Int
ern
at
i
onal
Journ
al of
P
ower
El
ectr
on
i
cs a
n
d
Drive
S
ystem
(I
J
PE
D
S
)
Vo
l.
11
,
No.
3
,
Septem
be
r 2020
, pp.
1132
~
1144
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v11.i
3
.
pp
1132
-
1144
1132
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
Implem
ent
ation o
f SVM f
or
five
-
l
evel
cascaded
H
-
bridg
e
multile
vel i
nvert
ers utiliz
ing
FPG
A
Maher
Ab
d
I
brah
im
A
l
-
je
wa
ri
,
Auza
ni
Jidin,
Siti
Azu
ra
Ah
m
ad
T
ar
usa
n
, M
oh
am
med
Ra
s
hee
d
Facul
ty
of Electr
ic
a
l
Eng
ineeri
ng
,
Univer
si
ti T
ekn
ika
l
Mala
ysi
a
M
el
ak
a
Hang
Tu
a
h
Jaya
,
Mal
aysia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
3
, 2
01
9
Re
vised
Dec
20
, 2
01
9
Accepte
d
Fe
b
16
, 20
20
The
Spa
ce
Vec
t
or
Modulation
S
VM
te
chni
qu
e
h
as
won
large
acc
ept
an
ce
for
AC
drive
appl
i
ca
t
ions.
How
ev
er
the
ut
il
i
za
t
i
on
of
multil
ev
el
inve
r
te
rs
conne
c
te
d
with
SV
M
by
Digit
al
signal
proc
essor
(DS
P)
rai
se
the
i
ntri
c
ac
y
of
cont
rol
al
gori
th
m
or
com
put
a
ti
onal
loa
d
,
in
cre
ase
s
of
th
e
obta
ini
ng
distort
ions
outp
ut
vo
lt
ag
e.
The
deve
lop
me
nt
of
SV
M
in
mu
lt
i
level
inve
r
te
rs
ma
y
off
er
h
igh
er
numb
ers
of
sw
it
chi
ng
vectors
for
a
cqui
r
i
ng
furthe
r
enha
nc
em
en
ts
of
outpu
t
vo
lt
ag
e
per
forma
n
ce
s
an
d
i
mpl
e
me
nt
by
using
Fiel
d
Programm
er
Gate
Array
(FP
GA
),
inve
stig
ate
lower
To
ta
l
Hormonic
Distorti
on
(THD).
Th
is
p
ape
r
r
ep
orts
th
e
p
erf
orm
anc
e
ev
al
u
at
ion
of
SV
M
for
five
-
l
eve
l
of
Casca
ded
H
-
Bridge
Mult
il
ev
el
Inv
ert
er
C
HM
I
using
MA
TL
AB/S
im
uli
nk,
whi
ch
is
sampled
at
the
mi
nim
u
m
sampling
ti
m
e,
i.e.
DT
=
5
μs
.
The
sw
it
chi
ng
signa
l
s
for
drivi
ng
ins
ula
t
ed
ga
te
b
ipolar
t
r
ansistor
(IGBTs)
which
are
stor
ed
in
MA
TL
AB
works
pac
es,
ar
e
th
en
used
to
b
e
progra
mmed
in
FP
GA
using
a
Quart
us
II
soft
ware
.
Which
c
a
n
be
st
at
ed
the
lower
THD
of
th
e
si
mul
a
ti
o
n
resul
t
is
about
14.
48%
for
f
ive
-
le
ve
l
CHM
I
and
expe
r
im
en
t
result
is
about
14.
3
1%
for
five
-
le
ve
l
CHM
I
a
t
modu
la
t
ion
inde
x
M_i=0
.
9.
The
err
or
p
ercent
ag
e
be
twee
n
the
si
mulati
on
result
s
an
d
expe
ri
me
nt
al
r
e
sults
of
the
fun
dam
en
ta
l
ou
tput
volt
ag
e
in
SV
M
is
smal
l
which
is
appr
oxi
ma
t
el
y
le
ss
th
an 1 %
.
Ke
yw
or
d
s
:
Ca
scaded H
-
bri
dg
e C
H
M
I
Five
le
vels
in
ve
rter
M
ulti
le
vel in
ve
rter
Sp
ace
vecto
r mo
du
la
ti
on
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
:
M
a
her
A
bd Ibra
him
A
l
-
je
wa
ri
,
Faculty
of
Ele
c
tric
al
Engineer
ing
,
Un
i
ver
sit
i Te
knikal
M
al
aysia
M
el
a
ka
,
Hang T
ua
h
Ja
ya
, 76100
D
ur
ia
n
T
unggal
,
M
e
la
ka,
M
al
aysia
Emai
l:
mah
e
r.
a
bd74@
ya
hoo.
c
om
1.
INTROD
U
CTION
M
a
ny
AC
dri
ve
a
pp
li
cat
ions
util
iz
e
V
oltage
S
ource
I
nvert
ers
(
VS
I
)
w
hic
h
ha
ve
ev
olv
e
d
as
the
m
os
t
popula
r
po
wer
conversi
on.
T
he
in
vo
l
veme
nt
of
V
SI
is
in
li
ne
with
the
de
velo
pm
e
nt
of
var
i
ou
s
Pu
lse
Width
M
od
ulati
on
(
P
W
M
)
al
gorith
ms
s
upporte
d
by
the
a
dv
e
nt
of
s
olid
-
sta
te
switc
hing
de
vi
ce
te
ch
no
l
ogie
s,
fast
dig
it
al
sign
al
process
or
s
,
Fi
el
d
Program
m
able
Gate
A
rrays
(
FPGA
)
in
orde
r
to
c
re
at
e
a
PWM
si
gn
al
f
or
the
real
-
ti
me
a
nd
micr
oc
on
tr
oller
s
ys
te
m
as
means
of
a
di
gital
processi
ng.
Since
a
fe
w
deca
des
a
go,
sever
a
l
PWM
al
go
rith
ms
ha
ve
bee
n
dev
el
op
e
d
t
o
i
mpro
ve
s
om
e
performa
nces
of
VSI
s
uc
h
as
high
-
powe
r
e
f
fici
ency
[1
–
5]
.
Few
st
ud
ie
s
in
dicat
e
d
that
the
y
ar
e
no
t
acc
ount
ed
f
or
P
W
M
con
t
ro
l
wit
h
high
f
reque
nc
y
an
d
div
e
rsified
te
c
hn
i
qu
e
s
of
sa
mp
li
ng,
pro
duce
gr
eat
e
r
ou
t
pu
t
volt
age
wi
th
opti
mum
f
unda
mental
value,
kee
p
total
har
m
onic
distor
ti
on
(T
HD)
lo
wer
within
the
crit
ic
al
range
with
lo
wer
rati
o
of
s
witc
hing
f
re
quency
t
o
fun
dame
ntal
frequ
e
nc
y
[
6].
