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
,
No.
3
,
Septem
be
r 2020
, pp.
1570
~
1578
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v11.i
3
.
pp
1570
-
1578
1570
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore.c
om
Op
timi
zation
of
SH
E
PW
M casc
aded m
ulti
level i
nver
t
er
switchi
ng patte
rns
Ayong
Hiendr
o,
I
smail
Yu
s
u
f,
Ju
n
aidi,
F. T
ri
as
P
on
ti
a Wi
gyaria
nto,
Yohannes
M.
Siman
ju
n
tak
Depa
rtment
o
f
E
le
c
tri
c
al E
ngin
eering,
Ta
n
jungpur
a
Univer
si
ty, Ind
onesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ja
n
2
7
, 2
0
20
Re
vised
A
pr
3
,
20
20
Accepte
d
Apr
24
, 20
20
Sele
c
ti
ve
har
mo
nic
elimi
n
at
ion
(SH
E)
is
an
eff
ic
i
ent
method
t
o
elimi
n
ate
low
-
orde
r
sel
ec
t
ed
har
mon
ic
s.
H
oweve
r,
du
e
to
n
onli
ne
ari
ty
in
th
e
problems
,
ma
ny
opt
im
i
zat
ion
techniqu
es
give
unsat
isfie
d
per
form
anc
es
in
findi
ng
opti
mum
sw
it
ch
ing
ang
le
s
for
t
he
SH
E.
Thi
s
pape
r
propos
es
a
mod
ifi
ed
mot
h
-
fl
am
e
opt
im
izati
on
al
gori
t
hm
to
el
i
mi
n
a
te
s
el
e
ct
iv
e
h
ar
moni
cs
in
ca
sca
d
ed
multil
eve
l
inv
erters.
The
optimizat
io
n
a
lgori
th
m
is
em
ploy
ed
to
find
sets
of
op
timum
sw
itching
angl
es
for
ca
sc
a
ded
5
-
l
evel,
7
-
l
e
vel
,
and
9
-
le
ve
l
inv
erters.
The
resul
ts
have
show
n
t
hat
modi
fi
ed
mot
h
-
fl
am
e
opti
mization
is
bene
fi
ci
a
l
in
f
in
ding
opti
mu
m
s
witc
hing
angl
es
.
It
per
for
ms
bet
t
er
tha
n
mot
h
-
fla
m
e
optimizat
ion
(MF
O)
and
diffe
ren
ti
a
l
evo
l
uti
on
(DE)
al
gorit
h
ms.
The
opti
mum
sw
it
ching
angles
ar
e
ap
pli
ed
to
g
ene
r
at
e
sw
it
chi
ng
pulses
for
a
c
asc
ade
d
9
-
le
v
el
inve
r
t
er
to
d
em
onstra
te
th
e
al
gori
thm
’s
ac
cur
ac
y
.
As
a
r
esult
,
the
low
-
or
der
har
moni
cs
ar
e
entire
ly
re
mov
ed
from
the
ac
ou
tput voltag
e
of the
inverter
.
Ke
yw
or
d
s
:
Ca
scaded H
-
bri
dg
e
Harmo
nic eli
m
inati
on
M
ulti
le
vel in
ve
rter
Op
ti
miza
ti
on t
echn
i
qu
e
Sw
it
chin
g
a
ngle
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
:
Ayo
ng H
ie
ndr
o,
Dep
a
rtme
nt of
Ele
ct
rical
En
gi
neer
i
ng,
Tanju
ngpura
Unive
rsity,
Jl. Jende
ral
A
hma
d Ya
ni,
P
on
ti
anak
Kali
ma
nt
an
Ba
rat,
I
ndonesi
a.
Emai
l:
ay
on
g.h
ie
ndro@ee.
un
t
an.
ac.i
d
1.
INTROD
U
CTION
M
ulti
le
vel
in
ve
rters
have
at
tr
act
ed
the
at
te
nt
ion
of
ma
ny
re
searche
rs
due
t
o
the
nee
d
for
high
power
app
li
cat
io
ns
i
n
in
du
st
ries
[1].
I
n
high
po
wer
ap
plica
ti
on
s
,
t
he
s
witc
hi
ng
f
reque
ncy
of
in
ver
te
r
s
is
li
mit
ed
by
switc
hing
los
s
es
an
d
el
ect
romag
netic
inter
fe
re
nces.
As
t
he
switc
hing
f
reque
ncy
a
nd
switc
hing
los
s
es
are
decr
ease
d, the
eff
ic
ie
nc
y of i
nverter
s is inc
re
ased s
i
gn
i
fican
tl
y
[
2]
.
The
ad
va
ntage
s
of
the
m
ulti
le
vel
i
nv
e
rters
are
a
)
the
y
ca
n
pro
duce
al
m
os
t
si
nuso
i
dal
wav
e
f
or
m
ou
t
pu
t
vo
lt
age
with v
e
r
y
lo
w
t
otal har
monic
distor
ti
on
(THD)
, b
)
the
y
can
o
pe
rate at
low switc
hing fre
quenc
y
[3],
an
d
c
)
t
he
y
a
re
a
ppli
cable
f
or
hi
gh
pow
er
a
nd
high
volt
age
syst
ems.
Ther
e
are
c
ommo
nly
three
ty
pes
of
mu
lt
il
evel
inv
e
rter
to
po
l
og
ie
s
:
flyin
g
ca
pacit
or
[4],
diode
-
c
la
mp
[
5,
6],
an
d
casca
de
d
m
ul
ti
le
vel
inv
erter
[
7]
.
Among
t
he
to
po
l
og
ie
s
,
the
casca
de
d
m
ulti
le
vel
inv
e
rter
bec
om
es
m
ore
at
tract
ive
f
or
me
dium
a
nd
hig
h
vo
lt
age
re
ne
w
able
ene
r
gy
s
ys
te
ms
due
to
it
s
modu
la
rity
a
nd
simpli
c
it
y
[
8]
.
