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.
4
,
Decem
be
r 202
0
, p
p.
218
3
~
219
3
IS
S
N:
20
88
-
8694
,
DOI: 10
.11
591/
ij
peds
.
v11.i
4
.
pp
218
3
-
219
3
2183
Journ
al h
om
e
page
:
http:
//
ij
pe
ds
.i
aescore
.c
om
Direct
predi
ctive spe
ed cont
ro
l
of
perm
anent m
agnet
synch
ronous mo
tor fed b
y matri
x conv
ert
er
Najmeh
Mo
vahhed Ney
a
1
,
S
ajad
Saberi
2
,
Babak
M
ozaf
ari
3
1,3
Facul
ty
of Elec
trica
l
Eng
ine
er
i
ng,
Sci
ence and Researc
h
Bran
ch
,
Isla
mic
Az
ad
U
nive
rsity
,
T
ehr
a
n,
Ir
an
2
Facul
ty
of Elec
tri
c
al
Engi
n
ee
rin
g,
Babo
l
Nos
hir
vani
Univ
ersit
y
of
Technol
ogy
,
Maz
anda
r
an, Babol,
Ir
an
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
M
a
y
2
, 2
0
20
Re
vised
Jun
10
, 20
20
Accepte
d
J
ul
9
, 20
20
Thi
s
pap
er
pro
poses
a
non
-
c
a
sca
de
-
sing
le
l
o
op
-
Dire
c
t
Sp
ee
d
Cont
rol
al
gorit
h
m
for
s
urfa
ce
moun
te
d
Perma
n
ent
Magne
t
Synchron
ous
Motor
(PM
S
M)
fed
by
Matri
x
Conv
erte
r.
The
proposed
me
thod
uses
Fin
it
e
Control
Set
Mode
l
Predi
ct
iv
e
Control
(F
CS
-
MP
C)
to
m
a
nipul
a
t
e
sys
te
m
spee
d
and
cur
ren
ts
si
mul
t
a
neously.
Also
,
for
better
per
fo
rma
nc
e
of
th
e
pre
di
ct
iv
e
me
thod
,
an
obs
erv
er
designe
d
to
est
im
a
te
m
e
cha
ni
ca
l
torqu
e
and
other
unce
rt
ai
n
par
amete
rs
of
the
m
echani
c
al
subs
ystem
as
a
lum
ped
disturba
nc
e.
Simul
ation
r
esul
ts
us
ing
Mat
la
b
/Sim
uli
nk
dem
o
nstrat
e
the
per
f
orma
nc
e
of
proposed
al
gor
ithm.
Ke
yw
or
d
s
:
Disturba
nce
O
bs
er
ve
r
FCS
-
M
PC
M
at
rix
Co
nver
te
r
PM
S
M d
rive
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
:
Corresp
o
nd
i
ng Auth
or,
Faculty
of Elec
tric
al
Engineer
ing
,
Scie
nce a
nd R
esearch
Bra
nch, I
sla
mic
Aza
d Un
i
ver
sit
y,
Dan
e
shga
h
Bl
vd, Si
mon B
uliv
ar
Bl
vd, T
eh
ra
n,
Ir
a
n.
Emai
l:
movahh
edn
e
ya
@gmai
l.com
1.
INTROD
U
CTION
In
rece
nt
ye
ars
Per
mane
nt
M
a
gn
et
Sync
hro
nous
M
ot
or
s
(PM
S
M
)
are
gett
ing
some
real
a
tt
ention
due
to
their
hi
gh
-
powe
r
de
ns
it
y,
fast
dyna
mi
c
respo
ns
e,
hig
h
e
ff
ic
ie
nc
y
and
ca
pa
bili
ty
of
work
i
ng
without
gearb
ox.
T
hes
e
cha
racteri
sti
cs
ma
kes
P
M
S
M
a
str
ong
ca
nd
i
dat
e
f
or
dif
fer
e
nt
a
pp
li
cat
ion
s
li
ke
wi
nd
energ
y
sy
ste
ms
,
el
ect
r
ic
ve
hicle
s
an
d
industrial
e
quipme
nt
[
1,
2].
M
at
rix
Co
nver
te
r
(
M
C)
is
a
ki
nd
of
c
onve
rter
tha
t
can
c
onne
ct
a
n
AC
dev
ic
e
di
r
ect
ly
to
a
thr
ee
phase
AC
s
ou
rce
by
us
in
g
9
bi
directi
onal
s
witc
hes
[
3].
U
nlike
conve
ntion
al
ba
ck
to
bac
k
ca
pacit
or
-
ba
sed
c
onve
r
te
rs,
M
C
s
achieve
ac
-
ac
co
nv
e
rsion
without
ene
rgy
st
or
a
ge
li
nk
, w
hich hel
ps
t
hem
t
o have
more
reli
abili
ty,
c
ompact
siz
e an
d
lo
nger
li
f
et
ime [3
-
6].
A
t
yp
ic
al
P
MSM
dri
ve
us
in
g
matri
x
c
onve
rter
with
th
re
e
-
phase
s
ource
,
in
put
filt
er
a
nd
m
otor
is
represe
nted
i
n
Figure
1.
It
ca
n
be
see
n
that
every
in
pu
t
phase
can
co
nnec
t
to
an
outp
ut
ph
a
se
us
i
ng
a
switc
h.
To
filt
er
input
curre
nts
an
d
avo
i
d
vo
lt
age
sp
ikes
duri
ng
switc
hing,
M
C
s
nee
d
a
n
LC
filt
er
com
bin
e
d
wit
h
par
asi
ti
c
resist
or.
More
detai
l
on
M
C
an
d
filt
er
will
b
e
giv
e
n
i
n
m
odel
in
g
sect
ion
.
T
her
e
are
diff
e
re
nt
c
on
t
rol
al
gorithms
f
or
P
M
S
M
s
bu
t
Fiel
d
Or
ie
nted
Co
ntr
ol
(FOC
)
a
nd
Direct
T
orq
ue
C
ontrol
(D
TC
)
are
the
m
os
t
popula
r
meth
ods
[
7
-
8].
I
ntera
ct
ion
bet
ween
co
nt
ro
l
va
riabl
es,
wind
up
pro
blem,
band
widt
h
li
mit
at
ion
d
ue
t
o
casca
de
struct
ur
e
an
d
mod
ulati
on
sta
ge,
m
akes
FO
C
a
non
-
ideal
c
ontr
oller
c
ho
ic
e
[
9].
DTC
us
es
a
look
up
ta
ble
ba
sed
on
switc
hing
sta
te
s
to
co
ntr
ol
mo
to
r
tor
que
directl
y.
Ou
t
put
vo
lt
ag
e
vec
tors
are
no
t
al
way
s
op
ti
mal,
as a c
on
s
eq
ue
nce
D
TC su
ff
e
rs fr
om
hi
gh to
rque
and stat
or f
l
ux
rip
ples [1
0,
11]
.
Simple
treat
m
ent
of
c
onstrai
nts,
Mult
ivaria
ble
struct
ur
e
a
nd
good
pe
rfo
rma
nce
in
the
pr
ese
nce
of
nonlinea
riti
es
make
M
PC
on
e
of
t
he
be
st
cho
ic
es
for
dr
i
ve
sy
ste
ms
[
12
-
13].
T
he
re
are
two
main
cat
e
gories
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
:
218
3
–
219
3
2184
for
M
PC
,
C
onti
nu
ous
Co
ntr
ol
Set
M
PC
(C
CS
-
MPC
)
an
d
Finit
e
C
ontrol
Set
M
PC
(
FC
S
-
M
PC)
[
9,
14
-
16].
