Indonesian J
ournal of Ele
c
trical Engin
eering and
Computer Sci
e
nce
Vol. 2, No. 1,
April 201
6, pp. 32 ~ 39
DOI: 10.115
9
1
/ijeecs.v2.i1.pp32
-39
ļ®
32
Re
cei
v
ed
Jan
uary 23, 201
5
;
Revi
sed Ma
rch 9, 2
016;
Acce
pted Ma
rch 2
0
, 2016
A New Electrode Regulator System Identification of Arc
Furnace Based
on Time-Variant Nonlinear-Linear-
Nonlinear Model
Jinfeng Wan
g
*
1
, Shoulin
Yin
2
, Xue
y
in
g Wang
3
Shen
ya
n
g
Nor
m
al Univ
ersit
y
No.25
3
, Hua
n
g
He Bei Street, Hua
ng Gu Dist
r
ict, Shen
yan
g
,
P.C 11003
4 - Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 1057
41
732
5
@
qq.com
1
, 35
272
02
14@
qq.c
o
m
2
, 75166
17
1
3
@q
q.com
3
A
b
st
r
a
ct
In this
pap
er,
w
e
express
ar
c furnac
e e
l
ec
tr
ode r
egu
lator
system as
a
time-v
aria
nt n
o
n
lin
ear-
line
a
r-n
onl
in
ea
r mode
l. On t
h
is b
a
sis, w
e
prop
os
e an on
line
id
entific
ati
on met
hod
ba
sed
o
n
non
lin
ear-
linear-nonlinear m
o
del system
. This
new schem
e solves
the
problem
of
m
o
del var
i
ation and predict
i
on
precisi
on d
e
cli
ne caus
ing by
time-vary
i
ng
of arc char
acteristic. In order to dispos
e
the difficulty o
f
para
m
eters se
parati
on i
n
the
onli
ne i
d
e
n
tific
a
tion pr
oc
ess,
this new
met
h
o
d
ado
pts the
mind of u
pdat
e the
parameters of linear parts
and nonlinear parts respectiv
e
ly. It reali
z
es
the param
e
ters s
e
paration of syste
m
effectively. Simu
lati
on resu
l
t
s show
that
this me
th
od c
an track the chan
ges of ar
c characteristi
cs
effectively. Tha
t
it achieves th
e ai
m of real
-ti
m
e
mo
nitor
i
ng
and co
ntroll
in
g system p
a
ra
meters.
Ke
y
w
ords
: Arc furnace el
ectrode re
gul
ator, onli
ne i
d
e
n
tific
a
tion, ti
me-var
ying, no
nli
n
e
a
r
-
line
a
r-n
onl
ine
a
r
m
o
del system
Copy
right
Ā©
2016 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Arc furna
c
e e
l
ectro
de
regu
lator sy
stem
remai
n
s
a co
nst
ant
a
r
c
re
sist
a
n
c
e
by
a
d
just
in
g
the length
of the arc bet
wee
n
the el
e
c
trod
es
and
the burden t
o
optimize o
u
tput po
wer
and
redu
ce
en
erg
y
con
s
um
ptio
n. Wh
en d
e
si
gning
ele
c
tro
de regul
ator
system, a
n
a
c
curate
sy
ste
m
model
ca
n p
r
ovide the
ba
sis for the
cal
c
ulation
of the
controlled
qu
antity and l
a
y the fou
ndati
on
for reali
z
ing t
he cont
rol go
als. Ho
weve
r, the
non-line
a
r of electrod
e regul
ator system and time-
varying of ele
c
tri
c
arc h
a
ve
increa
sed th
e difficu
lty in
modelin
g. Re
feren
c
e [1] p
r
opo
se
s a n
e
u
r
al
adaptive
PS
D
di
sp
ersive decouplin
g controlle
r wh
i
c
h
combi
n
ing
neural adapti
v
e
PSD algo
rithm
with disp
ersiv
e
deco
uplin
g netwo
rk. Yu F etc [2
] show a new IMC controll
er in
cluding two
RBF
neural netwo
rks. It adjusts cente
r
vecto
r
s a
nd
the sh
ape pa
ram
e
ters
of the networks onlin
e. P
Guan [3] re
prese
n
ts ele
c
trode re
gulat
o
r
system
s of indu
strial a
r
c
f
u
rna
c
e
with g
enetic al
gorit
hm
predi
ctive co
ntrol an
d de
signs a
detail
ed dynami
c
matrix cont
ro
ller to dimini
sh the p
r
edi
ctive
error a
nd g
e
t a desi
r
e
d
sy
stem outp
u
t. Ho
wever, the
y
donāt take ti
me-varyin
g
chara
c
te
risti
c
s of
electri
c
a
r
c in
to con
s
ide
r
ati
on. The est
a
blish
ed
mod
e
l
of the electri
c
arc furn
ace doe
s not refle
c
t
sy
st
em cha
r
a
c
t
e
ri
st
ic
s
a
c
c
u
rat
e
ly
.