A
noth
er
a
dv
a
ntage
of
t
he
casca
de
d
m
ulti
l
evel
inv
e
rter
is
that
it
requires
le
ss
nu
mb
e
r
of
c
omp
on
e
nts
tha
n
di
od
e
-
cl
am
p
a
nd
flying
capaci
tor
in
vert
ers.
On
e
of
the
cr
uc
ia
l
issues
in
mu
lt
il
evel
in
ve
rters
is
t
o
el
im
inate
ha
rm
on
ic
s
of
the
ac
ou
t
pu
t
volt
age.
Sele
ct
ive
har
m
on
ic
el
imi
natio
n
pulse
widt
h
mo
du
la
ti
on
(SHEP
W
M
)
[
9
-
12]
is
t
he
most
popula
r
te
ch
niq
ue
t
o
impro
ve
the
outp
ut
volt
age
wav
e
f
or
m
of
mu
lt
il
evel
inve
rters.
The
SHEPW
M
te
c
hn
i
qu
e
c
an
be
util
iz
ed
to
reduce
bo
t
h
th
e
switc
hing
fr
e
qu
e
nc
y
a
nd
T
HD
val
ue
of
t
he
m
ulti
le
vel
inv
e
rters
.
T
herefo
re,
it
w
oul
d
giv
e
a
sign
ific
a
nt a
dvantage
f
or
high
-
pow
er
mu
lt
il
evel in
ver
te
rs
to
operate
with l
ow sw
it
c
hing freq
ue
ncies [
13
].
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Op
ti
miz
atio
n o
f SHEPWM
c
asc
aded
mult
il
evel
inverte
r sw
it
ching
patt
ern
s
(
Ay
ong Hie
ndro
)
1571
Howe
ver,
acc
urat
e
est
imat
in
g
of
switc
hi
ng
pa
tt
ern
s
play
s
a
cru
ci
al
ro
le
in
SH
EP
W
M
in
ve
rters.
T
he
main
pro
blem
associat
ed
wit
h
t
he
SHEP
W
M
te
c
hniq
ue
is
the
s
olu
ti
on
of
nonlinea
r
eq
uations
unde
r
var
i
ou
s
op
e
rati
ng c
ondi
ti
on
s
of
t
he
i
nverter
s [1
4,
15]
.
Ther
e
are
tw
o
main
meth
ods
for
s
olv
i
ng
a
s
et
of
S
HEPW
M
t
ran
sce
nden
ta
l
equ
at
io
ns
:
determi
nisti
c
[16]
a
nd
stoc
hastic
[
17,
18
]
a
ppr
oach
es
.
O
ptimi
zat
io
n
te
ch
niques,
especial
ly
st
oc
hastic
meta
he
ur
ist
ic
al
gorithms
[19
],
hav
e
bec
ome
the
excit
in
g
too
ls
to
deal
with
co
m
plex
op
ti
miza
ti
on
pro
blems
i
n
S
H
EPW
M
inv
e
rters.
H
oweve
r,
due
to
nonlinea
rity
and
mu
lt
iple
var
ia
bles
exist
ed
in
the
iss
ue
s,
ma
ny
op
ti
miza
ti
on
a
lgorit
hm
s
gi
ve
unsati
sfied
perf
ormances
fo
r
their
premat
ure
or
slo
w
c
onverge
nce.
In
this
pa
pe
r,
a
m
od
i
fied
M
F
O
is
em
ploye
d
to
fi
nd
set
s
of
op
ti
m
um
s
witc
hing
a
ngle
s
f
or
S
HEPW
M
casca
de
d
5
-
le
vel,
7
-
le
vel,
a
nd
9
-
le
vel
i
nverters.
The
op
ti
miza
ti
on
pro
blem
for
th
e
casca
de
d
m
ulti
le
vel
inv
e
rters
is
co
mputed
by
us
i
ng
M
at
la
b.
Th
e
pulse
s
ge
nerat
ed
f
r
om
opti
mu
m
s
witc
hing
patte
rns
a
nd
powe
r
ci
rcu
it
s
a
re
bu
il
t
by
us
in
g
C
aden
ce
-
Ps
pice.
Fin
al
ly,
res
ults
are
obta
ined
an
d
ver
ifie
d
in
simulat
io
ns
for
a
casca
de
d 9
-
le
ve
l i
nv
e
rte
r.
2.
CASC
A
DED MULTIL
EVE
L IN
VERTE
R
A
ca
scade
d
m
ulti
le
vel
in
ver
t
er
c
onsist
s
of
N
sin
gle
-
ph
ase
H
-
bri
dge
in
ve
rters
with
se
par
at
e
N
dc
so
urces
.
T
he
si
ng
le
-
phase
H
-
br
i
dg
e
in
ver
te
r
s
are
co
nnect
ed
in
se
ries,
as
s
how
n
in
Fi
gur
e
1.
It
al
so
sho
ws
t
hat
the casca
ded m
u
lt
il
evel inv
e
rter
has
e
qu
al
dc
sour
ce
s,
1
=
2
=
⋯
=
(1)
Each
sin
gle
-
phase
H
-
br
id
ge
inv
e
rter
re
qu
i
re
s
one
dc
s
ourc
e,
or
in
ot
her
words,
t
he
ca
s
caded
(
2
N
+
1)
-
le
vel
in
ver
te
r
r
equ
i
res
N
dc
s
ources
[
20,
21]
.
Howe
ver
,
the
casca
de
d
m
ulti
le
vel
inv
e
rter
can
al
s
o
be
supp
li
ed
by
a
sin
gle
DC
sou
rce
[22,
23].
The
m
ulti
le
vel
inv
e
rter
co
ul
d
produce
me
diu
m
-
t
o
high
-
vo
lt
age
ou
t
pu
t
from
low
-
volt
age
in
pu
t.
T
he
ac
ou
tpu
t
volt
age
is
sy
nt
hesized
by
N
dc
v
oltage
so
urces
c
onnec
te
d
to
t
he
i
nd
i
vidual
H
-
br
i
dg
e
in
vert
ers.
