Int
ern
at
i
onal
Journ
al of
R
obot
ic
s
and
Autom
ati
on
(I
J
RA
)
Vol
.
7
, No
.
2
,
J
un
e
201
8
, pp
.
1
29
~
1
39
IS
S
N:
20
89
-
4856, DO
I: 10
.
11
591/ij
ra
.
v
7
i2
.
pp
1
29
-
1
39
129
Journ
al h
om
e
page
:
http:
//
ia
escore
.
c
om/j
ourn
als/i
ndex
.
ph
p/IJRA/i
ndex
Neural
Net
w
or
k Bas
ed MPP
T Con
troll
er
for So
la
r
PV
Con
nected
In
du
ction Mot
or
T
.
Sh
anthi
As
sistant
Profess
or
(Senior
Grad
e),
EEE
Dep
artm
ent
,
Kum
ara
gu
ru
Coll
ege of
T
e
chnol
og
y
,
Coim
bat
ore
,
T
a
m
il
Nadu, Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Feb
27
, 201
8
Re
vised
Ma
y
1
, 201
8
Accepte
d
Ma
y
17
, 201
8
In
thi
s
pap
er,
Maximum
Pow
er
Point
Tr
ac
k
ing
Control
le
r
is
des
igne
d
b
ase
d
on
Neura
l
Netw
ork
Contro
ll
er
(
NN
C)
.
Thi
s
con
t
roll
er
wil
l
sense
the
spee
d
of
a
singl
e
phase
i
nduct
ion
m
otor
which
is
f
ed
fr
om
solar
panel
.
Maximum
power
poin
t
trac
king
(MP
PT)
a
l
gorit
hm
ar
e
r
equ
ire
d
in
all
pho
to
volt
aic
(PV
)
s
y
stem
and
in
o
rde
r
to
inc
r
ea
se
th
e
eff
i
cienc
y
of
th
e
s
y
st
em,
I
ncr
emental
Conduct
an
ce
a
l
gorit
hm
which
i
s
an
eff
ec
t
ive
a
lgori
thm
is
use
d
to
ex
tract
m
axi
m
um
powe
r
from
the
sol
ar
panel
which
supplie
s
an
Induc
t
ion
m
otor
of
1HP
.
To
step
up
the
voltage
ava
i
la
bl
e
from
t
he
solar
p
anel,
t
he
SEP
IC
DC
–
DC
conv
e
rte
r
is
used
.
Th
e
m
ai
n
adva
nt
age
of
th
e
conve
r
ter
is
h
avi
ng
non
-
inve
rt
ed
out
put
.
The
conve
r
t
er
a
ct
s
as
an
in
terfac
e
bet
w
ee
n
P
V
arr
a
y
and
m
otor
loa
d
.
The
en
ti
r
e
s
y
stem
is
m
odel
ed
and
sim
ula
te
d
using
MA
TL
AB/S
imu
li
nk
softwar
e
.
Ke
yw
or
d:
In
c
rem
ental
co
nductanc
e
Neural
netw
ork
c
on
t
ro
ll
er
MPPT
Photo
vo
lt
ai
c
SEPI
C
Copyright
©
201
8
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
T
.
S
ha
nth
i,
Dep
a
rtm
ent o
f El
ect
rical
an
d
Ele
ct
ro
nics
E
nginee
rin
g,
Ku
m
aragur
u
C
ollege
of Tec
hnol
og
y,
Chin
nav
e
dam
patti
, Co
i
m
batore
–
64
1049, T
a
m
il
N
adu
,
In
dia
.
E
m
a
il
: shan
thit
s@g
m
ai
l
.
com
1.
INTROD
U
CTION
Am
on
g
al
l
re
new
a
ble
e
ne
rgy
sources
,
s
ol
ar
e
nergy
at
tr
act
s
m
or
e
at
te
ntion
because
they
pro
vi
de
excell
ent
oppo
rtu
nity
to
ge
ne
rate
el
ect
rici
ty
.
S
olar
e
ne
rg
y
is
a
cl
ea
n
re
ne
wab
le
res
ource
with
zer
o
em
i
ssion
.
Power
dem
and
is
i
ncr
ea
sin
g
day
by
da
y,
so
we
hav
e
t
o
s
witc
h
to
re
new
a
ble
e
nerg
y
sources
w
hi
ch
a
re
eco
-
fr
ie
nd
ly
a
nd
e
xist
ab
unda
nt
in
nat
ur
e
.
T
he
m
axi
m
u
m
po
we
r
point
tra
ckin
g
(MPPT
)
con
t
ro
ll
er
is
us
ed
to
i
m
pr
ove
the
ef
fici
ency
of
t
he
PV
syst
em
,
in
wh
ic
h
Pe
rturb
&
O
bs
er
ve
(P&O
)
an
d
I
nc
re
m
ental
Con
duc
ta
nce
(INC)
are
fr
e
quently
us
e
d
.
T
he
inc
rem
ental
cond
uc
ta
nce
a
lgorit
hm
determ
ines
the
gr
a
di
ent
of
the
P
-
V
curve
[1
-
2]
.
This
m
e
thod
has
overc
om
e
the
disad
van
ta
ge
of
the
P&
O
m
et
ho
d
to
trac
k
the
pe
ak
powe
r
unde
r
fa
st
var
yi
ng
at
m
os
ph
e
ric
co
ndit
ion
.
T
he
inc
rem
ental
cond
ucta
nce
can
deter
m
ine
wh
et
he
r
t
he
MPPT
has
rea
che
d
the
m
axi
m
u
m
powe
r
po
i
nt
(
MPP)
a
nd
st
op
per
t
urbin
g
t
he
operati
ng
point
or
el
se
t
he
relat
ionshi
p
betwee
n
dI
/
dV
&
-
I/V
can
be
us
e
d
to
dete
rm
ine
the
di
recti
on
in
w
hich
th
e
MPPT
op
e
r
at
ing
po
i
nt
m
us
t
be
per
t
urbed [
3
-
6]
.
A
dc
to
dc
c
onve
rter
is
ne
eded
to
bo
os
t
the
vo
lt
age
f
ro
m
PV
pan
el
an
d
c
omm
on
ly
avail
able
conve
rters
are
the
boost
,
buc
k,
bu
c
k
-
bo
os
t,
Cuk,
SEP
IC
[
7
-
8]
.
I
n
wh
ic
h
the
sin
gle
-
e
nd
e
d
pri
m
ary
ind
uc
ta
nce
conve
rter
(
SE
PI
C)
is
a
DC/DC
-
co
nverter
that
pro
vid
es
a
po
sit
ively
regulat
ed
ou
t
put
an
d
no
n
-
i
nver
te
d
ou
t
pu
t
.
B
uck
-
boos
t
co
nverte
rs
a
re
c
heap
e
r
be
cause
t
hey
re
quire
only
a
sin
gle
in
du
c
tor
an
d
a
cap
aci
tor
.