Appa
ren
tl
y,
th
e
researc
h
ab
out
VSI,
th
us
f
ar,
ha
s
not
rea
che
d
t
o
the
st
at
e
of
sat
ur
at
io
n,
as
novel
or
sim
plif
ie
d
P
W
M
methods
are
sti
ll
emer
ging
for
va
rio
us
t
opolog
y
inv
e
rter
ci
rcu
it
s
an
d
mu
lt
il
evel
in
ve
rters
[
7].
Th
rough
va
rio
us
ty
pes
of
m
odula
ti
on
strat
egies
or
P
W
M
meth
od
s
,
a
Sp
ace
Vecto
r
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow
Ele
c
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Impleme
nta
ti
on
of SVM fo
r fi
ve
-
le
vel
ca
sc
ad
ed H
-
br
id
ge m
ulti
le
vel
…
(
M
aher A
bd
Ib
r
ahim Al
-
je
w
ar
i
)
1133
M
od
ulati
on
(
S
VM)
te
c
hniq
u
e
has
gaine
d
wide
acce
pta
nc
e
beca
us
e
of
seve
ral
a
dv
a
nt
ages
s
uc
h
as
higher
ou
t
pu
t
volt
age
s,
lo
wer
T
HD,
high
-
ef
fici
enc
y
a
nd
fle
xib
il
it
y
to
be
imple
mented
in
vec
tor
c
ontr
ol
sys
te
ms
[8]
–
[
13].
Be
sid
es
that,
this
m
odulati
on
sche
m
e
al
so offers
in
opti
mizi
ng
t
he
use
d of
dc
volt
age
li
nk
util
iz
at
ion
wh
ic
h
mean
s
i
t
can
incre
ase
t
he
rati
on
of
outpu
t
mag
nitu
de
volt
age.
I
n
ge
ner
al
,
the
im
plementat
io
n
of
SVM
involve
with
t
he
us
e
d
of
D
S
P
bo
a
r
d
a
nd
re
qu
i
red
sect
or
i
den
ti
ficat
io
n
w
hich
br
in
gs
int
o
t
he
f
ormat
io
ns
of
ro
ta
ti
ng
s
pace
vecto
r
diag
ram
.
I
n
t
he
th
ree
phases
sy
ste
m
,
t
her
e
are
six
f
ra
ct
ion
s
i
n
the
s
pace
vecto
r
dia
gr
a
m
sp
in
ning
36
0°
wh
ic
h
eac
h
ha
s
e
qu
al
ly
div
id
ed
by
60°.
T
hi
s
s
pace
vecto
r
diag
ram
is
a
tr
ansfo
rmati
on
f
rom
a
balance
of
th
re
e
ph
ase
quantit
ie
s
into
two
phas
e
sy
ste
m
of
α
-
β
ref
e
rence
f
rame.
T
he
S
V
M
main
operat
ion
is
to
use
the
ne
ar
est
three
vect
or
rec
ogniti
on
of
the
re
fer
e
nce
vo
lt
age
a
nd
de
te
rmin
e
d
the
c
orres
pondin
g
on
-
ti
me
us
in
g
t
he prin
c
iples o
f vo
lt
se
cond
e
quivale
nt
[
14
-
17]
.
A
m
ulti
le
vel
inv
e
rter
offer
s
gr
eat
er
num
be
r
of
volt
age
ve
ct
or
s
as
c
ompare
d
to
ei
gh
t
vecto
rs
f
or
a
two
-
le
vel
in
ve
rter.
Fi
gure
1
il
lustrate
s
the
sp
ace
ve
ct
or
di
agr
ams
f
or
Se
ct
or
I
,
in
a
tw
o
-
le
vel
inv
e
rte
r
an
d
three
-
le
vel
in
ve
rter.
The
sp
ac
e
vecto
r
diag
ra
ms
can
be
use
d
to
co
m
pa
re
t
he
im
plementa
ti
on
of
SVM
i
n
tw
o
-
le
vel
an
d
t
hr
e
e
-
le
vel
in
ve
rters.
As
co
mp
a
r
ed
to
tw
o
-
le
ve
l
sp
ace
vect
or
diag
ram
,
th
e
sect
or
i
n
t
hr
ee
-
le
vel
inv
e
rter
is
di
vid
ed
int
o
f
our
identic
al
small
er
tria
ng
le
s
(i.
e.
∆_
0,
∆
_1,
∆_2
an
d
∆_
3).
To
re
duce
T
HD
(
or
dv⁄d
t)
an
d
s
witc
hing
los
ses
in
mu
lt
il
evel
in
ve
rter
(i
.e.
th
ree
-
le
vel),
it
is
ne
cessar
y
to
swit
chin
g
vecto
rs
wh
ic
h
are
the
near
est
to
the
re
fer
e
nc
e
vecto
r
v
̅
_s^
*.
He
nce
,
thr
ee
-
le
vel
S
VM
switc
h
the
vec
tors
v
̅
_2,
v
̅
_7
an
d
v
̅
_14fo
r
a
giv
e
n
re
fer
e
nce
vec
tor.
T
he
cal
c
ulati
on
of
on
-
dur
at
ion
in
mu
lt
il
evel
SVM
is
quit
e
com
plica
te
d
and
diff
e
re
nt
f
or
va
rio
us
tria
ng
le
s
due
to
small
tria
ng
le
s
in
t
he
s
pace
vecto
r
dia
gr
a
m
of
t
hr
ee
-
le
vel
in
ve
rter
in
Fig.
1
do
no
t
e
xa
ct
ly
imi
ta
te
th
e
ge
om
et
r
y
of
a
sect
or
of
tw
o
-
le
vel
in
ve
r
te
r.
I
n
tw
o
-
le
vel
S
VM,
t
he
cal
c
ulati
on
of
on
-
durati
on
is
strai
gh
t
for
ward
wh
ic
h
is
valid
f
or
eve
r
y
sect
or.
H
owever,
the
t
hr
ee
-
le
vel
S
VM
ne
eds
to
modif
y
the
re
f
eren
ce
vect
or
with
new
ori
gi
n
po
i
nt
to
ap
pl
y
the
t
w
o
-
le
ve
l
base
d
S
V
M
f
or
cal
c
ulati
ng
on
-
durati
on.