We p
r
op
ose
the online i
d
entification m
e
t
hod b
a
sed
on no
nlinea
r-linear-n
online
a
r (N-L-
N)
mod
e
l sy
stem for th
e first time. In thi
s
p
ape
r, we e
x
press th
e a
r
c furna
c
e
ele
c
trod
e
regul
a
t
or
system a
s
N-L-N
mod
e
l of
para
m
et
ers t
i
me-varyin
g
i
n
linea
r pa
rt
s for time-va
r
ying arc. And
we
prop
ose a on
line identifica
t
ion method t
o
identify
time-varyin
g
parameter of
system. To ha
n
d
le
the difficulty
on pa
ram
e
ters sepa
ratio
n
within id
entification process
[4,
5],
this pape
r
d
r
a
w
s the
identificatio
n idea of rel
a
xation iterativ
e me
thod [6
] and com
b
i
nes
N-L-N system stru
ct
ure
cha
r
a
c
teri
stic to divide
ev
ery mo
ment i
t
erative
cal
c
u
l
ation into
three
step
s. Th
en thi
s
sche
me
update
s
line
a
r
part
s
an
d n
online
a
r pa
rt
s in turn.
It achieve
s
the e
ffective sepa
ration of syste
m
para
m
eters a
nd reali
z
e
s
o
n
line identificat
ion N-L-N system time-v
arying pa
ram
e
ters.
2. Electrod
e
Re
gulating Sy
s
t
em Stateme
n
t
Ele
c
tr
ic
ar
c fur
n
ac
e
e
l
ec
tr
od
e
ad
jus
t
me
nt s
y
s
t
em
con
s
ists
of two
pa
rts, hyd
r
auli
c
system
and ele
c
tri
c
a
r
c. Its structu
r
e frame is a
s
figure 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ļ®
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 32 ā 39
33
Figure 1. Electri
c
arc furn
a
c
e ele
c
tro
de
adju
s
tment system stru
ctu
r
e
In figure 1,
u
is i
nput
co
ntrolled
q
uantit
y.
v
is p
r
op
o
r
tional i
nput
value.
d
i
s
el
ectro
de
positio
n.
R
is
arc
res
i
s
t
ance. Its
unit is
ā¦
.
2.1 H
y
draulic Sy
stem Model
The hyd
r
auli
c
sy
stem co
nsi
s
ts of a p
r
opo
rtion
a
l valve and hyd
r
auli
c
cylin
de
r se
rie
s
.
Becau
s
e
p
r
o
portion
al valv
e ha
s de
ad
zone fe
atures
.
At the
sam
e
time the
rest
riction
of
sp
o
o
l
displ
a
cement
re
sult
s in
up
per and
lo
we
r of valve o
u
tp
ut. So the
pro
portion
al valv
e cha
r
acte
ri
stic
can b
e
exp
r
esse
d a
s
a pro
portio
n
a
l
comp
one
nt betwee
n
d
ead zone
and saturation
cha
r
a
c
t
e
ri
st
ic
s.
