Int
ern
at
i
onal
Journ
al of
P
ower E
le
ctr
on
i
cs a
n
d
Drive
S
ystem
(I
J
PE
D
S
)
Vo
l.
11
,
N
o.
4
,
Decem
be
r 202
0
, p
p.
1785
~
1798
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v11.i
4
.
pp
1785
-
1798
1785
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
Modulat
ion
ind
ex
effect
on
i
nverter
based
i
nducti
on
mot
or
dr
i
ve
Ak
hil
esh
Sh
ar
ma
1
,
Anan
dh
N.
2
,
Sa
r
sing
G
ao
3
1,3
El
e
ct
ri
ca
l
Eng
ine
er
ing
Dep
artme
nt
,
North
Ea
ste
rn
Re
gional
In
sti
tute
of
Sci
ence
&
Tec
hnol
ogy,
Nirj
uli,
Arunachal
Pr
a
des
h,
India
2
Depa
rtment
of
El
e
ct
ri
ca
l
and
E
l
ec
tron
ic
s
Eng
ineeri
ng,
Manip
al
I
nstit
ute
of
Techn
ology,
Man
ipa
l
Aca
dem
y
of
Hig
her
Educ
a
ti
on,
Mani
pal
,
Karna
ta
k
a,
I
ndia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
hist
or
y:
Re
cei
ved
Ja
n
23
,
20
20
Re
vised
A
pr
26
,
20
20
Accepte
d
M
a
y
3
1
,
20
20
Due
to
a
subs
tantia
l
in
crease
in
the
use
of
inv
e
rte
r
for
num
ero
us
el
e
ct
ri
cal
appl
i
anc
es
starti
ng
from
dome
st
ic
s
to
industr
ia
l
drive
s
,
an
inverter
may
be
dire
c
tl
y
conn
ec
t
ed
to
the
power
grid
sys
tem.
Th
e
dep
ende
n
cy
on
an
inve
rt
er
has
bee
n
in
creased
over
th
e
y
ea
r
s.
Henc
e,
the
pr
oper
and
eff
icie
nt
design
of
the
inv
erter
will
lead
to
high
er
ef
fic
i
enc
y.
One
of
the
ma
jor
challenges
is
the
gene
ra
ti
on
of
sui
ta
bl
e
ga
te
pu
lses
for
power
sw
itc
hing
device
s,
wh
ic
h
in
turn
depe
nds
on
the
modul
ation
index.
Th
e
se
le
c
ti
on
of
prope
r
modu
la
ti
on
ind
ex
will
h
el
p
in
th
e
produc
ti
on
of
th
e
ra
te
d
voltage
.
If
the
modulati
o
n
inde
x
is
le
ss
,
the
dur
atio
n
of
on
-
ti
m
e
p
ulses
will
be
les
s
and
henc
e
,
t
he
device
's
conduc
t
ion
ti
m
e
is
al
so
l
ess,
th
ere
by
th
e
ou
tpu
t
vo
lt
ag
e
of
the
inv
ert
er
is
red
uce
d
.
A
r
edu
ce
d
vo
lt
ag
e,
when
app
li
ed
to
an
induc
t
ion
mo
to
r
will
h
ave
lower
spe
ed
an
d
eve
n
i
ts
per
fo
rma
nc
e
wi
ll
be
sluggish.
Th
e
s
pee
d
of
th
e
mot
or
i
mprove
s
when
it
is
oper
a
t
ed
in
a
c
losed
-
lo
op
for
the
same
modul
ation
inde
x.
Thi
s
rese
a
rch
pape
r
tries
to
bring
out
the
eff
ec
t
of
modul
a
ti
o
n
inde
x
on
spee
d
con
trol
of
an
induc
t
ion
m
otor
base
d
on
an
inve
r
te
r
for
b
oth
open
as
well
as
c
losed
-
loop
oper
ation.
The
simu
la
t
ed
result
s
indicate
th
at
th
e
modul
ation
index
in
the
vi
ci
ni
ty
to
un
it
y
wil
l
p
r
ovide
ra
te
d
vol
t
age
for
th
e
smooth
oper
at
io
n
of
th
e
mot
or
.
Ke
yw
or
d
s
:
Ind
uction
Mot
or
M
od
ulati
on
I
ndex
Neural
Netw
ork
Sp
ace
Vecto
r
PWM
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
:
Akhile
sh
S
harma,
Ele
ct
rical
Eng
i
neer
i
ng
De
par
t
ment,
North
East
er
n
Re
gional
I
ns
ti
tute
of
Scie
nce
&
Tec
hnolog
y,
Nirjuli,
Papum
Pare
,
Arun
ac
ha
l
Pr
a
desh
79
1109,
I
ndia
.
Emai
l:
as@n
e
r
ist
.ac.in
1.
INTROD
U
CTION
Fr
e
qu
e
nt
fau
lt
on
the
power
sy
ste
m,
un
a
vai
la
bili
ty
of
pow
er
to
rem
ote
l
ocati
on
a
nd
a
va
il
abili
ty
of
so
la
r
e
ne
rgy
at
su
ch
locat
io
ns
ha
ve
in
creas
ed
the
util
it
y
of
i
nverter
,
as
they
ser
ve
as
a
sta
ndby
ac
powe
r
su
ppli
es
f
rom
a
DC
source
[
1]
to
meet
the
load
dema
nd.
Eve
n,
in
ver
te
rs
are
ap
plied
for
the
in
dustria
l
dr
ive
durin
g
s
hu
t
do
wn
or
powe
r
f
ai
lure
due
to
f
ault.
He
nce,
the
op
e
rati
on
of
inv
e
rter
pla
ys
an
imp
or
ta
nt
r
ole
in
su
c
h
a
pp
li
cat
io
ns
.
T
he
in
ve
rter
ci
rc
uit
c
on
sis
ts
of
ma
ny
s
wi
tc
hin
g
dev
ic
es
.
To
tu
rn
on
th
ese
de
vices,
prop
e
r
trigg
e
rin
g
pulse
s
are
esse
ntial
.
The
du
rati
on
of
these
pul
ses
de
pe
nds
on
the
s
witc
hing
fr
e
qu
e
nc
y
a
nd
its
modu
la
ti
on
i
ndex
(MI)
[2
-
7].
MI
sel
ect
ion
be
comes
imp
ort
ant
pr
ov
i
ded
the
switc
hi
ng
f
re
qu
e
nc
y
is
c
ons
ta
nt.
With
the
a
dva
nceme
nt
in
the
desig
n
of
in
ve
rter,
the
y
a
re
c
apab
le
of
pro
vi
ding
the
powe
r
need
e
d
for
industrial
dri
ve
s.
Mostl
y,
the
se
dr
i
ves
c
onsist
of
i
nductio
n
mo
to
rs
(IM
)
because
the
y
ha
ve
s
pecial
fea
tures
li
ke
sim
ple,
rugg
e
d
in
co
ns
tr
uction
with
minimu
m
mainte
nan
ce
need,
et
c.
M
ost
im
port
antly,
the
y
ca
n
oper
at
e
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
.
11
, N
o.
4
,
D
ecembe
r
2020
:
1785
–
1798
1786
at
an
y
en
vir
onmental
c
onditi
on
s
.
The
se
mot
or
s
are
pr
e
ferred
for
c
onsta
nt
s
pee
d
opera
ti
on
.
T
he
c
ha
nge
in
sp
ee
d
f
r
om
no
loa
d
to
f
ull
lo
ad
is
small
.
H
ence
t
hese
mo
t
or
s
are
wi
dely
us
e
d
in
i
ndus
t
ries.
