Você está na página 1de 6

Auromarica, Vol. 32, No. 3, pp.

455-460, 19%
Copyright @I 1996 Elsevier Sciena Ltd
Printed in Great Britain. All ri@tts resewed
ooos-1098/96 315.00+0.00
0005-1098(95)00171-9

Brief Paper

On Speed Control of Induction Motors*

ROMEO ORTEGA,+ PER JOHAN NICKLASSON * and GERARD0 ESPINOSA-PhREZ 5

Key Words--Induction motors; passivity-based control; observerless scheme.

Abstract-In a recent paper [(Espinosa-Perez and Ortega,


1994). State observers are unnecessary for induction motor
control. Syst. & Control Lett., 23, 315-3231 Espinosa and Or-
tega presented an output feedback globally stable speed track- ,J~, = =Mqm) -sinkhA
ing controller for induction motors. The performance of the [ sin(h) co&d 1’
scheme is limited by the fact that the convergence rate of the
de := 14:. $lT is the current vector, 4;n is the rotor angular
speed tracking errors is determined by the natural mechani- velocity, Ls. L,, Ls, > 0 are the stator, rotor and mutual in-
cal damping of the motor. In this brief note we overcome this ductances, respectively, D, > 0 is the rotor inertia, R,s,R, > 0
drawback and show that by simple linear filtering of the speed are the stator and rotor resistances, the control signals u =
tracking error we can inject mechanical damping to the closed [ut, uzlT are the stator voltages, TL is the load torque, and 4
loop. This allows us to improve the transient performance of is the 2x2 identity matrix. The subscripts (.),, (o)~. (.),, (a),
the scheme in Espinosa-Perez and Ortega (1994) in position are used throughout the paper to denote electrical, mechan-
and speed tracking applications without significantly increasing ical, stator and rotor variables, respectively, while the super-
script (.)T denotes transposition. We also recall that the flux
the computational requirements, while preserving the global
vector h := [AT,hTIT is related with the current vector1 via
stability properties. Further, it is shown that if the inverter can
h = D,(q,l&.
be modelled as a current source and the desired speed is con- Following Espinosa-Perez and Ortega (1994) we propose to
stant, our passivity-based scheme exactly reduces to the well- rewrite (1) as
known indirect field-oriented control scheme, hence providing
a solid theoretical foundation to this popular control strategy. (4)

1. Introduction
The standard two phase o$ model 11of induction motors, ob-
tained in Ortega and Espinosa-Perez (1993) via direct applica-

1
tion of the Euler-Lagrange equations, is given by RJ2 0
R(qm.4m)= -L,V,3e-3qm drn R,12

Notice that zT {&(q,,& - 2C,(q,, &)} z = 0, Vz E a4. How-


ever, the symmetric part of the ‘damping’ matrix R(q,, &J,
which we denote [R(q,, 4;n)].vy,is not positive definite.
where In Espinosa-Perez and Ortega (1994) torque and speed
tracking passivity-based controllers, that do not rely on state
, reconstruction, were designed for (4) and (2) as follows.* *First,
1 we proved that the system dynamics can be represented as
the negative feedback interconnection of the following passive
aDe@tl) Ls,JeJqm (electrical and mechanical) subsystems (see Fig. I)
W(qm) = -
ah
=
[
-L,vr;e-lqm 1
0 ’ (3)

* Received 22 April 1994; revised 31 October 1994; revised Second, we designed an inner feedback loop which assures
2 June 1995; received in final form 22 September 1995. This that the electrical subsystem defines a strictly passive map from
paper was not presented at any IFAC meeting. This paper control signals to stator currents. This is achieved via the in-
jection of a nonlinear term to the ‘damping’ matrix. Third,
was recommended for publication in revised form by As-
the passivity-based technique was applied to the electrical sub-
sociate Editor Hassan Khalil under the direction of Editor system leaving the feedback subsystem as a ‘passive perturba-
Tamer Basar. Corresponding author Per J. Nicklasson. E-mail tion’. As explained in Ortega and Espinosa-Perez (1993) this
per.johan.nicklasson@itk.unit.no last step involves the definition of the desired closed-loop en-
t Universitt de Compibgne, URA C.N.R.S. 817, BP 649, 60206 ergy function whose associated (target) dynamics evolves on a
Compitgne, France; E-mail rortega@hds.univ-compiegne.fr. subspace of the state space insuring zero error tracking. The
* University of Trondheim, NTH, Dept. of Eng. Cybernetics,
7034 Trondheim, Norway.
8 Universidad de Mexico, Instituto de Ingenieria, PO. Box n Note that (I) is equivalent to h + R& = M,u.
70-472, D.F., Mexico; E-mail gerardoe@servidor.unam.mx. * * The model considered in Espinosa-Perez and Ortega (1994)
11In this model the axes for the stator have a fixed position is obtained by applying a change of coordinates to (4), (2) (see
while those corresponding to the rotor are rotating at the rotor Espinosa-Perez and Ortega, 1995)
angular velocity.
455
456 Brief Papers

