Internati
o
nal
Journal of P
o
wer Elect
roni
cs an
d
Drive
S
y
ste
m
(I
JPE
D
S)
V
o
l.
4, N
o
. 3
,
Sep
t
em
b
e
r
2014
, pp
. 29
0
~
29
8
I
S
SN
: 208
8-8
6
9
4
2
90
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
Speed Tracking of Field Orient
ed
Cont
rol Perm
anent Magn
et
Synchronous M
o
tor Using Neural
Network
Wah
y
u
M
u
l
y
o
Ut
om
o
*
,
N
o
o
r
adz
i
ani
e
M
u
hamm
ad
Z
i
n*, Z
a
i
n
al
Al
am
Har
o
n
*
,
Sy
Yi
Si
m*
, Az
uw
i
e
n
A
i
da Bo
ha
ri*
,
Ro
slina Ma
t
A
r
if
f**
,
D
i
rma
n
Hanaf
i
**
* Department of
Power El
e
c
tri
c
a
l
, Universi
ti
Tun
Hussein Onn Mala
y
s
ia
** Departmen
t
o
f
Mechatronic
and Robotic
, Universiti Tun Hussein Onn Malay
s
ia
Article Info
A
B
STRAC
T
Article histo
r
y:
Received
Ma
r 9, 2014
Rev
i
sed
Ap
r
28
, 20
14
Accepted
May 15, 2014
The fi
eld ori
e
nt
ed contro
l theor
y
and
space v
e
c
t
or pulse width
m
odulatio
n
techn
i
que make a permanent
magnet s
y
n
c
hro
nous motor can achieve th
e
performance as
well as a DC motor. However,
due to the nonlinearity
of th
e
permanent magn
et s
y
n
c
hronous
motor driv
e ch
aracteristics, it
is
difficu
lt to
control b
y
using
conventional pr
oportiona
l-integr
al-der
ivative con
t
roller. B
y
this reason
in
this paper
an
online
n
e
ural n
e
twork controller for th
e
permanent mag
n
et s
y
nchronou
s moto
r is pr
oposed. Th
e
controller is
designed to tracks variations of sp
eed references and also during load
disturbance.
The effectiv
eness of the
proposed m
e
thod is ver
i
fied
b
y
d
e
velop
sim
u
lation m
odel in MATLAB-sim
u
li
nk program. The simulation results
show that the pr
oposed controller can
r
e
duce the
overshoot
, settling time and
rise time. It can be concluded
that
the perfo
r
m
ance of the c
ontrolle
r is
improved.
Keyword:
Perm
anent Magnet
Syn
c
hro
nou
s Mo
to
r
Neu
r
al Netw
or
k
Copyright ©
201
4 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
:
Wahy
u
M
u
l
y
o Ut
om
o,
Depa
rt
m
e
nt
of
Po
wer
El
ect
ri
cal
,
Un
i
v
ersiti
Tu
n Hu
ssein
On
n Malaysia
8
640
0 Par
it Raj
a
, Bat
u
Pah
a
t,
Jo
hor
, Malaysia
Em
a
il: wah
yu@u
t
h
m
.ed
u
.
my
1.
INTRODUCTION
Perm
anent
m
a
gnet
sy
nc
h
r
o
n
ous m
o
t
o
rs
(
P
M
S
M
)
has
been
use ei
t
h
er i
n
l
o
w o
r
m
i
d powe
r
applications s
u
ch as a
d
justabl
e
spee
d
dri
v
es, com
put
er
peri
phe
ral
e
qui
pm
ent
,
r
o
botics a
n
d electric ve
hicles
l
a
t
e
l
y
. Per
m
anent
m
a
gnet
sy
nch
r
on
o
u
s m
o
tor h
a
s t
h
e feat
ures
densi
t
y
of
hi
gh
po
we
r, fr
ee
m
a
i
n
t
e
nanc
es and
h
i
gh
efficien
cy wh
ich
h
a
s b
e
en
u
s
ed
in
wi
d
e
ly ap
p
licatio
n
in
th
e v
a
riou
s electric d
r
iv
es [1
]. In
1
998
, Pillay &
Kri
s
hna
n
pre
s
ent
e
d t
h
at
PM
m
o
t
o
r dri
v
es
an
d cl
assi
fied
th
em
in
to
two categ
ori
e
s
w
h
i
c
h a
r
e
pe
rm
anent
m
a
gnet
sy
nc
h
r
on
o
u
s m
o
t
o
r
d
r
i
v
es
(PM
S
M
)
an
d
br
us
hl
ess
dc
m
o
t
o
r (B
D
C
M
)
d
r
i
v
es
[
2
]
.
The
PM
SM
has a
si
nus
oi
dal
bac
k
em
f and t
o
p
r
o
d
u
ce t
h
e co
n
s
t
a
nt
t
o
rq
ue,
it
n
eed
sinu
so
idal stato
r
cu
rren
ts wh
ile th
e BDCM
has a trapezoi
d
al back em
f and to produce the consta
nt
to
rqu
e
, it n
e
ed
rect
an
gu
lar
stator currents
.
The PMSM
has feat
u
r
es si
m
i
l
a
r t
o
a woun
d r
o
t
o
r sy
nc
hr
o
n
o
u
s
m
achine exce
pt the perm
an
en
t m
a
g
n
e
t of PMSM will
pr
o
duce
exci
t
a
t
i
on
i
n
st
ead
o
f
fi
el
d wi
n
d
i
n
g f
o
r
ser
v
o
a
p
pl
i
cat
i
ons
a
n
d
t
e
nd
t
o
not
have a
n
y
dam
p
er wi
n
d
i
ngs
.
