Internati
o
nal
Journal of P
o
wer Elect
roni
cs an
d
Drive
S
y
ste
m
(I
JPE
D
S)
Vol
.
7
,
No
. 2,
J
une
2
0
1
6
,
pp
. 56
8~
58
2
I
S
SN
: 208
8-8
6
9
4
5
68
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJPEDS
Predi
c
ti
ve Control f
o
r Reduced St
ructu
r
e Multilevel
Converters: Experimenting on
a Seven Level Packed U-Cell
Adel Naz
emi, Omid
Salari, Mo
hammad T
a
vak
o
li Bina,
Masou
d
Kaz
emi, Bah
m
an E
s
kan
d
ari
Departem
ent
of
Ele
c
tri
cal
and
C
o
m
puter Engin
e
ering,
K.
N
.
Too
s
i
University
of Techno
log
y
,
Tehran, Iran
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Feb 10, 2016
Rev
i
sed
May 21
, 20
16
Accepted
May 31, 2016
Recen
tl
y,
a br
a
n
ch of m
u
lti
lev
e
l conv
ert
e
rs
is
em
erged,
in
which the
i
r
‘reduced structu
r
e’ topologies use lowe
r number of devices compared to the
available topolo
g
ies. To get a
cost
efficient
converter
, lower
number of
com
ponents as
well as h
i
gh qualit
y
wav
e
form
s, m
u
ltilevel converters
with a
‘reduced stru
ctu
r
e’ (MCRS) are suitabl
e for high/medium power s
y
stems.
Also, utili
zing
the fast m
i
crop
rocessors avail
a
ble toda
y,
appl
ica
tions of
predic
tive
contr
o
l in power con
v
erters
ar
e of ver
y
power
ful an
d attra
c
tiv
e
alternatives to classical contro
llers.
This pap
e
r
proposes a finite contro
l set
model-based predictiv
e con
t
rol (
F
CS-MPC) for load curren
t
regu
lation
and
capacitor
voltag
e
balancing for
a ty
pi
ca
l M
CRS
. A c
a
s
e
s
t
ud
y
cons
idered
,
three-ph
ase sev
e
n level packed U-
cell (PUC)
,
which
is
among reduced
structure
m
u
ltil
evel convert
ers. A
discre
t
e
m
odel of
the s
y
st
em
is der
i
ved
,
and a predict
i
ve
m
odel-bas
e
d co
ntrol is
develope
d according to t
h
is
m
odel in
order to pr
edict
the futur
e
beh
a
v
i
or of
the s
y
stem for all possible switching
s
t
ates
;
then
, the
s
w
itching s
t
a
t
e
t
h
at opt
im
ized
th
e cos
t
fun
c
tion
i
s
s
e
lect
ed.
The f
easibility
o
f
the proposed
FCS-MP
C strategy
for
a sev
e
n level PUC is
evalu
a
ted
based
on simulations w
ith MATLA
B/ SIMULINK.
More
ove
r,
experimental validation of the pr
oposed
control sy
stem on a 5 k
VA
PUC is
exam
ined
throug
h DSP im
plem
entation
.
Keyword:
Cap
acito
r vo
ltag
e
b
a
lan
c
i
n
g
FCS-MPC
ML conve
rter
PUC m
u
ltilev
e
l co
nv
erter
R
e
duce
d
st
ruct
ure
Copyright ©
201
6 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Adel Nazem
i
Babadi,
Depa
rtem
ent of Electrical a
nd Co
m
p
u
t
er
Engin
eer
ing
,
K.
N. Too
s
i
Un
iv
ersity o
f
Tech
no
log
y
,
Sey
e
d Kha
n
da
n, Teh
r
an
, Ira
n
.
Em
a
il: a.nazemi@ee.knt
u.ac
.ir
1.
INTRODUCTION
Com
p
ared wit
h
conventional
two
lev
e
l conv
erters, m
u
lti
l
e
v
e
l con
v
e
rters can provide
an efficient
altern
ativ
e to hig
h
po
wer app
licatio
n
s
,
p
r
ov
id
ing
a
h
i
gh
quality o
u
t
p
u
t
voltag
e
, in
creasing
th
e efficien
cy and
robu
stn
e
ss, and
red
u
c
i
n
g
t
h
e electro
m
a
g
n
e
tic in
terferen
ce [1
-2
]. En
erg
y
efficien
cy, reliabilit
y, p
o
wer
d
e
n
s
ity,
sim
p
licity, cost effectiveness
,
re
duce
d structure with lower num
b
er
of active and passive ele
m
ents, hi
gh
p
o
wer qu
ality, an
d
app
licatio
n
field
are th
e
main
to
p
i
cs
for d
i
sting
u
i
sh
ing
d
i
fferen
t
topo
log
i
es of m
u
ltilev
e
l
con
v
e
r
t
e
rs [
3
]
.
R
e
searche
r
s a
l
l
over t
h
e w
o
rl
d are
spe
n
di
ng
g
r
eat
eff
o
rt
s t
o
i
n
t
r
o
duce
new t
o
p
o
l
ogi
es f
o
r
m
u
l
tilevel converters [4
-6].
Recently, a branch
of m
u
lti
l
e
vel conve
rt
ers is e
m
erged, in which t
h
eir ‘reduce
d
structure’ topologies
use lower num
b
er
of
devices com
p
ared to t
h
e a
v
aila
bl
e t
o
pol
ogi
es
[7
-
11]
. T
o
get
a cost
efficient c
o
nve
rter, l
o
we
r
num
ber of c
o
m
pone
nts as
well
as high quality
wave
form
s, m
u
ltilevel converters
with a ‘re
duce
d
structure’
(MCRS) are s
u
itable for hi
gh
or m
e
di
um
po
wer sy
st
em
s. In [
1
2]
, a cl
assi
fi
cat
i
o
n
and
revi
e
w
o
f
M
C
R
S
s i
s
pres
ent
e
d.
The c
o
n
v
ent
i
o
nal
t
o
pol
ogi
es a
r
e t
h
e
b
a
si
s fo
r ne
w re
searche
r
s t
o
de
vel
o
p
a no
v
e
l MCRS fam
i
ly o
f
m
u
l
tilev
e
l co
nv
ert
e
rs
with
lower nu
m
b
er o
f
power electron
i
c
d
e
v
i
ces. Th
is
p
a
p
e
r
considers the
m
o
st recent ‘mu
ltilevel conve
rters’ with a
‘r
educe
d
struct
ure’
(MCRS).