where fl > 0 and

RI(&) = z&, 0 < E < min(R,V,R,),

T~(z)=D,,,~~~-z+TL (7)

and controller state equations

&,I = %/(z)% &d(o) = , (8)


Fig. 1. Passive subsystem decomposition. 82 [: I
i = -az + 66,. z(O) = P,,,(O) (9)

overall procedure leads to an internally stable dynamic out- with e!,, := & - 4m~ and a, b > 0.
put feedback controller that assures global asymptotic torque Under these conditions asymptotic speed tracking and rotor
or speed tracking. In contrast to this paper, the mechanical flux regulation are insured, i.e.
part of the system was described in Espinosa-Perez and Ortega
(1994) by the equation

with all internal signals bounded, where 11. (I is the Euclidean


norm.
and unfortunately, to avoid acceleration measurement in speed
tracking, the stability proof requires the mechanical damping Remark 2. Notice that since &d is required for implementa-
to be strictly positive, i.e. R,,, > 0. Furthermore, it follows tion of the controller, id is needed. Thus, if Ed includes &,,
from the analysis in Espinosa-Perez and Ortega (I 994) that the acceleration must be measured. To overcome this problem, in
convergence rate of the speed tracking error is bounded from Espinosa-Perez and Ortega (1994) we assumed & > 0, and
below by the motor time constant D,,,/R,. Some simulations proposed the speed control strategy TJ = D,ij,,,d + R&,,d + TL.
that reveal the performance degradation for small mechanical Two drawbacks of this scheme are that it is open loop (in the
damping constants are presented in Espinosa-Perez (1993). speed tracking error), and that it can be shown that its conver-
The main objective of this paper is to remove this restric- gence rate is limited by the mechanical time constant D,,,/R,,,,
tion. To this end, we show that by simple linear (strictly proper) Defining the desired torque Ed as in (7) with z from (9) allows
filtering of the speed tracking error we can inject mechanical us to effectively feed back the speed tracking error (without
damping to the closed loop without acceleration measurement. acceleration measurement). Further, as will be shown below,
This renders the transient behaviour of the scheme indepen- we also make the convergence rate independent of the natural
dent of the natural mechanical damping, which can now be mechanical damping.
assumed to be zero, thus enhancing the controller performance
in position and speed tracking applications. The computational
Remark 3. Equation (6) represents a desired behaviour for the
requirements are not significantly increased, and the global sta-
currents, whose choice is ex lained as follows: from (2) and
bility properties are preserved. Interestingly enough, no condi-
tions on the filter bandwidth are imposed. (3) we have that T = L& ? 3dqmq,, and from the last two
Other recent efforts to address the speed control problem equations of h = D,(q,)Q, we obtain
of induction motors may be found in Marino et al. (1993)
Kanellakopoulos et al. (1992) Kim et al. (1990) and Chiasson h, = L,yre-3qm~,v+ L& (10)
(1993).
By solving (10) for & and substituting it into T above we see
that the torque can be expressed as
2. Main result
T = h;Jcj,. (11)
Proposition I. (Speed control) Consider the induction motor
model (4) and (2). Assume: To be consistent with these physical relationships we propose
A.I. Stator currents & rotor speed & and position qm are to define the desired fluxes and currents such that, for a given
available for measurement. TJ. we have
A.2. Motor parameters are exactly known.
A.3. The load torque TLC?) is a known * smooth and bounded
(12)
function with known bounded first-order derivate, such
that ITL(~)[s cl < Q), Vt E [O,m). (13)
A.4. The desired rotor speed &d(f) is a smooth bounded twice
differentiable function with known bounded first and Td = &!&d. (14)
second-order dertvatives, such that l&d(t) I 5 c2 < m, Vt E LO,~1.
Let the controller be the nonlinear dynamic output feedback Notice that (14) admits a solution of the form C& = fiJA,d.
Finally, solving (13) for &J gives the rest of (6). For the detailed
u = L,& + L.v,eJqm& + L.,Je3q~n&cj,~ + R,~.~~- KI @,,)ti,(5)
derivation of the controller see Espinosa-Perez and Ortega
with (1994).