Th
e PM m
o
to
r fam
i
ly d
i
v
i
d
e
d
in
to two
categ
ories wh
ich
are in
tern
al ro
t
o
r and
ex
tern
al ro
tor. B
o
th
desig
n
s
are
used in c
r
itical applications like
wind
po
wer
ge
nerat
o
rs
an
d el
evat
or
w
i
nches
.
There a
r
e t
w
o
m
e
t
hods t
o
ac
hi
eve zer
o st
eady
st
at
e error
:
swi
t
c
hi
ng an
d i
n
t
e
gr
at
i
o
n
.
To el
im
i
n
at
e
steady state error, a Proporti
onal
-
In
t
e
g
r
al
(PI
)
co
nt
r
o
l
l
e
r
sho
u
l
d
be em
pl
oy
ed
[3]
.
B
y
usi
n
g PI c
o
nt
rol
l
e
r
exact dq axis
reactance pa
ra
meters can be
obtaine
d.
M
o
reove
r, t
o
step
change
of c
o
mmand s
p
eed, param
e
ter
v
a
riation
s
and lo
ad
d
i
st
u
r
b
a
n
ces is v
e
ry sen
s
itiv
e. Si
n
c
e it is sl
ig
h
tly si
m
p
le to
i
m
p
l
e
m
en
t, Propo
rt
io
n
a
l-
In
teg
r
al-Differen
tial (PID) co
n
t
ro
ller
b
e
come
m
o
st wid
e
ly u
s
ed
for PM
SM. So
, a real ti
m
e
self au
to
mated
in
tellig
en
t h
a
rdware im
p
l
e
m
e
n
tatio
n
of PID
co
n
t
ro
ller
as
well as FOC is desired
[4
].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Spee
d
Tr
acki
n
g
of
Fi
el
d
Ori
e
nt
ed C
o
nt
rol
P
e
rma
n
e
n
t
M
a
g
n
et
Sy
nc
hr
on
o
u
s M
o
t
o
r…
(
W
ahy
u
Mul
y
o
Ut
om
o)
29
1
The
PM
m
o
t
o
r
s
can
be
use
d
i
n
Vect
o
r
C
o
nt
r
o
l
(
V
C
)
o
r
s
o
c
a
l
l
e
d Fi
el
d
Ori
e
nt
ed C
o
nt
rol
appl
i
cat
i
o
ns
[5]
.
I
n
or
der t
o
dri
v
e a
PM
S
M
sm
oot
hl
y
,
a
n
El
ect
ro
ni
c C
ont
rol
U
n
i
t
(E
C
U
) s
h
al
l
be d
e
si
gne
d s
u
ch t
h
at
t
h
e
stator
c
u
rrent space vector.
So,
FOC ca
n
achieve t
h
is goal. In FOC,
m
o
tor stator c
u
rrents
& volt
ages a
r
e
m
a
ni
pul
at
ed
i
n
t
h
e
di
rect
-
q
ua
drat
ure
(
d
-
q
)
re
fere
nce
fram
e
of
the
rot
o
r.
The a
r
t
i
f
i
c
i
a
l
neural
net
w
or
ks
(A
N
N
) a
r
e be
st
sui
t
e
d f
o
r so
l
v
i
ng t
h
e
pr
obl
em
s t
h
at
are n
onl
i
n
ea
r
i
n
nat
u
re.
In
A
N
N
we
can
use
paral
l
e
l
p
r
o
c
e
ssi
ng m
e
t
h
o
d
s
to sol
v
e s
o
me real-wor
ld
prob
lem
s
wh
ere it is
d
i
fficu
lt to
d
e
fi
n
e
a co
nv
en
tion
a
l algo
rith
m
s
. Th
e ab
ility o
f
ANN to
learn
l
a
rg
e classes of
n
o
n
lin
ear fun
c
tio
n
s
is w
e
ll kn
own [
7
]-[8
].
I
t
can b
e
tr
ain
e
d
to
e
m
u
l
ate th
e unk
now
n non
lin
ear
p
l
an
t
d
y
n
a
mics b
y
p
r
esen
tin
g a
sui
t
a
bl
e set
of
i
n
p
u
t
/
out
put
p
a
t
t
e
rns ge
nerat
e
d by
t
h
e pl
a
n
t
[6]
.
Once sy
s
t
em
dy
nam
i
cs
have
been i
d
e
n
t
i
f
i
e
d
usi
n
g a
n
A
NN,
m
a
ny
conve
nt
i
onal
c
ont
rol
t
e
chni
que
s ca
n
b
e
ap
pl
i
e
d t
o
ac
hi
eve t
h
e
desi
r
e
d
ob
ject
i
v
e
.
In
t
h
i
s
pa
per,
a
m
odel
of
AN
N cl
o
s
ed
-l
o
o
p
PM
SM
co
nt
r
o
l
sy
st
em
t
h
at
i
s
cont
rol
l
e
d
by
SVP
W
M
a
r
e
devel
o
ps and
presents an a
n
alysis of ANN speed c
ont
r
o
ller
for s
p
eed pe
rf
orm
a
nce
in FOC PMSM drive. The
effect
i
v
e
n
ess
o
f
t
h
e
p
r
o
p
o
se
d
m
e
t
hod i
s
ve
ri
f
i
ed by
de
vel
o
p
sim
u
l
a
t
i
on m
odel
i
n
M
A
TL
A
B
-si
m
ul
i
nk.
2.
R
E
SEARC
H M
ETHOD
2.1.