CCS
-
MPC
nee
ds
a
mod
ulato
r
to
ge
ner
at
e
ga
te
sign
al
s,
bu
t
FCS
-
M
PC
ta
kes
a
dv
a
ntag
e
of
t
he
fact
t
ha
t
the
numb
e
r
of
po
s
sible
switc
hi
ng
sta
te
s
is
li
mit
e
d,
s
o
it
us
e
s
th
e
mode
l
of
the
sy
ste
m
t
o
pre
dict
nex
t
sta
te
of
it
fo
r
each
possible
c
on
t
ro
l
act
io
n
a
nd
be
st
c
on
tr
ol
act
ion
will
be
chosen
by
min
imi
zi
ng
a
co
st
functi
on
[17].
Using
cost fu
nction h
el
ps
FCS
-
M
PC
to direct
ly c
on
trol m
or
e
tha
n on
e
object
ive a
t t
he
same
ti
me.
Ca
scade
st
ru
ct
ur
e
is
a
co
mm
on
strat
e
gy
f
or
s
pee
d/posit
io
n
c
ontrol
of
P
M
S
M
s,
bu
t
co
mp
a
rin
g
to
current
dy
nam
ic
,
mec
han
ic
al
dyna
mic
is
sl
uggis
h
a
nd
thi
s
dif
fer
e
nce
be
tween
ti
me
c
on
sta
nts
e
nforce
s
the
desig
ner
to
c
onside
r
lo
nger
predict
io
n
horiz
on
f
or
D
PSC
[
18].
More
o
ver,
el
imi
nation
of
oute
r
l
oop
c
on
trolle
r
no
t
on
l
y
bri
ng
s
sta
bili
ty
issu
es
bu
t
al
s
o
dec
rease
syst
em
pe
rformance
a
nd
cause
s
ste
ad
y
sta
te
error
[19].
I
n
[20
-
22]
some
methods
pro
po
sed
t
o
c
on
t
ro
l
sp
ee
d
of
a
P
M
S
M
fe
d
by
a
two
-
le
vel
ba
ck
to
back
co
nv
e
rter
direct
ly.
T
his topolog
y
el
imi
na
te
s both o
uter l
inear c
on
t
ro
ll
e
r
a
nd m
odulato
r.
In
this
pa
per w
e
pro
pose
a d
ir
ect
predict
ive
s
peed
c
on
t
ro
l
f
or
P
MSM
fe
d
by matrix
c
onve
rter.
A
n
ew
cost
f
un
ct
io
n
i
ntr
oduce
d
by
c
ombinin
g
s
pee
d
dy
namic,
c
urren
t
dynamic
a
nd
s
ys
te
m
c
on
s
trai
nts
to
have
high
performa
nce
w
it
ho
ut
ca
scade
structu
re.
To
ta
ckle
sta
bili
ty
is
su
e
a
nd
hi
gh
c
urren
t
disto
rtio
n,
due
to
dif
fere
nce
betwee
n
mech
anical
an
d
el
ect
rical
ti
me
c
on
sta
nts,
a
c
urre
nt
re
fe
ren
c
e
desi
gn
e
d
ba
s
ed
on
sli
di
ng
m
od
e
con
ce
pt
is
a
dded
t
o
co
st
f
unc
ti
on
.
T
his
te
r
m
gua
ran
te
es
oute
r
lo
op
sta
bili
ty
a
nd
decr
ea
s
e
cu
rr
e
nt
disto
rtion.
Also
,
to
hav
e
lowe
r
ste
a
dy
s
ta
te
error
a
nd
bette
r
perf
or
m
ance,
a
distu
rbance
obser
ve
r
desig
ne
d
t
o
es
ti
mate
load
t
orq
ue value a
nd o
t
her u
ncer
ta
inti
es
of
the mec
ha
nical
s
ubs
ys
te
m as
a lum
ped d
ist
urba
nce.
The
rest
of
th
is
pa
per
is
or
gan
iz
e
d
as
f
ollow
s:
I
n
sect
ion
2
a
model
of
the
s
ys
te
m
a
nd
matri
x
conve
rter
is
presented
.
I
n
sec
ti
on
3
direct
predict
ive
s
pee
d
co
ntr
ol
cost
f
un
ct
io
n
desc
ribed
an
d
i
n
sec
ti
on
4
sta
bili
ty
pro
of
f
o
r
cu
rr
e
nt
r
efere
nce
par
t
and
obse
rv
e
r
is
pro
vid
e
d.
I
n
sect
io
n
5
simulat
ion
re
su
l
ts
ar
e
discusse
d t
o de
monstrate
t
he per
forma
nce
of the c
on
t
ro
ll
er
.
S
ua
S
va
S
wa
S
ub
S
vb
S
wb
S
uc
S
vc
S
wc
C
f
R
f
L
f
V
u
V
v
V
w
N
I
np
ut
Fi
l
t
e
r
M
a
t
ri
x
C
o
n
v
e
rt
e
r
B
i
o
p
o
l
a
r
S
w
i
t
c
h
P
M
S
M
V
a
V
b
V
c
i
a
i
b
i
c
V
eu
V
ev
V
ew
i
eu
i
ev
i
ew
i
u
i
v
i
w
Figure
1
.
Sc
he
mati
c o
f
a
mat
r
ix con
ver
te
r
dr
ive for
P
M
S
M
2.
SY
STE
M MO
DEL
As
it
was
me
nt
ion
e
d,
al
l
MP
C
based
c
ontr
ol
al
go
rith
ms
us
es
sy
ste
m
m
od
el
to
pr
e
dict
ne
xt
sta
te
,
so
sy
ste
m
model
has
si
gn
i
ficant
impact
on
this
al
gorithm
.
In
the
f
ollo
wing
s
ect
ion
t
he
m
od
el
s
for
M
C,
P
M
S
M
and in
pu
t
filt
er
are desc
ribe
d.
2.1.
M
atri
x
co
nv
er
ter m
od
el
Figure
1
s
how
s
a
33
M
C
with
9
bid
irect
io
nal
switc
hes
.
Eac
h
s
witc
h
is
c
omp
os
e
d
of
t
wo
powe
r
transisto
rs
a
nd
two
pa
rall
el
di
od
es
.
T
he
MC
is
connecte
d
to
the
t
hr
ee
-
phase
s
ource
t
hro
ugh
t
he
in
pu
t
filt
er
that
it
s
in
du
ct
a
nce
i
s
,
it
s
ca
pacit
ance
is
and
par
asi
ti
c
re
sist
an
ce
is
.
T
his
filt
er
will
el
imi
nate
hi
g
h
fr
e
qu
e
nc
y
harmo
nics
in
in
put
cu
rr
e
nts
(
,
,
)
an
d
a
vo
i
d
vo
lt
a
ge
sp
i
kes.
The
mathemat
ic
al
r
el
at
ion
sh
i
p
betwee
n
in
put
and
outp
ut
vo
lt
age a
nd curre
nt
o
f
M
C
are [
23]:
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
Direct
p
re
dicti
ve spee
d
c
ontrol of
per
m
an
e
nt
mag
net sy
nchrono
u
s
m
oto
r f
ed by
…
(
N
ajme
h
M
ov
ahhe
d N
eya)
2185
[
]
⏟
=
[
]
⏟
[
]
⏟
,
[
]
⏟
=
[
]
⏟
[
]
⏟
(1)
Wh
e
re
an
d
a
re
insta
ntane
ous
tra
nsfer
mat
rix
a
nd
it
s
tra
nspose
re
sp
ect
i
vel
y
a
nd
=
{
,
,
}
,
=
{
,
,
}
.
=
1
if
t
he
switc
h t
hat
co
nnect
s
ℎ
i
nput
t
o
ℎ
ou
t
pu
t
p
ha
se
is
O
N
an
d
=
0
if
the
s
witc
h
is
OF
F
.