t (6) 3. Proof of main result


By substituting (5) into (4) and using (12), it follows that
the closed-loop system is fully described by

* We make this assumption for simplicity, the result can be


&(q,)& + G(q,. 4&r + [Rtqm &,I + Xkj,,] i, = 0,
extended for the case of unknown linearly parametrized load
torque (Ortega and Espinosa-Perez, 1993). (15)
Brief Papers 457

&&n = -z + T(& q,,,)- T,/(Z). (16)


i = --LIZ f bit,,,. (17)

&J = %l(zl3&/,
B2
(18) := 6x1 (t) + a2(t)z (2-5)
and noting that
where X(c),) := diag(Kt (cj,,,)l~, 0). For later convenience we
write (16) and (17) as

L*
f = .% + B(T - Td)

The matrix d. is Hurwitz for all positive values of a and b.


(19) IloczWII5
J ((----- )2 + ~mfll
Lr 82
LT;liiz < m

Under assumptions A.3 and A.4 the system (IS)-(lS) is yields II& II I LI + ii2 lzl . Replacing this bound, together with
II&II zz rn~li~~(O)ll, Vt E [O, T) in (24) it follows that:
locally Lipschitz in the state [e:, em, z, ALIT and continuous
in r. This condition assures that there exists a time interval
[O, T) where the solutions exist and are unique. First, taking IT - Tdl 1 ~cm:lle,oll* + zm,114(0)11~1)
the derivative of 4 J(hdl12 and using (18) it can be shown that
llh,~II = 8. Vt E [0, T). Now, consider the following quadratic +L.,~m,llP~~O~Ila2lzl,
Vt E [O,n.
function Vt = %;fD,(q,,,)t$, whose time derivative along the This last inequality proves, via Gronwab’s inequality, that x,
solutions of (1st for all times in [O, T), is given by and consequently &, cannot grow faster than an exponential
in the time interval [O, 2’). Moreover, since all the constants
4 = -6: [R(q,. &I) + K(&l)l,YY&* (20) in the above bound are independent of T, we can repeat this
argument to extend the time interval of existence of solutions
where we have used the skew symmetry property mentioned to the whole real axis. See Canudas de Wit et of. (1993) for a
above and similar approach.
Having proved that (22) holds as t - 00, we proceed by
proving that this implies that lim,-, x = 0, with &d bounded.
Inserting (23) into (19X and using (25) to express &, we
get the system

We now prove that this matrix is strictly positive definite,


uniformly in &, namely that

[R(q,,,, &I + ~t(im)].yr 2 ~24 > 0, (21)

where 5 is the 4 x 4 identity matrix. By using standard results


from matrix theory and the facts that Rr > E and Je-Tsm = 0
e_TemJ, it follows that (21) holds if and only if R,v+ KI (4,) -
g&l 2 E. Tbis motivates our definition of KI (&).
Therefore, f = Ax + B(f)X + c(t).
from (20) we have that for some constants mP > 0 and pe > 0
Taking norms, we obtain [l&t) II I &Lsr&m,IIQ(0) Ile-Pc’,
independent of T
and llc(t)II 5 z ~rn~ll~=(O~lle-~=~ + 2&t ~rn~l~~~(O)i~e-~~‘.
lh.%r)il 5 m~li~~~O~ile~~~r,Vt E lo, T). (22) From this (see Khahl, 1992, pp. 199-200) we can conclude
that the system i = [f4 + B(tf]x is globally exponentially
Notice that, unless &, escapes to infinity in finite time, we stable, and since Ilc(t)ll - 0 as t - 00, we have that x - 0.
have lim,,, & = 0. Thus, we will prove first that this is not Further, from (25) we establish that &d is bounded.
the case by showing that the input (T - TJ) to the linear filter To prove that the norm of the rotor Aux converges to ~3
(16x17) is linearly bounded by the filter state. To this end, notice that
notice that the desired torque r~ can be written as

