J Control Theory
Appl
2013
11
(4) 656-660
DOI
1O
.1007/s11768-013-2156-1
A predictive functional control algorithm for
multivariable systems with time delay
LeiZHOU
1
,针,
Shumin
FEI
1, Jiacai
HUANG
2,
Junyong
ZHAI
1
l.
School
of
Automation
,
South
巳
ast
University
,
Nanjing
Jiangsu
210096
,
China;
2.School
of
Automation
,
Na
町
ing
Institute
ofTechnology
,
Nanjing
Jiangsu
211167
,
China;
Abstract:
To
improve the dynamic characteristics and
th
巳
cOllpling
capability, a
new
predictive fllnctional control
algorithm is proposed for strong coupling multivariable systems with time delay
, which
combin
巳
s
predictive functional
control and
d
巳
coupling
contro
1.
First, a decoupling control algorithm
is
proposed, in which first-order
mod
巳
Is
with time
delay are established
by
analyzing the amplitude-frequency and
phas
巳
-frequency
charact
巳
ristics
of
th
巳
d
巳
coupled
sub-
j
巳
c
t.
Then, a
controll
巳
r
is
designed for the single-variable subjects after decoupling
bas巳
d
on the principles of predictive
functional contro
1.
Th
巳
simulation
results show that this proposed algorithm has less online computation time and faster
tracking.
1t
can provide a
more
巳仔
ectiv
巳
control
for complex multivariable systems.
Keywords:
Multivariabl
巳;
Time-delay; Decoupling control; Predictive functional control
1 Introduction
To
巳
ffectively
apply
predictive
control
to fast
systems
,
a
new
predictiv
巳
control
algorithm
-
predictive
functional
control
(PF
C)
was
propos
巳
d
by
Richalet
et
a
1.
[1
-3]
in
th
巳
1980s.
PFC
introduces
the
conc
巳
pt
ofbasis
functions,
which
makes
it
differ
巳
nt
from
traditional
predictive
contro
l. Tradi-
tional
predictive
control
does
not
tak
巳
into
account
th
巳
struc
ture
of
control
input
when
it
uses
optimization
algorithms
to
obtain
th
巳
future
control
inpu
t.
As
a
result
, an
unknown
con-
trol
input
may
occur
when
the traditional predictive
control
is
applied
to fast servo-systems.
In
contrast
,
PFC
structural
2 A predictive functional control algorithm
for multivariable systems with time delay
izes
the
control
input
at
any given
tim
巳,
in
which
the
input
is a
linear
combination
of
selected
basis
functions
and
the
output
is a
weighted
combination
of
these
basis
functions.
The
onlin
巳
optirnization
is
used
only
to
obtain
the
weight
coeffìci
巳
nts
of
the
basis
functions,
which
can
further
obtain
the
futur
,巳
control
input
and
thus significantly
r
巳
duces
th
巳
arnount
of
online
computation.
PFC
is a
simple
algorithm
with less
online
computation
,
fast
tracking
and
good
robustness.
Du
巳
to
its low require-
ments
for
process
modeling
,
PFC
can
be
applied
to various
systems
, e.g.,
singl
巳
-variabl
巳,
multivariable,
unstabl
巳,
time-
delay
and
constrained
syst
巳
ms.
To
date
,
most
of
PFC
al-
gorithms
focus
巳
d
on
single-variable
syst
巳
ms
and
thus they
have
been
well studied.
However
,
more
PFC
algorithms
have
been
conducted
for multivariable
systems
so
far
[4-
8].
Decoupling
control
is
an
effective solution to multivariable
systems
with
strong
coupling
and
time
delay.
It
splits
the
complicated
system
into
independent
subsystems
by
de-
coupling
[9].
This
paper
pr巳
S
巳
nts
a multivariable predictive
functional
control
(MPFC)
algorithm
bas
巳
d
on
decoupling
contro
l.
Th
巳
simulation
results
show
that
MPFC
is a practi-
cal
and
effective
solution
to multivariable systems.
Received
5
July
2012;
revised
1
August
2013.
2.1
Decoupling
control
algorithm
D
巳
coupling
control is to
eliminat
巳
the
coupling
by
adding
a
set
of
decoupling
device to the
system
,
and
then
to
divide
it
into
independent
single-loop
systems
[10].
A
dynamic
decoupling
control
method
was
used
for
mul-
tivariable
systems
in this
pap
旺
A
multivariable
system
with
time
delay
is
assum
巳
d
as follows:
G
l1
(s)e-
1llS
G
12
(s)e-
Z,
2
S
...
Gln(s)e-lJnS
I
G21(s)e-121S
G22(s)e-122s
...
G
2n
川(归
ωS
功)汩
e
一
-1
如
G(
伊
s)=
I
Gnl(s)e-lnlS Gn2(s)e-ln2S
...
Gnn(s)e-lnnS
(1)
The
decoupling
controller
is
1
d
叫
s)
...
d
川
s)
I d
21
( S ) 1 .
..
d
2n
( S )
D(s) = I -
.'
. I . (2)
dn1(s) d
n2
(s)
...
1
Aecordi
吨
to
the
d巳
co
叩
li
吨
principles
,
G(
s
)D(
s)
is
a
diagonal
matrix
,
which
is
expressed
as follows:
G
l1
(s)e-l
l1
Sd12(S)
+...
+
Gln(s)e-
Z,
nSdn2(S)
=0
,
G
l1
(s)e-l
l1
Sdln(S) + ... +
G
川市
-1
川
=0
,
Gnl(s)e-lnlS
+...
+ Gnn(s)e-lnnSdnl(S) = 0,
G
n1
(s
)e-lnlSdl(n
一月
(s)
+.
+Gnn(s)e-lnnSdn(n_
1)
(s) = 0
(3)
tCorresponding
autho
r.
E-mail:
seuzl@aliyun.com.
Te
L:
+86-25-86118320
This
work
was
supported
by
the
National
Natural
Science
Foundation
of
China
(Nos.
61104085
,
61104068
,
61273119)
,
the
Natural
Science
Foundation
of
Jiangsu
Province
(No.
BK2
0l
0200)
,
and
the
Natural
Science
Foundation
of
Jiangsu
Province
Department
of
Education
(No.
II
KJB51
0005).
@
South
China
University
of
Technology
and
Acad
巳
my
of
Mathematics
and
Systems
Science
,
CAS
and
Springer-
V
,巳
rlag
Berlin
Heidelberg
2013