The hydrauli
c
cylinde
r fe
ature
s
ca
n b
e
expr
e
s
sed
by 3-order transfe
r functi
on. Afte
r
discreti
zing, it can be o
b
tai
ned by:
3
3
2
2
1
1
3
3
2
2
1
1
1
1
1
)
(
)
(
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ«
ļ«
ļ½
z
a
z
a
z
a
z
b
z
b
z
b
z
v
z
l
y
y
y
(1)
2.2 Electric
Arc Mod
e
l
Arc an
d po
we
r system mo
d
e
l are stati
c
n
onlin
e
a
r varia
b
le ch
ara
c
te
ri
stics, so arc voltage
cha
r
a
c
teri
stics ca
n be expressed a
s
the
following approximate relation:
bl
a
U
arc
ļ«
ļ½
(2)
Whe
r
e
U
arc
is arc voltage, i
t
s unit i
s
V
.
l
i
s
a
r
c length, i
t
s unit i
s
cm
.
a
is
arc
catho
de an
d a
nod
e
voltage drop,
its unit is
V
.
We
can
co
nsider it a
s
a
consta
nt.
b
is
electri
c
a
r
c d
r
op g
r
a
d
ient,
its
unit is
V/c
m
. In the
begin
n
i
ng of
sm
eltin
g
,
b
ā
8
.
b
w
ill
d
e
c
r
e
as
e as
te
mp
er
a
t
u
r
e
r
i
s
e
. Elec
tr
ic
ar
c
furna
c
e temp
eratu
r
e drop
gradi
ent ch
an
ges ove
r
fu
rn
ace temp
erature which refl
ects the
syst
em
time variabilit
y.
Powe
r syste
m
model is e
quivalent to R-L ci
rcuit.
2
2
2
)
(
)
(
arc
d
arc
arc
d
p
I
X
U
I
R
U
ļ«
ļ«
ļ½
(3)
Whe
r
e
U
p
is
transfo
rme
r
seco
nda
ry voltage, its unit i
s
V
.
R
d
is
sh
ort-n
e
t re
sist
ance, its unit
is
ā¦
.
X
d
is sho
r
t-net impeda
nce.
I
arc
is
arc
c
u
rrent,its
unit is
A
.
U
arc
is arc voltage, its unit is
V
.
2.3 Electrod
e
Regula
t
ing
Sy
stem Model
Bec
a
us
e
ele
c
tric a
r
c fu
rn
a
c
e el
ectrode
adju
s
tment
system con
s
ist
s
two
pa
rts, h
y
drauli
c
system
an
d electri
c
arc; electri
c
arc
f
u
rna
c
e
ele
c
trode reg
u
lato
r
syste
m
can
con
s
i
s
t
of
th
ree
parts, d
ead
zon
e
nonli
n
e
a
r characte
ri
stic, 3-or
d
e
r
linear cha
r
acteri
stic an
d
arc
nonlin
ear
cha
r
a
c
teri
stic in serie
s
. It matche
s the
N-L
-
N model.
In the id
entification
process, it u
s
e
s
p
o
l
y
nomial fun
c
t
i
on to
app
roxi
mate an
d
rep
l
ace
the
nonlin
ear
cha
r
acte
ri
stics caused
by arc and po
we
r supply syste
m
. Time-varyin
g
paramete
r
s of
arc
ch
ara
c
te
ri
stics incre
a
se
the difficulty of ident
ificati
on in the pol
ynomial fun
c
tion co
efficien
ts.
To simplify the identificatio
n method, th
e
system is tra
n
sformed a
s
follows.
Let
U
d
=bl
be
arc
colum
n
v
o
ltage
dro
p
.
Putting it into
(1
) a
n
d
(3
)
and
co
mbinin
g (2), it
can g
e
t:
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
ļ®
A New Ele
c
trode Regulat
o
r
System
Ide
n
tification of Arc Fu
rna
c
e
Based o
n
ā¦
(J
infeng W
a
ng)
34
3
3
2
2
1
1
3
3
2
2
1
1
3
3
2
2
1
1
3
3
2
2
1
1
1
1
1
1
)
(
)
(
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ«
ļ«
ļ½
ļ
ļ
ļ
ļ«
ļ«
ļ½
z
a
z
a
z
a
z
b
z
b
z
b
z
a
z
a
z
a
z
b
z
b
z
b
b
z
v
z
U
y
y
y
b
(4)
2
2
2
)
(
)
(
)
(
arc
d
arc
d
p
I
X
bl
a
I
R
l
F
U
ļ«
ļ«
ļ«
ļ½
ļ½
(5)
From
form
ula
(4
) a
nd
(5
)
we
ca
n
kno
w
that t
he
equi
valent tran
sfo
r
mation
me
rg
es time
-varyi
ng
para
m
eters i
n
to linea
r m
o
d
e
l pa
rt. So th
e arc fu
rn
ace electrode
reg
u
lator syste
m
is equivale
nt to
N-L
-
N mod
e
l
of linear
pa
rt para
m
eters ti
me-varyin
g
. It simplifie
s the
identificatio
n
pro
c
e
ss
du
rin
g
the arc time
-v
arying. The n
e
w mod
e
l structure is a
s
figure 2.