But
the
draw
ba
ck
is
t
he
high
am
ou
nt
of
i
nput
cu
rr
e
nt
rip
pl
e
wh
ic
h
c
reate
ha
rm
on
ic
s,
i
n
m
any
ap
plica
ti
on
s,
these
ha
rm
on
ic
s
require
us
i
ng
a
la
rg
e
ca
p
aci
t
or
o
r
an
LC
filt
er
.
T
his o
ften
m
akes
the b
uc
k
-
boos
t
i
neffici
ent
or
exp
e
ns
i
ve,
a
nd
that
can
com
plica
te
the
us
ag
e
of
bu
c
k
-
bo
ost
conver
te
r
s
is
the
fact
that
th
ey
inv
ert
the
outp
ut
vo
lt
age
.
C
uk
c
onve
rters
so
l
ve
both
of
thes
e
pr
ob
le
m
s
by
us
i
ng
a
n
extr
a
in
du
ct
or
a
nd
capaci
to
r
.
Howev
e
r
,
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
IJ
RA
V
ol
.
7
,
No
.
2
,
June
201
8
:
1
29
–
1
39
130
bo
t
h
buc
k
-
boost
an
d
Cu
k
co
nverter
operati
on
c
ause
la
r
ge
a
m
ou
nts
of
el
ec
tric
al
stress
on
the
c
om
po
ne
nt
s,
t
his
can
resu
lt
in
de
vice ove
rh
eat
in
g or fail
ure
.
SE
PI
C c
onve
rters
so
l
ve bo
t
h of
t
hese
pro
blem
s
.
In
t
his
pa
pe
r,
SEPI
C
c
onve
rt
er
re
gula
te
s
th
e
dc
vo
lt
a
ge
obta
ined
f
ro
m
the
s
olar
panel
an
d
fee
ds
the
sin
gle
phas
e
inv
e
rter
.
Thi
s
sing
le
phase
inv
e
rter
r
uns
t
he
sin
gle
phas
e
Ind
uction
m
otor
of
1HP
ca
pacit
y
[9
-
10]
.
T
he
s
pe
ed
of
the
i
nduc
ti
on
m
oto
r
is
us
e
d
as
a
fee
db
ack
si
gn
al
fro
m
wh
ic
h
volt
age
is
der
i
ved
a
nd
er
ror
and
cha
nge
in
error
are
obta
ined
a
nd
giv
e
n
to
N
eu
ral
Net
work
co
ntr
oller
[
11]
.
T
he
ge
ner
at
e
d
pu
lse
s
from
the
c
on
tr
oller
a
re
c
om
bin
ed
w
it
h
the
pulse
s
obta
ined
f
r
om
t
he
In
c
rem
ental
Co
nductance
Algorithm
of
a
So
la
r
pan
el
a
nd
g
ive
n
to
SE
PI
C c
onve
rter a
nd
des
ired o
utput v
ol
ta
ge
is
produce
d
.
2.
PROP
OSE
D SYSTE
M
2
.
1
.
S
ola
r
pan
el
So
la
r
pa
nel
us
es
t
he
li
gh
t
energy
phot
on
from
the
sun
to
ge
ne
rate
el
ect
rici
ty
throu
gh
the
photov
oltai
c
(PV)
ef
fect
.
The
m
ajo
rity
of
m
od
ules
use
wa
fer
ba
se
d
cel
ls
or
thi
n
-
film
cel
ls
based
on
non
-
m
agn
et
ic
cond
uctive
tra
ns
it
ion
m
et
al
or
sil
ic
on
.
It
can
prov
i
de
ne
arly
per
m
anen
t
po
wer
at
lo
w
ope
rati
ng
cost
an
d
is
virt
ually
fr
ee
poll
ution
.
A
ty
pica
l
PV
cel
l
pr
oduces
le
ss
t
han
3
watt
s
at
ap
prox
im
at
el
y
0
.
5V
DC
.
A
PV
m
odule
consi
sts
of
se
ve
ral
PV
cel
ls
connecte
d
i
n
se
ries
or
paral
le
l
.
Series
co
nnec
ti
on
s
are
res
pons
ible
for
inc
reasin
g
the
volt
age
a
nd
the
pa
rall
el
connecti
ons
a
r
e
res
pons
ible
f
or
i
ncr
easi
ng
the
c
urren
t
.
Fi
gure
1
.
sh
ows
the
blo
c
k diag
ram
o
f
th
e
pro
posed
sys
tem
.
Figure
1
.
Bl
oc
k
dia
gram
o
f
th
e pro
posed
sc
hem
e
2
.
1
.
1
.
Desi
gn
of s
ola
r
panel
To
ge
ne
rate
1k
W
powe
r
fro
m
the
so
la
r
pa
nel,
f
our
25
0W
pa
nels
is
c
onnected
i
n
se
ries
.
D
ue
to
the series
con
ne
ct
ion
, t
he v
oltage
is
i
ncr
ease
d
a
nd the c
urre
nt r
em
ai
n
s
co
nst
ant
.
I
n Table
1
is s
olar
p
a
nel
.
Table
1
.
So
la
r Panel S
pecific
at
ion
S
.
No
SPECIFIC
ATI
O
N
RAN
G
E
1
Nu
m
b
e
r
o
f
cells
72
2
Op
en
cir
cu
it vo
lta
g
e (
V
OC
)
4
4
V
3
Sh
o
rt
circuit cur
re
n
t (
I
SC
)
7
.
5
8
A
4
Vo
ltag
e at
m
ax
i
m
u
m
p
o
wer
(
V
MP
)
35
.
5
5
V
5
Cu
rr
en
t at
m
ax
i
m
u
m
p
o
wer
(
I
MP
)
7
.
0
4
A
2
.
2
.
M
axim
u
m
po
w
er
p
oin
t
tr
ackin
g
In
wind
tu
rb
i
nes
an
d
P
V
s
olar
syst
em
s,
to
m
axi
m
u
m
powe
r
extracti
on
is
possible
under
al
l
conditi
ons,
if
MPPT
te
c
hn
i
que
is
use
d
.
PV
so
la
r
syst
em
has
m
any
different
co
nf
i
g
ur
at
ion
s
.
A
ty
pic
al
so
la
r
pan
el
c
onve
rts
only
30
-
40
(
%)
of
t
he
inc
ident
s
olar
ir
r
adiat
ion
i
nto
el
ect
rical
energy
.
MPP
T
in
creases
the
ef
fici
ency
of
the
so
la
r
pa
nel
r
api
dly
.
I
f
they
ope
rate
at
their
MPP
des
pite
the
i
nev
it
able
cha
nges
i
n
the
e
nviro
nm
e
nt,
m
ax
im
u
m
po
wer
f
r
om
the
s
olar
pa
nel
ca
n
be
ha
rv
est
e
d
.
O
ve
r
t
he
past
de
cades,
m
any
m
et
ho
ds
to
fin
d
t
he
M
PP
ha
ve
bee
n
publishe
d
a
nd
dev
el
op
e
d
.