Du
e
t
o
the
i
nductive
natur
e
of
the
loa
d,
sud
den
i
nter
rupt
in
cu
rr
e
nt
woul
d
cau
se
ove
rvo
lt
age
tha
t
ca
n
destr
oy
the
c
omp
on
e
nt.
Al
so,
switc
he
s
s
hould
not
short
c
ircuit
two
in
put
ph
ase
s;
beca
us
e
of
t
his
rest
rict
ion
,
fo
ll
owin
g
e
qua
ti
on
s
houl
d
al
way
s
be sat
isfi
ed wh
ic
h red
uc
es the
num
ber
of possible
sw
i
tc
hin
gs t
o 2
7.
+
+
=
1
∀
∈
{
,
,
}
(2)
2.2.
Inpu
t
filter
m
od
el
Inp
ut
filt
er
s
hown
in
Fi
gure
1
is
j
us
t
li
ke
a
n
RLC
ci
rc
uit,
a
nd
ca
n
be
desc
ribe
d
by
t
he
f
ol
lowing
sta
te
sp
ace
model [
23]:
̇
(
)
=
[
0
1
−
1
−
]
⏟
(
)
+
[
0
−
1
1
0
]
⏟
(
)
(3)
Wh
e
re:
(
)
=
[
(
)
(
)
]
,
(
)
=
[
(
)
(
)
]
(
)
=
2
3
(
+
+
2
)
(
)
=
2
3
(
+
+
2
)
(4)
Wh
e
re
=
2
3
.
To u
se
in FCS
-
MPC
, a discre
te
model f
or
i
nput
filt
er is n
ee
ded. Co
ns
i
der
i
ng a sam
ple ti
me
(5)
descr
i
bes
t
his
model [
24].
[
+
1
]
=
[
]
+
[
]
=
,
=
∫
(
−
)
0
(5)
To pr
e
dict
[
+
1
]
one
can use
equati
on (5) o
r
,
an
d
e
qu
at
io
n (
6)
.
[
+
1
]
=
(
2
,
1
)
[
]
+
(
2
,
2
)
[
]
+
(
2
,
1
)
[
]
+
(
2
,
2
)
[
]
(6)
Using
eq
uatio
n
(6),
a
te
rm
with
relat
ion
to
i
s
can
be
a
dded
t
o
obje
ct
ive
f
unct
ion
an
d
c
on
tr
ol
us
in
g
FCS
-
M
PC.
2.3.
PMS
M
m
od
el
The d
yn
a
mic
model f
or a P
M
S
M
i
n
the
dq
re
fer
e
nce
fr
a
me
can
be descri
be
d
as
foll
ows:
=
1
(
−
+
)
=
1
(
−
−
−
)
=
(
3
2
2
−
+
)
(7)
Wh
e
reas
s
R
is
the
st
at
or
resist
a
nce,
=
=
are
d
a
nd
q
axis
in
duct
ances
of
t
he
mo
t
or,
,
,
,
are
nu
mb
e
r
of
pole
pairs
,
mot
or
i
ner
ti
a,
f
rict
ion
coe
ff
ic
ie
nt
and
fl
ux
li
nka
ge
of
t
he
pe
rma
nen
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.
4
,
D
ecembe
r
2020
:
218
3
–
219
3
2186
mag
net
mo
t
or
resp
ect
ivel
y
a
nd
d
is
a
l
umpe
d
sum
distu
rbance
f
or
loa
d
t
orq
ue
a
nd
ot
he
r
uncertai
ntie
s
in
mecha
nical
subsyst
em
=
(
,
,
)
.
The
value
of
has
im
porta
nt
impa
ct
on
c
ontr
oller
pe
rforma
nce,
so
it
will
be
est
imat
ed
us
in
g
an
obse
rv
e
r
in
sect
io
n
5.
A
lso
an
d
are
volt
age
vect
or
s
in
ref
e
re
nce
fr
ame
that a
re
r
el
at
ed
to s
witc
hing
vo
lt
ag
es a
s Figure
2.
C
l
ar
k
e
Tr
an
sf
o
r
m
at
i
o
n
11
1
-
-
2
22
D=
3
33
0-
22
(
)
(
)
(
)
(
)
c
os
θ
-
si
n
θ
M=
si
n
θ
c
os
θ
P
ar
k
Tr
an
sf
o
r
m
at
i
o
n
V
a
V
b
V
c
V
α
V
β
θ
V
d
V
q
Figure
1
.
Par
k
-
Cl
ark
e tra
nsfo
r
mati
on
f
or
dq a
xis
vo
lt
ages
To
pr
e
dict
f
u
tu
re
values
of
c
urre
nt
an
d
sp
ee
d,
a
disc
rete
m
od
el
of
P
M
S
M
is
nee
de
d
f
or
FCS
-
M
PC.
To
re
du
ce
cal
culat
ion
bur
de
n,
a
sim
ple
discreti
za
ti
on
m
et
hod
with
go
od
performa
nc
e
sho
uld
be
se
le
ct
ed.
Her
e
the
mode
l discreti
zed
u
s
ing
f
orward E
ul
er m
et
ho
d
[
23
]
as
f
ollow, i
s
us
e
d,
whereas
is sampli
ng ti
me.
[
+
1
]
=
(
1
−
)
[
]
+
[
]
+
[
]
[
]
[
+
1
]
=
(
1
−
)
[
]
+
[
]
−
[
]
[
]
−
[
]
[
+
1
]
=
(
1
−
)
[
]
+
(
3
2
2
[
]
+
)
(8)
3.
COS
T
FU
N
C
TION
SEL
EC
TION
FCS
-
M
PC
ta
ke
s
a
dv
a
ntage
of
the
in
her
e
nt
discrete
f
orm
of
the
po
w
er
c
onve
rters.
T
he
c
ontro
l
ob
je
ct
ives
will
be
predict
ed
for
finite
num
ber
of
acce
pta
ble
switc
hi
ng
sta
te
s
of
the
powe
r
co
nverte
r.
T
he
pr
e
dicte
d
va
ria
bles
will
be
c
ompa
red
with
t
he
ir
re
fer
e
nce
v
al
ues
t
hroug
h
a
co
st
f
unct
ion.
T
he
switc
hi
ng
sta
te
that
minimi
z
es
the
c
os
t
f
unct
ion
will
be
ap
pl
ie
d
to
t
he
c
onve
rter
to
exe
rt
the
vo
lt
age
vect
or
t
o
the
loa
d
i
n
the
nex
t c
ontr
ol int
erv
al
,
there
f
or
e
the
pr
ese
nce
of a
mod
ulator
is not re
quire
d
[
16].
Accor
ding
to
t
he
FCS
-
MPC
scheme
c
os
t
f
unct
ion
desi
gn
i
s
a
key
point.
By
pro
pe
r
sel
ect
ion
of
c
os
t
functi
on,
FCS
-
M
PC
is
a
ble
to
co
ntr
ol
mu
lt
iple
ob
je
ct
ives
at
t
he
same
ti
me.
H
oweve
r,
prop
e
r
sel
ect
ion
of
it
ems and
weig
htings is a c
hal
le
ng
in
g
ta
s
k. I
n
this
sec
ti
on it
ems
of
c
os
t
fun
ct
ion
a
re
discu
ssed.
3.1.
PMS
M
c
ost t
e
rm
The
main
goa
l
of
the
co
ntr
oller
is
t
o
c
ontr
ol
the
sp
ee
d
of
P
M
S
M
so
that
t
he
e
rror
betwe
en
pr
e
dicte
d
sp
ee
d
a
nd mea
su
re
d
s
pee
d
c
ould
be
a
can
di
da
te
f
or s
pee
d
c
os
t.