Am
ong
di
f
f
ere
n
t
t
y
pes o
f
t
h
e
M
C
R
S
, Packe
d
U
-
cel
l
(P
UC
) can
be
fo
rm
ed by
m
a
ni
pul
at
i
ng ei
t
h
e
r
a
co
nv
en
tio
n
a
l cap
acito
r-clam
p
e
d
o
r
the cascad
ed
H-b
r
i
d
g
e
m
u
l
tilev
e
l to
p
o
lo
g
i
es [1
3
]
. Each
U-cell co
n
s
i
s
ts o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Predictive Control for
Reduce
d
Struct
ur
e M
u
ltilevel Converters … (Adel N
a
zemi)
56
9
two
po
wer
switch
e
s and on
e cap
acitor t
h
at produ
ces m
o
re ou
tpu
t
v
o
ltag
e
lev
e
l using a sm
all n
u
m
b
e
r
o
f
passi
ve
an
d act
i
v
e com
p
o
n
ent
s
. L
o
w
ha
rm
oni
c cont
e
n
t
s
, m
o
re
v
o
l
t
a
ge l
e
v
e
l
s
i
n
i
n
vert
e
r
out
put
si
de,
re
duce
d
filters ratin
g, an
d
po
ssib
ility o
f
co
m
b
in
atio
n o
f
GTOs an
d
IGBTs, are th
e
m
a
in
featu
r
es
o
f
th
e
PUC
[13
]
. For
the three
phase
PUC, t
h
e m
a
in disa
dv
an
tag
e
is th
e in
d
e
p
e
nd
en
t
DC power
supply for ea
ch phases
. Als
o
, the
max
i
m
u
m
p
h
a
se vo
ltag
e
is li
mited
to
p
r
i
n
ci
p
a
l DC link
voltag
e
. Nam
e
ly,
m
a
x
i
m
u
m
p
h
a
se vo
ltag
e
is eq
u
a
l t
o
th
e prin
ci
p
a
l DC lin
k
am
p
litu
d
e
.
Pre
d
ictive cont
rol a
ppea
r
s as
an attractive al
terna
tive
fo
r th
e co
ntrol
of
p
o
w
er c
o
nve
rters
and
o
ffe
rs
a num
b
er of a
dva
ntage
s
: no
need
for linear controllers a
n
d m
odulators,
easy
inclusion of nonli
n
earities
and
con
s
t
r
ai
nt
s, a
n
d fast
dy
n
a
m
i
c resp
on
se [1
4,
15]
. Se
ve
ral
c
ont
rol
al
g
o
ri
t
h
m
s
have bee
n
prese
n
t
e
d
un
d
e
r t
h
e
nam
e
of predi
c
t
i
v
e cont
r
o
l
,
as
prese
n
t
e
d i
n
[
16]
. Am
on
g t
h
ese cont
r
o
l
sch
e
m
e
s, fi
ni
t
e
cont
r
o
l
set
m
odel
-
base
d
p
r
ed
ictiv
e contro
l
(FCS-MPC) has
d
e
m
o
nstrated
t
o
b
e
a v
e
ry
p
o
werfu
l an
d attractiv
e altern
ativ
e
for t
h
e
cont
rol
of
p
o
w
e
r co
n
v
ert
e
r
s
a
nd
ha
ve
been
use
d
t
o
i
m
pl
em
ent
di
ffe
rent
cont
rol
st
rat
e
gi
es i
n
p
o
w
er c
o
nve
rt
e
r
to
po
log
i
es and ap
p
lication
s
[1
7-2
2
]
.
Th
e basic id
ea o
f
FCS-MPC is to p
r
ed
ict th
e futu
re
b
e
h
a
v
i
or
o
f
the
sy
st
em
based on i
t
s
di
scr
e
t
e
m
odel
.
Therea
ft
er, t
h
e s
w
i
t
c
hi
n
g
vect
or t
h
at
opt
i
m
i
zes a
gi
ve
n cost
f
u
n
c
t
i
on i
s
selected
an
d
ap
p
lied
t
o
th
e co
nv
erter. Th
e si
m
u
ltan
e
ous
cont
rol
o
f
t
h
e
PUC
o
u
t
p
ut
cur
r
ent
s
a
nd
D
C
-l
i
n
k
capacitors
volt
ages is
done
preferab
ly with
a m
o
d
e
l-b
a
se
d
pred
ictiv
e con
t
ro
ller (MBPC
)
.
In
th
is
p
a
p
e
r, essen
tial step
s fo
r im
p
l
e
m
en
ti
n
g
t
h
e pred
icti
v
e
con
t
ro
l alg
o
rith
m
fo
r a typ
i
cal MCRS
is p
r
esen
ted
.
Th
ereafter, a d
i
screte-tim
e
math
e
m
atic
al
m
odel
of t
h
e
t
h
ree-
p
h
ase P
U
C
i
s
deri
ved
and
a
pre
d
ictive m
odel-base
d control accordi
ng t
o
this disc
rete
m
odel is develope
d. T
h
is m
odel is use
d
to
pre
d
ict
t
h
e f
u
t
u
re be
h
a
vi
o
r
o
f
sy
st
e
m
for al
l
p
o
ssi
bl
e swi
t
c
hi
ng
st
at
e. The
pr
o
pos
ed
pr
edi
c
t
i
v
e c
ont
r
o
l
st
ra
t
e
gy
:
1)
regulates the l
o
ad c
u
rre
nts; and
2)
obtains capacitor voltage
bala
ncing
of each
phases. Perform
a
nce of t
h
e
pr
o
pose
d
FC
S
-
M
P
C
st
rat
e
gy
fo
r a se
ve
n l
e
vel
pac
k
e
d
U
-
cel
l
con
v
e
r
t
e
r
i
s
eval
uat
e
d
base
d o
n
si
m
u
l
a
t
i
on
st
udi
es i
n
M
A
TLAB
/
S
I
M
U
LIN
K
s
o
ft
ware
an
d ex
pe
ri
m
e
nt
al
val
i
d
at
i
o
n
t
h
r
o
u
g
h
real
-t
i
m
e im
pl
em
ent
a
t
i
on
on
the DSP
TMS320F28335.
Th
is p
a
p
e
r is o
r
g
a
n
i
zed
as fo
llo
ws: in
Sectio
n
2
essen
tial step
s fo
r im
p
l
e
m
en
tin
g
FCS-MPC for a
t
y
pi
cal
M
C
R
S
i
s
present
e
d;
i
n
Sect
i
on 3 t
h
e PUC
t
o
p
o
logy is introduced as a case study and its discrete
m
odel
i
s
deri
v
e
d;
i
n
Sect
i
o
n
4 t
h
e
t
h
e
o
ry
o
f
FC
S-M
P
C
i
s
prese
n
t
e
d;
an
d
sim
u
l
a
t
i
on an
d
ex
peri
m
e
nt
al
resul
t
s
are prese
n
ted
in
Section 5
a
n
d 6, res
p
ectivel
y.