Td = $dm (qm;n)&d. thus h, - &d, which has norm 8. l

hence we obtain that Remark 4. It is easy to see that choosing the desired torque
in the controller above as
1 .T
r - l-4 = :‘I % (qm)t + e:w, &z)&d (23)

and the following bound holds where em := q,,, - qd, yields gtobal asymptotic position track-
ing for all positive values of a, b, f. In this case, in the error
equation (19) we have
h---dI5 ~(~~~~~~2+2~~~~~t~l~q~d~I). (24)

On the other hand, writing the desired current (6) as

The proof of global convergence in this case follows verbatim


from the proposition.
458 Brief Papers

Remark 5. Choosing the norm of the Hux to be constant as m


the proposition, means that the drive system must be chosen
such that the output of the inverter does not saturate for
the speed range the motor is required to operate within. This
restriction has recently been removed in Espinosa-Perez et al.
(1995), where we propose a controller that allows (global)
tracking for both speed and flux norm.

4. Reduction to indirect field-oriented control


We will now prove that, in speed regulation applications,
and under some simplifying assumptions on the motor model,
our controller exactly reduces to the well-known indirect field-
oriented control (FOC).
In typical applications of FOC it is assumed that the inverter
can be modelled as an ideal current source, therefore the control
signals become the stator currents &. Using this assumption
our passivity-based control reduces to Fig. 2. Speed error & 4,,,d = 100 rad/s. a =
80, 100. b = IO.

(27)

-rd = D,,,ij,,,~ - a& •t TL (28)

Notice that in (28) there is no need for using the filtered speed
error, since id is no longer needed for the implementation. Also,
notice that the controller states (27) can be exactly integrated
as

(29)

where pd is the solution of

&J = R’Q, Pd(O)= 0.


82
Replacing this expression in (26) we get

Fig. 3. Controls ur. ~2. a = 100, b = 10.

CO”,,0lS
/
On the other hand, assuming that the desired speed is constant
(j,,,d = 0) and replacing the exact load torque cancellation by
an integral action we get from (28)

-rd=-
(a+l 1 y>
P
&, 0.

The controller (30)+32) exactly coincides with the classical


indirect FOC. See Ortega et al. (1995) for further details on
the comparison between these two schemes.

5. Simulation results
The performance of the control scheme was investigated
by simulations. We used the same motor parameters for the
squirrel cage induction motor as in Espinosa-Perez (1993)
R,C= 0.687Q R, = 0.842a, L,s = 84mH, L, = 85.2mH, L,,, = Fig. 4. Controls ur. uz. a = 80, b = 10.
81.3mH, D, = 0.03kgm2, and the controller parameters fi = 2
and E = 0.5R,. The mechanical damping R, was set to zero,
and the motor was initially at standstill with zero load torque torque without filtering, which gives rd = D,,,ij,,,~ - hi,, + TL
TL. At startup, the load torque was set to TL = 20Nm, and (see also (7)).
the speed reference was changed from zero. Figure 2 shows
the error in speed for a constant reference &,d = lOOrad/s and
filter values a = 80, 100, b = IO. Figures 3 and 4 show the 6. Concluding remarks
controls corresponding to these filter values. The simulations
were repeated for the values a = 100, b = 2, 5, and the resulting In this note we have extended the results on output feed-
speed errors and controls are shown in Figs 5-7. Finally, Figs 8 back globally stable speed tracking controllers for induction
and 9 show the speed error and controls when acceleration motors reported in Espinosa-Pkrez and Ortega (1994). We have
is measured, and the speed error can be used in the desired removed the dependence of the convergence rate of speed track-
Brief Papers

i-4 I-4

Fig. 5. Speed error I?,,,. &d = 100 radls. a = Fig. 8. Speed error 6,. 4m~ = 100 radls. Without
100, b=2, 5. filtering. b = 0.1.

Fig. 6. Controls at, ~2. a = 100, b = 5.

