Вы находитесь на странице: 1из 5

Control of Induction Motor using Poly topic LPV Models

Dalila KRAMAR, Abdesslem MAKOUF and Said DRD


LSPIE Laboratory, Depatmet of Electrical Engieering, University ofBatna, Email:
a makouf@yahoo.f and s did@yahoo.f
Abstac-A Gain scheduled control design for the stator
current is presented, the approach is novel in that the
gain scheduled design does not involve linearization
about operating points. Instead the motor dynamics are
brought to linear parameter varying form via state
transformation. A linear parameter varying system is
defned as a linear system whose dynamics depends on
unknown but measurable exogeneous parameter .The
current equations in the ( a,l) frame have a particular
structure, allowing to be written as an LPV system
because of afne dependence of rotational speed which
is taken as time varying parameter. This varying
parameter values can be measured on line during
control operations. The LMI based gain scheduled
controller using H-infnity synthesis and polytopic
representation is designed such that the robust
quadratic stability and robust quadratic performance
can be assured along the reference trajectory of the
varying parameter.
Keywords - LPV, Induction Motor, H-infnity, LMI, Gain
scheduling, polytopic representation.
I. INTRODUCTION
Induction motor modeling and contol have always
been one of the most challenging problems. Because
te dynamics of the induction motor are higly
nonlinear and quite uncertain. The trends i induction
motor control system design is to use effective robust
controlle design approaches such as Hocontrol [4]
ad other robust control approaches one of the recent
topics in control related felds is linear matrix
inequality (LMI) based contol system design
because of some advantages . For LPV systems a
taditional control metod is to design L TI controllers
for several points, ad te using interpolatig
technique to obtain te control law over the entire
operating range. The main drawback of this is a lack
of high peformance of robustness, even of stability
[8]. In the famework of LPV systems proposed in
[3], [4] [9],[ 10], and [ 1 1] te controller synthesis
problem is formulated as convex optimization
problem. Afer solving some linear matrix inequalities
(LMI) the so called self-scheduled controller is
obtained by a simple interpolation and ten stability
and certain performance bounds are guaranteed along
all possible trajectories of OCt). A self scheduled
LPV controller update its self on line using parameter
measurement, so tat te changing plants dynamics
are taken into account.
In tis pape, a gain scheduled curent control system
in te (a,) fame is designed for te whole
operating region using Linear Matrix Inequality
optimization and te Linear Paamete Varying (LPV)
systems famework. The system non linearity's
inheret to the rotation speed are taken into accout
by an afne LPV model with polytopic
representation. A optimal time-varying controlle for
feld -oriented (see[5] and [6]) control calculated
trough a LMI problem formulation to minimize the
L2 gain criterion and assure te stability of te closed
loop composed by the diferent contol objectives.
II. LINEAR PARAMETER VARYING PLANT
LPV plants are described by state space equations of
te form
x = A(O(t)x + B(O(t)u
y = C(O(t)x +
D
(O(t)u
(1)
Where x, y and u denote state vector, measured
output vector and control input, respectively. 0 is a
vector of time varying plant paametes ad plat
matrices are fxed fction of te O. in practice o. Ca
be the time varying physical parameter such as
velocity, damping, stifess and etc, and be given by
(2)
The time varying parameter 0 belongs to a parameter
polytope 8and varies wit vertices 0, O
l
f'''' O
r
of this
polytope
(3)
When OCt) udergoes large variations duing control
operation, it is ofen ipossible to achieve high
performance wit a single robust L TI controller if
te measuement of the OCt) are available in real time
during control operation , te designed controllers
have the same parameter dependence as te plant .
The controller form is :
XK = AK(O(t)x + BK(O(t)y
u = CK(O(t)x +
D
AO(t)y
(4)
Where y is the output measured vector and u is the
control input. According with the parameter variation
tis controller is continuously adjusted in the plat
dynamics ad maintains stability and good
performaces.
III. LMI BASED GAIN SCHEDULED CONTROL
FORMULATION
LMI based gain scheduled H contol formulation
given hee is aimed for practical use. For more detail,
consider state- space representation of the plat ad
t

c:n
[
t'
E
M_

]
(5)
Y C2 D21 0
U
(6)
Where x E Rn and .
.
E R
k
ae te plant ad te
controller state vector, respectively. x ad y denote
te contolled output and te measued output vectors.
u is the control input ad w is te distubace input
vector. Combining te two systems, te closed-loop
system can be obtained by
_.