Motors
use
d
in
industries
nee
d
co
ntin
uous
e
le
ct
rical
po
w
er
.
Disr
upti
on
in
the
el
ect
rical
powe
r
le
ads
to
the
co
mp
le
te
sh
ut
dow
n
of
the
i
ndus
t
ries
as
el
ec
tric
al
energy
c
onve
rsion
is
m
os
tl
y
based
on
the
co
nventio
na
l
so
urces
of
e
nerg
y.
The
dr
i
ve
sto
pp
a
ge
may
be
av
oid
e
d,
pro
vi
ded
t
he
m
otor
dri
ves
sy
ste
ms
are
base
d
on
both
i
nv
e
rt
er
an
d
conve
ntion
al
powe
r
plants
.
T
he
i
nv
e
rter
-
bas
ed
dr
i
ves
may
be
a
ble
to
provi
de
the
rate
d
volt
age
nee
ded
for
the
dr
i
ves.
The
in
ver
te
r
ou
t
pu
t
volt
age
de
pends
on
t
he
s
witc
hi
ng
pulse
s
[
8],
wh
ic
h
in
tu
rn
is
relat
ed
to
MI.
S
o,
there
is
a
ne
ed
for
a
pro
per
sel
ect
ion
of
MI.
Fo
r
a
pa
rtic
ula
r
MI,
t
he
trig
ge
rin
g
pulse
s
ar
e
gen
e
rated
t
ha
t
are
us
e
d
for
switc
hing
the
dev
ic
es
.
Th
e
ou
t
pu
t
volt
age
of
a
t
hr
ee
-
ph
as
e
inv
e
rter
is
a
ppli
ed
to
a
t
hr
e
e
-
phase
i
nduct
ion
mo
t
or
(IM
),
whose
par
a
mete
rs
are
giv
e
n
in
T
able
1
,
in
an
open
as
well
as
cl
os
e
d
-
l
oop
c
onditi
on.
In
the
cl
os
e
d
-
l
oop
op
e
rati
on
[10
],
ma
ny
con
t
ro
ll
ers
like
P
ID
,
fu
zz
y
-
log
ic
[
11]
or
ne
ural
netw
ork
co
uld
be
em
ployed
.
In
t
his
pa
per
,
a
NN
con
t
ro
ll
er
[
12
-
18]
has
bee
n
employe
d
to
c
on
t
ro
l
t
he
e
rro
r
in
the
sp
e
ed
of
the
m
otor,
ge
ner
at
e
d
fro
m
the
diff
e
re
nce
of
r
efere
nce
sp
e
ed
a
nd
m
otor
s
pe
ed.
T
hus,
co
nt
ro
ll
ing
the
m
otor
to
f
ollow
the
ref
e
rence
s
peed,
wh
il
e,
in
ope
n
-
loop
c
ontrol
,
t
he
s
pee
d
of
IM
is
in
dep
e
nd
ent
of
re
fer
e
nc
e
sp
ee
d.
In
this
pa
per,
an
at
te
mp
t
has
been
mad
e
to
fin
d
the
e
ff
ect
of
MI
on
the
s
peed
of
an
IM
and
t
he
re
su
lt
s
are
obta
ine
d.
The
re
su
lt
s
in
dicat
e
that
the
mod
ulati
on
in
dex,
ne
ar
the
unit
y,
ge
ner
at
es
the
tri
ggeri
ng
pu
lse
s
of
suffici
ent
width
to
pro
du
ce
rate
d
vo
lt
age
for
t
he
m
otor.
B
ut
w
hen
the
MI
is
l
ess
tha
n
0.8
or
m
or
e
tha
n
1.2
5,
the
wi
dth
of
the
switc
hi
ng
pu
lse
s
so
produce
d
is
le
ss
an
d
he
nce
the
in
ver
te
r
ou
tpu
t
vo
lt
age
re
du
ce
s.
T
her
e
f
ore,
the
m
oto
r
ta
kes
a
l
onger
ti
me
to
pick
up
the
re
f
eren
ce
sp
ee
d.
Table
1.
Para
m
et
ers
of
i
nduction m
otor
Nu
m
b
er
of
Po
les
4
Per
Ph
ase
Stato
r
&
Ro
to
r
Res
istan
ce
0
.90
Ω;
0
.66
Ω
Per
Ph
ase
Stato
r
&
Ro
to
r
Ind
u
ctan
ce
0
.00
4
5
7
H
each
Moment
of
in
ertia
0
.13
8
4
Kg
-
m
2
Frequ
en
cy
50
Hz
Load
Torq
u
e
10
N
-
m
DC
so
u
rce
Vo
ltag
e
540
V
2.
MO
DU
L
ATI
ON
I
N
DEX
Con
si
der
two
s
ign
al
s,
a
tria
ng
ular
wa
ve
an
d
a
DC
sig
nal,
a
ct
ing
as
a
car
ri
er
a
nd
mod
ulati
ng
sig
nals
resp
ect
ivel
y
as
show
n
in
Fig
ur
e
1
(i).
W
he
n
the
t
rian
gu
la
r
sig
nal
[
19]
is
com
par
e
d
with
the
DC
si
gnal
su
c
h
that,
the
ma
gnit
ud
e
of
t
he
tria
ngular
be
co
me
s
eq
ual
to
t
hat
of
DC
sig
nal,
then
the
ou
t
pu
t
will
be
co
ns
ta
nt
unti
l
it
beco
mes
le
ss
than
the
DC
si
gn
al
as
s
how
n
in
Fig
ur
e
1
(ii)
and
(iii
)
res
pe
ct
ively,
act
in
g
as
trigg
e
rin
g
pulse
s
for
c
ontr
olled
powe
r
el
ect
r
on
ic
de
vices
T
1
and
T
2
as
s
hown
in
Fig
ur
e
2.
T
he
wi
dth
of
the
pulse
s
dep
e
nds
upon
t
he
mag
ni
tud
e
of
t
he
DC
sig
nal.
The
width
of
the
pulse
for
T
1
dec
reases
as
t
he
mag
nitud
e
of
t
he
DC
so
urce
in
creas
es.
At
the
sam
e
ti
me,
the
pul
se
widt
h
f
or
T
2
increases
t
hereby
i
ncr
ea
sin
g
the
co
nductio
n
per
i
od
of
T
2
.
The
high
an
d
low
dura
ti
on
pulse
s
bec
om
e
e
qu
al
pro
vid
e
d
the
m
od
ulati
ng
si
gn
al
has
zer
o
ma
gn
it
ud
es.
Sin
ce
t
he
a
m
plit
ud
e
of
the
m
odulate
d
si
gn
al
deci
des
the
widt
h
of
t
he
pulse
s,
he
nce,
this
sche
me
of
modu
la
ti
on
is
cal
le
d
as
a
mp
li
tud
e
m
odulati
on
[20].
Th
us
,
t
his
is
de
fine
d
as
the
rati
o
of
the
am
plit
ud
e
of
the
modu
la
ti
ng
si
gnal
,
(V
m
)
to
t
ha
t
of
the
a
mp
li
tu
de
of
the
ca
rr
ie
r
w
ave
,
(V
c
),
i.
e.
=
(1)
If
T
c
is
the
ti
me
pe
rio
d
of
t
he
tria
ngular
wa
ve
,
the
n
c
onduct
ion
pe
rio
ds
for
T
1
an
d
T
2
a
re
e
xpresse
d
as
,
ℎ
=
2
(
1
−
)
(
2)
=
2
(
1
+
)
(3)
Wh
e
re
t
h
a
nd
t
l
are
t
he
c
onduc
ti
on
ti
me
of
T
1
and
T
2
resp
ect
i
vely.