As
s
how
n
by
Fig
ure
1
(b),
t
he
modifie
d
re
fer
e
nc
e
vecto
r
v
̅
_s
^
*
with
vector
v
̅
_2
as
or
i
gin
po
i
nt
is
determi
ned
suc
h
that
t
he
cal
culat
ion
of
on
-
durati
on
is
li
ke
that
of
tw
o
-
le
vel
based
S
VM.
T
he
cal
c
ulati
on
becomes
c
omp
li
cat
ed
if
the
r
efere
nce
volt
age
vect
o
r
v
̅
_s^
*
li
es
in
tria
ng
le
∆
_2,
w
he
re
the
ori
entat
ion
of
tria
ng
le
is
dif
f
eren
t
a
mon
g
ot
her
s;
as
can
be
seen
t
he
tria
ngle
s
∆_
0,
∆_
1
and
∆_
3
ha
ve
the
same
ori
en
ta
ti
on
with
a
si
ngle
tr
ia
ng
le
or
Sect
or
I
in
tw
o
-
le
vel
S
V
M
,
as
s
hown
in
Fig.
1.
T
he
co
mp
le
xity
inc
reases
as
numb
e
r
of
lev
el
s o
f
i
nverter b
ec
om
e
s
higher
, e.g.
in
f
ive
-
le
vel in
ve
rter,
the
re ar
e si
x
tria
ngle
s am
ong
sixtee
n
tri
ang
le
s
,
that ha
ve dif
fere
nt orientat
io
n
Figure
1
.
Com
par
is
on
bet
wee
n
(a)
t
wo
-
le
vel
sp
ace
vecto
r di
agr
a
m a
nd (b) t
hr
ee
-
le
vel s
p
ac
e
vecto
r diagr
am
, e.g. f
or
sect
or i
The
im
pleme
nt
at
ion
of
S
V
M
for
mu
lt
il
evel
inv
e
rters
re
quire
some
im
portant
par
ts
wh
i
ch
ar
e
a
s
fo
ll
ows;
(
1)
de
te
ct
ion
of
sect
or
,
(
9)
detect
i
on
of
tria
ngle
∆
,
(10
)
cal
culat
ion
of
on
-
du
rati
on
f
or
switc
hi
ng
the
ne
a
rest
vec
tors,
an
d
(
11)
determi
nation
the
s
witc
hing
seq
uen
ce
f
or
e
very
s
witc
hing
pe
rio
d.
As
f
ound
i
n
li
te
ratur
e,
t
here
are
t
wo
co
mmon
meth
od
s
to
cal
c
ulate
the
on
-
durati
on
s
.
T
he
first
meth
od
is
to
detect
the
tria
ng
le
an
d
s
olv
e
th
ree
s
imult
ane
ou
s
e
quat
ions
of
the
t
rian
gle
to
de
te
rmin
e
the
on
-
ti
mes
as
s
uggest
ed
i
n
[18].
The
sec
ond
met
hod
is
t
o
detect
t
he
tri
ang
le
an
d
us
e
on
-
durati
on
e
quat
ions
st
or
e
d
in
a
lo
okup
ta
ble
for
this
tria
ngle
,
as
pro
pose
d
in
[
19
,
20]
.
B
oth
m
et
hods
howe
ve
r
require
co
mpl
ex
c
ompu
ta
ti
on
s
a
s
the
numbe
r
of
le
vel
increase
s
.
Alte
r
nativel
y,
the
cal
culat
i
on
of
on
-
dur
a
ti
on
ca
n
be
obta
ined
us
i
ng
gen
e
ral
al
gorit
hm
s
.
Sp
eci
fical
ly,
use
s
a
E
uclidea
n
vect
or
s
ys
te
m
with
seve
ra
l
matri
x
tra
nsf
ormat
ion,
pro
vi
ded
that
it
does
no
t
pro
vid
e
a
s
ys
te
mati
c
appr
oac
h
f
or
real
ti
me
SVM
imple
m
entat
ion
.
O
n
th
e
oth
e
r
wa
y,
c
al
culat
ed
on
-
durati
on
and
obta
ined
s
witc
hing
sta
te
s
by
m
eans
of
c
oor
din
at
e
sy
ste
m,
w
her
e
t
he
a
xes
are
60
de
grees
apa
rt.
H
oweve
r,
the
60
degrees
trans
formati
on
le
ads
to
the
com
plexity
sin
ce
the
volt
a
ge
ref
e
ren
ce
is
c
om
m
only
def
i
ned
in
the
or
t
hogona
l
coord
i
nate
syst
em.
Re
centl
y,
a
sim
ple
S
VM
al
gorithm
for
mu
lt
il
eve
l
inv
erter
s
bas
ed
o
n
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.
3
,
Se
ptembe
r
2020
:
1132
–
1144
1134
sta
nd
a
rd
tw
o
-
l
evel
S
VM.
T
he
two
-
le
vel
ba
sed
S
V
M
c
on
c
ept
is
init
ia
te
d
by
[
21],
howe
ver,
the
cal
c
ul
at
ion
of
on
-
durati
on
is
base
d
on
or
i
gin
m
odific
at
ion
a
nd
60
de
gr
ees
c
oor
din
a
te
transformat
ion,
w
hich
can
no
t
be
exten
ded
in
i
mp
le
me
ntin
g
SVM
f
or
hi
gher
le
vels,
i.e
.
L
great
er
tha
n
three.
U
nlike
the
f
ormer
m
et
hods
,
the
impleme
ntati
on
of
S
VM
that
include
s
the
detect
io
n
of
sect
or
an
d
tria
ng
le
,
a
nd
cal
cu
la
ti
on
of
on
-
durati
on
wer
e
d
e
rive
d g
eometrical
ly a
nd syst
emat
ic
al
ly which
su
it
a
ble for
any le
ve
l of in
ver
te
r
[22
-
24].
2.
RESEA
R
CH MET
HO
D
Ba
sed
on
the
inv
est
igati
on
i
n
li
te
ratu
re
re
view
a
bout
ty
pe
of
P
W
M
,
t
he
S
V
M
is
pr
efera
ble
as
it
offer
s
se
ve
ral
adv
a
ntages
a
nd
recentl
y
know
n
as
the
mo
st
po
pu
la
r
te
chn
iq
ue
for
man
y
el
ect
rical
dr
iv
e
app
li
cat
io
ns
.
T
his
c
ha
pter
dis
cusses
the
devel
op
me
nt
of
spa
c
e
ve
ct
or
m
odulati
on
(SV
M)
al
gorithm
ba
sed
on
five
-
le
vel
casc
aded
H
-
bri
dge
mu
lt
il
evel
inv
erter
s
as
pro
po
s
ed
by
[
18
,
25].