A
l
BCI
[
_
.

Z Ccl Dcl
w
(7)
Closed loop matrices ACII B
cll Ccll Dl
ae :

ACI BCI
[
[
AD + AoE Bo + BOD21
]
Ccl DCI
=
Co + D210E Dll + D120D21
(8)
where
A 0 B1
0 B2
[A' B
o

]
=
0 0 0 Ik
0
C
o
Dll
C1 0
Dll
0
D12 (9)
E D21
0 Ik
0 A
.
_
.
C2 0
D21 C
.
_
.
Note that controllers matrices ae collected into a
single matrix o.
Te Lyapunov fnction V(x) = .T px > 0 etablishes
global asyptotic stability for the closed- loop system
(7) . The L
2
induced norm fom w to z for L TI
system is bounded as
II
z
lb
<
II
w
lb
( 10)
Finally, there exists a positive defnite Lyapunov
fnction V(x) = xT px, P > 0 tat satisfes
d
-
V(x) + ZTZ
- y
2
wTw < 0
dt
( 11)
The validity of equality ( 1 1) is proved in [5]. The
Hoo suboptimal contol problem is equivalent to te
existence of a solution to te following inequality
for
XcI
> 0
[A
IXCI
T
+ Xcl
A
cl Xcl
Bcl
B
clXcl
-
y
I
Ccl Dcl
cI
Dcl
< 0
-
y
I
Solution of the LMI ( 12) requires
symmetric matrices R ad S such tat
[AR + RAT RC[ B1
]
N"NR C1R -
y
I
Dll < 0
B[
D
[
l
-
y
I
( 12)
to fnd two
(l3)
[ATS + SA SB1 C[
]
N
I
N
s B[S -
y
I
DI1
< 0
C1 Dll
-
y
I
( 14)
Where NR and N
s
denote basis of te null spaces of
(BI
,
DI2) ad (C2
,
D12) , respectively.
The above Hoo contol problem is valid only for LTI
system and can be extended for LPV sytems.
Let's consider state-space representation of LPV plat
[ A(e)
z
= C1(e)
y
C2(e)
( 15)
e is a vector of time varying system parmeters
matrices A (. ), B1 (. ), C1 (. ) ad
Dll
fxed fnctions
of the . B2
, C2
,
D12
f
D21 matrix ae indepenent of te
paaeter e because of tactability reasons. Finally ,
te solutio of Hoo control problem for LPV system
has te same form of LTI system as follows :
[Ai R + RAf R
cI B1
.
]
N" Cli R -
y
I
D1,i NR < 0
B[i D[li -
y
I
[Af S + SAi SB1i
cI
]
N
I B[i S -
y
I D[li
N
s
< 0
Cli D11i
-y
I
_R I>
0
I S
-
( 17)
( 18)
( 19)
Where Au Bli , Cli , ad
Dlli denote te parameter
values of A(e), B1 (e), C1 (e) and
Dll
(e) at te
vertices e = ei of the paramete polytope.
Te solution of inequalities ( 17), ( 18), and ( 19) is
possible using advaced sofware such as convex
optimization algoritms. The construction of the
controller matrix 0 fom R ad S matrices can be
done by the same convex programs.
IV. MODELING OF INDUCTION MOTOR SYSTEM
The state space model of stator cur et is in te (a, f)
fam s as follows :