Figure 2. Electri
c
arc furn
a
c
e ele
c
tro
de
adju
s
tment system model
3. N-L
-
N M
odel
Param
e
ter I
d
entific
a
tion
3.1
N-L
-
N Sy
stem
Param
e
teri
zatio
n
Before
syste
m
identification, we mu
st
make
pa
ram
e
terization fo
r N-L-N
syste
m
[7, 8].
The dea
d zo
n
e
saturation n
online
a
r characteri
st
ics
ca
n be expre
ssed as the foll
owin
g equati
on:
ļ„
ļ½
ļ½
ļ½
p
i
i
i
k
u
g
c
k
u
G
k
v
1
))
(
(
))
(
(
)
(
(6)
Whe
r
e
u(
k)
is input controll
ed qua
ntity at
k
time.
v(
k)
i
s
propo
rtion
a
l valve output at
k
time.
c
i
is un
determined pa
ram
e
ter.
g
i
( Ā· )
i
s
polynomial fu
nction.
From fo
rmula
(5), we kno
w
that
F(l
)
is invertible function. Th
e i
n
vertible function is as
follows
[14-16].
ļ„
ļ½
ļ
ļ
ļ½
ļ½
q
i
i
i
k
R
f
d
k
R
F
k
l
1
1
1
))
(
(
))
(
(
)
(
(7)
Whe
r
e
l(k)
i
s
arc le
ngth at
k
time.
R(k
)
i
s
ar
c re
sist
a
n
c
e at
k
time.
)
(
1
ļ
ļ
i
f
is polynomi
a
l function.
The linea
r pa
rt can b
e
obta
i
ned by formu
l
a
(4). Combi
n
ing (4
), (6
) a
nd (7
) gets:
ļ„
ļ„
ļ„
ļ„
ļ½
ļ½
ļ½
ļ
ļ½
ļ½
p
i
i
i
m
m
q
j
j
j
n
n
k
u
g
c
z
b
k
R
f
d
z
a
1
3
1
1
1
3
0
))
(
(
)
(
))
(
(
)
(
(8)
Whe
r
e
a
0
=1
,
a
1
=a
1
z
-1
,
a
2
=
a
2
z
-2
,ā¦ā¦;
b
1
=b
1
z
-1
,
b
2
=b
2
z
-2
,ā¦ā¦
So, we can i
dentify the param
eters
a
i
,
b
i
,
c
i
,
d
i
in fo
rmula (8) onl
y by input-ou
t
put data.
Finally we obt
ain the syste
m
model.
Evaluation Warning : The document was created with Spire.PDF for Python.
ļ®
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 32 ā 39
35
3.2
N-L
-
N Sy
stem Param
e
ter
s
Identifi
cation
In the smeltin
g
process, el
ectr
ic
arc
e
x
ists
time
-
v
ar
ying
ch
a
r
acte
rist
ics. In
order to re
al-
time control system ch
ang
es an
d get the best cont
rol
signal, we m
u
st study the
N-L
-
N syste
m
's
online ide
n
tification meth
o
d
. (8) is a
bbreviated to:
0
)
(
ļ½
ļ·
k
ļ¦
ļ±
(9)
Whe
r
e
Īø
=[
d
1
,ā¦,
d
q
,
a
1
d
1
,ā¦,
a
1
d
q
,ā¦,
a
3
d
1
,ā¦,
a
3
d
q
,ā¦,
b
1
c
1
,ā¦,
b
1
c
p
,ā¦,
b
3
c
1
,ā¦,
b
3
c
p
].