It
is
ne
cessar
y
to
de
s
cribe
the
pr
i
nciple
or
f
or
m
ulati
on
of
the
SVM
al
gorithm
with
t
he
ai
d
of
s
uitable
diag
rams
a
nd
e
quat
ions
f
or
e
ve
r
y
le
vel
of
in
vert
er
is
gi
ven
t
o
de
velo
p
a
pro
per
m
odul
at
or
f
or
e
valu
at
ing
it
s
perf
ormances
T
H
D.
T
he
performa
nces
evaluati
on
will
be
pr
e
sente
d
in
res
ults
a
nd
di
scussion.
T
he
n,
the
sim
ulati
on
model
of
S
VM
for
every
le
vel
of
inv
e
rter
is
de
ve
lop
e
d
us
i
ng
M
A
TLA
B
-
Si
mu
li
nk.
T
he
de
velo
pm
e
nt
of
the
simulat
io
n
m
od
el
us
es
same
pa
r
amet
ers
value
s
,
e.
g.
sam
plin
g
ti
me,
th
ree
-
ph
a
se
loa
d,
D
C
volt
age,
et
c.,
as
im
pleme
nted
in
the
ha
rdwa
re
s
ys
te
m.
The
ve
r
ific
at
ion
a
nd
e
valuati
on
ar
e
a
lso
car
ried
out
via
ex
pe
rimen
ta
l.
In
t
he
h
a
r
dware
sy
ste
m,
t
he
S
VM
al
gorith
m
is
execu
te
d
usi
ng
a
Fiel
d
P
r
ogram
mable
G
at
e
Ar
r
ays
(FP
GA)
DEO
C
ontr
oller.
The
implem
en
ta
ti
on
of
t
he
hard
war
e
s
ys
te
m
e
mphasiz
e
s
on
s
om
e
i
mporta
nt
as
pe
ct
s
f
or
pro
per
S
V
M
op
e
rati
ons,
pro
vid
in
g
a
blan
ki
ng
ti
me
ge
ne
r
at
or
t
o
a
vo
id
s
hort
ci
rc
uit
co
nd
it
io
ns
,
en
sur
ing
t
he
DC
vo
lt
age
su
ppl
y
for
ea
c
h
H
-
B
rid
ge
i
nv
erter
ci
rcu
it
ha
s
pr
ov
i
des
is
ol
at
ion
a
nd
a
c
onsta
nt
DC
volt
age
a
nd
a
pplyi
ng
gate
dr
i
ver
ci
rc
uits wh
ic
h
ha
ve
suf
fici
ent p
owe
r
a
mp
li
ficat
ions
a
nd
is
olati
on
to
sw
it
ch
O
N/OFF
the
IG
BTs
o
f
V
SI
.
On
ce
t
he
simu
la
ti
on
an
d
ex
pe
rimental
res
ul
ts
are
ob
ta
i
ned,
al
l
the
reco
r
de
d
data
is
ta
bu
la
te
d
in
ta
bles,
and
gr
a
phs
for
c
ompa
rison
t
o
performa
nce
evaluati
on.
T
hen,
base
d
on
s
uitable
e
quat
ions
a
nd
t
heories,
the
anal
ys
is
of
performa
nces
is
ca
rr
ie
d
out
to
ve
rif
y
the
im
pro
veme
nts/ad
van
ta
ges
of
us
i
ng
dif
fer
e
nt
ty
pe
of
le
vels
of
in
ve
rters
wh
ic
h
is
five
-
le
vel
casca
ded
H
-
Bri
dg
e
mu
lt
il
evel
in
ve
rters.
The
de
ve
lop
me
nt
of
ha
r
dw
a
re
is
us
e
d
to
obta
in
the
e
xperim
ental
res
ults
w
hich
will
be
c
ompare
d
with
th
e
sim
ulati
on
resu
lt
s,
validat
ion
of
SVM alg
ori
th
m as
well
as th
e ad
van
ta
ges o
f
S
V
M
in
m
ulti
le
vel inverte
rs
.
2.1.
Spa
ce
vec
t
or m
od
ul
at
i
on
of
f
ive
-
le
vel c
as
c
ad
ed
H
-
brid
ge
multil
evel in
vert
er
This
sect
i
on
br
ie
fly
desc
ribes
the
pri
nciple
of
S
V
M
for
fi
ve
-
le
vel
CH
MI
base
d
on
tw
o
-
le
vel
SVM.
Since
t
he
SVM
co
ntr
ol
al
gorithm
is
base
d
on
tw
o
-
le
vel
S
VM,
it
can
be
pro
ven
that
m
os
t
par
ts
util
ize
sam
e
equ
at
io
ns
a
nd
appro
ac
hes
f
or
im
pleme
nting
t
he
switc
hi
ng
m
odulati
ons.
Fi
gure
2
dep
ic
ts
a
fi
ve
-
le
vel
Ca
scaded
H
-
B
rid
ge
Mult
il
evel
I
nv
e
rter
(C
HMI)
w
hich
c
on
sist
s
of
six
isolat
ed
DC
volt
ages.
T
he
CHM
I
sh
ow
n
i
n
Fi
gure
2
is
ref
e
rr
e
d
to
five
-
le
vel
i
nv
e
rter
is
due
t
o
the
fact
t
hat
the
in
ver
te
r
ca
n
pro
du
ce
fi
ve
le
vels
of
ou
t
put
vo
lt
a
ges,
wh
ic
h
are
2
5
,
5
,
0
,
−
5
and
−
2
5
.
The
five
ou
t
pu
t
vo
l
ta
ges
can
be
obta
ined
by
pro
vid
in
g
poss
ible
switc
hing
sta
te
com
bin
at
ion
s
on
the
simpli
fied
ci
rc
uit
of
the
fi
ve
-
le
vel
CH
MI
(i.e.
f
or
a
sing
le
-
phase) gi
ven
in Figu
re 2
. H
ence
a
si
ngle
-
ph
ase o
utput
volt
age
of
th
e
CHMI
(i.e
.
-
ph
a
se
w
her
e
=
-
,
-
, or
-
ph
ase
)
c
an be e
xpresse
d
as:
=
(
1
−
2
)
.
5
+
(
3
−
4
)
.
5
(1)
The
tria
ngle
f
or
five
-
le
vel
CH
M
I
as
s
how
n
in
Fig
ur
e
4
is
de
fine
d
us
i
ng
(2)
a
nd
(
3).
It
ca
n
be
s
how
n
that ea
ch
secto
r
will
hav
e
16 t
rian
gles (
i.e
.
∆
0
,
∆
1
,
∆
2
,…,
∆
15
).