=

Csa
I.
0
csp
0
T
MIT
T
w
11a * Ls
0
o
;Ia
o
* Ls [
(20)
o 0
Where (isa, isp) ae te two component of the stator
current ad (Csa, C sp) ae two component of the
magnetic fux, (Vsa,Vsp)is te stator voltage, ad W
is te rotational speed of the shaf acting as a
distubace.
The electromagnetic torque is given by:
M _
Te
=
P
L
r (I
s
;
)
and its associated equation of motion is:
dD
Te - Tl
=
J T
Where T, is the load torque
V. CONTROL STRATEGY
A. Vector control by fu orientation
This consists in orienting the rotor fux [6]. Thus, it
results the constraints given below in (24). The rotor
fux is oriented on the d axis.
rd
=
r
rq
= 0
B. Speed controller design
(24)
The PI speed contoller is designed with pole
placement method (= 0.707, fn= 17.3 rd/s).
C. Stator current controller design
The inner current controllers are designed in the (a,)
plant to avoid the discretization of the difeomorphism
related to the Park transformation and using the LMI
Gain scheduled Ho approach. This control is designed
independently of the speed controller.
VI. GAIN SCHEDULED Ho CONTROLLER DESIGN
The state space model of the system can be obtained
by
x = A(w) + Bu
y = Cx
(25)
Parameter dependence range of the plant due to
rotational speed is:
w E [0
,
wmaxl
(26)
Specifcally for this problem, the parameter vector
O(t) has te following convex decomposition
a 2:0 (27)
Where 0i gives the comer of polytopic parameter
range. The comer values of parameter range for this
problem are:
01= (0,0), 01 = (0, wmax)
At the comer values of 0
,
the plant matrix is:
G(O) = a1G(01) + a2G(02
)
Augmented plant can be given by

A(O) B1 B2

G (0) = C1 Dll D12


C2 D21 D22
(28)
(29)
(30)
Te desired performance specifcation on control
system is defned in terms of fequency shaping flters
W1 and W
2
placed at the output of the controller and
the plant, respectively.
Finally, using the parameter dependent augmented
plant, the gain scheduled controller is computed. The
structure of this controller is given by
A
k(O) Bk(O) _
L
2
.
A
k(oJ BCI (oJ
Ck(O) Dk(O) - t=
1 a
t CCI (OJ Dcl
(oJ
The LPV controller is carried out by using the LMI
toolbox fnction hinfgs
VII. Ho LOOP-SHAPING DESIGN
The loop-shaping design procedure described in this
section is based on Ho robust stabilization combined
with classical loop-shaping as proposed in [7]. It is
essentially a two stage design process. First, the open
loop plant is augmented by pre and post
compensators to give a desired shape to the singular
values of the open loop fequency response. Then the
resulting shaped plant is robustly stabil ized using H 0
optimization. A importat advantage is that no
problem-dependent parameter varying or weight
selection is required in the second step.
The loop -shaping criterion is:
< 1
Where the shaping flter is:
200
W
1 = S + 0.0002
200
W
-
---
2
-
S + 0.0002
The LPV controller is computed such as the
closed loop is stabilized. y witch guaranty the L
2
gain of the system is bounded and equal to 1
(y=I.0002).
The resulting LPV controller exploits all available
information on f to adjust to the current plant (see
Fig. 1 ). This provides smooth and automatic gain
scheduling with respect to the varying parameters f.
w
z
p (0))
w
Fig.I. LPV contol of LPV Systems
Fig. 2 is a geneal block diagram of the suggested
induction machine control scheme. As shown in this
fgure, we can see that only one PI speed contoller is
used.
Fig.2. Gain-scheduling controller structure
VIII. SIMULATIONS
In orde to validate our approach, simulation tests are
carried out using the proposed control scheme. The
testing conditions are as follows. The fgure 3
represents the speed reference. The speed changed
fom \00 rds to - \00 rds at 4 s, the machine is
loaded at 4Nm. For introducing the efects of
parameter variations, the rotor resistance is increased
of 100% compared to teir normal values at 2s.
100 r
,
--------------------------

I
i
i
\1
1
1
50
i
i

0
w
-50
-100
6
Tme (5)
Fig.3. Speed various time.
Fig.3 show speed response versus time; we observe
that a good tracking speed was achieved without any
effect of variation of parameters. Also we can see in
fgs.4, 5 and 6 the good tracking curent. Fig.7
represents the stator voltage input contol.
40 r
---------
3
0
--------