=
(
∗
−
)
2
(9)
Wh
e
re
is
pr
e
dicte
d
value
a
nd
∗
is
the
refe
ren
ce
val
ue
of
sp
ee
d.
I
n
c
on
t
rast
to
el
ect
rical
su
bsyste
m,
me
chan
ic
al
s
ub
s
yst
em
dyna
mic
is
slug
gis
h.
So
f
or
bette
r
performa
nce
M
PC
need
s
a
longe
r
horizo
n
t
o
dec
rease
c
urren
t
a
nd
to
rque
distor
ti
ons
w
hen
s
peed
er
ro
r
bec
om
es
small
[
18]
.
We
a
dd
a
current
te
rm
to
s
pee
d
cost
f
un
ct
i
on
t
o
not
on
l
y
el
imi
nate
lo
ng
e
r
horizo
n
necess
it
y,
but
al
so
ta
ke
care
of
out
er
lo
op
sta
bili
ty and c
urre
nt d
ist
or
ti
on
. So the
trac
king
pa
rt
of co
st f
un
ct
io
n sh
ould
b
e a
s foll
ow:
=
(
∗
[
]
−
[
+
1
]
)
2
+
(
∗
[
]
−
[
+
1
]
)
(10)
∗
=
1
(
̇
∗
+
−
̂
+
+
(
)
)
(11)
Wh
e
re
̂
is t
he e
sti
mati
on
of d
i
sturbance
and
a, b an
d
(
)
are
as
foll
ow
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
Direct
p
re
dicti
ve spee
d
c
ontrol of
per
m
an
e
nt
mag
net sy
nchrono
u
s
m
oto
r f
ed by
…
(
N
ajme
h
M
ov
ahhe
d N
eya)
2187
=
3
2
2
,
=
&
(
)
=
{
+
1
>
0
−
1
<
0
(12)
Stabil
it
y
pro
of
for
∗
in
pr
ese
nc
e of
̂
is disc
us
s
ed
in
secti
on
4.
3.2.
Constr
aints c
os
t
t
erm
Ther
e
are
tw
o
te
rms
t
hat
s
h
ou
l
d
be
c
onsidere
d,
one
for
cu
rr
e
nt
li
mit
at
ion
a
nd
an
ot
her
one
for
M
a
ximum T
orqu
e
Pe
r Am
pere (
M
TP
A) crit
eria [
24]:
(
)
(
)
2
2
2
2
m
a
x
m
a
x
2
2
2
2
0.
dq
ii
L
d
q
d
q
L
LL
dq
Z
id
d
d
q
Z
id
d
i
i
I
i
i
I
C
ow
LL
C
i
i
i
C
i
=
+
−
+
=
−
=
+
−
⎯
⎯
⎯
→
=
(13)
3.3.
Inpu
t
filter
cost ter
m
Wh
e
n
we
are
con
t
ro
ll
in
g
s
peed
of
P
M
S
M
usi
ng
di
rec
t
M
C,
co
ntr
ol
of
i
nput
cu
rrent
is
a
ve
ry
chall
eng
i
ng
ta
s
k
e
ven
by
FC
S
-
M
PC,
but
if
one
need
s
to
con
t
ro
l
i
nput
c
urren
t
as
pri
m
ary
c
ontr
ol
obj
ect
ive,
li
ke
unit
powe
r
fact
or
in
ge
ne
rati
ve
mode,
a
reacti
ve
powe
r
te
rm
s
hould
be
ad
de
d
t
o
t
he
cost
functi
on.
Using
equ
at
io
n (
6)
f
or
gr
id
curre
nt, r
eac
ti
ve
pow
er
would be
calc
ul
at
ed
us
i
ng foll
ow
i
ng equati
on:
=
−
(14)
Wh
e
re
and
a
re
real
a
nd
im
agina
ry
pa
rt
of
v
s
and
i
s
in
equ
at
ion
(4).
S
o
by
add
i
ng
i
Z
C
an
d
reacti
ve
te
rm
to
sin
gle
te
rm
we
will
ha
ve
a
te
rm
cal
le
d
ze
ro
te
rm
that
m
eans
this
pa
rt
of
co
st
f
un
ct
io
n
sh
oul
d
go to
ze
r
o.
=
(
)
2
+
(
)
2
(15)
As
i
n
this
w
ork,
our
pri
mar
y t
ask
c
on
t
ro
l
is
sp
ee
d
of
P
M
S
M
,
Z
C
do
e
s n
ot have r
eact
ive po
wer
te
rm
.
By
c
ombini
ng
T
C
as
t
rack
i
ng
t
erm,
L
C
as
li
mit
at
ion
te
rm
an
d
Z
C
as
zer
o
te
rm
t
he
c
os
t
f
unct
io
n
will
be
ready
for FC
S
-
M
PC.
=
+
+
(16)
It
is
w
or
t
h
me
ntion
i
ng
t
hat
f
ind
in
g
opti
mal
wei
gh
ti
ng
fac
tors
w
h
e
n
t
here
are
c
onstrai
nts
i
n
c
ost
functi
on
is
not a
strai
gh
t
ta
s
k.
Fo
r
this
w
ork
we
use
d
mu
lt
ip
le
simulat
ion
s
t
o
fi
nd
w
ei
ghti
ng
fact
or
s w
it
h
good
performa
nce
of
the s
ys
te
m.
4.
DIS
T
UR
B
ANCE OBSE
RVE
R
A
N
D CURRE
NT
REF
EREN
CE ST
ABILIT
Y
Disturba
nces
and
uncertai
nt
ie
s
hav
e
high
impact
on
F
CS
-
MPC
pe
rformance
s
o
a
n
accu
rate
disturba
nce
ob
serv
e
r
ca
n
imp
rove
prof
ic
ie
nc
y
of
the
s
ys
te
m.
I
n
this
sect
i
on
a
distu
rb
a
nc
e
obser
ver
de
sign
e
d
to
est
imat
e loa
d
to
r
qu
e
v
al
ue a
nd o
t
her u
nce
rtai
nties i
n me
chan
ic
al
s
ubsyst
em as a l
ump
ed
s
um distu
r
ba
nce.
Let
the moto
r
c
urren
ts
and s
pe
ed be mea
sura
ble and c
onsid
er m
ec
ha
nical
p
art
of (7)
as
f
al
low:
̇
=
−
+
,
=
3
2
2
,
=
=
̇
+
−
(17)
The desig
n p
rocedure
of the
di
sturb
a
nce
obs
erv
e
r
is as
foll
ow
:
Def
i
ne
the
d
ist
urba
nce
dynam
ic
as:
̂
̇
=
(
−
̂
)
(18)
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
:
218
3
–
219
3
2188
Wh
e
re
K
is
a
po
sit
ive
co
ns
ta
nt.
T
o
hav
e
gl
ob
al
ly
sta
ble
e
sti
mati
on
with
ou
t
ste
a
dy
sta
te
error,
a
n
aux
il
ia
r
y varia
b
le
is
def
ine
d l
ike the
one i
ntr
oduce
d
in
[2
5]
:
=
̂
−
(19)
By
diff
e
ren
ti
at
ing (
19)
a
nd
repl
aci
ng
(17) an
d (
18)
to
n
e
w
e
quat
ion we
h
a
ve
:
̇
=
̂
̇
−
̇
=
(
−
̂
)
−
̇
=
(
̇
+
−
−
̇
)
−
̂
=
(
−
)
−
̂
(20)
So
distu
rb
a
nce
observe
r
ca
n b
e d
esi
gn
e
d usi
ng equati
on
(21)
:
{
̇
=
(
−
)
−
̂
̂
=
+
(21)
The
or
em
1.