∆
=
1
2
+
2
2
(2)
∆
=
1
2
+
2
2
+
1
(3)
The
discussi
on
ab
ov
e
sho
ws
the
di
ff
e
ren
t
pa
rts
util
iz
ed
in
five
-
le
vel
C
H
M
I
as
c
ompar
ed
to
that
of
three
-
le
vel,
w
her
e
the
five
-
le
vel
CH
MI
in
vo
l
ves
gr
eat
e
r
num
ber
s
of
t
rian
gles,
le
vel
outp
ut
vo
lt
ag
es
an
d
vo
lt
age
vecto
rs
.
T
he
rest
pa
rts
will
util
iz
e
same
a
ppr
oach
es
an
d
eq
uatio
ns
.
F
or
exa
mp
le
,
the
on
durati
on
s
f
or
switc
hing
vect
or
s
in five
-
le
ve
l C
HMI
a
re cal
culat
ed usin
g (
4), (5
)
a
nd (6).
=
3
2
5
(
∗
−
1
√
3
∗
)
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow
Ele
c
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Impleme
nta
ti
on
of SVM fo
r fi
ve
-
le
vel
ca
sc
ad
ed H
-
br
id
ge m
ulti
le
vel
…
(
M
aher A
bd
Ib
r
ahim Al
-
je
w
ar
i
)
1135
Figure
2
.
The
c
ircuit
of
five
-
le
vel
CH
MI
Figure
3
.
Sim
pl
ifie
d
fiv
e
-
le
ve
l
H
-
br
i
dg
e
in
ve
rter
(for an
y Phase
)
Figure
4
.
De
fin
it
ion
of
tria
ngle
(
∆
=
∆
0
,
∆
1
,
∆
2
,…,
∆
15
)
f
or
e
very s
ect
or
2.2.
Simul
at
i
on
m
od
el
of sp
ace
vector m
odul
ati
on
This
sect
ion
presents
a
sim
ul
at
ion
model
of
Sp
ace
Vecto
r
Modula
ti
on
(S
V
M)
us
i
ng
M
A
TLAB
-
Simuli
nk.
Fi
gure
5
de
picts
th
e
simulat
ion
m
od
el
,
sp
eci
fical
ly
the
co
ntr
ol
al
gorithm
of
S
VM
w
hich
ge
ne
rates
switc
hing
sta
tu
s.
The
sim
ulati
on
model
sho
wn
i
n
this
fig
ure
is
us
e
d
to
ge
ner
at
e
s
witc
hin
g
sta
te
s
f
or
dri
vin
g
IG
BTs
of
fi
ve
-
le
vel
CH
MI
by
modif
ying
th
e
num
be
r
of
i
nputs
a
nd
DC
volt
age
te
rms
in
s
om
e
cal
culat
ion
s
.
The
simulat
io
n
is
performe
d
us
in
g
t
wo
set
s
of
sa
mp
li
ng
ti
mes,
s
uc
h
a
s
1
=
200
for
detect
ing
s
ect
or
s,
tria
ng
le
s
a
nd
com
pu
ti
ng
on
-
durati
on,
a
nd
2
=
5
for
ge
ne
rati
ng
a
ppr
opriat
e
pulse
width
with
highli
near
it
y
a
nd
accu
rac
y.
Gen
e
rall
y,
the
simulat
ion
m
odel
is
co
ns
tr
uc
te
d
us
i
ng
tw
o
t
yp
e
s
of
pro
gr
a
mmin
g
appr
oach
es
,
na
mely
the
gra
phic
al
pro
gr
a
m
ming
ap
proac
h
us
in
g
Sim
ul
ink
blo
c
ks
a
nd
the
c
-
pro
gr
a
mmin
g
appr
oach
wr
it
te
n
i
n
MATL
A
B
f
un
ct
io
n
bloc
ks
.
T
her
e
a
re
two
in
pu
ts
r
eq
uire
d
i
n
t
he
simulat
ion
mode
l
w
hic
h
are
the
dema
nds
of
ma
gnit
ude
an
d
f
re
qu
e
ncy.
T
hese
tw
o
in
pu
ts
w
il
l
pro
du
ce
a
ref
e
ren
ce
of
t
hr
ee
-
phase
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.
3
,
Se
ptembe
r
2020
:
1132
–
1144
1136
vo
lt
age
.
The
Si
mu
li
nk
blo
c
ks
inside
t
he
Subs
ys
te
m
1
for
ge
ne
rati
ng
the
re
fe
ren
ce
of
three
-
ph
a
se
a
re
pr
e
se
nted
in Figu
re
6.
It
can
be
see
n
t
ha
t t
he
re
fer
e
nce
is co
ns
tr
ucted usin
g (7).
∗
=
.
(
2
)
(5)
∗
=
.
(
2
−
2
3
)
(6)
∗
=
.
(
2
+
2
3
)
(7)
wh
e
re
is
the
mag
nitud
e
of
r
efere
nce
vo
lt
a
ge.
B
y
c
onside
rin
g
the
li
mit
at
ion
of
the
poss
ible
outp
ut
vo
lt
age
pr
oduc
ed
in
the
SVM
for
five
-
le
vel
CHMI
, t
he ma
gn
it
ude
can
be
calc
ulate
d
as
foll
ows
:
=
Μ
4
5
√
3
(8)
Figure
5
.
Sim
ul
at
ion
m
odel
of
sp
ace
v
ect
or
modu
la
ti
on
(SVM)
(e.
g.
f
or fi
ve
-
le
vel
C
H
MI)
wh
e
re
Μ
can
be
var
ie
d
bet
we
en
0
to
1
f
or
a
dj
us
ti
ng
t
he
ma
gn
it
ude
of
i
nput
an
d
hen
ce
ou
t
pu
t
vo
lt
age
s.
T
his
means
,
Μ
is
set
to
1
for
pro
duci
ng
t
he
maxi
m
um
ou
t
pu
t
volt
age.
T
he
th
ree
-
phase
vo
lt
age
is
then
tran
sf
orm
ed
i
nto
re
fer
e
nce
volt
age
c
ompone
nts
∗
a
nd
∗
,
as
t
he
i
nput
of
the
modu
la
to
r.
T
he
trans
formati
on
is o
btained
u
si
ng (9) an
d (
10)
.