Con
si
der
an
in
ver
te
r
,
as
s
how
n
in
Fi
gure
2.
The
ou
t
pu
t
vo
l
ta
ge
betwee
n
node
s
A
a
nd
B
will
hav
e
a
mag
nitud
e
of
e
it
her
±
0.
5
V
dc
or
0,
de
pendin
g
on
the
DC
m
odulati
ng
sig
nal
.
This
volt
age
has
been
re
ferr
ed
to
as
pole
vo
lt
age
,
V
AO
as
in
dicat
ed
in
Fig
ur
e
2.
The
DC
c
omp
on
e
nt
of
this
volt
age
is
e
valu
at
ed
as,
=
0
.
5
(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
Mo
du
l
ation i
ndex ef
fe
ct
on i
nv
ert
er base
d
in
du
ct
io
n m
oto
r
dr
iv
e (
Akhil
es
h S
ha
r
m
a)
1787
=
(5)
Eq
uations
(
2)
a
nd
(
3)
can
be
e
xpresse
d
in
te
r
ms
of
mod
ulati
on
in
dex
(
MI),
as
ℎ
=
2
(
1
−
)
(6)
=
2
(
1
+
)
(
7)
As
ca
n
be
see
n
f
r
om
e
xpres
sion
(
4)
,
the
mean
volt
age
becomes
ze
ro,
w
hen
the
ma
gn
it
ude
of
the
modu
la
ti
ng
sig
nal
bec
om
es
ze
ro
i.e
.
the
pu
ls
e
width
of
both
low
a
nd
high
s
ign
al
s
are
eq
ua
l
and
hen
c
e
the
are
a
unde
r
both
hal
ves
are
e
qu
al
,
making
the
a
ve
rag
e
value
of
the
DC
c
omp
onent
to
be
ze
r
o.
Along
with
the
DC
com
pone
nt,
the
outp
ut
will
al
so
ha
ve
harmo
nics
of
inte
gr
al
m
ulti
ples
of
t
he
car
rier
fr
e
qu
e
nc
y,
while
the
fr
e
qu
e
nc
y
of
lo
wer
orde
r
harmo
nics
becom
es
eq
ual
to
the
carrier
f
reque
nc
y.
A
m
p
l
i
t
u
d
e
o
f
C
a
r
r
i
e
r
S
i
g
n
a
l
A
m
p
l
i
t
u
d
e
o
f
M
o
d
u
l
a
t
i
n
g
S
i
g
n
a
l
M
o
d
u
l
a
t
i
n
g
S
i
g
n
a
l
C
a
r
r
i
e
r
S
i
g
n
a
l
A
m
p
l
i
t
u
d
e
o
f
S
i
g
n
a
l
s
T
i
m
e
G
a
t
e
P
u
l
s
e
f
o
r
T
1
A
m
p
l
i
t
u
d
e
T
i
m
e
T
c
0
.
5
V
dc
0
.
5
V
dc
T
i
m
e
T
h
T
l
A
m
p
l
i
t
u
d
e
G
a
t
e
P
u
l
s
e
f
o
r
T
2
T
i
m
e
V
AO
(
i
)
M
o
d
u
l
a
t
i
n
g
a
n
d
C
a
r
r
i
e
r
S
i
g
n
a
l
s
(
ii
)
G
a
t
e
P
u
l
s
e
f
o
r
T
1
(
iii
)
G
a
t
e
P
u
l
s
e
f
o
r
T
2
(
iv
)
O
u
t
p
u
t
V
o
l
t
a
g
e
,
V
ao
Fig
ure
1
.
Ge
ne
rati
on
of
m
odul
at
ing
an
d
car
rier
sig
nals,
gate
pulse
s
a
nd
pol
e
volt
age
The
DC
mod
ul
at
ing
si
gn
al
may
be
re
plac
ed
by
a
sin
usoidall
y
va
r
ying
si
gn
al
[
21]
with
a
certai
n
amplit
ude,
ph
a
se
an
d
f
reque
nc
y.
T
he
f
re
qu
e
ncy
of
this
sig
nal
is
quit
e
le
ss
than
t
he
car
r
ie
r
fr
e
quenc
y.
This
makes
the
modu
la
ti
ng
sig
nal
vi
rtuall
y
c
on
s
ta
nt
ov
e
r
the
high
car
rier
f
r
equ
e
nc
y,
ma
ki
ng
the
a
ve
rage
pole
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
.
11
, N
o.
4
,
D
ecembe
r
2020
:
1785
–
1798
1788
vo
lt
age
to
de
pe
nd
on
the
ma
gn
it
ude
of
t
he
modu
la
ti
ng
si
gnal
.
T
hus,
the
po
le
volt
age
w
avefor
m
will
ha
ve
a
low
-
f
reque
ncy
c
ompone
nt
w
ho
s
e
i
ns
ta
nta
ne
ous
mag
nitu
de
is
pro
portio
na
l
to
the
modul
at
ing
signa
l.
It
will
al
so
hav
e
a
higher
f
reque
ncy
harmo
nic
vo
lt
a
ge.
If
DC
is
us
ed
as
a
m
odulati
ng
sig
nal,
t
her
e
will
be
ha
rm
on
ic
fr
e
qu
e
ncies.
These
fr
e
que
nc
ie
s
will
be
an
i
nteg
ral
m
ul
ti
ple
of
the
c
arr
ie
r
f
re
qu
e
nc
y.
But
t
his
does
not
ha
pp
e
n,
if
DC
is
re
pl
aced
by
si
nus
oid
al
ly
varyin
g
si
gn
al
w
her
e
pulse
width
de
pe
nds
on
the
f
requ
ency
of
m
odul
at
ing
sig
nal.
This
is
visible
from
equ
at
io
ns
(
2)
a
nd
(
3)
res
pecti
vely.
D
ue
to
t
hi
s,
there
will
be
harmo
nics
in
the
pole
volt
a
ges
a
nd
t
her
e
e
xists
a
band
of
f
reque
ncies
in
t
he
vicinit
y
to
a
c
ar
rier
an
d
its
m
ulti
ple
f
reque
ncies
.
The
ba
nd
f
re
qu
e
ncies
a
re
f
orme
d
by
an
i
ntegr
al
mu
lt
iple
of
the
f
reque
ncy
of
modu
la
ti
ng
sig
nals.
It
is
usual
to
ha
ve
mod
ulati
ng
fr
e
quenc
y
quit
e
le
ss
in
co
mp
a
r
ison
with
the
c
arr
ie
r
fr
e
quenc
y,
hen
ce
the
frequ
e
nc
y
of
the
domina
nt
ha
r
monics
will
be
in
the
cl
os
e
vicinit
y
of
car
rier
fr
e
que
ncy
an
d
its
i
ntegr
al
mu
lt
iple.
Figure
2
.
A
simpli
fied
in
ver
t
er
to
polo
gy
3.
SPACE
VEC
TOR
P
ULSE
WIDTH
M
O
DU
L
ATIO
N
(
SV
PW
M
)
This
is
a
vecto
r
te
c
hn
i
qu
e
t
ha
t
is
a
pp
li
e
d
for
pulse
wi
dth
m
odulati
on
(PW
M
)
for
3
-
Φ
in
ve
rters.
T
his
scheme
is
wi
de
ly
use
d
f
or
ge
ner
at
in
g
gatin
g
si
gn
al
s
f
or
hi
gh
volt
age
with
low
harmo
ni
c
distor
ti
on,
s
uitable
for
var
ia
ble
f
r
equ
e
nc
y
in
dust
rial
dr
i
ves
s
uc
h
as
in
du
ct
io
n
mo
t
or.