I
n
that
m
os
t
su
it
a
ble
te
ch
niques
for
m
edium
and
,
la
rg
e
-
siz
e
ph
oto
volt
a
ic
ap
plica
ti
on
s
a
re
P&
O
an
d
IN
C
.
T
hese
te
ch
nique
s
ha
ve
the
m
erit
s
of
an
easy
im
plem
entat
ion
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
IS
S
N:
20
89
-
4856
Ne
ur
al Net
wo
r
k Ba
se
d MPPT
Co
ntro
ll
er for
So
l
ar
P
V Co
nn
ect
ed
I
nductio
n
M
oto
r
(
T
.
Sh
an
t
hi
)
131
2
.
2
.
1
.
I
ncreme
nt
al
conduc
tance al
go
ri
th
m
A
wide
range
of
MPPT
Algorithm
s
are
avail
able
.
O
f
al
l
the
a
vaila
bl
e
al
gorithm
s,
I
ncr
em
ental
Cond
uctance
Algorithm
lends it
sel
f
well
.
T
he
inc
rem
ental
Cond
uctance
m
et
ho
d
is
oft
en
c
on
si
der
e
d, due to
it
s
high
pe
r
form
a
nces
s
uch
as
easy
i
m
ple
m
e
ntati
on
,
hi
gh
t
rac
ki
ng
sp
ee
d,
bette
r
ef
fici
ency
an
d
it
get
s
easi
ly
adap
ta
ble
f
or
t
he
cha
ngin
g
en
vir
on
m
ental
con
diti
ons
th
us
i
ncr
ease
t
he
ef
f
ic
ie
ncy
of
P
V
syst
e
m
.
It
is
fo
und
to
be
the
best
t
echn
i
qu
e
an
d
easi
ly
adap
ta
ble
to
t
he
c
ha
ng
i
ng
en
vir
onm
ental
con
di
ti
on
s
.
O
n
c
om
par
ing
the
ef
fici
ency
res
ults
obt
ai
ne
d
f
r
om
Pertur
b
&
Obser
ve
(P
&
O
)
95
%
and
the
In
c
re
m
ental
Conduc
ta
nce
Algorithm
9
8%
.
This
al
gorithm
se
ns
es
t
he
out
pu
t
cu
rr
e
nt
an
d
volt
age
of
the
PV
ar
ray
us
i
ng
se
ns
ors
.
T
he
dem
erit
s
of
P&O
m
et
ho
d
to
trac
k
the
pe
ak
power
un
der
t
he
fast
va
ryi
ng
at
m
os
ph
eric
co
ndit
ion
is
ov
e
rc
om
e
by
IN
C
m
et
ho
d
.
T
he
i
nc
rem
enta
l
condu
ct
a
nce
ca
n
de
te
rm
ine
the
MPPT
a
nd
if
the
MP
P
is
reac
hed
it
sto
ps
pe
rturbin
g
the ope
rati
ng point
.
(d
P/
dV)
MPP
= 0
.
d(VI)/d
V
=
0
.
I(dV
/
dV)
+
V(
dI
/
dV)
=
0
I+V(dI/
dV)
MPP
= 0
(
dI/
dV)
MPP
=
-
I/V
The
te
rm
–
I/V
re
pr
e
sents
th
e
insta
ntane
ous
c
onduct
ance
of
the
P
V
pa
nel
a
nd
t
he
te
rm
(d
I/
dV
)
represe
nts
inc
r
e
m
ental
cond
uc
ta
nce
of
the
PV
m
odule
.
T
his
m
et
ho
d
is
base
d
on
the
f
act
that
the
slop
e
of
the
power
cu
r
ve
is
zer
o
at
the
MPP,
if
t
he
sl
op
e
is
decr
ea
sing,
MP
P
li
es
on
th
e
ri
gh
t
sid
e
an
d
if
the
slo
pe
is
increasin
g, MP
P lie
s on
t
he
le
ft side
.
T
his ca
n be
giv
e
n by,
(dI/d
V)
MPP
=
-
I/V, at
the MP
P
(dI/d
V)
MPP
>
-
I/V, o
n
t
he
le
ft
(dI/d
V)
MPP
<
-
I/V, o
n
t
he
ri
ght
The
per
t
urbati
on
is
r
epea
te
d
unti
l
the
MP
P
is
re
ache
d
.
Un
ti
l
a
c
ha
nge
in
c
urre
nt
is
m
easur
ed,
the
MPP
T
c
onti
nu
es
t
o
opera
te
at
the
sam
e
po
i
nt
.
In
Fi
gur
e
2
is
flo
w
c
ha
rt
I
nC
al
gorith
m
and
Ba
sic
c
on
ce
pt
of INC o
n a P
-
V
c
urve
as
s
hown in Fi
gure
3
.
Fig
ure
2
.
Flo
w
ch
a
rt for
INC
a
lgorit
hm
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
IJ
RA
V
ol
.
7
,
No
.
2
,
June
201
8
:
1
29
–
1
39
132
Fig
ure
3
.
Ba
sic
concept
of IN
C o
n
a
P
-
V
c
urve
2
.
3
.
Sepic c
onverter
The
ci
rc
uit
dia
gr
am
of
t
he
S
EPIC
c
onve
rter
is
s
how
n
in
Figure
4
.
A
S
EPIC
(single
-
e
nd
e
d
pr
im
ary
inducto
r
co
nve
rter)
is
on
e
ty
pe
of
DC
-
DC
c
onve
rter
.
It
co
ns
ist
s
of
bo
os
t
conve
rter
f
ollo
wed
by
a
buc
k
-
bo
os
t
conve
rter
.
The
m
ai
n
ad
van
ta
ge
of
t
his
c
onve
rter
is
capa
ble
of
pro
vi
ding
a
non
-
in
ver
te
d
outp
ut
(i
.
e
.
t
he
ou
t
pu
t
has
the
sam
e
po
la
rity
as
the
i
nput)
.
Its
outp
ut
volt
age
m
us
t
be
great
er
t
ha
n
or
le
ss
t
han
or
e
qual
to
t
he
input
vo
lt
age
and is
widely
used
in batt
ery o
perat
ed
a
pp
li
cat
io
ns
.
The
outp
ut vol
ta
ge
is con
t
ro
ll
ed
by adj
us
ti
ng
the d
uty cy
cl
e o
f
the c
ontr
ol sw
it
ch
.
Th
e co
nt
ro
l swit
ch
is
ty
pical
ly
a
MOSFET
,
w
hi
ch
offe
rs
m
uch
highe
r
i
nput
i
m
ped
ance
,
l
ow
volt
age
dr
op
a
nd
lo
wer
s
witc
hing
losses
.
A
SE
P
IC
c
onver
te
r
i
s
a
fou
rth
ord
er
c
onve
rter,
it
m
eans
these
conve
rters
ha
ve
f
our
ene
rg
y
stora
ge
el
e
m
ents
they
are
t
wo
in
duct
or
s
a
nd
tw
o
ca
pacit
or
s
,
a
nd
it
is
us
e
d
t
o
tra
ns
fe
r
t
he
ene
rgy
from
input
s
ide
to
ou
t
pu
t
side
.
