro
-
Fu
zz
y
meth
od over
com
e
these
li
mit
at
ion
s
t
hro
ugh
a
be
tt
er
de
finiti
on
of
t
he
m
od
el
co
mp
le
xity
ba
sed
on
the
f
uz
zy
r
ules.
The
re
a
re
sever
al
pe
rformance
crit
eria
in
th
e
li
te
ratu
re
of
t
he
A
NN
meth
od
as
m
entione
d
be
for
e.
I
n
t
his
stu
dy,
a
fter
evaluate
d
met
hods
only
in
t
erm
of
est
imat
ion
.
We
ha
ve
base
d
on
the
c
al
culat
ion
of
t
he
e
rror
betwe
en
t
he
measu
red
valu
es
an
d
the
the
oret
ic
al
value
s
of
eac
h
meth
od
treat
ed
i
n
t
his
arti
cl
e
P&
O,
A
NN
a
nd
neuro
-
fu
zz
y
appr
oach.
This
cal
culat
io
n
re
veals
t
he
mini
mal
er
r
or
of
the
ne
uro
-
f
uzz
y
a
ppr
oac
h
c
ompa
red
to
th
e
ot
her
methods
P&
O
an
d
A
N
N,
on
the
one
ha
nd.
O
n
the
oth
e
r
hand
,
we
will
cal
culat
e
the
e
ff
ic
ie
nc
y
to
ha
ve
the
performa
nce,
s
peed an
d
a
bili
ty to
r
es
pond t
o t
he
P
V
s
ys
te
m
in
a
releva
nt a
nd ef
fecti
ve w
ay.
Table
4.
Simul
at
ion
resu
lt
s
of Pma
x
c
heck
i
ng
for diffe
re
nt
consi
der
e
d
c
on
trol
G [
w/
m
²]
T
[°
C]
Pm
ax
(W)
Perturb
&
o
b
serv
e
Artif
ici
al
n
eu
ral
n
e
two
rks
Neu
ro
-
fuzzy
app
ro
ach
Theo
retical
1000
25
1
0
0
,3
1
0
0
,4
108
1
0
8
,2
900
22
9
0
,53
9
1
,06
9
7
,61
9
7
,80
800
20
8
0
,91
8
1
,31
8
6
,73
8
6
,97
700
19
7
0
,66
7
1
,13
7
5
,69
7
5
,81
600
15
6
1
,55
6
1
,63
6
5
,66
6
5
,95
500
12
5
1
,58
5
1
,61
5
5
,94
5
6
,95
400
10
4
1
,35
4
1
,25
4
4
,84
4
4
,97
Figure
13
a
nd
Figure
14
respec
ti
vely
sho
ws
the
PV
ou
t
pu
t
Power
f
or
dif
f
eren
t
c
on
si
der
e
d
co
ntr
ol
at
STC
weathe
r
conditi
ons
an
d
low
we
at
her
conditi
ons.
At
STC
weathe
r
conditi
ons
m
ean
un
der
the
so
la
r
irrad
ia
ti
on G
=
10
00
W/
m2
a
nd
P
V
cel
ls’
te
mp
e
ratur
e
TC
=
25
°C,
w
e
ca
n
see
that
the
pro
posed
h
ybri
d
m
od
el
Neur
o
-
F
uzz
y
a
ppr
oach
ac
hiev
ed
the
m
os
t
ac
cur
at
e
est
imat
ion
c
ompa
rin
g
to
the
A
N
N
an
d
P&
O
met
hods
.
At
ti
me
0.48
s
,
the
pro
po
se
d
hybr
id
m
od
el
e
xtra
ct
s
the
ma
xim
um
powe
r
of
t
he
s
ys
te
m
e
qual
Pout=1
08W,
wh
il
e
the
A
NN
m
et
hod
e
xtract
P
ou
t
=100,4
W
an
d
P&O
e
xtract
P
ou
t=
100,3
W
w
it
h
os
ci
ll
at
ion
arou
nd
M
P
P.