T
his
t
echn
i
qu
e
ca
n
be
e
xp
la
in
ed
with
the
help
of
Fi
gure
3.
T
he
ci
rc
uit
will
produce
t
wo
volt
age
le
ve
ls.
It
co
ns
ist
s
of
si
x
co
n
tr
olled
switc
hes,
S
1
to
S
6
and
a
DC
s
our
ce
vo
lt
age
of
“V
s
”.
T
her
e
ar
e
ei
ght
possibl
e
s
witc
hing
ve
ct
or
s
[
22]
,
as
dep
ic
te
d
in
Fi
gure
4.
The
sta
te
s
V
0
[
000]
a
nd
V
7
[
111]
a
re
nu
ll
ve
ct
or
s
w
hile
V
1
-
V
6
a
re
act
ive
vect
or
s
.
T
he
nu
ll
vecto
r
pro
du
ce
s
zero
volt
age
w
hile
an
act
ive
vecto
r
pr
oduce
s
non
-
zer
o
volt
age.
T
he
sp
ace
vecto
r
[23
-
25]
volt
age
gen
e
r
at
ion
,
oth
e
r
tha
n
t
he
se
sta
te
s,
is
s
how
n
in
Fig
ur
e
5.
T
he
re
qu
ired
volt
age
ve
ct
or
s
c
ould
be
obta
ined
ba
s
ed
on
equ
at
io
ns
(
8)
a
nd
(
9)
w
hich
c
ou
l
d
be
e
xten
de
d
to
3
-
dim
ens
ion
al
s
pace
ve
ct
or
[26]
as
we
ll
.
This
est
a
blishes
a
ro
ta
ti
ng
vect
or
[27]
a
nd
it
is
simi
la
r
to
a
ro
ta
ti
ng
fl
ux
de
velo
ped
in
the
sta
tor
of
an
i
nductio
n
m
otor.
T
he
ro
ta
ti
ng
flu
x
pa
sses
th
r
ough
t
he
ai
r
ga
p
to
t
he
r
oto
r
si
de.
This
te
c
hn
i
que
co
uld
be
e
xte
nd
e
d
to
a
mu
lt
i
-
le
vel
inv
e
rter
as
wel
l.
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
M
o
t
o
r
Figure
3:
I
nver
te
r
base
d
mo
t
or
loa
d
represe
nt
at
ion
V
dc
0
.
5
V
dc
0
.
5
V
dc
+
+
-
-
O
A
T
1
T
2
V
AO
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
Mo
du
l
ation i
ndex ef
fe
ct
on i
nv
ert
er base
d
in
du
ct
io
n m
oto
r
dr
iv
e (
Akhil
es
h S
ha
r
m
a)
1789
Fr
om
Fig
ure
5,
the
ref
e
re
nce
vo
lt
age
,
̅
̅
̅
̅
̅
,
is
exp
ress
ed
as
,
in
equ
at
io
n
(
8).
T
he
co
rr
es
pondi
ng
li
ne
vo
lt
age
s
are
ta
bu
la
te
d
in
Tabl
e
2
.
̅
̅
̅
̅
̅
=
1
(
1
⁄
)
+
2
(
2
⁄
)
+
3
(
3
⁄
)
(8)
1
⁄
+
2
⁄
+
3
⁄
=
1
(9)
Her
e
T
1
,
T
2
an
d
T
3
a
re
t
he
pe
rio
ds
th
r
ough
ve
ct
or
s
V
1
,
V
2
a
nd
V
3
res
pecti
vely,
T
s
is
sa
m
pling
pe
rio
d.
V
1
=[
1
0
0
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
ii
)
S
t
a
t
e
2
V
2
=[
1
1
0
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
i
i
i
)
S
t
a
t
e
3
V
3
=[
0
1
0
]
Vs
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
iv
)
S
t
a
t
e
4
Vo
=[
1
1
1
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
i
)
S
t
a
t
e
1
V
5
=[
0
0
1
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
ii
)
S
t
a
t
e
6
V
6
=[
1
0
1
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
i
i
i
)
S
t
a
t
e
7
V
7
=[
0
0
0
]
Vs
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
iv
)
S
t
a
t
e
8
V
4
=[
0
1
1
]
V
s
S
1
S
3
S
5
S
4
S
6
S
2
R
Y
B
(
i
)
S
t
a
t
e
5
1
i
n
d
i
c
a
t
e
s
O
N
s
t
a
t
e
s
o
f
u
p
p
e
r
s
w
i
t
c
h
e
s
S
1
,
S
3
o
r
S
5
w
h
i
l
e
0
"
s
h
o
w
s
O
N
s
t
a
t
u
s
o
f
l
o
w
e
r
s
w
i
t
c
h
e
s
S
2
,
S
4
o
r
S
6
,
V
o
-
V
7
r
e
p
r
e
s
e
n
t
o
u
t
p
u
t
v
o
l
t
a
g
e
s
;
V
o
a
n
d
V
7
r
e
p
r
e
s
e
n
t
z
e
r
o
v
e
c
t
o
r
s
;
V
1
-
V
7
r
e
p
r
e
s
e
n
t
a
c
t
i
v
e
v
e
c
t
o
r
s
Figure
4
.
S
witc
hing
sta
te
s
of
the
two
-
le
vel
in
ver
te
r
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
.
11
, N
o.
4
,
D
ecembe
r
2020
:
1785
–
1798
1790
Table
2.
V
oltage
vecto
r wit
h i
ts resp
ect
ive
li
ne vo
lt
age
Vo
ltag
e
Vector
Switch
in
g
Vector
Line Vo
ltag
es
a
b
c
V
RY
V
YB
V
BR
V
o
0
0
0
0
0
0
V
1
1
0
0
Vs
0
-
Vs
V
2
1
1
0
0
Vs
-
Vs
V
3
0
1
0
-
Vs
Vs
0
V
4
0
1
1
-
Vs
0
Vs
V
5
0
0
1
0
-
Vs
Vs
V
6
1
0
1
Vs
-
Vs
0
V
7
1
1
1
0
0
0
Figure
5. S
pac
e v
ect
or
repres
entat
ion
4.
DYN
AM
I
C
M
ODELL
IN
G
OF
I
NDU
CTI
ON
MOT
OR
An
in
duct
ion
mo
to
r
(IM
)
po
ssesses
man
y
s
pecial
feat
ur
es
li
ke
r
obus
t
in
c
on
st
ru
ct
io
n,
hig
h
to
rque
to
inerti
a
rati
o,
a
ble
to
be
util
iz
ed
in
a
ny
en
vi
ronme
nt.
Hen
c
e,
m
os
t
of
the
industrial
dri
ve
s
us
e
an
in
du
ct
io
n
mo
to
r.
T
he
sta
tor
of
the
in
du
c
ti
on
mo
t
or
r
eq
uires
a
t
hr
ee
-
phase
sup
ply
.
T
his
will
est
abli
sh
MMF
in
the
sta
to
r
wh
ic
h
in
tu
rn
pa
sses
th
rou
gh
the
ai
r
ga
p,
li
nked
with
the
r
ot
or
ci
rcu
it
.
Th
us,
an
em
f
is
in
duced
in
t
he
ro
t
or
a
nd
it
is
simi
la
r
to
a
tran
sf
ormer.
Hen
ce
,
t
he
e
quivale
nt
m
odel
of
the
tran
sf
ormer,
with
sli
gh
t
modific
at
io
n,
co
uld
be
us
e
d
f
or
IM.