))]
3
(
(
,
)),
3
(
(
,
)),
(
(
,
)),
(
(
,
ļ¼
))
3
(
(
,
)),
3
(
(
,
)),
1
(
(
,
)),
1
(
(
[
)
(
1
1
1
1
1
1
1
1
ļ
ļ¼
ļ
ļ¼
ļ¼
ļ¼
ļ
ļ¼
ļ
ļ¼
ļ
ļ¼
ļ
ļ½
ļ
ļ
ļ
ļ
k
u
g
k
u
g
k
u
g
k
u
g
k
R
f
k
R
f
k
R
f
k
R
f
k
p
p
q
q
ļ¦
The formula
(9)
sh
ows
that usin
g e
x
isting matu
re onlin
e ide
n
tification m
e
thod
s, su
ch
as
recursive lea
s
t squ
a
res m
e
thod (RLSM
)
, can g
e
t vector
Īø
o
n
line.
However,
Īø
is
co
mpo
s
ed of
prod
uct
of
a
i
,
b
i
,
c
i
a
nd
d
i
, it
can
not get sep
a
rate
p
a
rameters of
a
i
,
b
i
,
c
i
an
d
d
i
.
Papers [9] us
e
sing
ular val
u
e de
comp
osit
ion (SV
D
) m
e
thod to offlin
e
sep
a
rate fou
r
sets of p
a
ra
meters. That
is
not suitabl
e for the onlin
e identificat
io
n of time-varyin
g
N-L
-
N syste
m
.
Becau
s
e it is difficult to separate p
a
ra
m
e
te
rs.T
his pa
per draws th
e identificatio
n idea of
relaxation
iterative method.
According
to
N-L-N
sy
ste
m
spe
c
ific 3-parts st
ru
cture cha
r
a
c
teri
stic,
it first fixes two p
a
rts
pa
rameters a
nd
then adj
ust
s
the pa
ram
e
ters of the oth
e
r parts to
achi
eve
para
m
eter se
paratio
n.
Model e
rro
r is:
ļ„
ļ„
ļ„
ļ„
ļ½
ļ½
ļ½
ļ
ļ½
ļ
ļ½
p
i
i
i
m
m
q
j
j
j
n
n
k
u
g
c
z
b
k
R
f
d
z
a
k
1
3
1
1
1
3
0
))
(
(
)
(
))
(
(
)
(
)
,
(
ļ±
ļ„
(10
)
Erro
r functio
n
is defined a
s
:
2
)
,
(
2
1
)
,
(
ļ±
ļ„
ļ±
k
E
E
ļ½
(11
)
We con
c
lude
the identification pro
c
e
d
u
re with
i
n
a sampli
ng pe
ri
od. Firstly, the initial
iteration val
ues
are giv
en:
)
1
(
Ė
A
,
)
1
(
Ė
B
,
)
1
(
Ė
C
,
)
1
(
Ė
D
. Sett
ing
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
(
Ė
k
C
,
)
(
Ė
k
D
as t
he
estimation val
ue of para
m
e
t
ers at
k
sam
p
ling time.So
)]
(
Ė
),
(
Ė
),
(
Ė
),
(
Ė
[
)
(
Ė
3
2
1
0
k
a
k
a
k
a
k
a
k
A
ļ½
)]
(
Ė
),
(
Ė
),
(
Ė
[
)
(
Ė
3
2
1
k
b
k
b
k
b
k
B
ļ½
)]
(
Ė
,
),
(
Ė
),
(
Ė
[
)
(
Ė
2
1
k
c
k
c
k
c
k
C
p
ļ¼
ļ½
)]
(
Ė
,
),
(
Ė
),
(
Ė
[
)
(
Ė
2
1
k
d
k
d
k
d
k
D
q
ļ¼
ļ½
Step 1. Selecting input-o
utput data
R(
k
)
,
R(
k-1
)
,
R(
k
-
2
)
,
R(
k-
3)
,
u(k-1)
,
u(k-2)
,
u(k-3)
.
Step 2. Usi
n
g
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
(
Ė
k
C
,
)
(
Ė
k
D
as mo
del pa
ramete
rs. Calculatin
g error fu
ncti
on value
based on formula (1
0).