CantrcJs Fig. 9. Controls ut, ~2.Without filtering. b = 0.1.

in Espinosa-Perez and Ortega (1995). Especially, using the sta-


tor fixed ub frame, the controller can be implemented without
position measurement as in Espinosa-Perez and Ortega (1994).
We want to stress the fact that the passivity-based scheme
does not require the explicit implementation of an observer,
has no controller singularities (avoiding the need to take special
precautions at start-up), and gives global stability results. We
have further shown that under some simplifying assumptions
on the motor model and the inverter this scheme exactly reduces
to the well-known indirect FOC. Recent experimental evidence
(Nicklasson et al., 1995) has shown that our scheme provides
a significant performance enhancement to the indirect FOC.
A drawback of the scheme is that we have assumed known
motor parameters. This is a well-known important problem,
Fig. 7. Controls IQ, ~2. a = 100, b = 2. especially with respect to the changing rotor resistance. In-
terestingly enough, we have recently been able to prove that
for current-fed motors in speed regulation the controller is
ing errors upon the natural mechanical damping of the motor, actually highly insensitive to the time constant r,. Namely, we
something that is important for todays low friction motors. have shown that global stability is preserved for all estimates
This is done by using a linear filtering of the speed error in the 0 < ?, < 2T, (de Wit et al., 1995). Currently research is being
controller equations, avoiding the need for acceleration mea- carried out to extend this robustness result to speed tracking
surement. The performance is illustrated by simulations, and for the complete motor model.
it is close to the performance of the system with acceleration
measurement used in the controller equations. It is worth re- Acknowledgements-This work was supported in part by the
marking that the controller is given here in terms of the a@ Commission of European Communities under contract ERB
model, however, for the purposes of computation it is more CHRX CT 93-0380, and the Research Council of Norway
convenient to use an equivalent dq implementation as explained under contract 100651/410.
460 Brief Papers

References Control of induction motors in the field weakening region.


In Proc. 34th IEEE Conf: on Decision and Control, New
Canudas de Wit, C.. R. Ortega and S. 1. Seleme Jr. (1993). Orleans, LA.
Robot motion control using induction motor drives. In Kanellakopoulos, I., P. T. Krein and E Disilvestro (1992).
Proc. IEEEInt. Con/: on Robotics and Automation, Atlanta, Nonlinear flux-observer-based control of induction motors.
GA, pp. 533-538. In Proc. Am. Control Conf, pp. 17OC-1704.
Chiasson, J. (1993). Dynamic feedback linearization of the Khalil, H. K. (1992). Nonlinear Systems. Macmillan.
induction motor. IEEE Trans. Automat. Control, 38, 1588- Kim, D.. I. Ha and M. Ko (1990). Control of induction motors
1594. via feedback linearization with input-output decoupling.
de Wit, l?, R. Ortega and I. Mareels (1995). Indirect field ori- Int. J Control, 51, 863-883.
ented control of induction motors is robustly globally sta- Marino. R., S. Peresada and P Valigi (1993). Adaptive input-
ble. In Proc. 34th IEEE Conference on Decision and Con- output linearizing control of induction motors. IEEE
trol, New Orleans, LA. (Also submitted to Automatica.) Trans. Automat. Control, 38, 208-221.
Espinosa-Pirez, G. (1993). Control No Lineal de Motores de Nicklasson, P. J., R. Ortega and 0. Egeland (1995). Experimen-
Inducci6n. PhD thesis. DEPFI, UNAM. Mexico. (In Span- tal results from passivity-based control of a VSI induction
ish.) motor. Technical report. Dept. of Eng. Cybernetics, NTH.
Espinosa-Pkrez, G. and R. Ortega (1994). State observers are Norway. (Submitted to IFAC World Congress ‘96).
unnecessary for induction motor control. .Sysr t Control Ortega, R. and G. Espinosa-Ptrez (1993). Torque regulation
Lett., 23, 315-323. of induction motors. Automatica, 29, 621433.
Espinosa-PCrez, G. and R. Ortega (1995). An output feed- Ortega, R., D. Taoutaou, R. Rabinovici and J. P. Vilain (1995).
back globally stable controller for induction motors. IEEE On field oriented and passivity-based control of induction
Trans. Automat. Control, 40, 138-143. motors: Downward compatibility. In Proc. IFAC NOLCOS
Espinosa-PCrez, G.. P. J. Nicklasson and R. Ortega (1995). Conference, Tahoe City, CA, pp. 672-677.

Você também pode gostar