U
sing
the
obser
ver
desi
gn
e
d
with
(21)
an
d
from
(
11),
th
e
sp
ee
d
e
rror,
def
i
ned
a
s
̃
=
∗
−
,
a
nd
t
he
er
ror
of
dist
urba
nc
e,
def
i
ned
as
̃
=
−
̂
,
will
c
onve
r
ge
to
ze
ro
as
ymptoti
cal
ly.
Pr
oo
f.
C
onside
r
the
foll
ow
i
n
g Lya
puno
v
f
un
ct
ion
ca
nd
i
date
:
=
1
2
(
̃
2
+
̃
2
)
(22)
The deri
vatio
n of
V
resp
ect
to
ti
me is:
̇
=
̃
̃
̇
+
̃
̃
̇
(23)
By
s
ub
sti
tuti
ng
(17), (
18)
a
nd
(21) into
(2
3)
,
it
can
be
writ
te
n
that:
̇
=
̃
(
̇
∗
−
+
−
)
+
̃
(
̇
−
̂
̇
)
=
̃
(
̇
∗
−
+
−
)
+
̃
(
̇
−
̇
−
̇
)
(24)
Using
from
(
11),
̇
from
(
20)
an
d
co
ns
i
der
i
ng
t
his
fact
t
ha
t
in
pract
ic
al
eng
i
neer
i
ng
w
e
al
way
s
consi
der
̇
=
0
[26]
we have:
̇
=
̃
(
−
̃
−
(
̃
)
−
(
−
̂
)
)
+
̃
(
−
(
̇
−
)
+
̂
−
̇
)
=
−
̃
2
−
|
̃
|
−
̃
+
̃
(
−
(
̇
+
̇
−
)
⏟
+
̂
)
=
−
̃
2
−
|
̃
|
−
̃
̃
−
̃
2
≤
−
̃
2
−
̃
2
−
|
̃
|
+
|
̃
|
|
̃
|
(25)
Using
obse
r
ver
error
dyna
mic
̃
̇
=
−
̃
,
we
know
that
the
obser
ve
r
is
asym
pto
ti
cal
ly
sta
ble
a
nd
w
e
can ass
um
e
|
̃
|
≤
for
s
ur
e
.
S
o,
t
he f
ollow
i
ng ine
qual
it
y
for deri
va
ti
on
of Ly
a
punov f
unct
ion i
s s
at
is
fied:
̇
≤
−
̃
2
−
̃
2
−
|
̃
|
+
|
̃
|
=
−
̃
2
−
̃
2
−
(
−
)
|
̃
|
(26)
If
we
c
hoose
a
new v
a
riable a
s
̄
=
−
a
nd
>
:
̇
≤
−
̃
2
−
̃
2
−
̄
|
̃
|
̇
≤
−
̄
(
1
2
̃
2
+
1
2
̃
2
)
=
−
̄
(27)
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
Direct
p
re
dicti
ve spee
d
c
ontrol of
per
m
an
e
nt
mag
net sy
nchrono
u
s
m
oto
r f
ed by
…
(
N
ajme
h
M
ov
ahhe
d N
eya)
2189
Wh
e
re
̄
=
2
{
,
}
.
Using t
he
le
m
ma at [
26]
t
he soluti
on
of
̇
≤
−
̄
is:
(
)
≤
−
̄
(
−
0
)
(
0
)
(28)
So
the
sli
di
ng
mode
surface
conve
rg
e
to
ze
ro
e
xponentia
ll
y
a
nd
by
co
nv
erg
i
ng
V
to
ze
ro,
e
rror
of
sp
ee
d
a
nd d
ist
urba
nce esti
ma
ti
on
go
e
s to
zer
o
a
nd the
pro
of is co
mp
le
te
d.
5.
SIMULATI
O
N AND
RES
U
LT
S
To
pro
ve
t
he
eff
e
ct
iven
es
s
of
the
propo
sed
model,
a
P
M
S
M
fed
by
M
C
is
si
mu
la
te
d
us
i
ng
M
A
TLAB/Si
m
ulink.
Pa
ramet
ers
of
the
P
MSM
a
re
li
ste
d
in
Table
1.
Al
so
Fi
gure
3
s
hows
the
flo
wc
har
t
of
pro
po
se
d
met
hod
base
d
on
F
CS
-
MPC
an
d
V(n)
is
the
n
th
vecto
r
of
M
C
vo
lt
age
i
n
,
or
ref
e
re
nce
fr
ame
whe
n
it
i
s n
ee
de
d.
To
sho
w
t
he
e
ff
ect
ive
ness
of
the
pro
pose
d
method,
tw
o
ot
her
dif
fer
e
nt
s
peed
c
on
t
ro
l
s
chemes
are
us
e
d
in
sim
ulat
ion
.
First
sc
he
me
is
dir
ect
pr
edict
ive
s
peed
con
t
ro
l
(P
SC
)
with
out
c
urrent
te
rm
that
in
t
r
oduce
d
in [18] for
bac
k
to
back co
nv
erter and
oth
e
r scheme is a c
ontr
ol sy
ste
m
usi
ng
P
I wit
h
ant
i
-
wind
up
st
ru
ct
ur
e as
ou
te
r
lo
op c
on
t
ro
ll
er a
nd FC
S
-
MPC
as i
nn
e
r l
oop
c
urren
t c
ontr
oller.
Rem
ark
:
O
ne
of
t
he
dr
a
wbac
ks
of
FCS
-
M
P
C
is
t
he
ti
me
of
opti
miza
ti
on
procedu
re.
E
valuati
on
of
c
ost
functi
on
f
or
27
dif
fer
e
nt
a
vaila
ble
volt
ages
is
ti
me
co
nsumi
ng.
Also,
in
some
cases
m
or
e
than
27
cal
cul
at
ions
are
require
d.
T
he
meth
od
i
n
[
18]
us
e
d
th
ree
ste
p
horizo
ns
f
or
lo
wer
t
orque
osc
il
la
ti
on
a
nd
c
urren
t
distor
ti
on,
that
it
means
three
ti
mes
pr
e
dicti
on
for
each
volt
age
vecto
r
are
r
equ
i
red,
so
t
he
method
nee
ds
2
7
3
cal
culat
ion
s.
I
f
one
nee
ds
to
us
e
l
ow
e
r
num
ber
of
volt
age
vecto
rs
i
n
eac
h
sa
mp
li
ng
int
erv
al
,
the
re
is
so
me
methods
for
ba
ck
to
bac
k
c
onver
t
or
s
[27]
a
nd
matri
x
c
onve
rters
[
28].
In
pro
posed
meth
od
we
us
e
d
j
us
t
one
ste
p
to
pr
e
dict
fu
t
ur
e
values
of
var
ia
bles
a
nd
nee
d
27
ca
lc
ulati
on
of
c
ost
functi
on
ho
wev
e
r
we
co
ul
d
use
method i
n [28] t
o hav
e
ev
e
n
l
ow
e
r
cal
c
ulati
on
bur
den.
Table
1
.