Permane
nt Magnet Synchr
onous
Motor
Dynamic
Modeling
PMSM is essen
tially a th
ree p
h
a
se AC m
o
to
r
with
si
n
u
s
oid
a
l b
a
ck
EM
F driv
en
b
y
a
DC source,
wh
ich
is con
v
e
rted
t
o
three-ph
ase altern
ating
cu
rren
ts su
pp
lyin
g
t
o
t
h
e t
h
ree stato
r
wi
nd
ing
s
o
f
PMSM. The
m
a
t
h
em
at
i
c
m
odel
o
f
PM
S
M
i
dq sy
nch
r
o
n
o
u
s r
o
t
a
t
i
ng
r
e
fere
nce fram
e
can be o
b
t
a
i
n
ed fr
om
sy
nchro
n
ous
machine m
odel. Due t
o
the c
onsta
nt field
produced
by pe
rm
anent
m
a
gnets, the field
variatio
n
is zero. It is
also ass
u
m
e
d that saturation
and losses
of c
o
re a
r
e
ne
g
l
i
g
ib
le, th
e i
n
d
u
c
ed
em
f is sin
u
s
o
i
d
a
l an
d th
ere is no
dam
p
er wi
ndi
n
g
on
r
o
t
o
r.
Usi
n
g
t
h
ese
ass
u
m
p
t
i
o
n
s
, t
h
e
vol
t
a
ge e
quat
i
ons
can
wri
t
e
as
f
o
l
l
o
w:
ds
d
d
d
q
e
q
dd
vR
i
L
i
L
i
dt
dt
(1)
qs
q
q
q
d
e
d
e
P
M
dd
vR
i
L
i
L
i
dt
dt
(2)
The
pr
o
duce
d
t
o
r
q
ue
of t
h
e m
achi
n
e c
a
n
be
prese
n
t
e
d
as
fo
l
l
o
w:
3
[(
)
]
2
PM
q
d
q
d
q
e
TP
i
L
L
i
i
(3)
Wh
ile, t
h
e m
a
x
i
m
u
m
sp
eed
can
b
e
id
en
tified
fro
m
th
e relatio
n
s
h
i
p
:
L
fm
m
e
d
TT
K
J
dt
(4)
The
up
dat
e
f
r
e
que
ncy
o
f
t
h
e
cont
rol
l
o
o
p
s
m
u
st
be hi
gh e
n
o
u
gh a
n
d t
h
e
SVP
W
M
s
h
o
u
l
d
be
p
r
o
p
erl
y
configure
d
t
o
e
n
sure si
nusoi
d
al curr
en
ts applied
to
the stator wi
nd
ing
s
.
2.
2.
Fi
el
d Ori
e
nte
d
C
o
n
t
roller (FOC) Description
Fi
el
d ori
e
nt
ed
cont
rol
(
F
OC
) al
so k
n
o
w
n as
deco
u
p
l
i
ng
or vector control, ca
m
e
in
to
th
e field
of ac
d
r
i
v
es research
in
th
e late 19
60
s an
d
was
d
e
v
e
l
o
p
e
d
p
r
omin
en
tly in
th
e 1
980
s to
m
e
et th
e ch
alleng
es of
o
s
cillatin
g
fl
ux and
torq
u
e
resp
on
se i
n
i
n
v
e
rter fed
i
n
du
ction
an
d syn
c
h
r
on
ou
s m
o
to
r dri
v
e.
Th
e in
ex
p
l
i
cab
le
dy
nam
i
c beha
vi
o
r
o
f
l
a
r
g
e c
u
r
r
ent
t
r
a
n
si
e
n
t
s
and t
h
e res
u
l
t
i
ng fai
l
u
re o
f
i
nve
rt
ers
was
a curse
an
d
ba
rri
er t
o
the ent
r
y of i
n
verter
fed ac
d
r
iv
e
s
in
t
o
th
e
ma
r
k
e
t
.
FOC is a p
r
o
c
ess for h
a
nd
lin
g
m
o
to
r con
t
ro
l resu
ltin
g
in
energ
y
-efficien
t o
p
e
ration
and
fast d
y
n
a
m
i
c
response at all speeds
.
It commutates
m
o
tor
by calcu
l
a
t
i
ng t
h
e
vol
t
a
ge
and cu
rre
nt
v
ect
or base
d
on
m
o
t
o
r
cu
rren
t feed
b
a
ck
. It
m
a
in
tain
s
h
i
gh
efficiency
in
a wide operating rang
e
and
al
l
o
w
s
pre
c
i
s
e dy
nam
i
c cont
rol
of s
p
eed a
n
d torque. The
FOC control the stator curre
nts
repre
s
ente
d by a space v
ector. It trans
f
orm
s
the
three-phase sta
t
or c
u
rrents
(a,
b, c
)
into t
h
e two-phase
system variants (
α
,
β
) .
A two
tim
e
in
v
a
rian
t coord
i
n
a
te
syste
m
(d-q) is
obtaine
d from
the syst
e
m
v
a
rian
ts. Fo
r t
h
is syste
m
, d
(d
irect) p
a
rt is m
a
k
i
ng
th
e m
o
to
r flu
x
wh
ile q
p
a
rt (qu
a
drature) is
gen
e
rate th
e toqu
e. In
FO
C
,
m
o
t
o
r st
at
or c
u
r
r
e
nt
s an
d
voltages are m
a
nipulated in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-86
94
I
J
PED
S
Vo
l.
4
,
No
.
3
,
Sep
t
em
b
e
r
2
014
:
29
0 – 298
29
2
directq
u
a
d
rat
u
r
e
(d
-q
) re
fere
n
ce fram
e
of the
roto
r, a
way mu
st b
e
m
a
th
e
m
atically
tran
sform
e
d
u
s
in
g
t
h
e Park
and
C
l
arke
t
r
a
n
sf
orm
a
t
i
ons
b
e
fo
re t
h
ey
ca
n
be
used
t
o
o
u
t
put
SV
P
W
M
.