2.
MO
DELIN
G
RED
UCE
D S
T
RU
CTU
R
E MULTILEVE
L
CO
NVE
RT
ERS
In
t
h
is section, th
e essen
tial step
s for m
o
d
e
lin
g of lo
ad an
d conv
erter, an
d
im
p
l
e
m
en
tatio
n
o
f
p
r
ed
ictiv
e co
n
t
ro
ller
will b
e
p
r
esen
ted. For i
m
p
l
e
m
en
tin
g a pred
ictiv
e co
n
t
ro
l
b
a
sed
al
g
o
rith
m
o
n
a typ
i
cal
MCRS, th
e d
i
screte
m
a
th
e
m
a
tical
m
o
d
e
l o
f
th
e co
nv
erter sh
ou
ld
b
e
d
e
rived
.
Accord
i
n
g
to
th
is
m
o
d
e
l, o
u
t
p
u
t
vol
t
a
ge a
n
d capaci
t
o
r
vol
t
a
ge
s and c
u
r
r
ent
s
sho
u
l
d
be cal
cul
a
t
e
d f
o
r eac
h swi
t
c
hi
ng st
at
e of t
h
e co
n
v
ert
e
r
.
The key
st
ep
s of
t
h
i
s
pr
oce
d
u
r
e
t
h
at
ca
n be g
e
neral
i
zed
for every discrete syste
m
, are summarized as follows:
Deriving the discrete state sp
ace equations of the MCRSs.
Obtaining the predictions, according to di
screte m
a
them
atical m
odel of MCRSs.
M
i
ni
m
i
zat
i
on of a predefi
n
ed cost
funct
i
on.
It sh
ou
ld
b
e
no
ted
th
at, in
th
is p
a
p
e
r and
for po
wer conv
erter ap
p
licat
io
n
,
a sim
p
le
m
o
d
e
l o
f
th
e
in
v
e
rter
will be u
s
ed
(b
ecau
s
e o
f
d
i
screte natu
re an
d lo
w switch
i
ng
frequ
en
cy [17
]
).
To
illu
strate
th
e p
r
o
cess,
each ste
p
will be st
udy se
pa
rately as follows.
2.
1.
Discrete State
Space Model of
the
Conver
ter
Al
t
h
o
u
gh
va
ri
ous
cost
fu
nct
i
ons c
o
ul
d
be
con
s
i
d
er
ed
, o
n
l
y
l
o
ad c
u
r
r
e
n
t
re
gul
at
i
o
n
and
DC
l
i
n
k
cap
acito
r
vo
ltag
e
b
a
lan
c
ing
i
s
con
s
id
ered in
th
is
p
a
p
e
r.
By assu
m
i
n
g
th
at th
e
p
l
an
t i
s
a typ
i
cal MCRS (or
ot
he
r
po
wer
co
nve
rt
er
), t
h
e
di
scret
e
st
at
e sp
a
ce eq
uat
i
o
ns ca
n
be
descri
bed
by
(i
n
gene
ral
f
o
rm
) [
17]
:
1
(1
)
whe
r
e
is th
e in
pu
t v
a
riab
le
(co
n
t
ro
l action
s
o
r
switch
i
ng
state v
ecto
r
s),
is th
e p
l
an
t ou
tp
u
t
(ou
t
pu
t vo
l
t
ag
e
or
cu
rre
nt
vect
ors
)
, an
d
is the
state space
va
riable
vect
or
wi
th dim
e
nsion
.
The
differe
n
ce
of the state-s
p
a
ce equation is:
∆
1
∆
∆
(2
)
To
relate
∆
to
the ou
tpu
t
, a ne
w state s
p
ace
variable
vector i
s
chose
n
to be
[23]:
∆
(3
)
wh
ere sup
e
rscrip
t T ind
i
cates
matrix
tran
spose.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-86
94
I
J
PED
S
Vo
l.
7
,
No
.
2
,
Jun
e
2
016
:
56
8-5
82
57
0
Based
on E
qua
tions
(1), (2), a
n
d (3), t
h
e
foll
owi
n
g state s
p
ace m
odel is derive
d
(due t
o
t
h
e
pri
n
ciple
of
rece
di
n
g
ho
r
i
zon c
o
nt
r
o
l
,
0
in
th
e d
i
screte syste
m
m
o
d
e
l [2
3
]
):
1
∆
(4
)
w
h
er
e
1
,
,
1
,
00
…
0
Based on
the state
space m
odel
(
A
,
B
,
C
), t
h
e
fut
u
re out
put
vari
abl
e
s
are calculated
as follows: (By
assum
i
ng one-st
e
p-ahead predi
c
t
i
on):
1
∆
(5
)
Accord
ing
to
Eq
s. (4) an
d
(5),
for eac
h input
vect
or
(
s
w
itch
i
ng
state
v
ector
s i
n
po
wer
co
nv
er
ter
application), st
ate space
(ca
p
a
c
itor
voltage
a
n
d loa
d
c
u
rrent
)
a
n
d output
va
riables ca
n
be
pre
d
icted.
2.
2.
Prediction
of State Space Variables
Aft
e
r i
d
e
n
t
i
f
y
i
ng al
l
possi
bl
e swi
t
c
hi
n
g
st
at
es of a t
y
pi
cal
M
C
R
S
, a speci
fi
c vol
t
a
ge ve
c
t
or sh
o
u
l
d
be
related to e
ach switching stat
es. Act
u
ally, each s
w
itching
state vector ge
nerates a
speci
al voltage
vect
or
i
n
AC
si
de of t
h
e
con
v
ert
e
r a
nd
as a resul
t
a l
ook
-
up t
a
bl
e i
s
obt
ai
ne
d.
Al
so
,
a l
ook
-u
p t
a
bl
e t
h
at
det
e
rm
i
n
es t
h
e
rel
a
t
i
on bet
w
e
e
n fl
y
i
ng ca
pa
ci
t
o
r an
d l
o
ad
cur
r
ent
s
i
s
de
r
i
ved f
o
r eac
h
swi
t
c
hi
n
g
vect
or
. So
, t
h
ere a
r
e t
w
o
look-up tables that were
used
for pr
ediction
of state space
varia
b
les. By
using thes
e look-up tables, m
e
a
s
ure
d
and estim
ated
param
e
ters, flying capacitor voltages and lo
a
d
cur
r
e
n
t
can b
e
cal
cul
a
t
e
d and pre
d
i
c
t
e
d ba
s
e
d o
n
discrete state space m
odel of the conve
r
ter (Eqs.