Figure
1.
The
powe
r
ci
rc
uit o
f
casca
ded (
2
N
+1)
-
le
vel i
nv
e
r
te
r
V
d
c
1
S
1
3
S
1
1
S
1
2
S
1
4
V
d
c
2
S
2
3
S
2
1
S
2
2
S
2
4
V
d
c
3
S
3
3
S
3
1
S
3
2
S
3
4
V
d
c
N
S
N
3
S
N
1
S
N
2
S
N
4
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
:
15
70
–
15
78
1572
The
s
um
mati
on
of
vo
lt
ages
ge
ner
at
e
d
by
ea
ch
H
-
br
id
ge
in
ver
te
r
pro
duce
s
a
sta
ircase
ou
tpu
t
volt
age
wav
e
f
or
m
(as
seen
i
n
Fig
ure
2).
Eac
h
sin
gle
-
phase
H
-
bri
dg
e
inv
e
rter
ge
ne
rates
+
V
dc
,
0,
a
nd
–
V
dc
outp
ut,
an
d
the r
es
ulti
ng
outp
ut volt
age
of a casca
de
d (2
N
+1)
-
le
vel in
ve
rter
ranges
fro
m
–
NV
dc
t
o
NV
dc
.
Figure
2
prese
nts
the
outp
ut
vo
lt
age
wa
veform
of
the
cas
caded
(
2
N
+1
)
-
l
evel
in
ve
rter,
wh
ic
h
can
be
expresse
d
i
n
F
ourier
series as
foll
ow
s:
)
s
i
n
(
)
(
,...
5
,
3
,
1
t
n
V
t
v
N
n
n
=
=
(2)
Figure
2
.
The
ou
t
pu
t
volt
age
wav
e
f
or
m
of a
casca
de
d
(
2
N
+
1)
-
le
vel in
ve
rter
The ma
gnit
ude
of
har
m
onic
c
ompone
nts (
i
nc
lud
in
g
t
he fu
ndame
ntal)
for
e
qu
al
dc so
ur
ce
s i
s g
i
ven by
=
=
)
c
o
s
(
4
1
k
N
k
n
dc
n
n
V
n
V
V
(3)
The
main
goal
of
the
SH
E
PWM
is
t
o
el
imi
na
te
(
N
-
1)
ha
rm
on
ic
s
from
the
wav
e
f
or
m
.
The
re
mainin
g
equ
at
io
n
is
us
e
d
as
the
ma
gn
i
tud
e
of
t
he
fun
dame
ntal
com
pone
nt
at
the
de
sired
value
,
V
1
.
Ba
sed
on
(
3),
the
simult
ane
ou
s
e
qu
at
io
ns f
or d
e
te
rmin
in
g
N
s
witc
hing a
ng
le
s of the
S
HEPW
M
is
giv
e
n b
y
1
1
1
1
)
c
o
s
(
4
)
(
=
−
=
=
N
k
M
f
k
3
1
3
3
1
)
3
co
s
(
3
4
)
(
=
=
=
=
N
k
M
V
V
f
k
5
1
5
5
1
)
5
co
s
(
5
4
)
(
=
=
=
=
N
k
M
V
V
f
k
)
1
2
(
1
)
1
2
(
)
1
2
(
1
)
)
1
2
co
s
(
(
5
4
)
(
−
−
−
=
=
−
=
=
N
k
N
N
N
k
N
M
V
V
f
(4)
wh
e
re
f
3
, f
5
,
…
,f
(2N
-
1)
is t
he
nor
mali
zed m
a
gnit
ud
e
of
har
m
on
ic
s (
res
pect to
t
he fu
ndame
ntal), a
nd
M
=
V
1
/V
dc
is t
he
m
odulati
on in
dex. T
he ob
je
ct
ive
fun
ct
ion
of the
S
HE
PWM
pro
bl
em
is d
e
fine
d
as
=
|
1
|
+
|
3
|
+
|
5
|
+
⋯
+
|
(
2
−
1
)
|
(5)
The o
ptimum
s
witc
hing a
ng
le
s ar
e
obta
ined
by minimiz
i
ng
(5)
a
nd m
us
t sa
ti
sfy
the
foll
ow
ing
c
onstrai
nt:
0
<
1
<
2
<
⋯
<
(
2
−
1
)
<
90
(6)
The
first
of
(
4)
guara
ntees
the
desire
d
f
unda
mental
co
m
ponen
t,
an
d
t
he
ot
her
s
a
re
util
iz
ed
to
en
sure
the eli
minati
on
of
3
rd
, 5
th
, 7
th
, 9
th
, …, a
nd (2
N
-
1)
th
har
m
on
ic
s.
V
d
c
0
2
V
d
c
3
V
d
c
N
V
d
c
-
N
V
d
c
-
3
V
d
c
-
2
V
d
c
-
V
d
c
θ
1
θ
2
θ
3
θ
N
9
0
o
1
8
0
o
2
7
0
o
3
6
0
o
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
P
ow Elec
& Dri S
ys
t
IS
S
N: 20
88
-
8
694
Op
ti
miz
atio
n o
f SHEPWM
c
asc
aded
mult
il
evel
inverte
r sw
it
ching
patt
ern
s
(
Ay
ong Hie
ndro
)
1573
3.
MO
DIFIE
D MOTH
-
FLA
ME OPTI
MI
ZATION
A
L
GORIT
HM
The
mo
t
h
flam
e
opti
miza
ti
on
(
M
F
O)
al
gorithm
us
es
m
oth
s
as
searc
h
a
ge
nts
that
move
arou
nd
t
he
search
s
pace
[
24].
Flame
s
ar
e
con
si
der
e
d
a
s
the
best
pos
it
ion
of
mo
t
hs
in
the
searc
h
sp
ace.
Eac
h
mo
th
searche
s aro
und
a
flame a
nd
updates t
he discov
e
r
y
as it
s
be
st
so
luti
on.