Step 3. Fixin
g
pa
ramete
rs
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
(
Ė
k
C
and
adju
s
ting
)
(
Ė
k
D
to ma
ke th
e e
rro
r
function
minimum. Th
en we
can g
e
t
the new parameter
)
1
(
Ė
ļ«
k
D
.
Step 4. Usin
g
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
(
Ė
k
C
,
)
1
(
Ė
ļ«
k
D
as mo
del
param
eters. Cal
c
ulating
e
rro
r fun
c
tion
value
based on formula (1
0).
Step 5. Fixin
g
pa
ram
e
ters
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
1
(
Ė
ļ«
k
D
and a
d
ju
sting
)
(
Ė
k
C
to make
the e
r
ror fu
nction
minimum. Th
en we
can g
e
t
the new parameter
)
1
(
Ė
ļ«
k
C
.
Step 6.
Usi
n
g
)
(
Ė
k
A
,
)
(
Ė
k
B
,
)
1
(
Ė
ļ«
k
C
,
)
1
(
Ė
ļ«
k
D
a
s
m
odel pa
ramet
e
rs. Calculati
ng error fun
c
tio
n
value ba
sed
on formul
a (1
0).
Step 7. Fixin
g
param
eters
)
1
(
Ė
ļ«
k
C
,
)
1
(
Ė
ļ«
k
D
and adj
usting
)
(
Ė
k
A
,
)
(
Ė
k
B
to
mak
e
the er
ror
function mini
mum. Then
we can g
e
t the new p
a
ra
met
e
r
)
1
(
Ė
ļ«
k
A
,
)
1
(
Ė
ļ«
k
B
.
Step 8. Giving the new in
p
u
t signal
u(
k+1
)
, let
k=
k
+
1
, return s
t
ep1.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
ļ®
A New Ele
c
trode Regulat
o
r
System
Ide
n
tification of Arc Fu
rna
c
e
Based o
n
ā¦
(J
infeng W
a
ng)
36
Therefore,
according
to th
e N-L
-
N
syst
em st
ru
cture,
ea
ch ite
r
atio
n can b
e
divi
ded into
three
step
s. I
t
update
s
p
a
rameters of
1-linear pa
rt an
d 2-nonlin
ea
r part
s
in tu
rn,
finally, we
can
get the p
u
rpo
s
e
of pa
rame
ters
se
pa
rati
on. Some
pa
ramete
rs
a
d
ju
stment o
p
timi
zation
metho
d
s
can b
e
used i
n
step 3, 5, 7, such as g
r
ad
ient algorith
m
or least
squa
re metho
d
.
4. Simulation Results a
nd Analy
s
is
As
mention
e
d
earlier, we take ele
c
tri
c
arc fu
rna
c
e o
f
one factory
as an exam
ple and
con
d
u
c
t onlin
e identificatio
n simulatio
n
fo
r arc furnace electrode a
d
justme
nt system.
The p
r
op
ortio
nal valve d
e
a
d
zone l
ength
of ar
c fu
rna
c
e ele
c
tro
de a
d
justme
nt sy
stem i
s
1. Its output uppe
r an
d lo
wer
bou
nd
s is
ļ±
10. Sampling pe
riod T
=
0.02s. Th
e d
i
scretization
transfe
r fun
c
tion of hydrauli
c
cylind
e
r can
be expre
s
se
d by:
3
2
1
3
2
1
1
3011
.
0
552
.
1
251
.
2
1
0043
.
0
0254
.
0
0082
.
0
)
(
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ
ļ«
ļ
ļ«
ļ«
ļ½
z
z
z
z
z
z
z
P
(12
)
In the beginni
ng of the sme
l
ting, arc pressure dr
o
p
gra
d
ient is 8. At
the end of sm
elting,
arc p
r
e
s
sure drop g
r
a
d
ient
is 1.
In step 3, 5, 7, we use
gradie
n
t algorithm
t
o
a
d
just
ea
ch p
a
ram
e
t
e
r.
A
d
just
in
g
para
m
eters a
r
e as follo
ws.