--------------

-------

----------------

---------

------
,


0
.
-10 '
, ,
r , , ,

Tme (5)
FigA Stator current various time.
Tme (5)
2.02 2.04 2.0 2.08
Fig.5. Zoom Stator curent ant its reference various time
6
----------!---------------r---------------!---------------!---------------r-------------
5 -------------------------------------------------------------------------------------
4
-------------------------------------------------------------------------------------
3
_
__ ____ ___ _L___ ___ __
: 1
.
1 11HHHH1!
-1
- -
-
- - -
- -
- -
- - - - - -
r
,
,
,
1.6 1.8 2.2 2.4 2.6
Tme (5)
Fig.6. Zoom Stator curent according rotor resistace
variation.
600
400
200
o
-200
-400
-600
-800
o
, , , , , , ,
.... ........... ............ ............ ................................... ......... .
, . , , . , .
. , , . , .
, . , , . , .
, . , , . , .
. . , . , .
, , , , , , ,
. . . , . ,
---'---------_. _---------_. _---------_. _----------,-----------'--------------------- , , , , , .
. . ,
, , , ,
. . .
. . .
, . , .
. . .
, , , . . ,
------------------------:-----------:-----------:--------------------------------
. . .
2 3 4 5 6 7 8
Fig.7 Stator voltage various time.
IX. CONCLUSION
LMI Gain -scheduled t contoller design for
induction motor presented here has good properties in
many ways. Scheduling of controller can be realized
continuously and any variation in the plant dynamics
can be covered by scheduling of the controller.
Simulation results obtained are reasonable. One
drawback of this approach is especially for DSP
implementation due to the increasing degee of the
controller. The result obtained in this study will be
appreciated in the planned expeiental works.
X. REFERENCES
[I] J. Shamma and M. Athans, "Gain-scheduling
Potentiel Hazrds and possible Remedies", IEE
control Systems Magazine 12(3), 1992,ppI01-107
[2] P. Apkarian, P. Gahinet , " A Convex Caracterization
of gain scheduled Hoocontrollers", IEE. Trans. on
Automatic Control, 40(1995),pp.853-863.
[3] P. Apkaria, P. Gahinet , G. Becker ,"Self scheduled
Hoo Control of Linear Parameter -varying systems : a
Design Exemple", Automatica, vo1.31-
9,1995,pp.1251-1261.
[4] K.Trangbaek,"LMI-based Gain-scheduled Robust
Flux Observer for Induction Motor" ,Proceedings of
the 14th Interational Conferenceon Mathematical
Theory of Networks and Systems(MTNS).
[5] W. Leonhard I 990, " Control of Electrical Drives .
Springer-Verlag, Berlin, Heidelberg.
[6] Blaschke E., "The principle of feld orientation as
applied to the new trans vector closed-loop control
system for rotating feld machine," Siemens Review,
vo1.34, pp.217-220, 1972.
[7] McFarlane, D.ad Glover,K (1990) "Robust
Controllers Design Using normalized Comprime
Factor Plat Descriptions,"Vol. 138 of Lecture Note
in Control ad Information Sciences, Springer-Verlag,
Berlin .
[8] F.D. Biachi, RIMantz, and C.F. Christiansen, "Gain
Scheduling Control of Variable- Speed Wind Energy
Conversion Systems using Quai- LPV Models,"
Control Engineering Practice, vol. 13(2), pp.247-
255, 2005.
[ 9] P. Apkarian and R Adams, "Advanced Gain
Scheduled techniques For Uncertain Systems,
"IEE Trans. Control Systems Technology,
vo1.65(1 ),pp.21-32, 1998.
[IO] P. Apkarian, P. Gahinet, and G. Becker, "Self
scheduled H jnf control of linear parameter-varying
systems: A design example," Automatica, vol. 31, pp.
1251-1261,1995.
[ 11] P. Gahinet and P. Apkarian, "A linear matrix inequality
approach to H _inf control," Int. 1. Robust and
Nonlinear Control, vol. 4, pp. 421-448, 1994.
APPENDIX
The machine parameters ae a follows:
Resistace e of the rotor; R = 4 n
Resistace of the stator; R = 8 n
Inductance of the rotor; Lr = 0.47 H
Inductance of the rotor; Ls = 0.47 H
Mutual inductance; M = 0.44 H
Number of poles; p =2
Inertia; J = 0.04 kg.m2

Вам также может понравиться