In
s
olv
i
ng
the
SH
E
pro
blems
,
the
op
ti
m
um
switc
hing
a
ng
le
s
are
the
bes
t
po
sit
io
n
of
mo
th
s
in
t
he
M
F
O
al
gorith
m.
T
he
init
ia
l par
a
mete
rs of
t
he MFO are th
e mo
th
popula
t
ion
N
, t
he
num
ber
of
va
riable
s
D
, t
he
lowe
r
boun
d
l
b
,
the
uppe
r
boun
d
u
b
,
a
nd
t
he
m
axim
um
it
erati
on
M
ax
.
T
he
lo
wer
a
nd
uppe
r
bounds
of
var
ia
bles are
define
d
as
f
ollo
ws
=
[
1
2
…
],
=
[
1
2
…
]
(7)
The mot
h pop
ul
at
ion
is ra
nd
oml
y gen
e
rated
by
,
(
=
1
)
=
+
.
(
−
)
,
=
1
,
2
,
…
,
;
=
1
,
2
,
…
,
(8)
The mot
h pop
ul
at
ion
of t
he M
FO
ca
n be
pres
ented
a
s
,
(
)
=
[
1
,
1
2
,
1
3
,
1
1
,
2
2
,
2
3
,
2
⋮
,
1
⋮
,
2
…
…
…
1
,
2
,
3
,
⋱
⋯
⋮
,
]
(9)
The fit
ness val
ue of
the
mo
t
hs i
s car
ried o
ut
by u
si
ng
(
=
1
)
=
(
,
(
=
1
)
)
,
=
1
,
2
,
…
,
;
=
1
,
2
,
…
,
(10)
The fit
ness val
ue
ca
n be
pr
ese
nted
a
s
(
)
=
[
1
2
3
⋮
]
(11)
Durin
g
th
e
it
erati
on
proce
ss,
the
fla
me
fitn
ess
(at
I
=
1)
is
so
rte
d
of
in
it
ia
l
fitness
va
lues
of
t
he
mo
th
s,
w
he
rea
s
the
flam
es
ar
e
s
or
te
d
acc
or
ding
to
their
fitness
val
ues.
T
he
flame
fit
nes
s
is
sorte
d
fro
m
t
he
best to
the
w
orst values
,
(
)
=
(
(
)
)
=
[
1
2
3
⋮
]
(
)
(12)
,
(
)
=
[
1
,
1
2
,
1
3
,
1
1
,
2
2
,
2
3
,
2
⋮
,
1
⋮
,
2
…
…
…
1
,
2
,
3
,
⋱
⋯
⋮
,
]
(
)
(13)
Updati
ng the
mo
th
po
pu
la
ti
on is
us
in
g
a
lo
ga
rithmic s
piral
model as
fo
ll
ows
,
(
+
1
)
=
,
(
)
.
(
)
.
(
2
)
+
,
(
)
(14)
wh
e
re
S
is t
he dist
ance
betwe
en
the
m
oth
s
a
nd the
flames
,
b
is a
sh
a
pe
c
onsta
nt
of the
spi
ral p
at
h (
-
1 ≤
b
≤ 1),
t
is a co
ntr
ol
pa
rameter
to ma
intai
n
the
d
ist
a
nce
betwee
n m
oth
a
nd
flame i
n
the
s
piral pa
th
(
r
≤
t
≤
1),
and
r
is
a co
nv
e
r
gen
ce
const
ant
wh
ic
h i
s d
ec
r
easi
ng
f
rom
-
1
t
o
-
2 t
o est
imat
e the v
a
lues
of
t
.
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
:
15
70
–
15
78
1574
The dist
ance
bet
ween
t
he mot
hs
a
nd the
f
la
m
es is de
fine
d
as
,
(
)
=
(
,
(
)
−
,
(
)
)
,
=
1
,
2
,
…
,
;
=
1
,
2
,
…
,
(15)
wh
e
reas
r
a
nd t
are
as
foll
ows
=
−
(
1
+
)
,
=
.
(
−
1
)
+
1
(16)
The p
os
it
ion
of
upd
at
in
g mot
hs co
uld eve
ntua
ll
y
de
gr
a
de
a
f
te
r
pas
sin
g
it
er
at
ion
s.
In
order t
o
e
nsure
that t
he mot
hs
will
meet t
he fla
mes, t
he
m
oths sh
ould
updat
e their
posit
ions wit
h res
pect t
o
the
b
e
st flam
es at
the f
i
nal step
of th
e
it
erati
on. A c
on
t
ro
l
par
a
mete
r necessa
r
y for this
mec
ha
nism is
=
(
−
(
−
1
)
.
)
(17)
Accor
ding to
(17),
the
mo
t
h p
opulati
on is
updated b
y usin
g
,
(
+
1
)
=
,
(
)
.
(
)
.
(
2
)
+
,
(
)
,
≤
,
(
+
1
)
=
,
(
)
.
(
)
.
(
2
)
+
,
(
)
,
>
(18)
wh
il
e
updatin
g t
he fla
mes:
(
+
1
)
=
[
(
+
1
)
(
)
]
,
=
1
,
2
,
…
,
(19)
The fla
me c
orr
esp
onding t
o
it
s f
it
ness
v
al
ue i
s
,
(
+
1
)
=
[
1
,
1
2
,
1
3
,
1
1
,
2
2
,
2
3
,
2
⋮
,
1
⋮
,
2
…
…
…
1
,
2
,
3
,
⋱
⋯
⋮
,
]
(
+
1
)
(20)
Finall
y,
t
he bes
t posi
ti
on
of th
e mo
t
hs
a
nd th
ei
r
fitness
v
al
ue
s ar
e
sel
ect
ed i
n
he
re as
(
+
1
)
=
1
(
+
1
)
(
+
1
)
=
1
,
(
+
1
)
,
,
=
1
,
2
,
…
,
(
21)
Updati
ng the
mo
th
s a
nd the
flames
process
es are
re
peated
unti
l FFb
e
st(I
+1) meet
s th
e
crit
erion ɛ as
def
i
ned in
(5)
a
nd
/
or I
=
M
ax
.