ļ„
ļ½
ļ
ļ
ļ
ļ½
ļ¶
ļ¶
ļ
ļ½
ļ«
3
0
1
)
,
(
))
(
(
)
(
)
(
Ė
)
(
Ė
)
(
)
(
Ė
)
1
(
Ė
j
i
j
d
i
i
d
i
i
k
j
k
y
f
k
a
k
d
k
d
k
k
d
k
d
ļ±
ļ„
ļ¬
ļ„
ļ¬
(13
)
ļ„
ļ½
ļ
ļ
ļ½
ļ¶
ļ¶
ļ
ļ½
ļ«
3
0
)
,
(
))
(
(
)
(
)
(
Ė
)
(
Ė
)
(
)
(
Ė
)
1
(
Ė
j
j
j
c
i
i
c
i
i
k
j
k
u
g
k
a
k
c
k
c
k
k
c
k
c
ļ±
ļ„
ļ¬
ļ„
ļ¬
(14
)
ļ„
ļ½
ļ
ļ
ļ«
ļ
ļ½
ļ¶
ļ¶
ļ
ļ½
ļ«
p
j
i
j
a
i
i
a
i
i
k
j
k
u
f
k
d
k
a
k
a
k
k
a
k
a
1
1
)
,
(
))
(
(
)
1
(
)
(
Ė
)
(
Ė
)
(
)
(
Ė
)
1
(
Ė
ļ±
ļ„
ļ¬
ļ„
ļ¬
(15
)
ļ„
ļ½
ļ
ļ«
ļ
ļ½
ļ¶
ļ¶
ļ
ļ½
ļ«
q
j
j
j
b
i
i
b
i
i
k
j
k
y
g
k
c
k
b
k
b
k
k
b
k
b
1
)
,
(
))
(
(
)
1
(
)
(
Ė
)
(
Ė
)
(
)
(
Ė
)
1
(
Ė
ļ±
ļ„
ļ¬
ļ„
ļ¬
(16
)
Whe
r
e
a
ļ¬
,
b
ļ¬
,
c
ļ¬
and
d
ļ¬
is step len
g
th of param
eter adju
s
tme
n
t. Set
a
ļ¬
=
b
ļ¬
=
c
ļ¬
=
d
ļ¬
=1
0
-3
.
System ignite
signal i
s
sin
u
s
oid
a
l sig
nal. Each initial p
a
ram
e
ter valu
e is 0.9 times as the
true value of initial simulati
on paramete
r
s. Afte
r 4000
0 step
s (i.e.8
00s) si
m
u
lati
on identification,
the true value
and estimati
on value of e
a
ch p
a
ra
mete
r are a
s
figure 3, 4.
Evaluation Warning : The document was created with Spire.PDF for Python.
ļ®
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 32 ā 39
37
Figure 3. Tru
e
value and e
s
timated valu
e of param
eter A com
pari
s
on
Figure 3 sh
o
w
s that mo
de
l param
eter A
dose
n
o
t ch
ange d
u
rin
g
identificatio
n pro
c
e
ss.
Its estim
a
ted
value
A
Ė
co
nverges to a
true
value rapidly.
As
A
Ė
ha
s a
fa
st convergen
ce
sp
eed,
we gives the i
dentificatio
n result within 2
00s.
Figu
re 4
sho
w
s that the paramete
r
B change
s with
time in th
e i
dentificatio
n
pro
c
e
ss.
The
estim
a
ted v
a
lue
B
Ė
ap
pea
rs
som
e
d
e
viation at th
e
begin
n
ing of identificatio
n. But it
tracks the tr
ue value
in a short time. The sim
u
lation take
s
60.
3219
s le
ss th
an the sim
u
la
tion time 800
s. Thu
s
we
can draw a
co
nclu
sio
n
that this ne
w sche
me
can b
e
appli
e
d for online id
entification of
param
eters of the electri
c
arc furna
c
e systems.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
ļ®
A New Ele
c
trode Regulat
o
r
System
Ide
n
tification of Arc Fu
rna
c
e
Based o
n
ā¦
(J
infeng W
a
ng)
38
Figure 4. Tru
e
value and e
s
timated valu
e of param
eter B com
pari
s
on
5. Conclusio
n
s
This pap
er cl
assifies arc
f
u
rna
c
e
ele
c
trode re
g
u
lator system
a
s
p
a
ram
e
ter tim
e
-varyin
g
N-L
-
N syst
e
m
of linear
p
a
rt. And N-L-N mod
e
l
is
p
a
ram
e
teri
zed
base
d
on
arc furn
ace ele
c
trod
e
regul
ator
system. In order to solve the
difficult
prob
lem of online
identification
N-L
-
N
syste
m
para
m
eter
se
paratio
n, we
use th
e rel
a
xation iter
ative
identification
method reali
z
ing p
a
ramet
e
r
sep
a
ratio
n
in
N-L
-
N sy
ste
m
identificati
on p
r
o
c
es
s.