∗
=
2
3
⌊
∗
−
∗
2
−
∗
2
⌋
(9)
∗
=
1
√
3
⌊
∗
−
∗
⌋
(10)
Using
these
e
qu
at
io
ns,
the
t
ran
s
f
or
mati
on
is
con
st
ru
ct
e
d
us
in
g
Sim
ulink
blo
c
ks
as
pr
ese
nted
i
n
Figure
7. The
s
e Simulin
k bl
oc
ks
a
re
g
r
oupe
d
as
a
Subs
yst
e
m2
, as sh
own
in
Fi
gure
5.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow
Ele
c
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Impleme
nta
ti
on
of SVM fo
r fi
ve
-
le
vel
ca
sc
ad
ed H
-
br
id
ge m
ulti
le
vel
…
(
M
aher A
bd
Ib
r
ahim Al
-
je
w
ar
i
)
1137
Figure
6
.
The
s
imulat
ion m
odel
b
loc
k of t
hre
e
-
phase
gen
e
rato
r
(at
Subsy
ste
m
1
)
Figure
7
.
The
s
imulat
ion m
odel
b
loc
k of t
hre
e
-
to
two
-
phase t
ransformati
on
(at
Subs
yst
e
m2)
It
ca
n
be
noti
c
ed
from
t
he
c
omplet
e
sim
ulat
ion
m
od
el
s
hown
in
Fi
gure
5
that
the
re
fer
e
nce
volt
age
com
pone
nts
∗
a
nd
∗
are
sam
pled
at
1
=
200
us
in
g
Zer
o
-
Or
der
H
old
blo
c
ks
.
T
hese
two
in
pu
ts
are
us
e
d
t
o
pro
duce
re
fer
e
nce
volt
age
vect
or
into
a
pola
r
f
orm
us
in
g
(
11)
and
(
12)
.
The
n,
the
ma
gn
it
ude
a
n
d
ang
le
of
re
fer
e
nce
volt
age
ve
ct
or
,
i.e.
∗
a
nd
are
bein
g
fe
d
to
MA
TLAB
Fu
nction1
bl
oc
k
f
or
detect
in
g
sect
or
s
,
tria
ngle
s
Δ
,
an
d
cal
cul
at
ing
∗
an
d
∗
,
a
s
discusse
d
in
previo
us
sect
io
ns.
T
he
s
ource
c
od
e
or
c
-
pro
gr
am
min
g wr
it
te
n
i
n
the
MATLAB
F
un
c
ti
on
1
bl
ock as
sh
ow
n
in
Fi
gur
e 8
.
∗
=
√
̅
∗
2
+
̅
∗
2
(11)
=
−
1
(
̅
∗
2
̅
∗
2
)
(12)
Figure
8
.
S
our
ce
co
de
li
sti
ng
for
M
A
TL
AB
functi
on1
f
or de
te
ct
ing
secto
r
s,
tria
ngle
s,
and
cal
culat
ing
∗
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
&
D
ri
S
ys
t,
V
ol
.
11
, N
o.
3
,
Se
ptembe
r
2020
:
1132
–
1144
1138
Figure
9
.
S
our
ce
co
de
li
sti
ng
for
M
A
TL
AB
functi
on2 f
or c
al
culat
ing
on
-
durati
on
The
re
fe
ren
ce
vo
lt
age
co
mpo
nen
ts
∗
an
d
∗
produce
d
f
rom
MATLAB
Fu
nc
ti
on
1
bl
ock
a
re
the
n
us
e
d
as
the
in
pu
ts
of
MA
TL
AB
Fu
nctio
n2
blo
c
k.
T
his
blo
c
k
is
res
pons
ible
to
cal
c
ulate
on
-
durati
on
fo
r
switc
hing
vector
s
within
a
t
riang
le
for
fi
ve
-
l
evel
CH
M
I
.
T
he
sou
rce
co
de
for
cal
culat
in
g
on
-
durati
on
sh
ow
n
in
Fig
ure
9.
T
he
on
durati
ons
an
d
pro
duced
from
the
MAT
LAB
F
unct
ion2
blo
c
k
a
re
us
e
d
to
cal
culat
e
du
t
y
rati
os
o
f
pre
-
s
witc
hing
st
at
es
for
eac
h
phase,
,
an
d
.
T
he
cal
c
ulati
on
i
s
pe
rformed
by
Subsy
ste
m
3
blo
c
k,
as
ca
n
be
no
ti
ce
d
in
Fi
gure
5.
Fi
gure
10
sho
ws
Sim
ulink
bloc
ks
c
on
ta
ine
d
i
n
Su
bs
yst
em
3
blo
c
k,
wh
ic
h
are c
on
s
t
ru
ct
e
d usi
ng (13),
(1
4) an
d (15
)
f
or
five
-
le
vel CH
M
I
.
=
−
2
(
4
)
(13
)
=
−
2
(
4
+
2
)
(14)
=
−
2
(
4
+
2
+
2
)
(15)
Figure
10
.
T
he
simulat
io
n
m
odel
b
l
ock of
du
ty r
at
ios
calc
ul
at
or
(at
Su
bs
yst
em3)
Figure
11
s
ho
ws
a
sim
ulati
on
model
of
pr
e
-
switc
hi
ng
sta
t
es
ge
ne
rato
r
w
hich
is
dev
el
oped
based
on
the
c
omparis
on
betwee
n
dut
y
rati
os
an
d
tri
angular
wav
e
f
orm.
The
sim
ul
at
ion
model
or
Si
mu
li
nk
bl
oc
ks
are
gro
up
e
d
a
s
a
Subs
yst
e
m4
,
as
de
picte
d
in
Fi
gure
5.
T
he
t
riangular
w
ave
f
orm
is
ge
ner
at
ed
at
sam
plin
g
ti
me
2
=
5
by the
Re
pea
ti
ng S
e
quence
bl
ock,
wh
il
e th
e
du
t
y rati
os
a
re
updated
ev
e
r
y
1
=
200
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow
Ele
c
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Impleme
nta
ti
on
of SVM fo
r fi
ve
-
le
vel
ca
sc
ad
ed H
-
br
id
ge m
ulti
le
vel
…
(
M
aher A
bd
Ib
r
ahim Al
-
je
w
ar
i
)
1139
Figure
11
.
T
he
simulat
io
n
m
odel
b
l
o
ck
of
pre
-
switc
hi
ng sta
te
s g
e
ner
at
or
(
at
Subs
yst
e
m4
)
At
the
la
st
part
of
Fi
gure
5,
i
t
can
be
obse
r
ved
that
al
l
th
e
inf
ormat
io
n
su
c
h
as
sect
or
(
),
tria
ngle
(
∆
)
an
d
pre
-
swit
chin
g
sta
te
s
(
,
an
d
)
for
sel
ect
ing
ap
pro
pr
i
at
e
switc
hi
ng
s
ta
te
s
to
dri
ve
I
GBTs
o
f
five
-
le
vel
CH
M
I
.