Whi
c
h
m
ean
s that the
stator c
u
rrent
feed
bac
k
trans
f
o
r
m
e
d from
the three
-
p
h
ase
static refere
nce fram
e
of the stator windings to t
h
e two axis
rot
a
t
i
n
g
d-
q
ref
e
rences
f
r
am
e of
t
h
e
rot
o
r
.
2.3.
Permane
nt Magnet Sync
hr
onous
Motor
Drive Syste
m
Fi
gu
re
1 i
s
t
h
e
di
ag
ram
of ve
l
o
ci
t
y
/
c
urre
nt
cont
rol
l
o
o
p
u
s
i
ng F
O
C
t
ech
nol
ogy
base
d
on
pr
o
p
o
s
ed
AN
N s
p
ee
d c
o
nt
r
o
l
l
e
r an
d
PI
D s
p
ee
d c
ont
r
o
l
l
e
r.
Fi
gu
re
1.
PM
S
M
dri
v
e sy
st
e
m
wi
t
h
PID
co
nt
r
o
l
l
e
r
Thi
s
pa
per
pr
o
pos
ed a
n
A
N
N
cont
r
o
l
m
e
t
hod o
f
FOC
base
d o
n
SV
P
W
M
t
o
red
u
ce t
h
e
ove
rs
ho
ot
,
steady state error a
nd
rise time. The
ANN speed
con
t
ro
l is ad
d
e
d
to
th
e
s
p
eed controller to
produce the
spee
d
refe
rence
.
The
bl
oc
k di
a
g
ra
m
of t
h
e p
r
op
ose
d
A
NN s
p
eed co
nt
r
o
l
l
e
r
of F
O
C
f
o
r P
M
SM
dri
v
e
sy
st
em
i
s
sho
w
n i
n
Fi
gu
r
e
2.
Fi
gu
re
2.
PM
S
M
dri
v
e sy
st
e
m
wi
t
h
AN
N c
ont
rol
l
e
r
2.
3. Pro
p
ose
d
AN
N Speed
Controller Structure
To desi
gn t
h
e
neu
r
al
net
w
or
k
cont
r
o
l
som
e
i
n
f
o
rm
at
i
on about
t
h
e pl
a
n
t
i
s
req
u
i
r
e
d
. B
a
si
cal
l
y
, t
h
e
num
bers
of i
n
put
a
nd
o
u
t
p
ut
neu
r
on at
eac
h l
a
y
e
r are e
q
u
a
l
t
o
t
h
e n
u
m
b
er o
f
i
n
put
a
n
d
out
put
si
gnal
s
of t
h
e
sy
st
em
respect
i
v
el
y
.
Fu
rt
he
r t
h
e
num
ber
o
f
h
i
dde
n l
a
y
e
rs
an
d t
h
e
t
o
t
a
l
neu
r
ons
i
s
dep
e
n
d
e
d
on
t
h
e
com
p
l
e
xi
t
y
of t
h
e system
and the
required trai
ni
ng
ac
curacy [9]. To
im
ple
m
ent
s
earch e
fficie
n
c
y
optim
al control
of
PM
SM
d
r
i
v
e,
a
m
u
l
t
i
l
a
y
e
r per
cept
r
on
ne
ural
net
w
or
k c
o
nt
ro
l
i
s
devel
o
ped
.
B
a
sed
on
t
h
e t
y
pe o
f
t
h
e
t
a
sk
t
o
be
p
e
rf
or
m
e
d
,
th
e
str
u
ctur
e
o
f
th
e pr
opo
sed ANN
sp
eed
co
n
t
roller
is show
n in Figu
r
e
3[
10
].
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Spee
d
Tr
acki
n
g
of
Fi
el
d
Ori
e
nt
ed C
o
nt
rol
P
e
rma
n
e
n
t
M
a
g
n
et
Sy
nc
hr
on
o
u
s M
o
t
o
r…
(
W
ahy
u
Mul
y
o
Ut
om
o)
29
3
Fi
gu
re
3.
PM
S
M
dri
v
e sy
st
e
m
wi
t
h
AN
N c
ont
rol
l
e
r
Th
e con
t
ro
ller
co
nsists of inpu
t layer,
h
i
dd
en
layer
an
d ou
t
p
u
t
layer. Based
o
n
nu
m
b
er
of
th
e
n
e
u
r
on
i
n
t
h
e l
a
y
e
rs
, t
h
e
AN
N i
s
def
i
ned as
a
1-
3-
1
net
w
or
k st
ruct
ure
.
T
h
e fi
rst
n
e
ur
o
n
o
f
t
h
e o
u
t
p
ut
l
a
y
e
r i
s
u
s
ed
a
s
a torq
ue
refe
re
nce sig
n
al (
a
2
1
=m
f
). T
h
e c
o
n
n
ect
i
ons
wei
g
ht
param
e
t
e
r bet
w
een
j
th
and
i
th
n
e
ur
on
at
m
th
layer is
gi
ve
n by
w
m
ij
,
wh
ile
b
i
as p
a
rameter o
f
t
h
is layer at
i
th
neu
r
o
n
i
s
gi
ve
n
by
b
m
i
. Tr
an
sf
er
fun
c
tio
n of
th
e network
at
i
th
n
e
ur
on
i
n
m
th
l
a
y
e
r i
s
de
f
i
ned
by
:
1
1
1
m
S
mm
m
m
ii
j
j
i
j
nw
a
b
(5)
The
o
u
t
p
ut
f
u
n
c
t
i
on
of
ne
ur
o
n
at
m
th
l
a
y
e
r
i
s
gi
ve
n by
:
()
mm
m
ii
af
n
(6)
whe
r
e:
f
is activ
atio
n
fun
c
tio
n of t
h
e
n
e
uron
.