To
dra
w,
eq
ui
valent
ci
rcu
it
di
agr
am
of
IM,
t
he
a
ssumpti
ons
[
27,
28]
ma
de
are
;
(i)
T
he
re ex
ist
s
un
i
form
ai
r
ga
p,
(ii)
Si
nu
s
oi
dal
fl
ux
distr
ibu
ti
on
is
un
i
f
orm,
(iii
)
Ef
fe
ct
of
c
ha
nge
in
t
he
parame
te
r
is
neg
le
ct
e
d,
(iv
) Effect
of satu
r
at
ion
is
ne
glect
ed.
The
perf
or
ma
nc
e
of
the
IM
c
ou
l
d
be
easi
ly
ob
ta
ine
d
base
d
on
its
eq
ui
valent
ci
rc
uit
model
w
hich
is
sh
ow
n
in
Fig
ure
6.
T
his
help
s
to
stu
dy
t
he
ste
ady
-
sta
te
c
ha
racteri
sti
cs,
ne
glect
ing
the
t
ran
sie
nt
sta
te
wh
ic
h
occurs
due
to
change
in
t
he
load
or
e
ven
c
hange
in
fr
e
qu
ency,
espe
ci
al
ly
in
va
riable
s
peed
dr
i
ves.
M
ost
ly,
these
dr
i
ves
a
r
e
base
d
on
co
nv
e
rter
ci
rc
uits.
T
he
c
onve
rte
rs
a
re
fe
d
t
hro
ugh
de
finite
s
ource
volt
age
.
Hen
ce
,
there
is
a
ce
rtai
n
li
mit
to
draw
t
he
outp
ut
po
wer
from
t
hem.
More
ove
r,
filt
e
rs
of
de
finite
siz
e
ma
y
be
connecte
d
to
s
hap
e
the
outp
ut
vo
lt
ag
e
of
t
he
co
nv
e
rters
.
T
hi
s
li
mit
s
them
to
s
uppl
y
la
r
ge
transient
pow
e
r
a
nd
makes
esse
ntial
to
stu
dy
the
dyna
mic
m
odel
li
ng
of
s
uch
dri
ves.
0
s
e
q
u
e
n
c
e
e
q
u
i
v
a
l
e
n
t
c
i
r
c
u
i
t
Q
-
a
x
i
s
e
q
u
i
v
a
l
e
n
t
c
i
r
c
u
i
t
-
+
-
+
R
s
ω
e
Ψ
qs
L
ls
L
l
r
(
ω
e
-
ω
r
)
Ψ
qr
V
ds
I
ds
L
m
I
dr
R
r
+
-
V
dr
I
dr
V
o
r
+
L
l
r
-
V
o
s
I
o
s
-
R
s
L
ls
+
Rr
V
q
s
V
q
s
D
-
a
x
i
s
e
q
u
i
v
a
l
e
n
t
c
i
r
c
u
i
t
+
+
-
R
s
L
ls
L
m
L
l
r
R
r
(
ω
e
-
ω
r
)
Ψ
dr
+
-
+
-
-
I
q
s
I
qr
ω
e
Ψ
ds
Figure
6:
A
dq0
e
qu
i
valent
ci
rcu
it
of
an
I
nduction
M
ot
or
d
-
a
x
i
s
q
-
a
x
i
s
V
1
=[
1
0
0
]
V
2
=[
1
0
0
]
V
3
=[
1
0
0
]
V
4
=[
1
0
0
]
V
5
=[
1
0
0
]
V
6
=[
1
0
0
]
V
0
=[
1
1
1
]
V
7
=[
0
0
0
]
T
1
T
2
V
ref
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
Mo
du
l
ation i
ndex ef
fe
ct
on i
nv
ert
er base
d
in
du
ct
io
n m
oto
r
dr
iv
e (
Akhil
es
h S
ha
r
m
a)
1791
4.1
Equi
va
le
nt
ci
r
cuit
of
in
duc
tion
m
otor
The
dynamic
beh
a
vior
of
t
he
mo
to
r
is
due
to
the
ti
me
-
de
pende
ncy
of
vo
lt
age
an
d
to
rque.
He
nce,
they
ca
n
be
e
xpress
ed
in
differentia
l
eq
uations.
B
ut,
this
will
increase
s
ys
te
m
c
omplex
it
y.
Fig
ure
6
in
dicat
e
s
dq0
eq
ui
valen
t
ci
rcu
it
dia
gram
of
an
I
M
.
T
he
dyna
mic
eq
uatio
ns
of
the
mo
t
or
are
de
vel
op
e
d
f
r
om
its
equ
i
valent
ci
rc
uit
[
9,
11,
24,
27,
29
].
4.2
Arbitr
ary
refe
rence fr
ame
This
helps
in
r
epr
ese
ntin
g
a
s
inu
s
oid
al
ly
va
r
ying
qu
a
ntit
y
i
nto
its
DC
eq
ui
valent.
If
t
his
con
ce
pt
is
us
e
d
in
I
M
,
t
he
con
tr
ol
of
the
mo
to
r
bec
ome
s
easy
,
simi
la
r
to
a
DC
m
otor.
U
nder
t
his
fr
ame
,
it
is
assume
d
that
the
an
gu
l
ar
sp
ee
d
of
the
mo
to
r
is
r
otate
d
at
the
an
gu
la
r
sp
ee
d
of
the
ref
e
re
nce
fr
ame
.
U
nder
this
ci
rcu
msta
nce,
t
he
di
ff
e
ren
ce
in
an
gula
r
s
pee
d
bec
om
e
s
zer
o
an
d
th
us
a
si
nu
s
oi
d
sig
nal
looks
li
ke
DC
s
ign
al
.
This
m
akes
ea
sy
to
dev
el
op
small
-
sig
nal
e
qu
at
io
n
out
of
the
non
-
li
near
eq
uatio
n,
des
cribin
g
t
he
op
erati
ng
po
i
nt
by
DC
va
lues
on
l
y.
The
eq
uiv
al
e
nc
e
bet
wee
n
t
w
o
a
nd
t
he
thr
ee
-
phase
mac
hine
co
uld
be
ach
ie
ved
pr
ov
i
ded
their
MMF
pro
du
ce
d
a
re
equ
al
.
If
t
her
e
are
N
p
num
be
r
of
wi
nd
i
ng
s
pe
r
ph
a
se,
for
ide
ntica
l
MMF,
the
two
windin
g
machine
s
nee
d
3Np/2
t
urns
pe
r
phase.
The
MMF
of
th
ree
-
phase
c
ou
l
d
be
conve
rted
into
two
-
ph
ase
whos
e
axes
a
re
d
a
nd
q
re
sp
ect
ivel
y.
Con
si
der
i
ng
α
and
β
as
t
he
a
xi
s
of
the
a
rb
it
ra
ry
ref
e
re
nce
f
r
ame,
the
th
ree
-
ph
a
s
e
vo
lt
age
is
e
xpr
essed
a
s:
[
]
=
2
3
[
1
1
2
⁄
−
1
2
⁄
0
√
3
2
⁄
−
√
3
2
⁄
]
[
]
(
10)
The
direct
a
nd
qu
a
drat
ur
e
axis
volt
ages
a
re
e
xpresse
d
as
,
[
]
=
[
−
]
[
]
(11)
The
i
ns
ta
nta
ne
ou
s
r
otor
a
nd
s
ta
tor
c
urren
t
e
quat
ions
are
,
[
]
=
[
−
]
[
]
(12)
[
]
=
−
2
3
[
1
2
⁄
√
3
2
⁄
1
2
⁄
√
3
2
⁄
]
[
]
(13)
Wh
e
re,
V
R
,
V
Y
a
nd
V
B
are
t
he
sta
to
r
volt
ages
,
V
d
a
nd
V
q
are
the
direct
an
d
qu
a
drat
ur
e
a
xi
s
vo
lt
age
s.
i
rd
and
i
rq
are
t
he
insta
ntane
ous
dq
r
otor
cu
r
ren
ts,
V
α
an
d
V
β
are
a
r
bitrar
y
re
fer
e
nce
fr
a
me
volt
ages
wi
th
i
α
an
d
i
β
as
their
insta
ntane
ous
cu
rre
nts,
wh
il
e
i
R
,
i
Y
and
i
B
be
the
instanta
neous
st
at
or
c
urre
nts.