(4) and
(5)).
All the pre
d
icted contro
l param
e
ters inserted in
a pred
efin
ed
co
st fun
c
tion
and
o
p
tim
ized
for selecting
a
n
appropriate switching st
ate vector. T
h
ese st
eps i
s
summ
arized as follows:
Ident
i
f
y
i
ng al
l
possi
bl
e swi
t
c
hi
ng st
at
e vect
ors of t
h
e consi
d
ered M
C
R
S
.
Relating each switching state to a speci
fic voltage vector (look-up table 1).
Relating flying capacitor currents to the load cu
rrents for each switching v
ector (look-up table 2).
Capacitor voltage and load current m
easurem
ent.
Est
i
m
at
i
on of i
m
m
easurabl
e
param
e
t
e
rs (such as i
n
t
e
rnal
vol
t
a
ge vect
or of t
h
e l
o
ad).
Prediction of state variables
and future outputs of the syst
em
, based on state space m
odel.
C
o
st
funct
i
on opt
i
m
i
zat
i
on and sel
ec
t
i
on of appropri
a
t
e
cont
rol
act
i
on.
Based
on disc
rete state space
m
odel of the
s
y
ste
m
and by c
onsi
d
eri
n
g the
capacitor voltage a
n
d loa
d
current vector as
state
varia
b
l
e
s, f
u
t
u
re
beha
vi
o
r
o
f
t
h
e st
at
e vari
a
b
l
e
s ca
n
be
deri
ved
.
Na
m
e
l
y
, for al
l
p
o
ssi
bl
e
swi
t
c
hi
n
g
st
at
e
,
ne
xt
sam
p
l
i
ng val
u
e
of
sh
o
u
l
d
be cal
cul
a
t
e
d. F
o
r exam
pl
e, ne
xt
sam
p
l
i
ng t
i
m
e val
u
e of
fl
y
i
ng ca
paci
t
o
r
vol
t
a
ge
an
d l
o
ad
cu
rre
nt
ve
ct
or ca
n
be
wri
t
t
e
n as:
1
,
1
,
,
(6
)
whe
r
e
and
are m
easured ca
pacitor
voltage
and lo
a
d
c
u
rrent vect
or, res
p
ectively. Als
o
,
1
(flying ca
pacitor c
u
rrent
)
and
(o
ut
p
u
t
v
o
l
t
a
ge vect
or
) sh
o
u
l
d
be de
ri
ve
d
fr
om
l
ook
-
up
t
a
bl
es 1 an
d
2.
Fin
a
lly,
de
not
e
s internal volt
age
of a
gene
ra
l load
a
n
d esti
mated according to system
equations.
2.
3.
Cos
t
F
unc
tion
Op
timiz
a
tion
Th
e b
a
sic pur
po
se
of
t
h
e
p
r
edictiv
e con
t
ro
l
syste
m
is to
b
r
i
n
g th
e
p
r
ed
icted
ou
tpu
t
as cl
ose a
s
pos
si
bl
e
t
o
t
h
e re
fere
nce
si
gnal
s
. T
h
is
objective is
the
n
t
r
anslated
into
a co
s
t
fu
n
c
tio
n
.
Th
e
r
e are
di
ffe
re
nt
f
o
rm
s fo
r c
o
st
f
unct
i
ons
. C
o
nsi
d
eri
ng
fast
sam
p
l
e
peri
ods
, as i
s
u
s
ual
i
n
po
we
r
con
v
e
r
t
e
rs,
di
f
f
ere
n
t
t
y
pes o
f
c
o
st
f
u
nct
i
o
n
s
have
sa
m
e
resul
t
s
[
1
4]
. He
nce,
t
h
e c
o
st
fu
nct
i
o
n
g i
s
cal
cul
a
t
e
d as
p
r
esent
e
d
by
(
7
)
,
|
1
1
|
(7
)
whe
r
e
is a
vec
t
or t
h
at contains the
refe
renc
e
signals.
To
f
i
nd
th
e op
t
i
m
a
l
(co
n
t
ro
l actio
n
)
th
at
will min
i
mize
, t
h
e
bel
o
w e
quat
i
on
i
s
use
d
:
∆
0
(8
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Predictive Control for
Reduce
d
Struct
ur
e M
u
ltilevel Converters … (Adel N
a
zemi)
57
1
From
whi
c
h t
h
e o
p
t
i
m
a
l
cont
rol
si
gnal
∆
(s
w
i
t
c
hi
ng
st
at
e v
ect
or i
n
po
wer
co
nve
rt
er a
p
p
l
i
cat
i
on)
coul
d
be
deri
v
e
d.
Di
f
f
ere
n
t
t
a
sks
per
f
o
rm
ed
by
t
h
e
p
r
edi
c
t
i
v
e c
ont
rol
al
go
ri
t
h
m
,
i
s
sho
w
n i
n
Fi
g
u
r
e
1.
3.
PACKED U-CELL CO
NVERTER MODEL
No
w,
t
h
ese
st
eps c
o
ul
d
be e
x
pl
oi
t
e
d
f
o
r m
o
del
i
n
g
t
h
e
PU
C
as a M
C
R
S
.
ST
A
R
T
M
e
a
s
u
r
em
en
t
s
o
f
:
Li
n
e
C
u
r
r
en
t
V
e
c
t
o
r
An
d
C
a
p
a
c
i
t
o
r
Vo
l
t
a
g
e
E
s
ti
m
a
t
ion
O
f
I
m
me
as
u
r
ab
l
e
P
a
r
a
me
t
e
r
s
I
n
i
t
ia
l
i
z
a
t
i
on
O
f
Co
n
t
r
o
l
L
o
o
p
i=
i
+
1
D
i
sc
r
e
t
e
St
a
t
e
Sp
a
c
e
M
o
d
e
l
f
o
r
L
o
ad
C
u
r
r
e
n
t
&
C
a
p
.
V
o
l
t
ag
e
P
r
e
d
i
c
t
i
o
n
Co
s
t
F
u
n
c
t
i
o
n
O
p
ti
m
i
z
a
ti
o
n
i=N
s
?
A
ppl
y
O
p
t
i
mal
S
w
it
c
h
in
g
S
t
a
t
e
V
e
c
t
or
Se
l
e
c
t
i
o
n
O
f
O
p
ti
m
a
l V
o
lt
a
g
e
Ve
c
t
or
Lo
ok
-
u
p
T
a
b
l
e
s
1
a
n
d
2
Ye
s
No
Figure 1. Flow d
i
agram of
the pr
ed
ic
tive
con
t
rol
for a
t
y
pi
cal
MCRS (
is the number of
switching
state vectors).