In th
is
d
e
sign th
e activ
ation fu
n
c
tion
of t
h
e
o
u
t
p
u
t
layer is
un
ity an
d fo
r th
e
h
i
dd
en
layer is a t
a
n
g
e
n
t
hy
pe
rb
ol
i
c
f
u
n
c
t
i
on gi
ve
n by
:
2
2
()
1
1
m
i
mm
i
n
fn
e
(7)
Up
dat
i
n
g
of
t
h
e co
nnect
i
o
n
w
e
i
ght
a
n
d
bi
as
param
e
t
e
rs are
gi
ve
n
by
:
()
(1
)
(
)
mm
ij
i
j
m
ij
Fk
wk
wk
w
(8)
()
(1
)
(
)
mm
ii
m
i
F
k
bk
bk
b
(9
)
After t
h
e ne
u
r
a
l netwo
r
k
arc
h
itecture is m
odelled
,
the
n
e
xt stage
defi
nes the lea
r
ni
ng m
odel to
update net
w
ork param
e
ters. By this
learning capability, it
makes the ANN su
itable to
be i
m
ple
m
ented for t
h
e
sy
stem
with m
o
to
r
param
e
ters w
h
ich
are
dif
f
icult to
de
fine
an
d
vary
a
g
ai
nst
with e
nvi
ro
nm
ent. The t
r
a
i
ning
pr
ocess m
i
nim
i
zes the err
o
r
out
put
of t
h
e
netw
or
k th
ro
u
gh a
n
o
p
tim
i
zation m
e
thod.
Gene
rally
, in l
earni
ng
m
ode of the n
e
ural net
w
o
r
k
contr
o
ller a suf
f
icient tr
aining
data inp
u
t-
out
put m
a
ppin
g
data of a pl
ant is
required.
Since the m
o
tor
para
m
e
ters
of the
PMSM drive vary with tem
p
e
r
ature and m
a
gnetic saturation, t
h
e
online lear
nin
g
B
ack pr
o
p
agat
ion alg
o
rithm
is devel
ope
d. B
a
sed o
n
fir
s
t or
der
optim
ization schem
e
, up
d
a
ting
of
the
netw
or
k
param
e
ters are
determ
ined. T
h
e pe
rf
orm
a
nce
inde
x s
u
m
of s
qua
re e
r
r
o
r
is
g
i
ven
by
:
2
1
()
(
)
2
i
i
Fk
e
k
(1
0)
()
()
(
)
ii
i
ek
t
k
a
k
(1
1)
whe
r
e:
t
i
i
s
ta
r
g
e
t
s
i
g
n
a
l
a
i
out
put si
gn
al o
n
last lay
e
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN:
2
088
-86
94
I
JPEDS
Vo
l.
4
,
No
.
3
,
Sep
t
em
b
e
r
2
014
:
29
0 – 298
29
4
The
gra
d
ient
d
e
scent
of t
h
e
p
e
rf
orm
a
nce in
d
e
x a
g
ainst t
o
th
e co
nnecti
o
n
w
e
ight is
give
n
b
y
:
m
n
FF
i
mm
m
wn
w
ij
i
i
j
(1
2)
The sensitivity param
e
ter of
t
h
e network
is defined
as:
m
i
m
i
F
s
n
(1
3)
m
m
i
i
mm
ii
a
F
s
an
(1
4)
Gra
d
ient t
h
e tr
ansfe
r
fu
nctio
n
again
to t
h
e c
o
nnectio
n
wei
g
h
t
param
e
ter is give
n
by
:
1
m
m
i
i
m
ij
n
a
w
(1
5)
From
substitution Equation
(13) and
(15) int
o
(8) the updatin
g connection
param
e
ter is gi
ven by:
11
(1
)
(
)
(
)
(
)
mm
i
m
m
ij
i
i
i
wk
w
k
s
k
a
k
(1
6)
W
i
t
h
the
sam
e
technique t
h
e
upda
tin
g bias pa
ram
e
ter
is
give
n by
:
1
(1
)
(
)
(
)
mm
i
m
ii
i
bk
b
k
s
k
(1
7)
3.
R
E
SU
LTS AN
D ANA
LY
SIS
The sim
u
lation result
of t
h
e PMSM
drive syste
m
is presented to
verify the
feasibilit
y of t
h
e
p
rop
o
s
ed
ANN sp
eed
co
n
t
ro
ller
un
d
e
r
v
a
r
i
ou
s op
er
ating
speed
con
d
ition
s
. I
n
th
is sectio
n th
e d
y
n
a
m
i
c
m
o
d
e
l
of a
three
-
pha
se PM
SM
, s
p
a
ce vecto
r
P
W
M
and
ne
ural
netw
or
k c
ont
r
o
l m
odel ha
ve
bee
n
de
velo
p
e
d. T
h
e
sim
u
lation is devel
ope
d
usi
ng B
o
rla
n
d C
++, an
d the
n
em
bedde
d as
S-f
u
n
ction
in
Sim
u
link-M
a
tl
ab. T
h
e
param
e
ters fo
r
the m
o
tor ar
e
g
i
ven i
n
Ta
ble
1
.
The sim
u
lation is obse
r
ved
d
u
r
in
g start u
p
re
spo
n
se
w
h
ich i
s
by
usi
ng c
o
n
s
tant refe
ren
c
e
s
spee
d 1
0
0
rad/s
with no load condition
as show
n in Fi
gure 4.
W
ith t
h
e sam
e
refe
rence speed, t
h
e si
m
u
lations of bot
h
PID
an
d
A
N
N
spee
d c
ontr
o
lle
rs a
r
e r
u
n
sim
u
ltaneou
sly
.
Table 1. Param
e
ters
of PMSM
Variables
Para
m
e
ter
Value
P Poles
4
J
M
o
m
e
nt of I
n
er
tia
0.