4.
RESU
LT
S
A
ND AN
ALYSIS
4.1.
Sw
itchi
ng pat
terns
A
chie
ving
t
he
fun
dame
ntal
c
ompone
nt
at
a
desire
d
of
M
a
nd
s
uppr
e
ssin
g
sel
ect
e
d
harm
on
ic
s
to
zer
o
as
descr
i
bing
in
(
4)
a
re
ve
ry
diff
ic
ult
to
be
so
lve
d
nume
ri
cal
ly.
I
n
this
work,
a
m
od
i
fied
MFO
al
gori
thm
is
app
li
ed
to
al
le
viate
the
c
omp
utati
on
al
pro
blems.
T
he
opti
mu
m
s
witc
hing
a
ng
le
s
f
or
ca
scade
d
5
-
le
vel,
7
-
le
vel,
and
9
-
le
vel
in
ve
rte
r
a
re
ex
plor
ed
by
us
in
g
the
m
od
i
fied
MF
O
for
al
l
val
ue
s
of
M
.
As
the
resu
lt
s,
the
opt
imum
switc
hing
a
ngle
s
f
or
the
casc
aded
5
-
le
vel,
7
-
le
vel,
a
nd
9
-
l
evel
in
ver
te
r
a
re
pr
ese
nted
in
Fig
ur
e
3,
4,
a
nd
5
,
resp
ect
ivel
y.
All
opti
mum
s
witc
hing
a
ngle
s
m
us
t
meet
t
he
re
qu
ir
eme
nts
of
(
6),
a
s
me
ntion
e
d
earli
er
.
T
hu
s
,
there
are
no
possible
op
ti
m
um
s
witc
hing
a
ng
le
s
within
M
<
1.1
027
f
or
a
casca
ded
5
-
le
vel
in
ve
rter.
U
sing
the
same
c
onstrai
nt
w
hich
m
us
t
s
at
isfy
(
6)
,
the
r
e
are
al
so
no
opti
mu
m
s
witc
hi
ng
an
gles
wit
hin
M
<
2.0
9
74
f
or
a
casca
de
d
7
-
le
ve
l
inv
erte
r
an
d
within
M
<
3.0
930
for
a
casca
ded
9
-
le
vel
in
ve
rter.
It
is
note
d
that
no
t
al
l
M
ha
s
it
s
so
luti
ons
f
or
the
opti
m
um
switc
hing
a
ng
l
es
since
the
opti
mu
m
s
witc
hin
g
a
ngle
s
will
on
l
y
be
a
ppli
cable
if
they f
ulfill
the
requireme
nt as
d
e
fine
d by (
6).
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
Op
ti
miz
atio
n o
f SHEPWM
c
asc
aded
mult
il
evel
inverte
r sw
it
ching
patt
ern
s
(
Ay
ong Hie
ndro
)
1575
Figure
3
.
O
ptimum
s
witc
hing
patte
rn
s
for c
ascade
d 5
-
le
ve
l i
nv
e
rter
Figure
4
.
O
ptimum
s
witc
hing
patte
rn
s
for c
ascade
d 7
-
le
ve
l i
nv
e
rter
Figure
5
.
O
ptimum
s
witc
hing
patte
rn
s
for c
ascade
d 9
-
le
ve
l i
nv
e
rter
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
:
15
70
–
15
78
1576
4.2.
Modifie
d
MF
O per
fo
rm
an
c
e
The
modifie
d
MFO
al
gorit
hm
is
c
ompar
ed
t
o
M
F
O
a
nd
DE
al
gorithms
in
orde
r
to
ver
if
y
it
s
performa
n
ce. Th
ese
al
go
rith
ms
are
a
nalyze
d
by u
sin
g
the same
obje
ct
ive
functi
on,
as
m
entione
d
in (4
-
6).
The
appr
oach
us
in
g
is
to
stop
it
era
ti
on
s
wh
e
n
the
searchi
ng
pro
cesses
ha
ve
re
ached
ɛ
<
10
-
7
.
Twe
nty
in
de
pe
nd
e
nt
exp
e
rime
nts
ar
e
carrie
d
out
f
or
eac
h
al
go
rithm.
A
s
t
he
res
ults,
the
a
ver
a
ge
num
be
r
of
i
te
rati
on
s
requi
red
by
the
modifie
d
M
F
O
is
11
5
it
erati
on
s
,
w
hile
the
MFO
a
nd
the
DE
nee
d
131
a
nd
182
it
erati
on
s
,
res
pe
ct
ively.
Com
par
is
on
to
both
MFO
an
d
DE
al
gorith
ms,
it
is
rev
eal
ed
that
the
modifie
d
MFO
gi
ves
bette
r
perf
ormance
in
fi
nd
i
ng
op
ti
mu
m
s
witc
hing
a
ngle
s
(a
s
s
een
i
n
Fi
gure
6).
T
he
modifi
ed
M
F
O
is
co
nv
e
r
gen
ce
fast
er
tha
n
M
F
O
a
nd DE a
lgorit
hm
s.
Figure
6
.
Fit
ne
ss v
al
ue
c
urves
for
ɛ
<
10
-
7
4.3.
Ca
sc
ad
e
9
-
le
ve
l i
nv
erter
In
orde
r
to
ge
ner
at
e
ou
t
pu
t
a
c
vo
lt
ag
e
f
rom
a
casca
de
d
m
ulti
le
vel
inv
ert
er,
the
pa
rame
te
rs
us
e
d
in
simulat
ion
s
ar
e:
casca
de
d
9
-
le
vel,
V
dc
=
10
0V,
M
=
3.2
,
and
f
=
50
Hz
(
T
=
20
ms).
The
set
of
op
t
imu
m
switc
hing
a
ng
l
es
fou
nd
f
or
M
=
3.2
a
re:
θ
1
=
10.81
69
o
,
θ
2
=
26.35
46
o
,
θ
3
=
53.01
06
o
,
an
d
θ
4
=
88.
0910
o
.