The
e
quat
ion
s
(10)
to
(13)
re
pr
e
sents
the
ass
ociat
io
n
of
cu
rr
e
nts
a
nd
volt
ages
in
diff
e
re
nt
f
rame
s.
The
dyna
mic
model
of
IM
may
be
ac
hiev
ed
by
est
a
blish
ing
an
e
qu
i
valence
betwee
n
3
-
Φ
a
nd
2
-
Φ
machine
s.
T
hi
s
is
base
d
on
t
he
ma
gnit
ud
e
of
MMF
c
reat
ed
in
windin
gs
of
both
mach
ines
w
he
n
the
equ
a
l
mag
nitud
e
of
c
urren
ts
is
al
lo
wed
to
fl
ow
t
hro
ugh
t
hese
wi
nd
i
ngs.
S
o,
N
p
of
3
-
Φ
will
pro
du
ce
t
he
same
MMF,
pro
vid
e
d
two
windin
g
mac
hi
nes
hav
e
3N
p
/2
num
ber
of
tur
ns
per
phase.
T
he
direct
(
d)
a
nd
qua
dr
at
ur
e
(q)
axe
s
MMF
are
fou
nd
by
res
olv
i
ng
M
MF
of
the
t
hr
ee
-
phase
al
ong
t
hese
axes
.
The
eq
uatio
n,
thu
s
dev
el
op
e
d,
will
hav
e
s
om
e
c
ommo
n
facto
r
t
hat
c
ou
l
d
be
e
li
minate
d,
e
xc
ept
c
urren
ts
in
windin
gs.
F
r
om
Fig
ure
6,
the
dq0
sta
tor
a
nd
ro
t
or
vo
lt
age
s,
unde
r
balance
d
c
onditi
ons,
can
be
ex
pr
e
ssed
ea
sil
y
[
27,
31
].
Fr
om
t
he
e
quivale
nt
ci
rcu
it
,
the
volt
age
e
qu
at
i
on
s
are,
=
+
+
(14)
=
+
−
(
15)
=
+
+
(
−
)
(
16)
=
+
−
(
−
)
(
17)
=
+
(
18)
=
+
(
19)
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
.
11
, N
o.
4
,
D
ecembe
r
2020
:
1785
–
1798
1792
Wh
e
re,
=
+
(
+
)
(20)
=
+
(
+
)
(21)
=
+
(
+
)
(22)
=
+
(
+
)
(23)
=
(24)
=
(25)
and
are
the
zer
o
-
se
quence
fl
ux
li
nkage
s.
Wh
e
re,
and
are
ar
bitrar
y
a
ngular
a
nd
sta
to
r
angular
f
reque
ncy
res
pecti
vel
y.
The
el
ect
r
om
a
gn
et
ic
to
r
qu
e
is
ex
pr
e
ssed
as,
=
3
2
2
(
−
)
(
26)
Also
,
=
+
2
(
27)
T
l
an
d
J
re
pr
es
ent
loa
d
to
r
qu
e
an
d
m
om
e
nt
of
ine
rtia
res
pec
ti
vely.
5.
CLOSE
D
-
LO
OP
CONTRO
L
The
cl
os
ed
-
loop
op
e
rati
on
usual
ly
pro
vid
es
more
acc
urat
e
res
ults
at
the
cost
of
inc
re
asi
ng
ci
rcu
it
com
plexity
.
I
ndus
t
rial
dri
ves
empl
oy
diff
e
r
ent
ty
pes
of
mo
to
rs
by
varyi
ng
ac
sup
ply
.
The
s
peed
of
thes
e
dr
i
ves
c
hanges
,
due
to
facto
r
s
li
ke
cha
nge
in
the
s
uppl
y
vo
lt
age
or
c
ha
ng
e
in
loa
d
t
orq
ues.
U
nd
e
r
t
hese
var
ia
ti
ons,
the
dr
i
ve
sp
ee
d
vari
es.
this
le
ads
to
de
velo
pm
e
nt
of
s
pee
d
co
ntr
ol
by
ma
ny
me
thod,
no
ta
ble,
direct
tor
qu
e
of
V/
F
methods
[
30,31].
T
her
e
a
re
c
ertai
n
a
pp
li
cat
ion
s
,
wh
e
re
t
he
sp
ee
d
of
t
he
in
du
st
rial
dr
i
ves
sh
oul
d
be
co
ns
ta
nt
un
der
a
ny
ci
rc
umst
ances.
Unde
r
su
c
h
a
sit
uation,
the
e
ff
ect
of
a
ny
of
the
pa
rameters
sho
ul
d
not
reduce
t
he
s
pe
ed
of
s
uch
dri
ves.
This
is
possible
onl
y
if
the
dr
i
ve
s
ys
te
m
is
op
e
rated
in
a
cl
os
ed
-
lo
op.
The
cl
os
ed
-
lo
op
op
erati
on
will
ha
ve
no
ef
fect
of
change
in
par
a
mete
rs
on
the
industrial
dr
i
ve
s.
A
simple
dia
gram
of
a
cl
os
e
d
-
lo
op
s
ys
te
m
is
s
how
n
in
Fi
gure
7.
As
see
n
from
t
he
Fig
ur
e
,
t
hat
an
error
si
gn
al
,
E(
s),
is
gen
e
rated
by
subtract
in
g
plant
out
pu
t,
Y(
s
),
from
t
he
ref
e
ren
ce
si
gna
l,
R(s).
A
c
on
t
r
oller
of
gain
,
G
c
(s
),
mu
st
be
a
ble
to
co
ntr
ol
the
error.
If
G
p
(s
)
is
the
gai
n
of
the
pla
nt,
the
n
the
ov
e
rall
tr
ansf
e
r
functi
on
is
e
xpresse
d
in
(
30).
Fig
ure
7
.
Cl
os
e
d
-
l
oop
represe
ntati
on
of
a
s
yst
em
(
)
=
(
)
−
(
)
(
28)
(
)
=
(
)
∗
(
)
∗
(
)
(
29)
The
overall
tra
ns
fe
r
functi
on,
TF,
is
e
xpres
se
d
as
,
=
(
)
(
)
=
(
)
∗
(
)
1
+
(
)
∗
(
)
(30)
5.1
Neur
al
ne
two
rk
(
NN)
archi
tectu
re
It
is
base
d
on
t
he
interce
ptio
n
of
a
ne
uron,
a
basic
unit
of
the
nerv
ou
s
sy
st
em,
on
wh
ic
h
t
he
messa
ge
is
trans
mit
te
d
from
t
he
brai
n
to
dif
fer
e
nt
pa
rts
of
t
he
hum
an
body
a
nd
ba
ck
to
the
br
ai
n.
Each
neuron
has
a
transmitt
er
a
nd
a
receive
r
thr
ough
wh
ic
h
these
mes
sag
es
are
tra
ns
mi
tt
ed
an
d
recei
ved.