T
hat
pe
rfo
rme
d
by
MA
T
LAB
F
un
ct
io
n3
bl
ock,
as
s
how
n
in
Fig
ur
e
5,
the
sou
rce
cod
e
a
s
sh
ow
n
in
Fi
gur
e 12
is
giv
e
n first sect
or
(
)
at
f
irst t
rian
gle (
∆
).
Figure
12
.
Sou
rce
c
od
e
li
sti
ng for
MATL
AB
FUNCTI
ON3 f
or
pro
duci
ng s
witc
hing stat
es
Finall
y,
t
he
s
witc
hing
sta
te
s
are
us
e
d
to
dri
ve
IG
BTs
of
the
in
ver
te
r.
Figure
2
dep
ic
ts
simulat
ion
models
f
or
fi
ve
-
le
vel
CH
MI.
N
ote
that,
t
he
switc
hi
ng
of
IG
BT
s
is
determine
d
by
th
e
switc
hi
ng
sta
te
s,
pro
vid
e
d
that
the
s
witc
hing
of
uppe
r
a
nd
lo
wer
I
GBTs
f
or
eac
h
le
g
m
ust
be
co
mp
li
m
ented
to
eac
h
oth
e
r.
The
ou
t
pu
t
of
each
in
ver
te
r
is
connecte
d
to
an
i
den
ti
cal
three
-
phase
a
nd
series
c
onnected
resist
iv
e
and
inducti
ve
l
oad
s
.
2.3.
The
descri
p
tio
n of
ex
peri
me
nt
al
set
u
p
In
this
sect
io
n
desc
ribes
t
he
assignme
nts
of
the
ci
rcu
i
ts
or
c
ompon
ents
em
ployed
to
set
up
the
e
xperime
nt
al
platf
orm
f
or
ve
rif
ying
the
eff
ect
ive
ness
of
SVM
te
ch
nique
f
or
five
-
le
ve
l
casca
de
d
H
-
br
i
dg
e
mu
lt
il
evel
in
ve
rters.
T
he
str
uctu
re
of
e
xpe
rimental
platf
orm
in
dicat
ed
t
he
c
omp
on
e
nt
s
use
d,
a
s
s
ho
wn
in
Figure
13.
Th
e
Fiel
d
Pro
gr
a
mmable
Gate
Array
s
(FPG
A
)
Co
ntr
oller
w
as
util
iz
ed
to
perform
t
he
ta
sk
s
of
SVM tec
hn
i
que.
2.4.
FPGA
co
nt
r
ol
le
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.
3
,
Se
ptembe
r
2020
:
1132
–
1144
1140
The
Fiel
d
Pro
gramma
ble
Gat
e
A
rr
a
ys
(
FP
G
A)
Co
ntr
oller
i
s
kn
own
to
ha
ve
high
-
sp
ee
d
cl
ock
w
hich
is
superi
or
to
execu
te
lo
gical
or
dig
it
al
ope
rati
on.
It
feat
ured
a
powe
rful
Alte
ra
C
yclo
ne
I
V
E
P42
2F1
7C6N
FPGA
with
22
,320
l
og
ic
el
e
ments,
32
M
B
of
SD
R
AM,
2
kb
EE
PRO
M
,
a
nd
a
64
M
b
serial
co
nfi
gurati
on
memor
y
de
vic
e.
The
FP
GA
bo
a
r
d
40
-
pi
n
Head
e
rs
(GPI
Os)
pro
vid
e
d
72
I/O
pi
ns
,
5V
po
wer
pins,
two
3.3V
powe
r
pi
ns
a
nd
f
our
gro
und
pin
s
,
15
3
ma
xi
mu
m
F
PGA
I
/O
pins
an
d
56
em
be
dd
e
d
m
ulti
pliers.
T
he
DE
0
-
Nano
boar
d
in
cl
ud
e
d
a
bu
il
t
-
i
n
US
B
Bl
ast
er
f
or
F
PGA
pro
gr
a
mmi
ng,
a
nd
the
boa
rd
co
ul
d
be
po
wer
e
d
ei
ther
from
this
U
SB
port
or
a
n
exte
rn
al
po
wer
s
ource.
I
nputs
an
d
outp
uts
inclu
de
d
2
pu
s
hbutto
ns
,
8
us
e
r
LED
s
a
nd
a
set
of
4
dip
s
witc
hes.
T
he
boar
d
incl
ud
e
d
exp
a
ns
i
on
hea
der
s
t
o
at
ta
ch
va
rio
us
te
rasic
da
ughter
ca
r
ds
or
o
t
her
dev
ic
es
.
Fi
gure
13
s
hows
the
FPGA
receives
inf
ormat
io
n
of
Q
uar
t
us
II
pro
gr
a
m
a
nd
pre
-
s
witc
hing
sta
te
s
,
an
d
.
T
he
FP
GA
is
a
cco
unta
ble
to
perf
orm
the
assi
gnm
ent
of
sel
ect
in
g
s
uitable
swit
chin
g
sta
te
s
ba
sed
on the
date
rec
ei
ved
.
T
his
ta
s
k
is si
mil
ar
to
t
he
MA
TLAB
F
un
ct
io
n3
blo
c
k as
dep
ic
te
d i
n Fi
gure
5.
Figure
13
.
T
he
ex
per
ime
ntal s
et
up
2.5.
Qua
r
tu
s
II
pr
og
r
am
f
or
FP
GA
c
ontr
oller
board
The
Q
uar
t
us
II
abs
orbe
d
t
he
data
ob
ta
ine
d
f
rom
the
s
witc
hi
ng
sig
nals
f
or
dri
ving
I
GBT
s
sto
red
in
M
A
TLAB
wor
ks
pa
ces.
T
hen,
the
data
we
re
save
d
in
E
xce
l
file
.
Ne
xt,
t
he
data
we
re
t
ra
ns
fe
rr
e
d
to
mif
file
.
The
fir
st
par
t
in
pro
gr
a
mmi
ng
c
reated
a
project
name
(
N
ew
Proje
ct
),
f
a
mil
y
a
nd
de
vice
set
ti
ngs
for
FPGA.
The num
ber
of wor
d
(
4001
) were
sel
ect
ed wh
ic
h
c
ou
l
d
tr
ansf
e
r
the
d
at
a
from 1
2
s
witc
hi
ng
stat
es
of E
xcel file
to
12
s
witc
hi
ng
sta
te
s
of
mif
.f
il
e.
T
his
file
consi
ste
d
of
0
-
4000,
t
hen,
dis
playe
d
the
mif
file
.
at
Qu
a
rtus
II
.