3.
1.
Modeling
of
PUC
A single-phase
packe
d
U-cell conver
te
r wit
h
connected ac
tive load is
conside
r
ed i
n
[13]. He
re the
sin
g
l
e-ph
ase co
nfigu
r
ation
is ex
tend
ed
to three-p
h
a
se co
nverter, a
s
depict
ed in Figure
2, which is
com
pos
ed
o
f
18
p
o
wer
switch
e
s,
th
ree p
r
i
n
cip
a
l DC-lin
k
cap
acitors
as well as t
h
ree aux
iliary DC-lin
k
.
For three-p
h
a
se
co
nf
igu
r
ation, th
er
e
ar
e 5
12
(
8
) s
w
i
t
c
hi
ng
st
at
es. T
h
e
vol
t
a
ge
i
n
a
n
y
pha
se
o
f
t
h
e i
n
vert
er
i
s
ex
p
r
esse
d
bas
e
d
o
n
switch
i
ng
states an
d
capacito
r vo
ltag
e
s (p
ri
n
c
ip
al and
aux
iliary DC lin
k
)
as th
e fo
llowing
(B
ased
on
switch
i
ng
tab
l
e illu
strated
i
n
[1
3
]
):
(9
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-86
94
I
J
PED
S
Vo
l.
7
,
No
.
2
,
Jun
e
2
016
:
56
8-5
82
57
2
whe
r
e
(
,
,
) is th
e ph
ase t
o
n
e
u
t
ral v
o
ltages
o
f
t
h
e co
nv
erter
(j-ph
a
se).
Also,
and
are t
h
e
pri
n
ci
pal
(
) and
au
x
iliary (
) DC-link
v
o
ltage,
resp
ectiv
ely.
Wh
ile th
e
sing
le-p
h
a
se PUC in
trodu
ces
7
d
i
ffere
n
t
vo
ltag
e
lev
e
ls, th
e ex
ten
d
e
d th
ree-ph
ase
co
nf
igu
r
ation pr
ov
id
es 13-
levels f
o
r
lin
e vo
ltag
e
an
d
7
-
l
e
vel
s
f
o
r
co
n
v
ert
e
r
pha
se t
o
ne
ut
ra
l
(n
)
vol
t
a
ge
.
Based
o
n
Eq
.
(9
), th
e d
e
riv
a
ti
v
e
o
f
lo
ad
cu
rren
t
in
an
y
p
h
ase-j of th
e converter can
b
e
w
r
itten
as:
R
(1
0)
whe
r
e
L
is t
h
e
load i
n
ductance,
R
the load resistance,
and
are th
e
prin
cip
a
l and
aux
iliary DC-link
capacitors
, a
nd
is th
e in
tern
al
vo
ltag
e
v
ector of th
e lo
ad
for ph
ase
j
.
B
y
consi
d
eri
n
g
t
h
e i
n
ve
rt
er
o
u
t
put
vol
t
a
g
e
ve
ct
or as
Eq
.
(
11)
, t
h
er
e ar
e
51
2 vo
ltag
e
v
ectors in
i
n
v
e
r
t
er
out
put
,
ba
sed
o
n
51
2 s
w
i
t
c
hi
n
g
st
at
es.
2
3
.
.
,
(1
1)
Th
e equ
a
tio
ns
o
f
a
g
e
n
e
ral load
for each
p
h
a
se (as d
e
p
i
cted
in
Figure
2
)
can
b
e
written
as:
(1
2)
whe
r
e
is th
e
vo
ltag
e
b
e
tween lo
ad and
con
v
erter
n
e
u
t
ral poin
t
s.
Accord
ing
to [1
7
]
, and
sub
s
titu
tin
g (12
)
i
n
to
(11
)
, th
e d
y
n
a
mic b
e
h
a
v
i
o
r
of a
g
e
n
e
ral l
o
ad
can
b
e
descri
bed
by
,
(1
3)
whe
r
e
i
s
vect
o
r
of t
h
e c
o
n
v
e
r
t
e
r
out
put
v
o
l
t
a
ge,
is th
e lo
ad
cu
rrent vect
or, and
th
e in
ternal v
o
ltage
vector
of t
h
e l
o
ad.
B
y
consi
d
eri
n
g
vect
o
r
pre
s
ent
a
t
i
on,
t
h
r
e
e phase sp
ace vo
ltag
e
v
ector
is as fo
llo
ws:
(1
4)
whe
r
e
,
,
,
,
,
, and
,
,
.
The ca
pacitor
voltage
s are
e
x
press
e
d base
d
on capacit
o
r c
u
rre
nts as
follows:
1
,
1,
2
(1
5)
Th
e
p
r
i
n
cip
a
l
DC link
cap
a
cito
r cu
rren
ts
(i
.e.,
,
, a
n
d
)
are pre
d
i
c
t
e
d base
d on
t
h
e
t
h
ree
-
pha
se loa
d
c
u
r
r
e
nts (i.e
.,
,
, and
) as
fo
llo
ws:
(1
6)
Also
, th
e aux
iliary cap
acitor cu
rren
ts (i.e.,
,
, a
n
d
) are
predict
e
d as
follows:
(1
7)
Based
on
eq
uatio
n
s
(1
0)
, (1
6)
and
(
17)
, th
e lo
ad currents and the
cap
acito
r vo
ltag
e
s can
be
cont
rol
l
e
d
by
c
h
o
o
si
ng
a
pr
op
er s
w
i
t
c
hi
n
g
st
at
e vect
o
r
.
3.
2.
Discrete
Model for
Predicti
on
Th
e
d
i
screte mo
d
e
l
will b
e
used
to
p
r
ed
ict th
e
fu
t
u
re
v
a
l
u
e of co
n
t
ro
l
p
a
ra
m
e
ters at ti
me in
stan
t
of
(
1
. The l
o
ad current
deri
vative
is re
placed
by
(Eule
r
a
p
proxi
m
ation),
1
(1
8)
whe
r
e
i
s
t
h
e
sam
p
l
i
ng peri
o
d
.
Acc
o
r
d
i
n
g t
o
E
q
.
(
1
3
)
,
by
usi
n
g
Eul
e
r
a
p
p
r
oxi
m
a
t
i
on,
t
h
e ex
pr
essi
o
n
fo
r
p
r
ed
ictio
n of
fu
ture lo
ad
cu
rren
t
is (i
d
eal case)
[17
]
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Predictive Control for
Reduce
d
Struct
ur
e M
u
ltilevel Converters … (Adel N
a
zemi)
57
3
1
1
(1
9)
whe
r
e
d
e
no
tes
th
e esti
m
a
ted
in
tern
al vo
ltag
e
v
ector
o
f
t
h
e lo
ad and
is th
e ou
tpu
t
vo
ltag
e
v
ector
o
f
t
h
e
i
nve
rt
er.