T
he
i
nput
in
duct
or
L
1
is
t
og
et
he
r
with
t
he
M
OS
FE
T
c
on
t
rol
swi
tc
h
t
o
be
li
ke
a
boos
t
t
opol
og
y,
w
he
re t
he
in
du
ct
or
L
2
locat
io
n
is si
m
il
ar to
a buck
-
boos
t t
opology
.
Fig
ure
4
.
P
ow
e
r
ci
rc
uit o
f
S
E
PI
C c
onve
rter
2
.
3
.
1
D
esi
gn c
alcula
tion
V
IN
= 142
.
2V
I
IN
= 7
.
04A
f
=
15
KHZ
V
OUT
= 23
0V
I
OUT
= 4A
Wh
e
re,
V
IN
=
INP
UT
VOLT
AGE
I
IN
= INPUT
CURR
ENT
f
=
FRE
QUENCY
V
OUT
= O
UTPUT
VOLT
AGE
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
IS
S
N:
20
89
-
4856
Ne
ur
al Net
wo
r
k Ba
se
d MPPT
Co
ntro
ll
er for
So
l
ar
P
V Co
nn
ect
ed
I
nductio
n
M
oto
r
(
T
.
Sh
an
t
hi
)
133
V
IN
= I
NPUT
VOLT
AGE
CALCUL
AT
I
ON DUT
Y
C
Y
CLE:
We
know that,
D=
T
ON
/T
Wh
e
re,
T = T
ON
+ T
OFF
Si
m
il
arly
,
D=
V
OUT
/ V
OU
T
+ V
IN
D=
230
/
230
+
142
.
2
D=
0
.
6179
INDU
CT
OR C
ALCU
LAT
I
O
N:
Fo
r
In
du
ct
or L
1
,
L
1
= V
IN
* D /
2*I
IN
*f
L
1
= 14
2
.
2 *0
.
6179 /
2*7
.
04*15*e^
3
L
1
= 0
.
416e
-
3H
Fo
r
In
du
ct
or L
2,
L
2
= V
OUT
*(1
-
D)
/
2*I
OUT
*f
L
2
= 230*
(1
-
0
.
6179)
/ 2*
4
.
5*15*e^
3
L
2
= 0
.
6457e
-
3H
CAPA
C
IT
OR
CALCUL
AT
I
ON
:
Fo
r
Capaci
to
r C
1
,
C
1
= I
OUT
/ V
*f
Wh
e
re,
V=
V
OUT
–
V
IN
C
1
= 3
.
447e
-
6F
Fo
r
Capaci
to
r C
2,
C
2
= I
OUT
/ V
OUT
*f
C
2
= 4
.
54 / 23
0*15
*e^3
C
2
= 1
.
31e
-
6F
2
.
4
.
Ne
ural
n
etworks
A
ne
ural
netw
ork
is
a
m
at
hem
at
ic
al
m
od
el
insp
ire
d
by
bi
ol
og
ic
al
ne
ur
al
netw
orks
as
li
ke
in
a
brai
n
.
Hu
m
an
br
ai
n
le
arn
s f
r
om
exp
erience
a
nd
st
ores
inf
orm
ation
'
s
as
patte
rn
s
.
In
s
om
e
sit
uatio
n,
these p
at
te
r
ns
are
ver
y
c
om
plica
t
ed
an
d
al
lo
w
us
the
abili
ty
to
recog
nize
ind
i
vidual
faces
.
S
o
the
process
of
sto
rin
g
in
for
m
at
ion
as
patte
r
ns
has
a
new
fiel
d
in
com
pu
ti
ng
.
T
his
c
om
pu
ti
ng
us
e
as
m
assive
pa
rall
el
netw
orks
an
d
trai
ni
ng
of
this net
w
ork
don
e
to sol
ve
a
sp
eci
fic
pro
blem
[
11
]
.
A
ne
ur
al
netw
ork
co
ns
ist
s
of
an
interc
onne
ct
ed
set
of
a
rtific
ia
l
neu
r
ons,
and
it
proce
ss
inf
or
m
at
ion
us
in
g
a
c
on
ne
ct
ion
ist
ap
proa
ch
to
com
pu
ta
ti
on
.
I
ntegr
at
e
d
ci
rc
uits
are
two
-
dim
ensional
dev
ic
es
wit
h
le
ss
nu
m
ber
of
la
ye
rs
for
i
ntercon
necti
ons
.
I
n
r
e
al
it
y,
arti
fici
al
neural
netw
ork
s
can
be
im
ple
m
ented
in
sil
ic
on
an
d
basical
ly
,
al
l
arti
fici
al
neural
netw
orks
ha
ve
a
sim
il
ar
struc
ture
.
I
n
that
s
tructu
re,
s
om
e
ne
urons
inter
f
ere
to
receive
in
pu
t,
oth
e
r
ne
urons
i
nterf
e
re
with
t
he
netw
ork
ou
t
pu
t
an
d
rest
of
the
neur
on
has
a
r
ole
but
will
no
t
be
disp
la
ye
d
.
Ba
s
ic
al
ly
,
the
ne
ural
net
wor
k
ha
s
th
ree
la
ye
rs
nam
ely
input
l
ay
er,
th
e
ou
t
put
la
ye
r
a
nd
hi
dd
e
n
la
ye
r
.
T
he
hidden
la
ye
r
m
ay
or
m
ay
no
t
be
present
de
pe
ndin
g
on
the
ap
plica
t
ion
us
e
d
.
This
ne
ural
ne
twork
for
sp
ee
d
co
nt
ro
l
of
in
du
ct
io
n
m
oto
r
has
two
la
ye
rs
na
m
el
y
inp
ut
la
ye
r
a
nd
outp
ut
la
ye
r
.
I
n
Fi
gur
e
5
is
t
he
la
yout
of a
r
ti
fici
al
n
eur
al
ne
twork
.
Fig
ure
5
.
Lay
out o
f
a
rtific
ia
l neural
netw
ork
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
IJ
RA
V
ol
.
7
,
No
.
2
,
June
201
8
:
1
29
–
1
39
134
2
.
4
.
1
.
Tr
ainin
g
an
d
a
r
tifici
al
neura
l
netw
or
k
If
a
netw
ork
ha
s
bee
n
f
ram
e
d
for
a
par
ti
cul
ar
a
pp
li
cat
io
n
t
hen
the
ne
xt
st
ep
is
trai
ning
the
a
rtific
ia
l
neural
net
wor
k
.
T
o
beg
i
n
th
is
process
i
niti
al
weig
hts
are
assum
ed
rand
om
l
y
.
Trainin
g
of
a
rtific
ia
l
neural
netw
ork
can
be
a
ppr
oac
hed
in
tw
o
ways
nam
el
y
su
pe
rv
ise
d
a
nd
un
su
pe
r
vised
.
In
s
up
e
rv
ise
d
tr
ai
nin
g
,
the
outp
ut
is
know
n
so
i
nput
is
giv
en
as
pe
r
the
de
sired
ou
t
pu
t
.