0008
[kgm
²]
Ld
,Lq
Rs
λ
pm
d-
q axis I
nductance
Ar
m
a
ture Resistan
ce
Flux
0.
0085 [H]
2.
875 [
Ω
]
0.
175 [W
b]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
JPEDS
I
S
SN:
208
8-8
69
4
Spee
d
Tr
ackin
g
of
Field Orie
nted
C
o
ntrol
P
e
rma
n
e
n
t M
a
g
n
et Sy
nc
hr
on
o
u
s M
o
tor…
(
W
ahy
u
Muly
o
Utom
o)
29
5
Figur
e 4
.
Star
t up r
e
sponse for
ANN
and
PID
sp
eed
con
t
ro
ller
From
the result
s, it show that
by using
ANN
spee
d
con
t
ro
ller
pr
odu
ced
a
better
star
t-
up
per
f
o
r
m
a
n
ce
com
p
are to the PID sp
eed controller where
the oversho
ot is totally re
m
oved and the settling ti
m
e
m
o
re
faster
than P
I
D spee
d co
ntr
o
ller in
achievin
g
desir
e
d o
u
tp
ut
s
p
ee
d . Sec
o
nd test
ing is
obse
r
ve
d d
u
r
in
g step
pi
ng
-
u
p
and ste
p
pin
g
-
d
o
w
n
re
sp
o
n
se
applicatio
n. F
o
r ste
ppi
ng
-
u
p
res
p
onse
,
th
e spee
d re
fere
nce is va
ry
ing
fr
om
6
0
r
a
d
/
sec to
120
r
a
d
/
sec is shown in
Figu
r
e
5.
Fig
u
r
e
5
.
Steppin
g
-
u
p
r
e
sp
on
se fo
r
ANN and PID sp
eed
co
ntr
o
ller
For
step
pi
ng
-d
ow
n
res
p
o
n
se
,
the s
p
eed
re
fer
e
nce is
va
ry
ing
fr
om
12
0r
ad/s
ec to
4
0
ra
d/sec
is s
h
o
w
n in
Fi
gu
r
e
6.
Fig
u
r
e
6
.
Steppin
g
-
u
p
r
e
sp
on
se fo
r
ANN and PID sp
eed
co
ntr
o
ller
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN:
2
088
-86
94
I
JPEDS
Vo
l.
4
,
No
.
3
,
Sep
t
em
b
e
r
2
014
:
29
0 – 298
29
6
The last testin
g, t
h
e sim
u
lation
is ve
rify
by
usin
g t
r
ian
gul
ar s
p
eed
re
fere
nce
fo
r
bot
h P
I
D a
n
d
AN
N s
p
e
e
d
cont
rollers
as s
h
o
w
n in
Fig
u
r
e
7.
Figu
re
7.
O
u
tp
ut s
p
eed
w
h
e
n
usin
g tria
ng
ula
r
s
p
eed
re
fere
n
c
e
Both system
PID a
nd
AN
N s
p
eed c
o
nt
roller is also tested on t
h
e effect
o
f
n
o
m
i
nal load distur
ba
nce.
Figu
re
9 s
h
o
w
s the
loa
d
d
i
sturba
nce a
p
p
lied an
d Fi
gu
r
e
8 s
h
ows
the
spee
d trac
kin
g
re
sp
o
n
se
fo
r
b
o
th
system
s. The s
y
stem
is tested with
value
of
5 Nm
load
at 1.0s with 80ra
d/sec
spee
d refe
re
nce.
Figu
re
8.
Loa
d
distu
r
ba
nce
f
o
r
P
I
D
an
d
A
N
N
s
p
ee
d c
ontr
o
ller
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
PED
S
I
S
SN
:
208
8-8
6
9
4
Spee
d
Tr
acki
n
g
of
Fi
el
d
Ori
e
nt
ed C
o
nt
rol
P
e
rma
n
e
n
t
M
a
g
n
et
Sy
nc
hr
on
o
u
s M
o
t
o
r…
(
W
ahy
u
Mul
y
o
Ut
om
o)
29
7
Fi
gu
re
9.
O
u
t
p
ut
s
p
eed
w
h
e
n
l
o
ad
di
st
u
r
ba
nc
e i
s
ap
pl
i
e
d
Refer t
o
th
e Fig
u
re
8
,
at
th
e in
itial b
o
t
h
syste
m
s are run
n
i
n
g
with
n
o
lo
ad.
Fro
m
Fig
u
re
9
,
it is
o
b
v
i
ou
s th
at the ANN co
n
t
ro
l
l
er d
r
o
p
th
e speed
fro
m
8
0
r
ad
/sec to
7
9
rad
/
sec with
g
r
eat
settlin
g
ti
m
e
.
Wh
ile,
PID con
t
ro
ller
d
r
op
th
e sp
eed
fro
m
8
0
r
ad
/sec to
71
.3
rad
/
sec
with
larg
est set
tlin
g
ti
m
e
.
So,
it show that t
h
a
t
th
e
pr
o
pose
d
A
N
N
s
p
eed
co
nt
r
o
l
l
e
r p
r
od
uce
si
gni
fi
ca
nt
i
m
p
r
ovem
ent cont
rol
perform
a
nce com
p
are to t
h
e PID
co
n
t
ro
ller
su
ch as redu
ce
o
v
e
rsh
o
o
t
, settli
n
g
t
i
m
e
an
d
rise time sp
eed respo
n
s
e.
4.
CO
NCL
USI
O
N
Thi
s
pa
pe
r ha
s
prese
n
t
e
d
t
h
e
m
odel
i
ng an
d
sim
u
l
a
t
i
on o
f
t
h
e fi
el
d
o
r
i
e
nt
e
d
co
nt
r
o
l
f
o
r
P
M
SM
dri
v
e
usi
n
g o
n
l
i
n
e
n
e
ural
net
w
or
k
cont
rol
l
e
r.