T
his
sc
he
me
of
R
(
s
)
G
p
(
s
)
G
c
(
s
)
Y
(
s
)
+
-
E
(
s
)
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
Mo
du
l
ation i
ndex ef
fe
ct
on i
nv
ert
er base
d
in
du
ct
io
n m
oto
r
dr
iv
e (
Akhil
es
h S
ha
r
m
a)
1793
transmitt
in
g
an
d
recei
ving
ca
n
al
so
be
ap
plied
to
so
l
ve
rea
l
-
li
fe
series
pro
blems
base
d
on
certai
n
set
of
ru
le
s
.
Hen
ce
,
it
can
be
sai
d
that
an
NN
has
a
se
qu
e
nce
of
set
of
r
ules
that
is
pro
fici
ent
to
i
den
ti
f
y
fun
da
mental
associat
ions
in
a
set
of
inf
or
mati
on
in
s
uch
a
man
ne
r
t
hat
is
simi
la
r
to
t
he
huma
n
brai
n.
An
NN
c
ou
ld
be
natu
ral
li
ke
m
essage
tra
ns
mi
ssion
a
nd
recei
ving
in
human
bein
gs
or
a
rtific
ia
l
to
so
lve
r
eal
-
li
fe
pro
blems
by
human
bein
gs
.
It
has
a
f
eat
ure
to
a
da
pt
it
sel
f
with
t
he
c
hange
in
in
puts
i.e
.
the
net
wor
k
c
an
pro
du
ce
the
m
ost
su
it
able
re
su
lt
without
c
hangi
ng
its
str
uctur
e
.
Neural
netw
orks
le
ar
n
it
sel
f
from
the
prel
oad
ta
sk
by
a
nalyzin
g
trai
ni
ng
sam
ples
from
t
he
ta
s
k
wh
ic
h
ha
ve
be
en
pre
-
lo
ade
d.
F
or
e
xam
ple,
an
obje
ct
rec
ogniti
on
sy
ste
m
ma
y
co
ntain
thousa
nds
of
la
bele
d
images
of
thin
gs
li
ke
glass
,
t
ables,
et
c.
It
w
ou
l
d
fin
d
gr
a
phic
al
patte
r
ns
of
t
he
ima
ges
that
are
c
onsis
te
ntl
y
correla
te
d
with
s
pecific
la
bels
.
In
ge
ner
al
,
an
NN
c
on
ta
in
s
t
hous
a
nds
or
ev
en
m
or
e
sim
ple
processin
g
node
s.
These
nodes
a
r
e
cl
os
el
y
li
nke
d.
T
hese
nets
a
re
orde
re
d
into
strat
a
of
no
de
s,
na
mely,
i
nput,
hi
dden
a
nd
ou
t
pu
t
nodes
.
T
he
net
s
co
ul
d
be
“fe
ed
-
for
ward,”
or
“
feed
-
backw
ard
”
.
In
feed
-
f
orward
arc
hitec
ture,
the
data
move
s
on
l
y
in
the
for
ward
di
recti
on.
A
sin
gle
node
may
recei
ve
and
tra
nsmi
t
data
from
se
ve
ral
nodes
in
a
la
yer
unde
rn
eat
h
an
d
ab
ove
it.
W
hile,
in
fee
d
-
ba
ckw
a
r
d,
s
ome
portio
n
of
the
outp
ut
is
f
ed
-
bac
k
to
tra
in
the
netw
ork.
In
the
ne
tw
ork,
eac
h
node
has
a
s
pecific
weig
ht,
in
dicat
ed
by
“W
i
”.
W
he
n
the
net
wor
k
is
li
ve
,
t
he
node
acce
pts
a
lt
ered
data
t
ha
t
gets
mu
lt
ipli
ed
by
its
weig
ht
“W
i
”,
sum
med
up
with
bi
as
“b”
res
ulti
ng
in
a
sing
le
numb
e
r.
If
the
num
ber
is
le
ss
than
a
t
hr
es
hold
value,
the
node
does
not
pass
a
ny
da
ta
to
the
subs
equ
e
nt
la
yer
.
B
ut,
if
t
he
num
ber
is
gr
eat
er
t
han
the
th
reshol
d
va
lue,
t
he
no
de
t
ran
s
mit
s
the
num
ber.
A
sim
plifie
d
mu
lt
i
-
la
ye
r
NN
struct
ur
e
[
32]
is
show
n
in
Fi
gure
8.
Figure
8
.
A
simpli
fied
m
ulti
-
la
yer
NN
str
uc
ture
In
it
ia
ll
y,
weig
hts
an
d
t
hr
es
holds
f
un
ct
io
n
a
re
ra
ndoml
y
c
on
si
der
e
d,
the
n,
t
he
trai
ni
ng
of
NN
sta
rts.
The
data
to
be
trai
ne
d
is
f
ed
t
hro
ugh
in
pu
t
la
ye
rs,
th
en,
it
passes
t
hro
ugh
su
c
ces
sive
la
yer
s
.
At
each
su
ccessi
ve
la
ye
r,
it
gets
m
ulti
plied
to
it
s
weig
hts.
Th
e
res
ultant
ge
ts
ad
de
d
t
og
e
ther
unti
l
ra
di
cal
ly
trans
forme
d
da
ta
is
avail
able
at
the
outp
ut
la
yer.
T
rainin
g
is
co
ntin
ued
by
co
ntinuo
us
ly
adjustin
g
t
he
weig
hts
and
th
reshold
s
un
ti
l
co
ns
ist
e
nt
res
ults
are
obta
ined
at
t
he
outpu
t.
A
blo
c
k
dia
gr
a
m
re
prese
ntati
on
of
in
ver
te
r
connecte
d
to
a
mo
to
r
loa
d
has
bee
n
sho
wn
in
Fig
ur
e
9.
In
this
case,
the
dynamic
loa
d
consi
der
e
d
is
an
I
M
.
The
m
otor
s
pee
d
is
measu
red
a
nd
it
is
co
nv
e
rted
into
its
equ
i
valent
fr
e
quenc
y.
E
ven,
t
he
ref
e
ren
ce
s
peed
in
r
pm
ha
s
bee
n
c
on
vert
ed
to
its
f
requ
ency
domain.
The
n,
the
fr
e
que
ncy
error
is
m
easu
r
ed.
T
hese
valu
es
are
sto
re
d
to
tun
e
t
he
NN
c
on
t
ro
ll
er.
I
niti
al
ly,
the
weig
hts
ha
ve
been
rand
om
ly
ass
um
e
d.
A
sigm
oid
has
bee
n
us
e
d
as
an
ac
ti
vation
f
unct
ion.
The
c
ontro
ll
er
tu
nes
up
the
er
r
or
so
that
it
is
mi
nimize
d.
T
he
ou
t
pu
t
of
the
c
on
t
ro
ll
er,
t
hu
s
ob
ta
ine
d,
is
a
f
reque
ncy
si
gn
a
l.
T
his
si
gn
al
,
al
ong
with
a
par
ti
c
ular
mod
ulati
on
ind
e
x,
is
us
e
d
to
generate
gat
e
pulse
s
for
t
he
in
ver
te
r
w
hich
c
ontr
ols
t
he
RMS
ou
t
pu
t
volt
age
app
li
ed
to
t
he
I
M
.
Figure
9.
Cl
os
e
d
loop
of
in
du
c
ti
on
m
otor
O
ut
put
,
Y
I
N
P
U
T
S
X
1
X
2
W
1
W
2
W
n
B
i
a
s
,
b
S
i
gm
oi
d
A
c
t
i
va
t
i
on F
unc
t
i
on
W
e
i
gh
t
s
,
W
i
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
.