This
proces
s
invol
ve
s
the
com
par
ison
of
resu
lt
in
g
outp
ut
a
nd
desire
d
outp
ut
so
that
er
ror
val
ue
w
ou
l
d
be
ge
ne
ra
te
d
.
The
e
r
r
or
value
s
o
obta
ined
is
pro
pag
at
in
g
ba
ck
a
nd
wei
ghts
are
a
djust
ed
a
nd
ab
ove
sai
d
process
co
ntin
ue
un
ti
l
desire
d
ou
t
pu
t
is
obt
ai
ned
.
The
set
of
data
wh
ic
h
ena
ble
this
w
ho
le
pro
cess
is
cal
le
d
“t
rainin
g
set
”
.
On
oth
e
r
un
s
uper
vised
t
raini
ng
will
be
prece
de
d
with
ra
ndom
input
as
desi
re
d
outp
ut
is
not
pr
e
dicta
ble
.
S
o
this
proc
ess
is
al
so
ca
ll
ed
as
sel
f
-
organ
iz
at
i
on [1
2
]
.
I
n Fi
gure
6 is l
ay
ou
t
of Typica
l Ne
ural
N
et
work
.
Fig
ure
6
.
Lay
out
of ty
pical
ne
ur
al
netw
ork
2
.
4
.
2
.
C
ompo
nent
s
of
art
i
fici
al neur
al ne
t
w
ork
Ther
e
a
re
seve
n
m
ajor
com
po
ne
nts
in
vo
l
ve
d
i
n
the
arti
fici
al
ne
ur
al
netw
ork
.
T
hese
c
om
po
nen
ts
are
valid
wh
et
her the
neur
on is us
ed fo
r
in
put,
outp
ut or i
n
the
hidden
lay
er [
13
-
14
]
.
a
.
Dete
rm
inatio
n o
f wei
ght
A
ne
uro
n
receives
m
uch
in
put
si
m
ultaneou
s
ly
.
Each
in
put
is
prov
i
de
d
wi
th
co
rr
es
pondi
ng
weig
ht
s
so
that
pro
vid
e
s
a
n
im
pact
on
the
processi
ng
el
e
m
ent
.
Pr
i
or
i
ty
to
t
he
i
nputs
de
pe
nds
on
t
he
wei
gh
t
.
So
w
ei
gh
t
can
be
de
scrib
ed
a
s
a
da
ptive
coe
ff
ic
ie
nts
wh
ic
h
determ
i
ne
t
he
i
ntensit
y
of
t
he
i
nput
sig
nal
a
nd
m
easur
e
the in
pu
t c
on
ne
ct
ion
stre
ngth
.
T
hese s
tre
ngths ca
n be m
od
i
fied d
ur
in
g
t
rainin
g
set
.
b
.
Su
m
m
ation
functi
on
The
first
proce
ss
is
t
o
c
om
pu
te
the
wei
gh
te
d
su
m
of
al
l
the
inputs
m
at
he
m
at
ic
al
ly
inp
ut
a
nd
weig
hts
are
in
vect
or
f
or
m
and
dot
products
a
re
ca
rri
ed
out
betwee
n
them
.
Ge
ome
tric
al
ly
the
dot
pr
oducts
re
pr
ese
nt
si
m
il
arity
betw
een
i
nputs
a
nd
weig
hts
.
T
he
resu
lt
ant
of
t
he
vect
or
pro
du
ct
s
de
pende
d
upon
t
he
directi
on
of
vecto
r
poi
nts
.
The
c
om
bin
at
ion
of
neural
inpu
ts
dep
e
nds
upon
netw
ork
arc
hitec
ture
so
m
et
i
m
e
act
i
vation
functi
on is als
o d
on
e
and its
pur
pose is t
o
al
low t
he o
utput
ob
ta
ine
d
t
o var
y wit
h resp
ect
t
o
ti
m
e
.
c
.
Tra
nsfer
func
ti
on
The
ob
ta
in
ed
weig
hted
s
um
is
trans
form
ed
to
w
orkin
g
outpu
t
t
hroug
h
th
e
trans
fe
r
f
un
c
ti
on
.
D
ur
in
g
the
process
,
the
wei
gh
te
d
s
um
is
com
pared
with
a
th
r
esh
old
val
ue
to
determ
ine
t
he
net
work
outp
ut
.
If
t
he
co
ndit
io
n
is
sat
isfie
d,
then
it
ge
ne
rates
sig
nal
othe
r
wise,
it
w
on'
t
pro
du
ce
a
ny
s
ign
al
.
T
he
t
ra
ns
fe
r
functi
on
is
ge
ner
al
ly
a
non
-
l
inear
syst
em
because
li
nea
r
f
un
ct
io
ns
are
not
use
f
ul
.
S
om
eti
m
es
ra
m
ping
functi
ons, si
gm
oid
o
r S
-
s
ha
pe
d
c
urve a
re
use
d
as
transfe
r
f
un
ct
io
ns
.
d
.
Scal
ing an
d l
i
m
it
ing
The
res
ults
so
ob
ta
ine
d
a
re
proces
sed
th
r
ough
scal
in
g
factor
s
a
nd
so
m
et
i
m
es
of
fset
value
is
ad
ded
.
Li
m
iting
is t
he m
echan
ism
w
hi
ch
is
done w
he
n
it
ex
c
eeds
the
bounda
ry c
onditi
ons
.
e
.
O
utput
funct
ion
Like
the
bio
lo
gical
ne
uro
n,
there
a
re
m
any
inputs
a
nd
on
ly
one
ou
t
put
.
T
he
outp
ut
is
di
rectl
y
pro
portion
al
to
trans
fer
f
unct
ion
res
ult
.
C
ompeti
ti
on
ca
n
oc
cur
at
one
or
both
of
tw
o
le
ve
ls
.
First,
the
ac
ti
vity
of
the
a
rtific
ia
l
ne
uro
n
is
de
te
rm
ined
a
nd
seco
nd
c
om
petit
ion
determ
ines
the
par
ti
ci
pa
ti
on
of
proce
ssing
el
e
m
ent
.
f
.
E
rro
r
f
un
ct
io
n
a
nd b
ac
k p
ropag
at
e
d value
In
t
his
process
dif
fer
e
nce
bet
ween
cu
r
re
nt
ou
t
pu
t
a
nd
des
ired
ou
t
put
is
cal
culat
ed
.
T
he
cha
ng
e
in
error
is
t
hen
tr
ansfo
rm
ed
to
error
f
un
ct
io
n
t
o
m
at
ch
the
ne
twork
a
rc
hitec
ture
.
T
he
er
ror
value
is
pro
pa
gated
into
the
le
arn
i
ng
f
un
ct
io
n
of
an
oth
e
r
pro
cessi
ng
el
em
e
nt;
this
er
ror
te
rm
is
called
as
a
c
urren
t
error
.