Para
m
et
er
s of sim
ula
ti
on
s
ys
te
m
Para
m
eter
Valu
e
Un
it
Para
m
eter
Valu
e
Un
it
Stato
r
resistan
ce
R
s
1
Visco
u
s d
am
p
in
g
B
v
0
.00
9
3
N
m
.
s
Stato
r
in
d
u
ctan
ce
L
s
,L
d
,L
q
3
.2
mH
Filter
resistan
ce R
f
1
Nu
m
b
er
o
f
po
le pairs Z
p
4
Filter
in
d
u
ctan
ce L
f
m
H
4
Flu
x
link
ag
e ψ
mg
0
.12
6
web
Grid v
o
ltag
e/Freq
V/Hz
1
0
0
/5
0
Moment o
f
inertia
J
m
0
.12
6
g
r/m
2
I
m
ax
A
7
S
t
a
r
t
M
e
a
su
r
e
i
s
,
v
s
,
ω
e
E
sti
m
a
te
d
u
si
n
g
(
19
)
C
a
l
c
u
l
a
t
e
i
q
R
ef
er
en
c
e
u
si
n
g
(
11
)
F
o
r
n
=
1
:
27
I
n
i
ti
a
l
i
z
e
g
o
p
t
=
i
n
f
&
x
opt
=
0
P
r
ed
i
c
t
i
d
,
i
q
,
ω
e
a
n
d
a
n
d
Q
u
si
n
g
(
8
)
&
(
14
)
C
a
l
c
u
l
a
te
g
u
si
n
g
(
10
)
,
(
13
)
,
(
15
)
,
(
16
)
I
f
g
<
g
opt
g
opt
=
g
&
x
opt
=
n
n
=
n
+
1
S
e
l
e
c
t
V
(
n
)
I
f
n
<
=
27
T
r
u
e
F
a
l
s
e
n
=
n
+
1
T
r
u
e
F
a
l
s
e
A
p
p
l
y
V
(
x
opt
)
Figure
2
.
Flo
w
char
t
of FCS
-
M
PC
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
:
218
3
–
219
3
2190
Figure
4
il
lust
rates
re
fer
e
nce
value
s
of
sp
e
ed
a
nd
loa
d
t
orq
ue
d
ur
in
g
simulat
ion.
Also
est
imat
ed
value
of
loa
d
t
orq
ue
usi
ng
disturbance
obse
rv
e
r
is
s
how
n
and
it
can
be
s
een
that
t
he
observ
e
r
has
very
goo
d
performa
nce
duri
ng step
c
ha
nges
of loa
d
to
r
qu
e
and
ste
ad
y st
at
e error i
s z
ero.
Figure
3
: R
efe
r
ence s
pee
d
,
act
ual
an
d
e
sti
mate
d value
of loa
d
to
r
qu
e
Figure
5
s
how
s
re
fer
e
nce
s
pe
ed
a
nd
P
M
S
M
s
peed
us
in
g
3
di
ff
e
ren
t
c
on
tr
ol
sc
heme
s.
It
can
be
see
n
that
t
wo
-
meth
od
base
d
on
predict
ive
c
on
tr
ol
ha
ve
faster
dynamic
in
c
ontrast
to
casca
de
str
uctur
e
w
it
h
P
I
con
t
ro
l.
It
is
w
or
t
h
to
me
ntio
n
that
PI
c
ontr
oller
desig
n
is
op
ti
mal
a
nd
hi
gh
e
r
gains
f
or
higher
sp
ee
ds
causes
more
ov
e
rs
hoot
in
s
ys
te
m.
Fi
gure
6
il
lustrat
es
zo
om
e
d
a
re
as
mar
ke
d
with
gr
ee
n
rectan
gle
in
Fig
ur
e
5
for
more
detai
ls.
Subp
l
ot
(a)
s
hows
v
e
ry
l
ow
ov
e
rs
hoot
f
or
pro
po
se
d
met
hod
i
n
co
mp
a
ri
so
n
t
o
ot
her
m
et
hods
even
PSC
an
d
sub
plo
t
(
b)
s
uppo
rt
this
op
i
ni
on
.
S
ubplo
t
(
c)
s
hows
sp
ee
d
var
ia
ti
on
dur
ing
l
oad
to
r
qu
e
ste
p
change
a
nd
it
can
be
see
n
th
at
pr
op
os
e
d
m
et
hod
does
no
t
cause
os
ci
ll
at
ion
a
nd
track
t
he
re
fer
e
nce
s
peed
as
fast
as
possi
ble.
Finall
y,
sub
plo
t
(
d)
sho
ws
a
ste
ady
sta
te
performa
nce
and
it
dem
on
s
trat
es
that
pro
po
s
ed
method
has
lo
wer
ste
ad
y
sta
t
e
erro
r
a
nd
Ta
ble
2
co
ntains
M
ea
n
S
quare
Error
(
M
S
E)
va
lues
f
or
this
par
t
of
simulat
ion j
us
t
for
c
ompa
rison.
Durin
g
s
peed
transient,
c
urre
nt
pe
rform
ance
is
an
i
mporta
nt
issu
e.
With
highe
r
sp
ee
d
e
rror
weig
hting
in
t
r
ackin
g
functi
on
(
10)
,
s
pee
d
dyna
mic
will
be
fas
te
r
but
it
w
ou
l
d
ca
us
e
more
c
urren
t
disto
rtion.
Current
var
ia
ti
on
duri
ng
sim
ulati
on
is
il
lustrate
d
i
n
Fi
gure
7
for
.
It
ca
n
be
see
n
t
hat
no
n
-
casca
de
struc
ture
schemes
h
a
ve mo
re s
imi
la
r b
ehav
i
or an
d
ca
scade c
ontrolle
r
wi
t
h
P
I have
diff
e
re
nt r
es
po
ns
e.
Fo
r
m
or
e
deta
il
s
Figure
8
s
hows
z
oome
d
areas
ma
r
ked
with
gr
ee
n
r
ect
ang
le
in
Fi
gure
7.
It
is
no
ti
ceable
that
al
l
c
on
tr
ol
sc
hemes
e
xer
t
c
urren
t
li
mit
at
i
on
a
nd
PI
co
nt
ro
ll
er
ha
ve
lo
wer
c
urren
t
va
lue
i
n
transient
mode
.
Sub
plo
t
(c
)
s
hows
ste
p
do
w
n
in
q
a
xis
cu
r
re
nt
f
or
PSC
a
nd
pro
posed
m
et
hod
du
rin
g
s
peed
change.
It
is
cl
ear
that
pr
opose
d
meth
od
has
lo
wer
un
der
s
hoot
with
no
osc
il
la
ti
on
an
d
s
ubplo
t
(
d)
il
lus
trat
e
this fact
durin
g st
ep u
p
c
hang
e.
Figure
4
: R
efe
r
ence s
pee
d
a
nd PMS
M
s
pee
d
us
in
g
t
hr
ee
dif
f
eren
t c
ontrol
m
et
hods
,
Pr
e
dicti
ve
S
peed co
nt
ro
l
(P
SC)
[1
8]
(
da
sh
, bla
c
k)
,
P
I
c
on
t
ro
l as
outer
loop
(so
li
d,
red),
pro
po
s
ed
me
thod
(d
as
h
-
dot,
mag
e
nta)
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
1
.
4
-
5
0
0
50
e
R
e
f
e
r
e
n
c
e
S
p
e
e
d
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
1
.
4
-4
-2
0
T
i
m
e
[
S
e
c
]
L
o
a
d
T
o
r
q
u
e
(
T
L
)
A
c
t
u
a
l
E
s
t
i
m
a
t
e
d
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
1
.
4
-
6
0
-
4
0
-
2
0
0
20
40
60
T
i
m
e
[
S
e
c
]
e
R
e
f
PI
P
r
o
p
o
s
e
d
PSC
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
Direct
p
re
dicti
ve spee
d
c
ontrol of
per
m
an
e
nt
mag
net sy
nchrono
u
s
m
oto
r f
ed by
…
(
N
ajme
h
M
ov
ahhe
d N
eya)
2191
Figure
5
.
Zo
om vie
w of Fig
ure
5
ma
rk
e
d wi
th gree
n
recta
ngl
e
Table
2
.
M
ea
n square e
rro
r of
PM
S
M spee
d f
or t =
0.951s
t
o t
= 0.95
3
Prop
o
sed
m
eth
o
d
PSC metho
d
PI
m
eth
o
d
M
SE
3
.45
5
3
e
-
09
2
.61
7
7
e
-
06
8
.77
0
8
e
-
07
Figure
6
.