Th
e estim
ated
in
tern
al
vo
ltage v
ector
o
f
th
e
lo
ad
is calcu
lated
thro
ugh
th
e
fo
llowing
formu
l
a [16
]
:
1
1
1
(2
0)
Th
e freq
u
e
n
c
y
o
f
th
e i
n
tern
al v
o
ltag
e
v
ector
(b
ack
-
EMF)
o
f
th
e lo
ad
is m
u
ch
less th
an
the sa
m
p
lin
g
fre
que
ncy
.
I
n
t
h
i
s
pa
per
,
t
h
e
sam
p
l
i
ng f
r
eq
uency
use
d
wi
t
h
t
h
e
FC
S-M
P
C
st
rat
e
gy
w
a
s
10
.
So
, for
si
m
p
licit
y o
f
calcu
lation
s
, th
e v
a
riatio
ns of
can b
e
n
e
g
l
ected
in
o
n
e
sam
p
lin
g
in
terv
al; th
u
s
:
̂
̂
1
.
Alth
oug
h, th
i
s
app
r
ox
im
a
tio
n
is
no
t v
a
lid
wh
en
t
h
e sam
p
lin
g
freq
u
e
n
c
y
is lo
w.
a
n
+
-
C2
a
S1a
S2a
S3a
b
+
-
C2
b
S1b
S2b
S3b
c
+
-
C2c
Vd
c
Vd
c
Vd
c
N
Ea
Eb
Ec
R
L
S1c
S2c
S3c
ic
ib
ia
Figure 2. Sugges
t
ed
configur
atio
n for 7-level three-phase PUC (
).
Acco
r
d
i
n
g t
o
Eq.
(
1
5
)
a
n
d
usi
n
g
Eul
e
r
a
p
p
r
oxi
m
a
tio
n
,
th
e ex
pression fo
r pred
iction
o
f
fu
ture
capacitor voltage is
(ideal cas
e):
1
(2
1)
whe
r
e
is the m
easure
d
voltage of the
capacit
o
r and
cal
cul
a
t
e
d base
d on
Eq
s.
(
1
6) an
d (1
7
)
.
3.
3.
T
w
o-Ste
p
-
A
h
e
ad Predi
c
ti
o
n
The ideal case
in whic
h the c
ont
rol pa
ram
e
ters can
be m
e
asure
d
,
pre
d
i
c
t
e
d, an
d c
ont
rol
l
ed i
n
st
ant
l
y
in
is im
possibl
e in real
-tim
e appli
cation
s
. Non
e
th
eless,
th
is p
r
ob
lem
can be can
be
overc
o
m
e
if a two-
step
-ah
ead pred
ictio
n
is consid
ered
. Th
is
way, a co
m
p
lete sa
m
p
le p
e
riod
is av
ailab
l
e to
p
e
rfo
rm th
e
cont
rol
al
go
ri
t
h
m
s
[14]
.
Th
us
, eq
uat
i
o
ns
(1
9
)
a
n
d
(
2
1
)
ca
n
be m
odi
fi
ed
as
fol
l
o
ws:
2
1
1
1
1
(2
2)
2
1
1
(2
3)
The
fut
u
re
re
fe
rence
cu
rre
nt
s
are
obt
ai
ne
d
b
y
[2
4]
(
f
o
r
hi
g
h
l
y
dy
nam
i
c sy
st
em
s),
2
1
0
2
0
1
1
5
2
3
(2
4)
4.
FCS
-
MPC
4.
1.
Co
ntr
o
l S
t
ra
t
e
gy
The p
r
o
p
o
se
d pre
d
i
c
t
i
v
e co
nt
rol
st
rat
e
gy
i
s
devel
ope
d in a
way to track the refe
re
nce load curre
nts
an
d regu
late the cap
acitor
vo
ltag
e
s.
For t
h
is
p
u
rp
o
s
e, it
is
ne
c
e
s
s
a
r
y
to me
a
s
u
r
e th
e thr
e
e-
ph
a
s
e
lo
ad
cur
r
e
n
t
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-86
94
I
J
PED
S
Vo
l.
7
,
No
.
2
,
Jun
e
2
016
:
56
8-5
82
57
4
an
d vo
ltag
e
s in
th
e cap
acitors,
p
r
ed
ict th
e state sp
ace variab
les an
d ev
alu
a
te a co
st
fu
n
c
tion
fo
r each
switching states. The switchi
ng state th
at
minim
i
zes the cost function is se
l
ect
ed and ap
pl
i
e
d d
u
ri
ng t
h
e nex
t
sam
p
l
i
ng pe
ri
o
d
.
Pr
o
pose
d
F
C
S-M
P
C
c
ont
r
o
l
sc
hem
e
for
a t
h
ree
-
phase
7-l
e
vel
P
U
C
c
o
n
v
e
r
t
e
r i
s
sh
o
w
n
i
n
Fi
gu
re 3.
Her
e
5
1
2
s
w
i
t
c
hi
ng
vect
or
s
at
t
h
e
sam
p
lin
g
ti
m
e
are used to
pred
ict t
h
e conv
erter ou
t
p
u
t
vol
t
a
ge
. T
h
e
p
r
edi
c
t
i
on
o
f
l
o
a
d
c
u
r
r
ent
s
an
d
DC
-l
i
n
k ca
paci
t
o
r
v
o
l
t
a
ge a
r
e
do
ne
usi
n
g
(
2
2
)
a
n
d
(
2
3
)
.
F
i
gure 3.
F
C
S
-
M
P
C for thre
e-p
h
as
e P
U
C conv
e
r
ter.
4.
2.
C
o
st
Fu
n
c
t
i
o
n
Based
o
n
th
e co
n
t
ro
l
requ
iremen
ts (lo
a
d
curren
t
referen
ce
track
ing
and
au
x
iliary cap
acito
r
vo
ltag
e
b
a
lan
c
e), t
h
e co
st fun
c
tio
n to
b
e
o
p
tim
ized
, is rep
r
esen
ted at th
e
sam
p
lin
g
ti
m
e
as fo
llows [1
7
]
:
∗
1
1
∗
1
1
∑
,
,
(2
5)
whe
r
e
∗
, a
n
d
∗
,
are the
real a
n
d im
aginary parts of
refe
re
n
ce cu
rre
nts,
re
spectively
.