The
c
urren
t
er
ror
pro
pa
gated
back
a
nd
sca
le
d
by
the
le
a
rn
i
ng
functi
on
m
ulti
ply
by
i
nco
m
ing
wei
ght
to
m
od
ify
the lear
ning cycl
e
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
IS
S
N:
20
89
-
4856
Ne
ur
al Net
wo
r
k Ba
se
d MPPT
Co
ntro
ll
er for
So
l
ar
P
V Co
nn
ect
ed
I
nductio
n
M
oto
r
(
T
.
Sh
an
t
hi
)
135
g
.
Lear
ning
functi
on
:
The
pur
pose
of
le
a
rn
i
ng
is
to
m
od
ify
th
e
weig
ht
of
the
in
puts
in
each
processi
ng
el
em
ent
.
The wei
ght i
s c
hange
d
to
obta
in d
e
sired
outp
ut which
is cal
le
d
as
ad
a
ptio
n functi
on
.
2
.
4
.
3
.
St
e
ps
in
vo
l
ved
in
f
r
am
ing
ANN
St
ep
1:
I
nput t
o t
he Artific
ia
l
Neural
Netw
ork
Chan
ge
i
n vo
lt
age
or er
ror vol
ta
ge
from
the
m
oto
r
is give
n as i
nput t
o
the
arti
fici
al
n
eu
ral n
et
w
ork
.
St
ep
2:
Process
ing
of In
pu
t
Fr
om
m
ult
iple
inputs,
a
si
ng
l
e
in
pu
t
is
ch
ose
n
a
nd
ass
um
e
d
m
ini
m
u
m
value
is
s
ubtract
e
d
a
nd
m
ulti
pli
ed
by
the
gain
wh
ic
h
is
t
he
rati
o
of
out
pu
t
ra
nge
t
o
i
nput
ra
nge
and
i
n
s
om
e
sit
uation,
a
ddit
ion
al
bias
is
gi
ven
t
o
the in
pu
t
.
T
his
form
s the f
irst
la
ye
r
of an
arti
fici
al
n
eu
ral
ne
twork
.
St
ep
3:
Ou
t
pu
t
Processi
ng
The
outp
ut
of
the
la
ye
r
1
is
f
ed
bac
k
as
an
input
to
this
la
ye
r
2
a
nd
the
ou
t
pu
t
w
ould
be
a
tw
o
-
dim
e
ns
io
nal
el
e
m
ent
.
This
el
e
m
ent
is
m
ad
e
do
t
pro
duct
s
with
the
weig
ht
s
assum
ed
and
then
giv
e
n
to
t
he
tra
ns
fe
r
f
unct
ion
and the
proces
s of ste
p 2 is
done
in
a
r
e
vers
e m
ann
er
.
3.
SIMULATI
O
N MO
DELIN
G
3
.
1
.
M
PPT
m
od
el
ing
usin
g
Neur
al N
e
two
rk Con
tro
ll
er
(NNC)
Fig
ure
7
s
how
s
the
m
od
el
in
g
ci
rcu
it
of
Ne
ural
Net
work
C
o
nt
ro
ll
er
m
et
ho
d
an
d
it
s
s
ub
syst
e
m
with
increm
ental
co
nductance
al
gorithm
.
It
sho
ws
that
t
he
1k
W
so
la
r
pa
nel
co
nnect
ed
to
the
S
EPIC
c
onve
rter
al
ong
with
sin
gle
phase
i
nduc
ti
on
m
oto
r
(
1HP)
as
a
loa
d
.
The
s
pee
d
of
the
sin
gle
-
ph
as
e
inducti
on
m
otor
i
s
us
e
d
as
a
fee
db
ack
sig
nal
f
rom
wh
ic
h
volt
age
is
de
rive
d
a
nd
e
rro
r
an
d
c
ha
ng
e
i
n
er
r
or
a
re
obta
ine
d
an
d
gi
ven
to
Neural
Network
C
on
tr
oller
.
The
gen
e
r
at
ed
pu
lse
s
f
r
om
the
co
ntr
ol
le
r
are
com
bin
ed
with
t
he
pu
lse
s
ob
ta
ine
d
from
the
In
c
rem
ental
Con
duct
a
nce
Algorithm
of
a
So
la
r
pan
el
and
giv
e
n
to
S
EPIC
co
nverte
r
an
d
desire
d ou
t
pu
t
vo
lt
age
is
produced [
15]
.
Fig
ure
7
.
Mo
de
li
ng
of
ci
rc
uit for
Neural
Network Co
ntr
oller (N
NC)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
IJ
RA
V
ol
.
7
,
No
.
2
,
June
201
8
:
1
29
–
1
39
136
3
.
2
.
M
od
el
in
g o
f
N
eur
al N
e
t
w
ork Contr
oller
subs
ystem
Figure
8
s
how
s
the
m
od
el
ing
dia
gr
am
of
th
e
ne
ural
netw
ork
s
ub
syst
em
.
This
c
om
par
es
the
outp
ut
vo
lt
age
value
with
giv
e
n
re
f
eren
ce
value
.
Accor
ding
t
o
t
ho
s
e
values
of
er
ror
an
d
cha
ng
e
in
e
rror,
t
he
outp
ut
vo
lt
age
is
gen
e
rated
from
the trained ne
ural
netw
orks
.
Fig
ure
8
.
S
ub
s
yst
e
m
ci
r
cuit fo
r
N
e
ur
al
Net
work Co
ntr
oller
4.
SIMULATI
O
N
RESU
LT
S
AND DIS
C
USSION
S
4
.
1
.
1
.
Par
ame
ters
of PV
m
odel
The
fo
ll
owin
g
Fig
ur
e
9
s
ho
w
s
the
sim
ula
ti
on
r
esult
of
outp
ut
volt
age
a
nd
outp
ut
curre
nt
of
the
ph
otovo
lt
ai
c
syst
e
m
.
The
PV
pa
nel
ha
s
been
m
od
el
le
d
us
i
ng
MATL
AB/Si
m
ulink
and
the
outp
ut
vo
lt
age
and
cu
rr
e
nt
wa
vefor
m
s
hav
e
been
obse
rv
e
d
as
143
.
8V
a
nd
6
.
9A
res
pecti
ve
ly
.
These
a
re
dep
ic
te
d
in
Fig
ur
e
9
.
The
outp
ut po
wer f
ro
m
the PV
pan
el
is
obse
rv
e
d
t
o be
1000w
which
h
a
s
been s
how
n
in
Fig
ure
10
.
Fig
ure
9
.
O
utput
volt
age
& c
urren
t
of P
V w
it
h
N
NC
Fig
ure
10
.
O
utp
ut
powe
r of P
V wit
h NN
C
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
RA
IS
S
N:
20
89
-
4856
Ne
ur
al Net
wo
r
k Ba
se
d MPPT
Co
ntro
ll
er for
So
l
ar
P
V Co
nn
ect
ed
I
nductio
n
M
oto
r
(
T
.
Sh
an
t
hi
)
137
4
.
1
.
2
.