Cu
rr
e
nt tra
ns
ie
nt
dur
ing
simulat
io
n t
ime usin
g
t
hr
e
e d
if
fer
e
nt c
on
t
ro
l
meth
od, Pr
edict
ive S
pee
d
con
t
ro
l
(P
SC
) [
18] (d
a
sh,
black), P
I
c
on
t
ro
l a
s outer l
oop (s
olid, re
d)
,
P
rop
os
e
d
met
hod (
das
h
-
do
t,
ma
ge
nta)
Figure
7
: Z
oomed view
of
Figure
7 mar
ked
with
gr
ee
n rect
ang
le
i
n
Fi
gur
e
7
6.
CONCL
US
I
O
N
In
t
his
pa
per
a
direct
predict
i
ve
spe
ed
c
ontr
ol
f
or
P
M
S
M
fed
by
matri
x
conve
rter
is
in
tro
du
ce
d.
A
new
s
pee
d
tra
ckin
g
c
os
t
f
unct
ion
is
desi
gn
e
d,
by
c
ombinin
g
s
peed
and
c
urren
t
dyna
mics
with
sy
ste
m
const
raints
to
hav
e
h
ig
h
performa
nce
with
l
ow
osc
il
la
ti
on
in
s
peed
wi
thout
casca
de
structu
re.T
o
re
so
lve
0
.
0
7
5
0
.
0
8
0
.
0
8
5
0
.
0
9
4
9
.
5
50
5
0
.
5
0
.
5
7
0
.
5
9
0
.
6
1
-
4
0
.
5
-
4
0
-
3
9
.
5
0
.
3
1
0
.
3
1
0
4
0
.
3
1
1
4
9
.
9
9
4
4
9
.
9
9
6
4
9
.
9
9
8
50
5
0
.
0
0
2
0
.
9
5
1
0
.
9
5
1
5
0
.
9
5
2
0
.
9
5
2
5
0
.
9
5
3
-
4
0
.
0
0
0
2
-
4
0
-
3
9
.
9
9
9
8
(
a
)
(
b
)
(
c
)
(
d
)
0
0
.
2
0
.
4
0
.
6
0
.
8
1
1
.
2
1
.
4
-5
0
5
T
i
m
e
[
S
e
c
]
C
u
r
r
e
n
t
[
A
]
PI
P
r
o
p
o
s
e
d
PSC
0
.
0
7
5
0
.
0
8
0
.
0
8
5
0
.
0
9
-4
-2
0
2
0
.
5
7
0
.
5
8
0
.
5
9
0
.
6
0
.
6
1
0
.
6
2
-4
-2
0
2
4
6
0
.
0
7
3
0
.
0
7
3
5
0
.
0
7
4
-4
-2
0
2
4
0
.
5
6
8
5
0
.
5
6
9
5
0
.
5
7
1
0
2
4
6
(
a
)
(
b
)
(
c
)
(
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.
4
,
D
ecembe
r
2020
:
218
3
–
219
3
2192
sta
bili
ty
issue
and
el
imi
nate
high
c
urre
nt
distor
ti
on
due
t
o
dif
fere
nce
be
tween
mec
ha
nical
an
d
el
ect
ric
al
ti
me
const
ants,
a
c
urre
nt
ref
e
re
nce
const
ru
ct
e
d
ba
sed
on
sli
di
ng
m
o
de
co
nce
pt
is
bee
n
ad
d
ed
to
trac
king
te
rm
of
cost
functi
on.
This
te
rm
guar
antees
oute
r
l
oop
sta
bili
ty
an
d
us
in
g
sim
ulati
on
res
ults
it
can
be
seen
th
at
this
te
rm
can
dec
r
ease
curre
nt
di
stortion.
Als
o,
to
ha
ve
l
ower
ste
ad
y
sta
te
error
a
nd
be
tt
er
performa
nce
a
disturba
nce
ob
serv
e
r
is
desig
ned
t
o
est
imat
e
loa
d
to
rque
value
a
nd
othe
r
uncertai
ntie
s
of
the
mec
ha
nical
su
bsyste
m
as
a
lum
pe
d
distu
rb
a
nce.
T
his
obser
ve
r
permi
ts
el
imi
nation
of
loa
d
to
r
qu
e
sens
or
a
nd
in
creases
reli
abili
ty of the
sy
ste
m
.
REFERE
NCE
S
[1]
Barr
adi.
Y,
K.
Za
z
i,
M.
Z
a
zi,
N.
Khal
i
,
“
Cont
rol
of
PM
SG
ba
sed
var
ia
bl
e
spe
ed
wind
ene
rgy
conve
rsion
sys
tem
conne
c
te
d
to
the
grid
wi
th
PI
and
AD
RC
appr
oa
c
h,
”
Int
ernati
onal
Journal
o
f
Pow
er
E
lectronic
s
a
nd
Dr
iv
e
S
yste
m
s
(IJ
PE
DS)
,
vol
.
1
0,
1
,
pp
.
128
,
20
19
.
[2]
Yuhendri
,
M.,
A.
Ahyanu
ard
i,
and
A.
As
wardi
,
“
Dir
ect
torque
cont
rol
stra
te
gy
of
PM
SM
em
pl
oying
u
lt
ra
spar
se
ma
tri
x
conv
erter
,
”
In
te
rnat
ional
Journal
of
Power
El
e
ct
ronics
an
d
Dr
iv
e
S
yste
ms
,
vol. 9, no. 1, pp. 64, 2018
.
[3]
N.
Fazl
i
and
J.
Siahba
l
ae
e
,
“Dir
ec
t
torque
cont
r
ol
of
a
wind
en
e
rgy
convers
ion
sys
te
m
with
p
e
rma
nen
t
ma
gn
et
synchronous
gen
era
tor
and
matri
x
conv
erter,”
Pr
oce
ed
ing
of
8th
Powe
r
Elec
troni
cs,
Dr
iv
e
Syst
ems
&
Techno
logi
e
s
Confe
renc
e
(
PED
STC
2017)
,
pp.
166
-
171,
2017.
[4]
T.
Shi
,
X.
Zha
n
g,
S.
An,
Y.
Y
a
n
and
C.
Xia
,
“
Harm
onic
su
ppr
ession
modulation
strategy
fo
r
ult
ra
-
sparse
m
atrix
conve
rt
er,”
I
ET
Powe
r E
le
c
troni
cs
,
vol
.
9
,
pp
.
58
9
–
599,
2016
.
[5]
Nagga
r
H
.
Saad,
Ahmed
A
.
El
-
S
at
t
ar,
Moham
ed
I.
Ma
rei,
“
A
cur
ren
t
cont
ro
lled
ma
tri
x
conve
r
te
r
for
wind
en
erg
y
conve
rsion
sys
t
em
s
base
d
on
per
ma
n
en
t
ma
g
net
synchronous
gene
r
at
or
,
”
Jo
urnal
of
El
e
ct
ri
cal
S
yste
ms
an
d
Information
Tec
hnology
,
v
ol
.
3
,
no.
1
,
pp
.
108
-
1
18,
2016
.
[6]
P.
B.
Shind
e
a
nd
T.
N.
Dat
e,
“
Puls
e
width
modul
ation
control
of
3
ph
ase
AC
-
AC
ma
tri
x
con
ver
t
er,”
201
7
Inte
rnational
Co
nfe
renc
e
on
Co
mputing
Me
thod
ologi
es
and
Comm
unic
ati
on
(ICC
MC)
,
Erod
e, pp.
992
-
997,
2017
.
[7]
J.
Holtz
,
“Adva
nce
d
PWM
and
pre
dictive
control
—
An
over
view
,
”
IEEE
Tr
ans.
Ind.
El
e
ct
ron.
,
vol.
63,
no.
6
,
pp.