I
n
o
r
de
r t
o
i
n
vest
i
g
at
e effect
i
v
en
ess of t
h
e p
r
op
ose
d
co
nt
r
o
l
l
e
r
,
som
e
refe
rence
spee
d m
odel
wa
s t
e
st
ed. T
h
e
si
m
u
l
a
t
i
on
st
u
d
y
i
s
realized i
n
M
A
TL
AB progra
m
.
The
res
u
lts of the
pr
o
pose
d
a
n
d c
o
n
v
e
n
t
i
onal
PI
D c
ont
rol
l
e
r a
r
e pl
ot
t
e
d
i
n
t
h
e
sam
e
speed
gr
aph
wi
t
h
i
n
t
e
nt
i
on t
o
m
a
ke de
t
a
i
l
s
com
p
ari
s
on
ba
sed o
n
vi
s
u
al
o
b
ser
v
at
i
o
n on s
t
eady state error, rise tim
e
and o
v
ers
h
oot
s
p
eed res
p
onse
s
.
It
can
be c
oncl
ude
t
h
at
t
h
e
spee
d
pe
rf
orm
a
nce
of
t
h
e
PM
SM
can
be i
m
pr
ove
d
by
a
ppl
y
i
ng t
h
e
o
n
l
i
n
e A
N
N
cont
rol
l
e
r sche
m
e
.
ACKNOWLE
DGE
M
ENTS
All th
e au
tho
r
s wo
u
l
d
lik
e t
o
ex
press a sin
c
ere
ack
nowledgmen
t to
Un
iversiti Tu
n
Hu
ssei
n
Onn
Malaysia fo
r t
h
e v
a
lu
ab
le supp
ort
d
u
ring
com
p
le
tio
n
th
is
research and
m
a
n
u
scrip
t
.
REFERE
NC
ES
[1]
K Song, W Liu, G Luo. Fuzzy
Logic B
a
sed On
line
El
ectromag
netic Loss Minimization of
PMSM Drive.
IEEE
Transactions on
V
ehicle Pow
e
r a
nd Propulsion
. 2
008: 1.
[2]
P Pilla
y,
R Kri
s
hnan. Model
i
n
g
, Sim
u
lat
i
on,
and Anal
ys
is o
f
Perm
anent-Ma
gnet Motor
Dri
v
es, Part
1:
Th
e
Permanent-Mag
n
et S
y
nchronous
Motor Driv
e.
IEEE Transactions
on Industry App
lications
. 1989
;
25: 265-273.
[3]
LK Wong, FHF Bung, PKS Tam. Combination
of Sliding
M
ode Controller
and
PI Controller
using Fuzzy
Lo
gic
Controller
.
IEEE International C
onference on
Fu
zzy System
. 1998
; 1: 296-301.
[4]
M Marufuzzam
a
n
, MBI Re
az
, M
A
Mohd Ali. FP
GA Im
plem
entat
i
on of
an Int
e
l
l
i
g
ent Curr
ent dq
PI Controlle
r for
FOC PMS
M
Drive
.
I
E
EE Transactions on
Computer Ap
p
lica
tions and
Industrial Electronics
(
I
CCAIE)
. 2010: 602
.
[5]
ES
S
e
rgaki, S
P
Georgilak
i
s
,
A
G
Kladas
, GS
S
t
avrakak
i
s
.
F
u
zz
y Log
i
c B
a
s
e
d Online E
l
ec
tr
om
agnetic
Los
s
Minim
i
zation
of
PMSM Drives.
I
EEE Transactio
ns on
El
ec
trical
Machines
. 2008; 1.
[6]
J Weidong, W
Qunjing, C Qu
an, S Xi
aof
e
ng.
SVPWM Strateg
y
for
Thr
ee-
Level
Inverter b
a
sed on SVPWM
Strateg
y
for Two-Level
Inverter
.
Transactions of China
Electrotechnica
l Society
.
2009; 24: 108-1
14.
[7]
Y Yi, D M
a
h
i
n
d
a, M
A
R
a
hm
an. Im
plem
ent
a
ti
on of an
Ar
tif
icial Neur
al N
e
twork based R
e
al
Time Adap
tive
Controller
for
an
Inter
i
or PMSM.
IEEE Transaction On Industry
Applica
tion
. 200
3; 39: 96-103.
[8]
KS Narenda, K
Parthasarath
y
.
Identif
ication
an
d Cont
rol of D
y
namical S
y
stems using Neural Networks,"
IEEE
Transaction Neu
r
al Network
. 199
0: 4-27.
[9]
I Choy
,
SH Kwon,
JY Choi,
JW Kim,
KB
Kim.
On-Line
Effi
ci
ency
Opti
mization Contro
l of a
Slip
Ang
u
lar
Frequency Controlled Induction
Moto
r Drive Using Neural Networks.
IECON
Proceedings 13 annual Confer
en
ce
.
1996; 1216-122
1.
[10]
AHM
Yatim
, W
M
Utom
o. Online Optim
al
Control of Variable Speed Co
m
p
ressor
Motor Drive Sy
st
em
using
Neural Con
t
rol
Model.
I
E
EE Transaction On
Po
wer and En
ergy
. 2004; 83-
87.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN:
2
088
-86
94
I
JPEDS
Vo
l.
4
,
No
.
3
,
Sep
t
em
b
e
r
2
014
:
29
0 – 298
29
8
BIOGRAP
HI
ES OF
AUTH
ORS
Wah
y
u Muly
o
Utomo was born in
Pati, Indo
nesia, in 1969
.