So we
achiev
e the pu
rp
ose of sy
stem o
n
line
identificatio
n. The
simul
a
tion results
sh
ow that
th
e
new
metho
d
can i
dentify the ele
c
tri
c
a
r
c
furna
c
e
elect
r
ode
adju
s
tm
ent syste
m
with pa
ram
e
ter time
-varyi
ng effectively
and e
n
sure
the
accuracy of system model
in the smeltin
g
pro
c
e
ss.
Ackn
o
w
l
e
dg
ment
This wo
rk
i
s
partially
supp
orted by
the Na
tion
al Natural S
c
ien
c
e
Found
ation o
f
Chin
a
(609
701
12).
Referen
ces
[1]
He J, Li
u Y, Yu
S, et al. Ne
ura
l
ad
aptiv
e PSD
deco
u
p
lin
g co
ntroll
er an
d its
app
licati
on
in t
h
ree-
phas
e
electro
de adj
u
s
ting
s
y
stem of
submer
ge
d ar
c furnace.
J
our
nal of
Ce
ntral South
U
n
iv
ersi
ty.
2013, 2
0
:
405-
412.
[2]
Yu
F, Ma
o
Z.
Interna
l
mo
del
control f
o
r el
ec
trode i
n
e
l
ectric
arc furn
ace
ba
sed o
n
rbf
ne
ur
al n
e
tw
orks.
Contro
l and D
e
cision C
onfer
e
n
ce (CCD
C).
2012 2
4
th Ch
ine
s
e. IEEE. 2012
: 4074-4
0
7
7
.
[3]
Guan P, Liu
X, Gao Y.
Predictive control of
arc furnace b
a
sed o
n
ge
net
ic alg
o
rith
m
. C
ontrol a
nd
Decisi
on C
onfe
r
ence (2
01
4 CCDC). T
he 26t
h Chi
nese. IEE
E
. 2014: 33
85-
339
0.
[4]
Nordsj
o AE, Z
e
tterberg L
H
. Identif
ic
atio
n of certain time-v
ar
yin
g
non
lin
e
a
r W
i
ener a
nd
Hammerstei
n
s
y
stems.
IEEE Transactions on Signal Proc
essing
. 20
01; 49
(3): 577-5
92.
Evaluation Warning : The document was created with Spire.PDF for Python.
ļ®
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 1, April 2016 : 32 ā 39
39
[5]
Voros J. Identification of Ha
mmers
tein s
y
s
t
ems
w
i
th time
-var
ying p
i
ec
e
w
i
s
e-l
i
ne
ar ch
aracteristics.
IEEE Transactions on Circuits
and System
s II: Express Briefs
. 2005; 52(
12)
: 865-86
9.
[6]
Crama P, Schouke
n
s J. Hammerstein
āW
i
ener s
y
stem estimator initi
a
lizatio
n.
Autom
a
tica
. 200
4;
40(9): 15
43-
15
50.
[7]
Ding F
,
Li
u
X
P, Liu G. Iden
tification m
e
th
ods for H
a
mmerstein
non
lin
e
a
r s
y
stems.
D
i
g
ital Si
gn
al
Processi
ng.
20
11; 21(2): 2
15-
238.
[8]
Bai EW. A
n op
tima
l tw
o-stage ide
n
tificati
on
algor
ith
m
for Ha
mmerstei
nā
W
i
ener N
onl
in
ear Syste
m
s
.
Block-ori
ente
d
Nonl
in
ear S
y
stem Identificati
o
n. Spring
er Lo
ndo
n. 201
0: 27
-34.
Evaluation Warning : The document was created with Spire.PDF for Python.