The
ti
me
delay
an
d
pulse
width
duri
ng
the
PWM
c
ontrol
of
t
he
in
ve
rter
a
re
de
rive
d
from
t
he
op
ti
m
um
s
witc
hing
ang
le
s
. T
he
s
w
it
ching
patte
r
ns are
presente
d i
n
Fi
g
ure
s
7
a
nd
8
.
The
sim
ulati
on
re
su
l
ts
at
a
modu
la
ti
on
in
dex
M
=
3.2
a
re
gi
ve
n
in
Fi
gure
9
.
From
the
fr
e
qu
e
nc
y
sp
ect
ra,
it
ca
n
be
see
n
that
the
ma
gn
it
udes
of
lo
w
-
orde
r
harmo
nics
su
c
h
a
s
t
he3
rd
,
5
th
,
a
nd
7
th
ha
ve
bee
n
om
it
te
d from
the
wav
e
f
or
m
e
ntirel
y.
The
resi
du
al
hi
gh
e
r
-
order
harmo
nics
ar
e:
V
9
/V
1
=
3.666
5%
,
V
11
/V
1
=
4.4
569%
,
V
13
/
V
1
=
4.468
4%,
V
15
/
V
1
=
0.954
5%,
V
17
/
V
1
=
3.325
1%,
V
19
/
V
1
=
4.0
786%
,
V
21
/
V
1
=
0.3430%.
The
mea
su
re
d
val
ue
of
V
1
is
319.8
43
V.
It
matc
hes
with
t
he
cal
c
ulate
d
V
1
f
or
M
=
0.8
,
wh
ic
h
is
V
1
=
31
0
V.
Eac
h
higher
-
or
der
ha
r
m
on
i
c
com
pone
nt
is
le
sser
tha
n
5%
with
t
otal
harmo
nic
disto
rtio
n
(T
HD)
of
9.9
5%.
T
he
TH
D
of
ou
t
pu
t
volt
age
c
an
be red
uce
d by
app
l
ying a
n
L
C passi
ve
filt
er
to
mi
nimize
th
e h
ig
h
-
orde
r ha
rm
on
ic
s
[
25].
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
Op
ti
miz
atio
n o
f SHEPWM
c
asc
aded
mult
il
evel
inverte
r sw
it
ching
patt
ern
s
(
Ay
ong Hie
ndro
)
1577
Figure
7
.
S
witc
hing
pu
lse
s
for a casc
a
ded
9
-
l
evel in
ver
te
r: S
11
, S
12
, S
21
,
S
22
,
S
31
, S
32
,
S
41
, S
42
Figure
8
.
S
witc
hing
pu
lse
s
for a casc
a
ded 9
-
l
evel in
ver
te
r: S
13
, S
14
, S
23
,
S
24
,
S
33
, S
34
,
S
43
, S
44
Figure
9
.
The
ou
t
pu
t
volt
age
wav
e
f
or
m
and
the corre
spo
nding
FFT
an
al
ysi
s
5.
CONCL
US
I
O
N
A
m
odifie
d
MFO
al
gorithm
has
bee
n
pro
pose
d
in
or
der
t
o
ov
e
rc
om
e
c
omp
utati
on
al
di
ff
ic
ulti
es
of
SH
EP
W
M
ca
s
caded
m
ulti
le
vel
inv
e
rters.
T
he
m
odifie
d
M
F
O
al
gorith
m
pr
ov
i
des
preci
se
com
puta
ti
on
of
op
ti
m
um
s
witc
hing
an
gles
for
casca
de
d
m
ulti
le
vel
inv
erte
rs
.
It
is
faste
r
co
nv
e
r
gen
ce
t
han
bo
t
h
M
F
O
an
d
D
E
al
gorithms.
Simulat
ion
res
ults
are
pr
ese
nte
d
f
or
a
casca
de
d
9
-
le
vel
i
nv
erter
i
n
sin
gle
-
ph
a
se
c
onfi
gur
at
ion
.
The
res
ults
al
so
s
how
that
the
lo
w
-
orde
r
ha
rm
on
ic
s
ar
e
utterl
y
el
imi
nated
from
th
e
ac
outp
ut
volt
age
wav
e
f
or
m
.
It
r
eveals
that
the
modifie
d
M
F
O
is
an
e
ff
ect
i
ve
al
gorith
m
to
s
olv
e
S
HE
pro
blems
of
cas
cade
d
mu
lt
il
evel in
ve
rters.
REFERE
NCE
S
[1]
R.
S.
Alishah,
S
.
H.
Hos
seini,
“
A
New
Multi
le
v
e
l
Inv
erter
Structure
For
High
-
P
ower
Appli
catio
ns
Us
ing
Mul
ti
-
ca
rri
er
PWM
S
witc
hing
Stra
tegy,”
Inte
rnat
ion
al
Journal
of
Powe
r
El
e
ct
ron
ic
s
and
Dr
ive
Syste
m
(
IJPEDS)
,
vol.
6
,
no
.
2
,
pp
.
318
-
325,
2015
.
[2]
E.
Isen,
A.F.
B
a
kan,
“Highl
y
Eff
ic
i
ent
Three
-
Pha
se
Grid
-
Conne
cted
Para
llel
Inve
r
te
r
Sys
tem,
”
Jou
rnal
of
Mod
ern
Pow
er
Syst
em
a
nd
Cle
an
En
ergy
,
vol
.
6
,
no
.
5
,
pp
.
1079
-
1089,
201
8.
[3]
M.
Ahmed,
E
.
Henda
wi,
M.
K.
Metwa
ly,
“Sin
gle
Phase
As
y
mm
et
r
ical
C
asc
ade
d
MLI
wi
th
Ext
re
me
Ou
tp
ut
Volta
ge
L
eve
ls
to
Sw
itch
R
at
i
o,
”
Inte
rnat
iona
l
Journal
of
P
ower
Elec
troni
c
s
and
Dr
iv
e
Sy
stems
(I
JPEDS)
,
vol.
9
,
no
.