The
M
e
g
a
Wiz
ard
Plu
g
-
I
n
Ma
nag
e
r
w
ould
i
ns
ert
(sa
1…s
a
4,
sb1…
sb4
an
d
sc
1…sc
4)
un
der
(
mif
file
).
VHD
L
file
was
create
d
a
nd
sa
ved
a
s
‘s
vm5level
’.vhd
f
or
co
mpu
ta
ti
on
al
pur
pos
es.
I
n
ad
diti
on,
ot
her
V
H
DL
Fil
es
save
d
as
‘clk
_div
’.vhd,
‘
blan
king_
c
ompa
rator’.
vhd,
‘
blan
ki
ng
_lower
_c
ounter’.v
hd,
‘
blank
i
ng_mo
d18’.
vhd,
‘b
la
nk
i
ng_upp
er_
c
ounter
’.v
hd
a
nd
‘
blan
ki
ngte
me_
main
’.v
hd,
e
dit
co
ding
wer
e
create
d
as
s
how
n
i
n
Figure
13.
T
he
n,
the
project
w
as
a
na
lysed
afte
r
ve
rify
i
ng
the
s
uc
cess
of
the
ana
lyses.
The
assi
gn
i
ng
pin
on
F
PGA
sel
ec
te
d
pi
n
pl
ann
e
r
for
2
i
np
uts
a
nd
24
out
pu
t
pins
12
lo
wer
switc
hi
ng
ou
t
pu
ts
an
d
up
per
switc
hi
ng
ou
t
pu
ts
.
Hen
ce
,
a
f
ull c
ompil
at
ion
des
ign
wa
s
s
ucces
sfu
l
to
pr
ogra
mme
t
he
FP
G
A
ha
rdwar
e
i
n
order
to
see
th
e
dis
play
of outp
ut
wav
e
form o
n oscil
lo
sco
pe,
as
s
how
n
in
Fig
ure
14.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow
Ele
c
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Impleme
nta
ti
on
of SVM fo
r fi
ve
-
le
vel
ca
sc
ad
ed H
-
br
id
ge m
ulti
le
vel
…
(
M
aher A
bd
Ib
r
ahim Al
-
je
w
ar
i
)
1141
Figure
14
.
Flo
wch
a
rt
of
a
ppla
id
FP
G
A
in
quart
us
ii
pr
ogra
m
2.6.
Po
wer
in
vert
e
r a
n
d
gate
driver ci
rcui
ts
Figure
15
t
he
FPGA
for
ci
rc
uit
casca
de
d
H
-
bri
dge
m
ulti
le
vel
in
ver
te
r
C
HMI
a
nd
gate
dr
i
ver
e
.g.
f
or
five
-
le
vel.
The
FP
G
A
c
ontr
oller,
it
a
ppli
ed
the
date
f
rom
Qu
a
rtus
II
P
rogr
a
m
t
o
a
bsor
b
t
he
gate
dri
ve
r
a
n
d
powe
r
in
ver
te
r
ci
rcu
it
s
th
ey
a
re
us
ed
to
pro
du
ce
d
e
sired
outp
ut
volt
ages
.
The
ai
m
of
use
gate
dr
i
ver
c
ircuit
s
was
t
o
pro
vide
isolat
ion
between
the
el
ect
ronic
c
on
t
ro
l
ci
rcu
it
s
a
nd
powe
r
i
nv
e
rter
ci
rcu
it
s
a
nd
pr
ov
i
de
enou
gh
powe
r
ex
pa
ns
io
n
f
or
s
witc
hing
I
G
BTs.
He
nce,
the
same
s
witc
hing
sta
te
s
we
re
pro
duced
f
r
om
the
ou
t
pu
t
of
gate
dr
i
ver
ci
rc
uits,
yet,
with
diff
e
r
ent
gro
unding
po
i
nts.
T
he
i
de
ntica
l
switc
hing
sta
te
s
wer
e
use
d
to
dr
i
ve
I
GBTs
i
n
CH
M
I
with
e
nough
volt
age
and
cu
rr
e
nt.
Figure
15
s
howe
d
the
six
unit
s
of
H
-
Brid
ge
i
nverter
ci
rcu
it
s
to
plan
t
five
-
le
vel
C
HMI
as
s
how
n
in
Fi
gure
2.
E
very
H
-
B
rid
ge
inv
e
rter
ci
rcu
i
t
was
s
upplied
by
a
n
isolat
ed DC v
ol
ta
ge
sup
ply
.
Figure
15
.
T
he
FPGA
for
ci
rc
uit
CH
M
I
and
gate dri
ve
r
e
.g.
for
five
-
le
vel
3.
RESU
LT
S
AND DI
SCUS
S
ION
In
this
pa
rt
e
va
luate
s
the
tot
al
ha
rm
on
ic
di
stortio
ns
(T
H
D)
of
outp
ut
volt
age
a
nd
t
he
f
unda
mental
ou
t
pu
t
volt
age
f
or
fi
ve
-
le
ve
l
inv
e
rter
of
S
pace
Vect
or
M
od
ulati
on
(SVM)
te
ch
niqu
e.
T
he
eval
uat
ion
is
base
d
on the si
mu
la
ti
on r
es
ults i.e. the
value
s o
f TH
D
a
nd
f
undame
ntal o
ut
pu
t v
oltage a
r
e o
btaine
d usin
g
Fast
Four
ie
r
Tra
ns
f
orms
(
FFT
)
an
al
ys
is
an
d
validat
ed
with
the
exp
e
rime
ntal
resu
lt
s.
T
he
re
su
lt
s
are
ob
ta
i
ned
a
t
Start
Generat
o
r F
P
GA
O
n
Switchi
n
g
Sta
te
Wo
rk
spa
ce
Exc
el fil
e
Start
Qua
rtus II
N
ew
pro
jec
t
Ap
p
li
ed
wi
th
N
am
e
M
e
m
o
r
y
In
itial
iz
ati
o
n
Fil
e
m
if.
fil
e
M
ega
Wiz
ard
P
lu
g
-
In
M
an
a
g
er
p
ro
ces
s
es
u
n
til fin
i
sh
VH
D
L
Fil
e
C
o
m
p
u
tat
i
o
n
o
f
VHDL
Fi
les
Fu
ll
c
o
m
p
il
atio
n
D
e
sig
n
P
ro
g
ra
m
m
er
H
ard
ware
S
e
tup
B
y
-
US
B
Outp
u
t
Wa
v
ef
o
r
m
D
isp
laye
d
A
ssig
n
m
ents
P
in
P
lan
n
er
Y
es
Un
successful
No
Y
es
Un
successful
No
P
ro
g
ra
m
m
er
is
succ
essful
End
Evaluation Warning : The document was created with Spire.PDF for Python.