Ou
tp
u
t
current
&
vo
l
tage
of
SEPI
C
The
outp
ut
f
rom
the
PV
pa
ne
l
is
fed
to
the
SEPI
C
c
onve
rter
in
or
der
to
boos
t
the
volt
ag
e
.
The
du
ty
cy
cl
e
of
the
S
EPIC
co
nverte
r
is
ad
j
us
te
d
with
the
us
e
of
neural
net
work
c
ontr
oller
in
or
der
to
s
ta
bili
ze
the
vo
lt
age
at
the
outp
ut
of
the
c
onve
rter
.
The
neural
ne
twork
co
ntr
ol
le
r
is
em
bedd
ed
with
i
ncr
e
m
enta
l
cond
uctance
MPPT
al
gorit
hm
wh
ic
h
e
xtracts
the
m
axi
m
u
m
po
we
r
from
the
PV
pan
el
unde
r
va
ryi
ng
irrad
ia
nce
co
ndit
ion
s
.
T
he
vo
lt
age
at
the
ou
t
pu
t
of
t
he
SEP
IC
c
onve
rter
is
obse
rv
e
d
to
be
23
0
.
6V
a
nd
c
urrent
is
4
.
06A
w
hich
h
as
b
ee
n
s
how
n
in
Fig
ure
11
.
Fig
ure
11
.
O
utp
ut
volt
age
&
current
of
SE
P
IC
co
nvert
er
wi
th NNC
4
.
1
.
3
.
Speed
of
t
he
Mot
or
The
ou
t
pu
t
from
the
SEPI
C
conve
rter
is
f
ed
to
t
he
sin
gl
e
-
phase
bri
dg
e
inv
e
rter
wh
i
ch
co
nverts
the
volt
age
to
AC
in
or
der
t
o
fee
d
it
to
the
sing
le
phase
i
nductio
n
m
oto
r
.
The
outp
ut
s
pped
of
t
he
i
nduction
m
oto
r
is
ob
se
r
ved
.
It
is
f
ound
that
the
s
pee
d
of
the
m
oto
r
gr
a
dual
ly
incre
ases
an
d
at
ta
ins
1000
r
pm
and
after
the
i
ns
ta
nt
of
t=
0
.
5
sec
onds
,
the
s
peed
of
the
m
oto
r
rea
ches
it
s
ste
ad
y
sta
te
val
ue
of
1
50
0
r
pm
.
Ou
t
put
par
am
et
er
valu
es are
f
urnishe
d
in
Ta
ble 2, a
nd in Fi
gure
12
is outp
ut s
pee
d of t
he
m
oto
r wit
h NN
C
.
Table
2
.
O
bs
er
ved sim
ulati
on
p
a
ram
et
ers
Ou
tp
u
t vo
ltag
e
Ou
tp
u
t curre
n
t
Sp
eed o
f
the
m
o
to
r
230
.
6
V
4
.
0
5
A
1
5
0
0
r
p
m
Fig
ure
12
.
O
utp
ut
sp
ee
d of t
he
m
oto
r
with
NNC
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2089
-
4856
IJ
RA
V
ol
.
7
,
No
.
2
,
June
201
8
:
1
29
–
1
39
138
Using
N
NC,
t
he
vo
lt
age
a
nd
current
from
pan
el
is
ob
ta
i
ne
d
as
14
3
.
8V
a
nd
6
.
98
A
an
d
outp
ut
powe
r
as
10
00W
.
T
he
obta
ine
d
volt
age
an
d
c
urr
ent
is
boos
te
d
to
230
.
6V
a
nd
4
.
05
A
us
in
g
S
EPIC
co
nverter
.
The
i
nv
e
rted
volt
age
is
fed
to
the
in
du
ct
io
n
m
oto
r
at
no
-
load
c
onditi
on
w
hich
runs
at
1500
r
pm
.
Co
m
par
ing
with
the
res
ults
of
the
syst
em
with
c
onve
ntion
al
c
ontr
oller
s
li
ke
P
,PI
an
d
PID
c
on
t
ro
ll
er
s,
it
is
obse
rv
e
d
tha
t
the
pro
pose
d
s
chem
e
with
ne
ur
al
netw
ork
c
on
t
ro
ll
er
at
ta
in
s
the
ste
a
dy
sta
te
value
of
s
pe
ed
f
ast
er
perf
or
m
s
bette
r
[16]
.
Ta
ble
3
s
hows
th
e
com
par
iso
n
c
har
t
of
the
resul
ts
from
the
co
nv
e
ntio
nal
c
ontrolle
r
a
nd
t
he
neural
netw
ork
c
ontro
ll
er
.
Table
3
.
C
om
par
iso
n of co
nv
entional m
et
hod
a
nd pr
opos
e
d schem
e
Para
m
eters
Co
n
v
en
tio
n
al con
t
roller
Prop
o
sed
con
trolle
r
Stead
y
stat
e
sp
eed
1
4
9
0
r
p
m
1
500
rp
m
Settlin
g
ti
m
e
0
.
7
5
seco
n
d
s
0
.
5
seco
n
d
s
Ris
in
g
ti
m
e
0
.
1
0
.
0
5
seco
n
d
s
5
.
CONCL
U
SION
This
p
a
per
hi
ghli
gh
ts
the
de
s
ign
of
t
he
ne
ural
net
wor
k
c
ontr
oller
f
or
t
he
sp
ee
d
co
ntr
ol
of
si
ng
le
ph
a
se
i
nductio
n
m
oto
r
of
the
rate
d
ca
pacit
y
of
1HP
wh
ic
h
is
dri
ve
n
by
s
olar
ene
r
gy
sy
stem
.
The
m
axi
m
u
m
powe
r
from
the
so
la
r
pa
nel
i
s
ext
racted
usi
ng
I
nc
rem
ental
Co
nduct
a
nce
Algorithm
.
Th
e
volt
age
an
d
curren
t
ob
ta
ine
d
f
ro
m
the
s
olar
pa
nel
is
re
gula
te
d
by
SEP
IC
c
onve
rter
w
hich
fee
ds
t
he
si
ng
le
phase
i
nv
e
rter
a
nd
that
dr
i
ves
t
he
in
duct
ion
m
oto
r
.
T
he
pulse
s
f
ro
m
the
r
especti
ve
con
t
ro
ll
er
a
nd
In
c
rem
ental
Cond
uctance
al
gorithm
base
d
MPP
T
C
on
t
ro
ll
er
is
co
m
bin
ed
the
n
a
pp
li
ed
to
the
S
EPIC
c
onve
rter
to
bo
os
t
the
vo
lt
age
obta
in
ed
from
so
la
r
pan
el
.
T
he
sp
ee
d
of
th
e
Indu
ct
io
n
M
otor
is
giv
e
n
as
feedbac
k
to
the
con
t
ro
ll
er
.
The
sim
ulatio
n
for
the
neural
net
work
c
ontrolle
r
has
bee
n
car
ri
ed
out
.