2
,
pp
.
712
-
721,
2018
.
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
:
15
70
–
15
78
1578
[4]
A.
Khos
hkbar
-
Sadigh,
V.
Darg
a
hi,
M.
Abarz
ad
eh,
S.
Darhga
hi
,
“
Redu
ce
DC
Volta
ge
Sour
ce
Flying
Capaci
to
r
Multi
cell
Multi
le
ve
l
Inv
erter:
Analysis
and
Implem
ent
a
ti
o
n,
”
IET
Power
Elec
troni
cs
,
vol
.
7
,
no.
2,
pp.
439
-
450,
201
4.
[5]
N.
Sus
hee
l
,
P.S
.
Kumar
,
“
Perfor
ma
nc
e
Ana
lysis
of
FP
GA
base
d
Diode
Cl
am
p
ed
Multi
le
v
el
Inve
r
te
r
Fed
Indu
ct
io
n
Motor
Drive
usi
ng
Phase
Opposi
ti
on
Dispos
it
ion
Multi
carrie
r
B
a
sed
Modulat
ion
Strat
egy
,
”
Int
ernati
onal
Journal
of
Pow
er
Elec
tronic
s and
Dr
ive System
(I
JP
EDS
)
,
vol
.
8
,
no
.
4
,
p
p.
1512
-
1523,
20
17.
[6]
R.
Cast
il
lo
,
B.
Diong,
P.
Bigg
e
rs,
“
Singl
e
-
Phas
e
Hybrid
Casc
ad
ed
H
-
Bridg
e
and
Diode
-
Cl
am
p
Multi
le
v
el
Inve
r
te
r
with
Cap
ac
i
tor V
olt
age Ba
l
anci
ng,
”
I
ET
Pow
er
El
e
ct
ronics
,
vol
.
11
,
no.
4,
pp.
70
0
-
707,
2
018
.
[7]
M.
Mali
nows
ki,
K.
Gopaku
ma
r
,
J.
Rodrigu
ez,
M.A.
Pére
z,
“
A
Survey
on
C
asc
ade
d
Multi
le
v
e
l
Inve
r
te
rs
,
IE
E
E
Tr
ansacti
ons on Indus
trial
E
le
c
tronic
s
,
vo
l. 57
,
n
o.
7
,
pp
.
2197
-
2
206,
2010
.
[8]
A.
Nam
i,
F.
Z
are
,
A.
Ghos
h,
F.
B
laabje
rg
,
“A
Hybrid
Cas
ca
de
Conver
te
r
Topol
ogy
with
Seri
es
Conne
cted
Symme
trica
l
an
d
As
ymm
et
ri
ca
l
Diode
Cl
am
pe
d
H
-
Bridge
Ce
l
ls,”
I
EE
E
Tr
an
sacti
ons
on
Po
wer
El
e
ct
ronics
,
vol.
26
,
no
.
1
,
pp
.
51
-
65
,
2011
.
[9]
W.
Fe
i,
X.
Rua
n,
B
.
Wu
,
“
A
Gene
raliz
ed
For
mul
ation
of
Qu
art
er
-
Wa
v
e
Sym
me
try
SH
E
-
PWM
Problem
s
for
Multi
le
v
el Inve
rt
ers
,”
IEEE
Tr
ansacti
ons on Pow
er
Elec
troni
cs
,
v
ol.
24(7)
,
pp
.
17
58
-
1766,
2009
.
[10]
W.
Fe
i,
X.
Du
,
B.
Wu,
“A
G
e
ner
alize
d
Half
-
Wa
ve
Symmetr
y
SH
E
-
PWM
Formula
t
ion
for
Multi
le
v
el
Voltage
Inve
rte
rs,
”
I
EEE
Tr
ansacti
ons on
Industrial Elect
ronics
,
vol
.
57(9
),
pp
.
3030
-
3038
,
2010
.
[11]
T.
R.
Sumi
thi
r
a,
A.N.
Kum
ar,”
Elimi
na
ti
on
of
Ha
rmoni
cs
in
Mult
il
ev
el
Inve
r
te
rs
Connec
t
ed
to
Solar
Photovolt
aic
Sys
te
ms
Us
ing
AN
FIS
:
An
Expe
rimental
Case
Study
,”
Journal
of
Appl
i
ed
R
ese
arch
and
Te
ch
nology
,
vol
.
11(1
),
pp.
124
-
132,
201
3.
[12]
L.
K.
Haw,
M.
S.A.
Dahida
h
,
H.A.F.
Almuri
b
,
“
SHE
–
PWM
Casca
ded
Multi
level
Inv
ert
er
Wit
h
Adjustabl
e
DC
Volta
ge
Le
v
el
s
Control
for
STA
TCOM
Applic
ations
,”
I
EE
E
Tr
ansacti
ons
on
Po
wer
El
e
ct
roni
cs
,
vol.
29
,
no
.
12,
pp.
6433
-
6444
,
2014.
[13]
M.S.A.
Dahid
ah,
G.
Kons
ta
nt
ino
u,
V.G.
Ageli
d
is,
“
A
R
evi
ew
of
Multi
le
v
el
S
el
e
c
ti
ve
H
arm
oni
c
E
li
mi
n
at
ion
PWM:
Formula
ti
ons
,
Solving
Algori
th
ms,
Implem
ent
a
t
ion,
and
Applica
ti
ons,”
I
EEE
Tr
ansacti
ons
on
P
ower
El
e
ct
ronic
s
,
vol.
30
,
no
.
8
,
pp
.
4091
-
4106
,
20
15.
[14]
H
.
Behba
h
ani
f
a
rd,
S.
Aba
za
r
i,
A.
Sadoughi,
“
New
Schem
e
o
f
SH
E
-
PWM
Te
chni
qu
e
for
C
asc
ade
Multi
le
v
el
Inve
rte
rs wi
th
R
egul
a
ti
on
of
DC
Volta
ge
Source
s
,
”
IS
A
Tr
ansaction
,
vol
.
97
,
pp
.