He
receiv
ed th
e B. S Degr
ee
in
electrical eng
i
n
eering
from the Universitas Br
awijay
a
Malang
, in
1993, th
e
M.S. degr
ee in
electri
cal
engin
e
ering from
the I
n
stitute Sepu
lu
h
Nopem
b
er Surabay
a
,
in 2000,
and the Dr.
Eng.
Degree from Un
iversiti Teknolo
g
i Malay
s
ia
in 2
007. He
is curr
ently
a Senior
Lecturer
in
Electr
i
cal
Power Engineering Departmen
t
,
Facult
y of Ele
c
tr
ica
l
and
E
l
ec
tron
ic Engine
ering
,
Universiti Tun
Hussein Onn
Mala
ysi
a
. His cur
r
ent
resear
ch int
e
rests includ
e t
h
e area of pow
er
electronics and
motor drive control.
Nooradzian
ie M
uhammad Zin was born
in Kedah, Malay
s
ia, in 1
987.
She received the B.S. degr
ee
in electrical eng
i
neer
ing from U
n
iversiti Tun
Hussein Onn Malay
s
ia, in
2011.
She is curr
ently a
Master student
in the Electr
i
cal Power Engineer
ing Department , Facult
y of
Elec
tric
al and
Ele
c
troni
cEngin
eering
,
Univ
ersi
ti Tun
Hussein
Onn Mala
ysi
a
.
His curren
t
r
e
search
inter
e
sts
includ
e th
e
are
a
of power
ele
c
tro
n
ics
and
m
o
tor d
r
ives
con
t
rol
.
Zain
al Alam Har
on was born
in
Malacca, Malaysia,
in 1956
. He
received
the B.S
.
degr
ee in
electrical eng
i
n
eering
from the University
of
Glasgow, U.K., in 1981
,
the
M. S Degree in
Ionisation
Ph
y
s
ics from the Univ
ersity
of
Wales
,
U.K., in
1985
, and
the
Dr. Eng. Degree
from
University
of
Newcastle Upon Ty
ne, U
.
K., in
19
92.
He is cu
rrently
an
Associate
Professor in th
e
Ele
c
tri
cal
P
o
wer Engin
eering
Departm
e
nt,
F
a
cult
y of
El
ec
tri
cal
and
Ele
c
tro
n
ic Eng
i
ne
ering
,
Universiti
Tun
Hussein Onn Mala
y
s
ia
. His
curre
nt rese
arch
int
e
r
e
sts inc
l
ude
the
area
of powe
r
ele
c
troni
cs
and
e
l
ec
tric
m
o
tor dri
v
es
.
Sim
Sy
Yi was born in Johor,
Mala
y
s
i
a
, in 1
988. She receiv
e
d the B.S. deg
r
ee in ele
c
tr
ica
l
engine
ering fro
m
Universiti Tu
n Hussein Onn Mala
y
s
ia
, in 201
1. She is curr
ent
l
y
a PhD student
in
the Electr
ical Power Engineer
in
g Department , F
aculty
of Electrical and Electr
onicEngin
eer
ing
,
Universiti Tun
Hussein Onn
Mala
ysi
a
. His cur
r
ent
resear
ch int
e
rests includ
e t
h
e area of pow
er
ele
c
troni
cs
and
m
o
tor drives
con
t
rol.
Azuwien Aida
Bohari was bor
n in
Johor, Malay
s
ia,
in 1987.
She received
th
e B.S. d
e
gree in
electri
cal
eng
i
neering from
Uni
v
ersiti
Tun Husse
in Onn Malaysia, in
2011. She is curr
ent
l
y
a
Master student
in the Electr
i
cal Power Engineer
ing Department , Facult
y of
Elec
tric
al and
Ele
c
troni
cEngin
eering
,
Univ
ersi
ti Tun
Hussein
Onn Mala
ysi
a
.
His curren
t
r
e
search
inter
e
sts
includ
e th
e
are
a
of power
ele
c
tro
n
ics
and
m
o
tor d
r
ives
con
t
rol
.
Roslina Mat Ariff was born in Perak, Malay
s
ia,
in
1987. She received th
e B.
S. d
e
gree in
electrical
engine
ering fro
m
Universiti Tu
n Hussein Onn Mala
y
s
ia
, in 201
1. She is curr
ent
l
y
a PhD student
in
the Electr
ical Power Engineer
in
g Department , F
aculty
of Electrical and Electr
onicEngin
eer
ing
,
Universiti Tun
Hussein Onn
Mala
ysi
a
. His cur
r
ent
resear
ch int
e
rests includ
e t
h
e area of pow
er
ele
c
troni
cs
and
m
o
tor drives
con
t
rol.
Dirman Hanafi The author was born in Agam
Re
gency
,
West
Sumatera, Indon
esia, in 1967. H
e
received th
e B.
S Degree in
electrical eng
i
neer
in
g from Universitas Bung Hatta, Padang, Indon
esia,
in 1994, the M.S
.
degr
ee in
el
ectr
onic
engineering
from the Institute Technolog
y
B
a
ndung,
in 1997
,
and th
e Dr. Eng
.
Degr
ee f
r
om Universiti Tekn
ologi
Malay
s
ia in 2006. H
e
is
currently
a Senio
r
Lecturer in Mechatron
i
cs and
Robotics Engin
eering Dep
a
rtm
e
nt, F
a
c
u
lt
y of
Elec
tric
al and
Ele
c
troni
c Eng
i
neering
,
Univer
siti Tun
Hussein
Onn Mala
ysi
a
. His cu
rrent
research
int
e
rest
s
includ
e th
e
are
a
of Inte
llig
ent
S
y
stem
ident
i
fic
a
t
i
on and In
te
llig
en
t Control
S
y
s
t
em
.
Evaluation Warning : The document was created with Spire.PDF for Python.