From
sim
ulati
on
re
su
lt
s
ob
ta
in
ed
,
it
is
obser
ve
d
that
Neural
Netw
ork
c
on
t
ro
ll
er
shows
bette
r
pe
r
form
ance
com
par
e
d
t
o
the
c
onve
ntion
al
c
ontr
oller
f
or
the
sp
ee
d
con
t
ro
l
of in
duct
ion
m
oto
r
.
REFERE
NCE
S
[1]
Gwo
-
Bin
W
u
a
nd
Chin
-
Si
en
Moo,
“
Maximum
Pow
er
Poi
nt
Tr
ac
k
ing
with
Ripp
le
Curre
n
t
Ori
entati
on
f
or
Photovolt
aic
Ap
pli
c
at
ions”
,
IE
E
E
Journal
of
Emerging
and
Se
lec
te
d
Top
ic
s
in
Po
wer
El
e
ct
ron
ic
s,
Dec
ember
2014,
Vol
.
2
,
No
.
4
.
[2]
Gom
at
hi
.
B,
and
Sivaka
m
i
.
P,
“
A
n
Inc
r
emental
C
onduct
an
ce
Alg
orit
hm
base
d
sol
ar
m
axi
m
um
po
wer
poin
t
tracki
ng
s
y
stem”,
Int
ern
a
ti
onal Journal of Electrica
l
Eng
in
ee
ring
.
ISS
N 09
74
-
2158
Volum
e
9,
Num
ber
1
(2
016),
pp
.
15
-
24
.
[3]
Sara
vana
Selva
n
,
Prat
ap
Nair
,
U
m
a
y
a
l,
“
A
Review
on
Photo
Volta
i
c
MP
PT
Algorit
hm
s,
Inte
rn
ational
Journa
l
of
El
e
ct
ri
ca
l
and
C
om
pute
r
Engi
n
e
eri
ng
(IJEC
E)
IS
SN
:
2088
-
8708
Vol
.
6
,
No
.
2,
April
2016
,
pp
.
56
7~582
.
[4]
Mr
.
Partha
Sar
athiMaji,
Mr
.
S
.
Dikshit,
Prof
.
S
.
Mohapa
tra,
“
Modell
ing
and
Sim
ula
ti
on
of
Photovolt
a
ic
Mode
l
Us
ing
Inc
remen
ta
l
Condu
ctanc
e
Algorit
hm
”
In
t
ern
ational
Journal
of
Engi
n
ee
ri
ng
and
Man
agem
ent
Resea
r
ch,
Volum
e
-
4,
Iss
ue
-
2,
April
-
20
14
.
[5]
Nur
Moham
m
a
d,
Md
.
As
iful
I
slam,
Ta
r
equl
Kari
m
and
Qua
zi
Delwa
r
Hos
s
ai
n,
“
Im
prove
d
Solar
Photovolt
ai
c
Arra
y
Model
wi
t
h
FLC
Based
Maximum
Pow
er
Point
Tr
ac
king
”,
Inte
rn
at
ion
al
Journal
of
E
l
ectrical
and
Com
pute
r
Engi
ne
eri
ng
(
IJE
CE)
ISS
N:
2088
-
8708
_
717
Vol
.
2
,
No
.
6,
Dec
e
m
ber
2012,
pp
.
7
17~730
[6]
T
.
Shanthi,
and
N
.
Am
m
asa
iGounden,
“
Pow
er
e
le
c
tronic
inte
rfa
ce
for
gr
id
-
c
onnec
t
ed
PV
ar
ra
y
using
boost
conve
rt
er
and li
n
e
-
comm
uta
te
d
in
ver
te
r
wi
th
MP
PT”,
Inte
rn
at
ion
al Confe
ren
ce
on
I
nte
lligent a
nd
A
dvanc
e
s
y
stems
,
page
s: 882
-
886
,
DO
I:
10
.
1109/I
CIAS
.
2007
.
4658513
.
[7]
Jam
es
Dunia,
B
a
kar
i
.
M
and
M
.
Mw
in
y
iwiwa
,
“
Perform
anc
e
Co
m
par
ison
bet
w
een
ĆUK
and
SEP
IC
Conver
te
rs
fo
r
Maximum
Powe
r
Poin
t
Tr
ac
ki
ng
Us
ing
Inc
re
m
ent
al
Conduc
t
anc
e
Techni
que
in
Solar
Pow
er
Applic
a
ti
ons”,
Inte
rna
ti
ona
l
Jo
urna
l
of
El
e
ct
r
i
ca
l
,
Com
pute
r
,
Ene
rge
ti
c
,
E
lect
ronic
and
Com
m
unic
at
ion
Eng
ine
er
ing
Vol
.
7
,
No:
12,
2013
.
[8]
Moham
ed
Ta
h
ar
Makhloufi,
Y
assine
Abdess
eme
d,
Moham
ed
Sal
ah
Khir
eddi
ne
,
“
A
Feed
forw
ard
Neura
l
Network
MP
PT
Control
Strat
eg
y
Appli
e
d
to
a
Modifie
d
Cuk
Conver
te
r
”,
Int
ern
a
ti
ona
l
Journal
of
Elec
trica
l
and
Com
pute
r
Engi
ne
eri
ng
(
IJE
CE)
ISS
N:
2088
-
8708
Vol
.
6,
No
.
4
,
Augus
t
2016
,
pp
.
1421~1433
.
[9]
Sangee
th
a
S
,
an
d
Jith
a
joseph
“
Design
an
d
Im
pl
ementa
t
ion
o
f
S
epi
c
Conver
te
r
Based
PV
S
y
ste
m
Us
ing
Modifi
ed
Inc
rement
a
l
Co
nduct
an
ce
Algo
rit
hm
”,
Int
ern
a
t
iona
l
Conf
ere
n
c
e
on
Elec
tri
c
al,
El
e
ct
roni
cs,
a
nd
Optimiza
t
io
n
Te
chn
ique
s (IC
E
EOT)
-
978
-
1
-
46
73
-
9939
-
5/16/
$3
1
.
00
©2016
IEEE
.
[10]
F
.
Lftisi,
G
.
H
.
George
,
A
.
Aktai
bi
,
C
.
B
.
B
utt
,
and
M
.
A
.
Rahman,
“
Artifica
l
n
eur
al
n
et
w
ork
base
d
spee
d
cont
roller
fo
r
in
duct
ion
m
otors”, 978
-
1
-
5090
-
347
4
-
1/16/
$31
.
00©
2016
IEEE
.
[11]
W
hei
-
Min
Li
n,
Chih
-
Ming
H
ong,
and
Chiu
ng
-
Hs
ing
Chen,
“
Neura
l
-
Netw
ork
-
Based
MP
PT
Contro
l
of
a
Stand
-
Alone
H
y
brid
Pow
er
G
ene
ra
ti
on
S
y
s
tem
”,
IEEE
Tr
an
sac
ti
ons
on
Po
wer
El
e
ct
ron
ic
s,
Vol
.
26,
No
.
1
2,
Dec
ember
-
2011
.
Evaluation Warning : The document was created with Spire.PDF for Python.