44
-
52,
2020
.
[15]
S.
Ahmad
,
M.
Mera
j,
A.
Iqba
l,
I.
As
hra
f,
“
S
el
e
c
ti
ve
Harm
oni
cs
El
i
m
ina
t
ion
in
Multi
le
v
el
Inve
r
te
r
by
a
Deri
v
ati
ve
-
Free
It
erati
ve
Method
und
er
Var
ying
Volt
age Condi
ti
on,
”
IS
A
Tr
ansacti
on
,
vol
.
9
2,
pp
.
241
-
256
,
2019.
[16]
W.
A.
Ha
li
m
,
N.
A.
Rahi
m
,
M.
Azri
,
“
Se
lecti
v
e
Harm
onic
El
i
m
ina
ti
on
fo
r
a
Sin
gle
-
Phase
13
-
L
e
vel
TCHB
Base
d
Cas
ca
ded
Multi
l
eve
l
Inve
r
te
r
Us
ing
FP
GA
,
”
Jour
nal
of
Powe
r
Ele
ct
ronics
,
vol
.
14
,
no.
3,
pp.
488
-
4
98,
2014
.
[17]
R.
Taleb
,
M.
H
el
a
im
i
,
D.
B
eny
ouce
f,
Z.
Boudj
em
a
,
“Ge
n
et
i
c
Algorit
hm
Appl
ic
a
ti
on
in
As
ym
me
trica
l
9
-
l
eve
l
Inve
rte
r
,
”
Int
ernati
onal Journal of
Pow
er
Elec
tronic
s and
Dr
ive Systems
(IJ
PE
D
S)
,
vol
.
7
,
no
.
2
,
pp.
521
-
530,
201
6.
[18]
M.
E
te
sam
i,
N.
Ghasem
i,
D.M.
Vila
thg
a
muwa,
W.
L
.
Mala
n
,
“P
art
i
cl
e
Sw
arm
O
pti
mi
sa
ti
on
-
Bas
e
d
Modifi
ed
SH
E
Method
for
C
asc
ade
d
H
-
bridg
e M
ult
il
ev
el Inve
rt
er,
”
IET
Powe
r
El
e
ct
ronics
,
vol
.
10
,
no.
1,
pp.
18
-
28,
2017
.
[19]
M.A.
Memon,
S.
Mekhil
ef
,
M.
M
ubin,
M.
Aa
mi
r,
“
Sele
c
ti
ve
Ha
rm
onic
E
li
minat
ion
in
Inve
rt
ers
Us
ing
Bio
-
inspire
d
Inte
lligen
t
Algo
rit
hms
fo
r
R
enew
abl
e
Ene
rgy
C
onver
sion
Appl
i
ca
t
ions:
A
Review,”
R
ene
wabl
e
and
Sustainable
Ene
rgy
Revie
ws
,
vol.
82,
pp.
2235
-
2253,
2018
.
[20]
M.F.
Kanga
r
lu,
E.
B
aba
e
i,
“
A
G
ene
ra
li
z
ed
C
asc
a
ded
Multi
le
v
el
I
nver
te
r
Us
ing
Se
ri
es
Conne
ct
ion
of
Submult
i
le
v
el
Inve
rte
rs
,
”
I
EEE
Tr
ansacti
ons of
Powe
r E
le
c
troni
cs
,
vol
.
28
,
no
.
2
,
pp
.
625
-
636
,
2
013.
[21]
M.D.
Siddique,
S.
Mekhilef,
N.
M.
Shah,
A.
Sa
rwar,
M.
A.
Me
mon,
“
A
N
ew
S
ingl
e
-
P
hase
C
as
ca
ded
M
ult
i
le
v
e
l
I
nver
te
r
T
opolog
y
with
R
edu
ce
d
N
umbe
r
of
S
wit
che
s a
nd
V
ol
ta
g
e
S
tre
ss
,
”
Inte
rn
ati
onal
Tr
ansact
ions o
n
Elec
tri
c
al
Ene
rgy
S
yste
ms
,
vol.
30
,
no
.
2
,
20
19.
[22]
A.
K.
Panda,
Y.
Suresh,
“
Rese
ar
ch
on
Casca
d
e
Multi
le
v
el
Inv
er
te
r
with
Sing
le
DC
Source
by
Us
ing
Thre
e
-
Ph
ase
Tra
nsfor
me
rs
,
”
I
nte
rnational
Jou
rnal
of El
e
ct
ri
ca
l
Pow
er
&
Ene
r
gy
Syst
ems
,
vol
.
40
,
no
.
1
,
pp
.
9
-
20,
2012
.
[23]
R.
A.
R
ana,
S.A
.
Pa
te
l
,
A
.
Mut
husamy
,
C.
W
.
Le
e
,
H
.
J.
Ki
m,
“
Review
of
Multi
l
eve
l
Vol
ta
ge
Source
Inve
r
ter
Topol
ogie
s
and
Analysis
of
Har
moni
cs
Distor
ti
o
ns i
n
FC
-
MLI
,
”
El
e
ct
ronics
,
vol
.
8
,
no.
11,
pp.
1
-
37,
2019
.
[24]
S.
Mirjalili,
Mo
th
-
Flam
e
Opti
m
iz
a
ti
on
Algo
rit
h
m:
A
Novel
Natur
e
-
Inspir
ed
H
eur
isti
c
Par
adi
g
m,
”
Knowl
edge
-
Based
Syst
ems
,
vol.
89
,
pp
.
228
-
249,
2015
.
[25]
F.
Chabni
,
R
.
Ta
l
eb,
A.
La
kh
eda
r,
M
.
Boun
adj
a
,
“
New
Mo
difi
ed
CHB
Multi
l
eve
l
Inve
rt
e
r
Topol
ogy
wit
h
El
imination of
L
ower
and
Higher
Order
Har
monics”,
Aut
omat
ik
a
,
vol.
59
,
no
.
1
,
pp
.
1
-
10,
2018.
Evaluation Warning : The document was created with Spire.PDF for Python.