T
h
e
, and
are
pre
d
icted c
u
rrents
,
and
are
weighting fa
ctors.
The
l
a
st
t
e
rm
i
s
t
h
e
num
ber
of
c
o
m
m
ut
ati
ons
(
) re
qui
red
t
o
c
h
a
nge
fr
o
m
t
h
e
p
r
esent
swi
t
c
hi
n
g
v
ector to
the fu
ture switch
i
ng
state
v
ect
o
r
.
By u
s
in
g th
is t
e
rm
, th
e switc
hing freque
ncy
of t
h
e c
o
nvert
e
r can
be decreased. Each term
in the cost function
has a sp
ecific weig
h
ting facto
r
wh
ich
is u
s
ed
to
tu
ne th
e
im
port
a
nce
o
f
t
h
at
t
e
rm
i
n
rel
a
t
i
on t
o
t
h
e
ot
he
rs c
ont
r
o
l
t
a
rget
s. T
h
es
e param
e
t
e
rs have t
o
be p
r
ope
rl
y
desi
g
n
e
d
i
n
or
der t
o
achi
e
ve
t
h
e desi
r
e
d
p
e
rf
orm
a
nce. U
n
f
o
rt
unat
e
l
y
, t
h
ere a
r
e n
o
a
n
al
y
t
i
cal
or nu
m
e
ri
cal
m
e
t
hods o
r
co
nt
r
o
l
desi
g
n
t
h
eori
es t
o
ad
j
u
s
t
t
h
ese param
e
t
e
rs [2
5]
. In t
h
i
s
paper
,
t
h
e pr
oced
u
r
e t
o
det
e
rm
i
n
e
the weighting
factor is
base
d on t
h
e em
pirical m
e
thod.
In
[26
]
a clear and
o
b
j
ectiv
e m
e
th
od
is
u
s
ed
t
o
select
an a
p
propriate
weight fact
or
range
.
5.
SIMULATIONS
To
v
e
rify th
e flex
i
b
ility o
f
th
e
propo
sed con
t
ro
l techniq
u
e
, a
po
wer circu
it
was si
m
u
lated
in
MATLAB
/SIMULINK. T
h
e system
consists of a t
h
re
e-phase PUC
conve
r
ter,
lo
ad
,
and
con
t
r
o
l un
its
(p
redictive
co
ntr
o
ller).
T
h
e
sam
p
ling fre
q
u
ency
use
d
wi
th
th
e pred
icti
v
e
strateg
y
was 10
kHz.
For stead
y
st
at
e condi
t
i
o
n
i
n
t
h
e sim
u
l
a
ted sy
st
em
, onl
y
238 s
w
i
t
c
hi
n
g
st
at
e vect
ors
am
ong
51
2 p
o
ssi
bl
e st
at
es i
s
use
d
fo
r sy
nt
hesi
zi
n
g
i
nve
rt
er
out
p
u
t
vol
t
a
ge. As
a resul
t
,
l
o
t
s
o
f
swi
t
c
hi
n
g
st
at
e i
n
l
o
o
k
-
up t
a
bl
e can be
rem
ove
d
and t
h
e c
o
m
p
u
t
at
i
onal
bu
rde
n
i
s
decreased
. I
n
ge
neral
,
a s
p
ecific switching state vector c
a
n be rem
ove
d from
l
o
o
k
-
u
p
t
a
bl
e d
e
pen
d
i
n
g o
n
:
1.
Transi
ent
or st
eady
st
at
e operat
i
on m
ode of t
h
e i
nvert
er.
2.
Num
b
er of cont
rol
param
e
t
e
rs t
h
at
i
n
cl
uded i
n
t
h
e cost
funct
i
on.
3.
Val
u
e of wei
ght
i
ng fact
ors.
5.
1.
Steady-State Performance
Figure
4 s
h
ows that the loa
d
currents t
r
acks
refe
re
nce
val
u
es, accurately, and its T
H
D is
less than
1%
.
Negat
i
v
e
cy
cl
e of
p
r
i
n
ci
pal
D
C
-l
i
nk c
u
rre
nt
as sh
o
w
n i
n
Fi
gu
re
4, i
ndi
cat
es t
h
at
bi
di
rect
i
onal
po
we
r s
u
ppl
y
i
s
n
eed
ed
(b
ecause of trin
ary
m
o
d
e
, n
a
m
e
ly
cap
acitor vo
ltag
e
is eq
u
a
l t
o
3
). M
o
re
vol
t
a
ge
lev
e
ls an
d
low THD is ob
tained
in
trin
ary case, bu
t it
is harm
ful for DC
-l
ink p
o
w
er s
u
p
p
ly
(or
battery
)
.
This
pr
o
b
l
e
m
can b
e
rem
oved
by
usi
n
g
2
(bina
r
y
case).
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Predictive Control for
Reduce
d
Struct
ur
e Multilevel Converters … (Adel N
a
zemi)
57
5
Figure 4. Thr
ee-
phase lo
ad
currents
(
p
.
u.
)
,
a
n
d pr
i
n
c
i
pa
l
DC
-
l
i
n
k c
u
r
r
e
nt
(p. u.)
.
F
i
gure 5.
S
a
m
p
li
ng tim
e
effe
ct
on
refe
renc
e
curren
t
tr
acking
of ph
a
s
e a
(per
unit
val
u
e).
Real and im
aginary
parts
of t
h
e pre
d
icted l
o
a
d
curre
nts
(re
d
)
an
d thei
r re
fe
rences
(
b
lue
)
a
r
e sh
o
w
n i
n
Fi
gu
re
5 (
v
ert
i
cal
pl
ot
).
Fi
g
u
r
e
5 i
n
di
cat
es t
h
e sam
p
l
i
ng f
r
e
que
ncy
ef
fect
on
re
fere
nce c
u
r
r
ent
t
r
ac
ki
n
g
.
As i
t
can
be see
n
,
m
o
re sa
m
p
ling fre
quency
res
u
lts in acc
ur
at
e refe
renc
e curre
nt tracki
n
g (beca
use of
na
t
u
re of
p
r
ed
ictiv
e con
t
ro
l), and
low switch
i
ng
frequen
c
y cau
se a
on
e sam
p
le d
e
lay in
cu
rren
t tra
c
king and inc
r
eases
out
put
v
o
l
t
a
ge THD
[
1
6]
.
Real an
d
esti
mated
in
tern
al
vo
ltag
e
v
ect
o
r
of th
e lo
a
d
are s
h
own in Figure 6.