11
, N
o.
4
,
D
ecembe
r
2020
:
1785
–
1798
1794
6.
RESEA
R
CH
METHO
D
Ba
sed
on
t
he
s
ect
ion
s,
disc
usse
d
a
bove,
Simuli
nk
m
odel
s
hav
e
bee
n
de
velo
ped
in
MATL
AB
16a.
The
im
pleme
nt
ed
m
od
el
is
de
picte
d
in
Fig
ur
e
10.
F
or
a
par
ti
cula
r
MI,
the
switc
hi
ng
pu
lse
s
f
or
an
i
nv
e
rter
us
in
g
s
pace
ve
ct
or
m
odulati
on
ha
ve
been
ge
ner
at
e
d
by
c
onsiderin
g
the
sinu
s
oi
dal
wa
ve
of
50Hz
as
refe
ren
ce
sign
al
.
A
tria
ngula
r
wa
ve,
use
d
as
a
ca
rr
ie
r
sign
al
,
has
a
carrier
fr
e
quen
cy
of
720
Hz.
The
c
omparis
on
of
these
two
sig
na
ls
is
us
ed
to
ge
ner
at
e
tri
gg
e
ring
pulse
s
f
or
t
he
po
wer
el
ect
ronic
co
ntr
olled
switc
hes,
pro
du
ci
ng
a
ste
p
al
te
rn
at
ing
volt
age.
T
hi
s
al
te
rn
at
ing
volt
age
has
bee
n
sh
a
pe
d
into
a
sinu
s
oid
al
ly
varyin
g
volt
ag
e
us
in
g
a
low
pass
filt
er
(LPF).
It
is
,
the
n,
a
ppli
ed
to
the
sta
to
r
of
a
th
ree
-
ph
a
se
in
duct
ion
mo
to
r,
m
odel
le
d
in
a
sta
ti
on
ar
y
ref
e
ren
ce
frame
w
ho
s
e
pa
ramete
rs
are
gi
ven
in
Table
1
.
This
pro
vid
es
an
op
en
-
l
oop
operat
ion
of
the
in
ver
te
r
-
ba
sed
c
ontrol
of
i
nductio
n
m
ot
or.
In
the
cl
os
e
d
-
l
oop,
f
or
a
known
MI,
meas
ured
sp
e
ed
of
IM
is
co
nverte
d
into
its
f
reque
ncy
do
main.
This
is
c
ompar
ed
with
the
eq
uiv
al
ent
freq
ue
ncy
of
t
he
re
fe
ren
ce
sp
ee
d.
T
he
e
rror
th
us
ge
ner
at
e
d
is
min
imi
zed
with
a
fee
d
-
forw
a
rd
ne
ural
ne
twork
(
NN)
with
15
la
ye
rs
,
one
i
nput,
one
ou
t
pu
t
an
d
a
sig
mo
i
d
act
ivati
on
functi
on.
T
he
NN
t
un
es
t
he
error
si
gn
al
so
that
it
beco
me
s
zero,
pro
vid
i
ng
an
ou
t
pu
t
f
r
equ
e
nc
y
eq
ual
to
the
ref
e
ren
ce
f
re
quenc
y.
T
his
frequ
e
nc
y
is
us
e
d
f
or
ge
ne
rati
ng
a
sin
usoida
l
ref
e
re
nce
si
gnal
,
w
hile
the
carrier
fr
e
qu
e
nc
y
r
em
ai
ns
unc
hange
d.
C
omparisio
n
of
t
he
two
s
ign
al
s
pro
du
ce
s
trig
ger
in
g
pu
lse
s
us
in
g
SVPW
M
te
chn
iq
ues
,
whose
width
of
t
he
pu
lse
s
de
pe
nd
up
on
the
re
fer
e
nce
fr
e
qu
e
ncy,
th
us
c
ontr
olli
ng
t
he
on
ti
me
of
the
s
witc
hes.
T
his
est
ablis
hes
a
cl
os
e
d
-
loop
operati
on.
Figure
10:
Cl
ose
d
-
l
oop
simul
ink
model
of
in
du
ct
io
n
mo
t
or
7.
RESU
LT
S
A
ND
DI
SCUS
S
ION
The
var
i
ou
s
re
su
lt
s
obta
ine
d
are
de
picte
d
in
Fig
ur
e
11
to
Figure
13
f
or
bo
t
h
op
e
n
-
l
oop
a
nd
cl
os
e
d
-
loop
co
ntr
ol.
Figure
11
i
nd
i
cat
es
the
li
ne
vo
lt
age
of
one
ph
as
e
with
it
s
corres
pondin
g
RMS
val
ue.
These
values
a
re
35
9.9
V,
38
8.8
V,
415.6
V,
427.8
V,
406.9
V,
a
nd
384.1
V
r
es
pecti
vely
for
MI
va
riat
ion
s
f
rom
0.6
to
1.5.
The
de
viati
on
in
the
RMS
volt
age
is
due
to
the
w
idth
of
the
pu
l
ses.
For
l
ow
e
r
MI,
the
wi
dth
of
t
he
pu
lse
s
for
uppe
r
powe
r
s
witc
hes
dec
reases
wh
il
e
t
he
widt
h
of
the
pulse
s
f
or
lo
wer
s
witc
h
i
ncr
eases
th
ereb
y
pro
du
ci
ng
lo
w
er
R
M
S
volt
ag
es.
Under
this
case,
it
is
te
r
m
ed
as
un
der
-
m
odulati
on.
As
t
he
m
odulati
on
ind
e
x
reaches
unit
y,
the
maxi
mum
RMS
vo
lt
age
of
427.8
V
is
ob
ta
ine
d.
At
t
his
m
odulati
on
ind
e
x,
the
wi
dth
of
pu
lse
s
for
bo
t
h
lower
a
nd
up
pe
r
switc
hes
a
re
equ
al
.
T
he
reverse
occ
urs,
w
hen
MI
is
gr
e
at
er
tha
n
unit
y.
Unde
r
this
case,
it
is
te
rmed
as
over
-
m
odulati
on.
T
her
e
fore,
the
RMS
vo
lt
age
r
edu
ce
s
to
38
4.1
V
for
1.5
MI.
T
he
per
ce
ntage
dro
p
in
volt
age
f
r
om
un
it
y
MI
to
1.5
MI
is
a
10.
22%.
F
or
dif
fer
e
nt
MI,
li
ne
vo
lt
age
s
are
ta
bu
la
te
d
in
Ta
ble
3
.
Figure
12
in
di
cat
es
an
open
-
loop
s
pee
d
c
ontr
ol
of
the
I
M
with
differe
nt
MI.
It
is
vis
ible
from
the
Figure
that
th
e
maxim
um
s
pee
d
of
IM
is
1476
r
pm
w
hich
is
achieve
d
at
a
un
it
y
MI
f
or
a
load
t
orqu
e
of
10
N
-
m.
The
s
pee
d
decr
ease
s
with
e
it
her
a
n
inc
re
ase
or
a
dec
rea
se
in
M
I
from
un
it
y.
I
n
t
he
unde
r
-
mod
ulati
on,
t
he
per
ce
ntage
s
pe
ed
va
riat
ion
is
2.3
7
%
w
hile
in
ove
r
-
m
odul
at
ion
,
t
his
va
r
ia
ti
on
is
1.0
1%
.
T
his
is
du
e
to
le
ss
Evaluation Warning : The document was created with Spire.PDF for Python.