3837
–
3844
,
2016.
[8]
Y.
Li,
Y.
Qu,
H.
Shi
and
X.
Men
g,
“
An
opt
im
a
l
sw
it
chi
ng
ta
bl
e
fo
r
PM
SM
DTC
s
y
stem
using
z
ero
volt
ag
e
ve
ct
or,”
2017
20th
In
te
rn
ati
onal
Con
fe
ren
ce
on
E
le
c
tric
al
Mac
hine
s and
S
yste
ms
(ICEM
S)
,
Sydney,
NS
W,
pp.
1
-
5
,
2017
.
[9]
Li
nder
,
R
.
Kanc
han,
P.
Stolze,
a
nd
R.
Kenn
el,
“
Mo
del
-
Based
Pr
edi
c
ti
ve
Control
of
E
le
c
tr
i
c
Dri
ves,
”
Göt
ti
ngen
:
Cuvi
llier
Ve
rlag
,
2010.
[10]
C.
Zha
ng
,
X.
W
ang,
D.
Wa
ng
,
Q.
Sun
an
d
G.
Ma,
“
Comp
arati
ve
Analysis
of
El
e
ct
rom
agne
t
ic
Force
Inve
rt
er
Fed
PM
S
M
Drive
U
sing
Fiel
d
Ori
en
te
d
Control
(FO
C)
a
nd
Dire
ct
T
orque
Con
trol
(
DTC),
”
2019
22
nd
Int
ernati
onal
Confe
renc
e
on
E
le
c
tric
al
Mac
h
in
es
and
Syst
ems (
ICEMS)
,
H
arb
in, Chi
na
,
pp
.
1
-
4
,
2019
.
[11]
D.
Ma
jc
hrz
ak
an
d
P.
Siwek
,
“
Co
mpa
rison
of
FO
C
and
DTC
me
t
hods
for
a
Ma
trix
Conver
te
r
-
f
ed
per
m
an
ent
m
agn
et
synchronous
mot
or,
”
2017
22nd
Inte
rnationa
l
C
onfe
renc
e
on
M
et
hod
s
and
Mod
el
s
in
Au
tomation
and
Robot
i
cs
(MMA
R)
,
Mied
z
yzdr
oje,
pp
.
525
-
530,
2017
.
[12]
J.
Bocker,
B.
Freude
nber
g
,
A
.
T
he,
and
S.
Die
ck
erh
off,
“E
xp
erime
nt
al
com
p
ari
s
on
of
mod
el
pre
dic
ti
v
e
cont
ro
l
a
nd
ca
sca
d
ed
cont
r
ol
of
the
mod
ula
r
mul
t
ilevel
converte
r
,
”
I
EE
E
Tr
ans.
P
ower
Elec
tron.
,
vol.
30
,
no
.
1,
pp.
422
–
430
,
20
15
.
[13]
Y.
Zh
ang,
Z
.
Yin,
W.
L
i,
X.
Tong
and
Y.
Zhong,
“
Speed
Sensorless
Model
Pr
edi
c
ti
ve
C
ontrol
B
ase
d
o
n
Disturbanc
e
Obs
er
ver
for
Indu
ct
i
on
Motor
Driv
es,
”
201
9
I
EE
E
In
te
rnational
Sym
posium
on
Pred
i
ct
i
ve
C
ontrol
o
f
El
e
ct
rica
l
Dr
ive
s and
Powe
r
Ele
ct
ronics
(
PR
EC
EDE)
,
Qu
anz
ho
u,
Chin
a, pp. 1
-
4
,
2019
.
[14]
F.
Wa
ng
,
X.
Me
i
,
J.
Rodrigu
ez
an
d
R.
Kenne
l,
“
Model
pre
d
ictive
c
ontrol
for
el
e
ct
r
i
ca
l
dr
ive
sys
tem
s
-
an
over
vi
ew,
”
in
CES
Tr
ansacti
ons on
E
lectrical
Mac
h
ine
s and
Sy
stems
,
vol. 1,
no.
3
,
pp
.
219
-
2
30,
2017
.
[15]
E.
Gar
aya
ld
e,
I.
Aizpur
u,
U
.
Ira
ola
,
I.
Sa
n
z,
C
.
Berna
l
and
E
.
O
yar
bide,
“
Fin
it
e
Control
Set
MP
C
vs
Conti
nuou
s
Control
Set
MP
C
Perform
ance
Compa
r
ison
for
Synchronous
Buc
k
Conver
te
r
Contro
l
in
Ene
rgy
Stor
age
Applic
a
t
ion,”
2019
Int
ernational
Conf
ere
n
ce
on
C
le
an
Elec
tric
al
P
ower
(ICCEP)
,
Otra
n
to,
It
aly,
pp.
490
-
495
,
20
19
.
[16]
A.
A.
A
hm
ed,
B
.
K
.
Koh
and
Y.
I
.
Lee,
“
A
Co
m
par
ison
of
Fin
it
e
Contro
l
Set
and
Cont
inuous
Co
ntrol
Set
Mo
del
Predic
ti
v
e
Con
t
rol
Sche
me
s
fo
r
Spe
ed
Con
tr
o
l
of
Indu
ct
ion
Motors,”
in
IE
EE
Tr
ansacti
on
s
on
Industrial
Informatic
s
,
vo
l. 14, no. 4, pp. 13
34
-
1346,
2018
.
[17]
O.
Sandre
-
Hern
ande
z
,
J.
Rang
el
-
Mag
da
le
no
a
nd
R.
Mora
le
s
-
Capora
l
,
“
A
C
ompa
rison
on
Finit
e
-
Set
Model
Predic
ti
v
e
To
rq
ue
Con
trol
Sche
me
s
for
PM
SM
s,”
in
IE
EE
Tr
ansacti
ons
on
Po
wer
E
lectronic
s
,
vol
.
33,
no
.
10
,
pp.
8838
-
8
847
,
2018.
[18]
P.
Kakosimos
a
nd
H.
Abu
-
Rub,
“
Predictive
Spe
ed
Control
wi
th
Short
Predic
t
io
n
Horiz
on
for
P
erm
an
ent
M
agn
et
Synchronous Motor
Drive
s,
”
in IEE
E
Tr
ansacti
o
ns on
Powe
r
Ele
ct
ronics
,
vol
.
33
,
no.
3,
pp.
2740
-
2750,
2018
.
[19]
L.
W
ang,
S.
Ch
ai
,
D
.
Yoo,
L.
Gan,
and
K
.
Ng,
“
PID
and
Pr
edi
c
ti
ve
Con
trol
of
Elec
tr
ical
D
rive
s
and
Pow
e
r
Convert
ers
Us
in
g
Matlab/Sim
u
link
,”
Hobo
ke
n
,
N
J,
US
A:
Wi
l
ey,
201
5
.
[20]
E.
J.
Fuente
s
,
C.
A.
Silva,
and
J.
I.
Yuz,
“Pre
d
ic
t
i
ve
Speed
Contro
l
of
a
Two
-
Mass
Sys
te
m
Drive
n
by
a
Perm
ane
n
t
Magne
t
Synchro
nous Motor,
”
IE
EE
Tr
ans.
Ind
.
E
le
c
tron.
,
vol. 59, no. 7, pp. 2840
-
2848,
2012
.
[21]
M.
Preindl
,
and
S.
Bologna
ni
,
“
Model
Predictiv
e
D
ire
c
t
Speed
Control
with
Fi
nit
e
Con
trol
Set
of
PM
S
M
Driv
e
Sys
te
ms,”
I
EEE
Tr
ans.
Powe
r E
l
ec
tron.
,
vo
l. 28,
no.
2
,
pp
.
1007
-
1015,
2013
.
Evaluation Warning : The document was created with Spire.PDF for Python.