At
low
conditi
ons
,
me
an
under
the
so
la
r
ir
ra
diati
on
G
=
600
W
/m2
a
nd
P
V
c
el
ls’
te
mp
e
ratur
e
TC
=
15°
C,
the
simulat
ion
res
ults
that
the
N
euro
-
Fu
zz
y
hy
br
i
d
m
odel
gi
ve
s
the
bes
t
res
ults
of
t
he
ma
ximum
powe
r
at
ti
me
0.5s,
al
th
ough d
uri
ng
ev
olu
ti
on,
t
he
t
wo MPPT meth
ods
a
re b
e
ginnin
g
be
fore
t
he
hybri
d
model Ne
uro
-
Fuzz
y.
Howe
ver,
the
la
st
on
e
co
ntri
bu
te
s
the
best
value
of
P
owe
r
in
sho
rt
ti
me
with
long
ste
ady
reg
im
e
wi
thout
os
ci
ll
at
ion
a
round
t
he MP
P.
Figure
15
pre
sents
t
he
si
m
ulati
on
outp
ut
of
t
he
PV
s
ys
te
m
(e
xtract
ed
powe
r)
dur
ing
va
riat
ion
weathe
r
c
ondit
ion
s
usi
ng
the
Neur
o
-
F
uzz
y
a
ppr
oach
an
d
the
si
ng
le
A
N
N
c
ompare
d
to
c
onve
ntio
nal
MPP
T
method
P&
O.
The
Ne
uro
-
Fuzz
y
M
P
PT
met
hodolo
gy
acc
ompli
sh
e
d
bette
r
performa
nces
the
n
t
he
si
ng
le
A
NN
or
the
P&
O
al
gorithms
that
c
an
fail
to
trac
k
the
MPP
or
osc
il
la
te
s
around
it
unde
r
rap
i
dly
cha
ngin
g
c
li
mati
c
conditi
ons.
Th
e
pe
rforma
nce
of
t
he
M
P
P
T
can
be
det
ect
ed
acc
ordin
g
t
o
t
he
e
ff
ic
ie
ncy
[
31
]
-
[
34]
.
The
eff
ic
ie
nc
y
cal
c
ulate
d by the
foll
ow
i
ng
(
13)
:
=
1
−
ℎ
−
ℎ
(13)
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 n
e
uro
-
f
uz
zy
approac
h f
or
t
ra
cki
ng
m
axim
um power
poin
t of ph
oto
v
oltai
c so
l
ar
syste
m
(
Ao
uati
f Ibnelo
uad
)
1261
Figure
13. PV
ou
t
pu
t
Powe
r
f
or d
if
fer
e
nt c
onside
red
con
t
ro
l at
G=
1000
W/m² a
nd
T=2
5°
C
Figure
14. PV
ou
t
pu
t
Powe
r
f
or d
if
fer
e
nt c
onside
red
con
t
ro
l at
G
=
600W/
m² a
nd T
=15°C
Figure
15. PV
ou
t
pu
t
Powe
r
f
or d
if
fer
e
nt c
onside
red co
ntr
ol un
der va
riat
ion o
f
ir
ra
diati
on
a
nd tempe
rat
ur
e
The
ef
fici
enc
y
of
P&
O,
ANN
an
d
Neur
o
-
Fu
zz
y
co
ntr
ollers
s
hows
that
the
Ne
uro
-
Fuzz
y
co
ntr
olle
r
can
gen
e
rate
up
to
99%
of
th
e
act
ual
ma
xi
mu
m
po
wer
c
ompare
d
t
o
th
e
A
NN
co
ntr
ol
le
r
can
ge
ne
rate
up
to
93%
an
d
P&
O
can
gen
e
rate
up
to
92%
of
i
t
[14]
a
s
s
how
n
in
Fig
ur
e
16
.