In
tern
al voltag
e
v
ecto
r
o
f
t
h
e lo
ad
is esti
m
a
ted
u
s
in
g
Eq. (2
2)
with
a sam
p
lin
g
ti
m
e
100
. Because of l
o
w
order Eule
r
app
r
oxi
m
a
t
i
on of cu
rr
ent
de
ri
vat
i
v
e an
d ot
h
e
r assum
p
t
i
ons
, a l
o
w pass fi
l
t
er i
s
requi
re
d
fo
r bet
t
e
r est
i
m
at
i
on
o
f
in
tern
al vo
lt
ag
e
vector
of t
h
e loa
d
.
Figure 6. Estimated
and r
eal in
ter
n
al vo
ltag
e
v
ect
or of th
e
load
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-86
94
I
J
PED
S
Vo
l.
7
,
No
.
2
,
Jun
e
2
016
:
56
8-5
82
57
6
Fi
gu
re
7 s
h
o
w
s sy
st
em
vol
t
a
ges
(p
hase
-a)
whe
r
e
13
-l
evel
i
n
l
i
n
e-t
o
-l
i
n
e
v
o
l
t
a
ge ca
n
b
e
seen.
L
o
a
d
an
d
lin
e-to-line v
o
ltag
e
THDs, are 8.6
%
and 7.96%,
res
p
e
c
tively. THD
decreases c
onsi
d
era
b
ly as well as the
in
crem
en
t in
t
h
e nu
m
b
er o
f
o
u
t
p
u
t
vo
ltag
e
lev
e
ls an
d
it's
su
itab
l
e fo
r
po
wer qu
ality
a
p
p
lication
s
. As far as
si
m
u
latio
n
resu
lts ind
i
cate, t
h
e
nu
m
b
er
o
f
v
o
ltag
e
leve
ls
is related to the cap
acit
o
r voltag
e
b
a
lan
c
ing
g
a
in
(
). By i
n
creasi
n
g
or
decreasi
n
g
, t
h
e
n
u
m
b
er
of
o
u
t
p
ut
vo
l
t
a
ge l
e
vel
s
c
o
ul
d
b
e
c
h
a
nge
d.
T
h
i
s
i
ndi
cat
es
t
h
at
t
h
ere i
s
an i
ndi
rect
rel
a
t
i
ons
hi
p bet
w
ee
n m
odul
at
i
on i
nde
x an
d t
h
e c
a
paci
t
o
r
vol
t
a
g
e
bal
a
nci
n
g ga
i
n
(o
r
t
h
e
n
u
m
b
er of out
put
v
o
l
t
a
ge l
e
vel
s
).
Figure 7. Line v
o
ltag
e
(
),
and lo
ad voltage (
).
As
s
h
o
w
n
i
n
Fi
gu
re 8, bef
o
re
0
.
1
, fo
r
0
.
5
, t
h
e pr
o
pose
d
m
e
t
hod
re
gul
at
e ca
paci
t
o
r
vol
t
a
ge
, an
d af
t
e
r
0.
1
with
0
, as expect
e
d
, ca
pac
i
t
o
r v
o
l
t
a
ges
q
u
i
c
kl
y
be
gan t
o
i
n
cre
a
se.
So
,
th
e
propose
d
c
o
ntrol strategy
suc
ceeded in
ba
la
ncing a
uxiliary
DC link ca
pac
itor
voltage
.
Figure 8. Eff
e
cts of
on cap
acitor
voltage balancing.
5.
2.
Perfor
mance During
transient
Dy
nam
i
cs perf
orm
a
nce o
f
t
h
e
FC
S-M
P
C
sc
h
e
m
e
wa
s invest
igated, usi
n
g S
I
MUL
I
NK a
nd com
p
ared
with
a m
o
d
i
fied
Lev
e
l-sh
ifte
d PW
M
(LS-PWM) con
t
ro
l sch
e
m
e
th
at p
r
esen
ted in
[13
]
.
For eval
uation of re
fere
nce c
u
rrent trac
ki
ng, the
re
fe
rence
of
t
h
e l
o
ad c
u
r
r
ent
s
was c
h
ange
d
fr
om
rated cond
itio
n to zero
at tim
e
-
in
stan
t
245
. At
445
, the l
o
ad c
u
rrent
re
fere
nce
was
change
d
back t
o
rat
e
d
val
u
e.
As s
h
o
w
n i
n
Fi
g
u
re
9
and Fi
gu
re
10
, FC
S-M
P
C
ac
hi
eves a
very
f
a
st
refere
nce c
u
r
r
en
t
track
ing
at bo
t
h
cond
itio
n
.
It
tak
e
s less th
an
3
to
d
e
liv
er th
e rated
lo
ad
cu
rren
ts.
On
th
e
oth
e
r
h
a
nd
, t
h
e
LS-PWM sch
e
me p
r
ov
id
es a
slo
w
resp
on
se. Fig
u
re
1
0
shows th
at the lo
ad
curren
ts
o
v
ersh
oo
t
b
e
fo
re
settlin
g
at their steady
state values.
W
i
t
h
LS-PWM technique, t
h
e re
fere
nce c
u
r
r
ent
t
r
ac
ki
n
g
m
i
ght
n
o
t
be
fu
rt
he
r
i
m
p
r
ov
ed b
ecau
s
e
o
f
PI gain
s li
m
i
tatio
n
s
th
at will resu
lt i
n
o
s
cillatio
n
s
i
n
t
h
e lo
ad
curren
t
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Predictive Control for
Reduce
d
Struct
ur
e Multilevel Converters … (Adel N
a
zemi)
57
7
Figure 9. Lo
ad
currents during
th
e power-down
tr
ansient, (
a
) FCS-MPC, and (b)
LS-PWM.
Figure 10. Lo
ad
currents dur
ing
the power-up
transie
n
t, (a
) FCS-MPC,
a
nd (b) LSP-W
M
.
Figure 11. Cap
a
citor
voltages du
ring th
e power-d
ow
n transien
t, (a) FCS-MPC, and (b) LS-PWM.
Figure 12. Cap
a
citor
voltages du
ring th
e power-u
p transient, (a) F
C
S-MPC, and (b
) LS-PWM.
W
i
t
h
the FCS-MPC schem
e
,
the capacitor voltages, as
show
n
i
n
Figu
r
e
11
and
Fi
g
u
r
e
12
, ar
e
w
e
ll
regu
lated
at b
o
th
co
nd
ition
.
After p
o
wer-
up
, th
e cap
acitor
vo
ltag
e
s settle with
i
n
5%
of the ave
r
a
g
e
voltage
value in less than three funda
mental cycles.
Whe
r
eas,
with
th
e LS-PWM sch
e
m
e
, th
e cap
acito
r vo
ltag
e
s sh
ow
larg
e o
s
cillatio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.