In
fact,
t
he
propose
d
Neur
o
-
Fu
zz
y
ap
pr
oach
-
ba
se
d
meth
od
at
ta
ined
t
he
highes
t
power
e
ff
ic
ie
ncy
with
6%
of
extra
-
gen
e
rat
ed
po
wer
c
ompari
ng
to the si
ngle
ANN an
d mo
re t
han 3%
to
the
P&O al
gorith
m b
eca
us
e
of it
s o
sci
ll
at
ion
s
a
rou
nd the
M
P
P
.
To
de
velo
p
t
he
ne
w
Ne
uro
-
F
uzzy
c
on
tr
oller
ap
proac
h,
we
reli
ed
on
se
veral
arti
cl
es
in
t
he
li
te
ratur
e
amo
ng
t
hem
[
12
]
,
[
30]
.
A
kin
d
of
c
ompari
so
n
i
n
sta
te
of
the
art
betwe
en
our
a
ppr
oac
h
a
nd
tw
o
ref
e
ren
ces
[12
]
,
[
30
]
in
ta
bu
la
r
format.
I
n
Ta
ble
5,
a
s
um
ma
r
y
of
the
pow
er
ef
fici
ency
betwee
n
our
ap
proac
h
a
nd
the
re
fe
ren
ce
[
30],
w
hich
is
base
d
on
To
olbo
x
ANFIS
un
der
M
A
TLAB/
sim
ulink
,
in
one
ha
nd.
I
n
t
he
othe
r
ha
nd
,
su
m
marizes
th
e
error
est
imat
e
betwee
n
our
appr
oach
a
nd
t
he
re
fer
e
nce
[12].
This
ta
ble
s
hows
t
hat
the
powe
r
eff
ic
ie
nc
y
of
t
he
A
NF
I
S
met
hod
reac
hes
10
0%
un
de
r
the
STC
c
onditi
ons
an
d
ou
r
a
ppr
oach
reaches
a
val
ue
up
t
o
99%.
U
nd
e
r
the
var
ia
t
ion
s
of
at
m
osph
e
ric
co
nd
it
i
on
s
,
the
powe
r
eff
ic
ie
ncy
of
our
ap
proac
h
al
way
s
remains
up
to
99%,
w
hich
shows
the
rele
va
nce
of
our
ne
uro
-
fu
zz
y
ap
proa
ch
c
ompare
d
to
t
he
ANFIS
meth
od,
wh
ic
h
is
al
re
ad
y
pr
e
def
i
ned
in
the MATL
AB/
simuli
nk
to
olbox
. A
fter
that, it
il
lustrate
s
that
our
ne
w
a
ppro
a
c
h
has
hi
gher
perce
ntages
of
e
rror
s
for
P
&O
or
A
NN
meth
ods,
c
ompare
d
t
o
the
c
omparat
ive
meth
od.
I
n
oth
e
r
words,
th
e
pe
r
centage
of
the
error
is
la
r
ge
in
our
ap
proac
h
that
the
er
ror
is
minimal
c
ompare
d
to
th
e
oth
e
r
ref
e
ren
ce
.
Table
5.
T
he
percenta
ge
est
i
mati
on
ne
uro
-
f
uzzy ap
proac
h met
hods
Refere
n
ces
Ef
fic
ien
cy
(
%)
Er
ror
P&O
Er
ror
ANN
This
stu
d
y
9
9
.82
,
9
9
.81
,
9
9
.7
2
,
9
9
.86
,
9
9
.71
3
.74
%
6%
Ch
ao
u
achi
et al.
,
2
0
1
0
-
2
.73
%
5
.86
%
Ay
m
en
Je
m
aa e
t al
.
2
0
1
6
1
0
0
,
9
9
.99
,
9
9
.9
5
,
9
9
.78
,
9
9
.68
-
-
Evaluation Warning : The document was created with Spire.PDF for Python.