Professional Documents
Culture Documents
better exploit their speed and the functional capabilities of their software. The concept of "instantaneous torque control" is introduced as
an objective worth pursuing in the application of such digital IC's to
drive systems. Instantaneous torque control would in principle permit
the fastest possible response and the elimination of torque ripple, along
with many other advantages not possible with conventional control algorithms, most of which are set up to control a time-averaged torque.
Some of the fundamental principles of instantaneous torque control are
developed for the switched reluctance motor, which is used as an example because, like the brushless dc permanent-magnet motor with
concentrated windings, it has the potential for rapid response, but it
can have appreciable torque ripple with unfavorable firing angles. A
reference frame transformation that would eliminate the rotor position
from the voltage and torque equations is not known for either of these
machines. This opens up a number of interesting questions as to the
generality of instantaneous torque control algorithms, and whether they
can be incorporated into the general or unified theory of electrical ma-
mutator motors and ac sine-wave motors because the instantaneous torque is ideally constant in the steady state
and is therefore indistinguishable from the average
.
trol.
This suggests the possibility of controlling the instantaneous currents or flux linkages in such a way as to produce constant instantaneous torque. In this paper the possibility for such
is
a conventional variable-reluctance stepper motor provided with a shaft position transducer. The SR motor is
chosen because of the relative simplicity of the elementary theory describing torque production in the absence of
magnetic saturation and mutual coupling between phases.
INTRODUCTION
r HE SPEED and functionality of digital integrated cir- It is known to be susceptible to torque ripple if the firing
I cuits now suggests the development and evaluation of angles are unfavorable.
The control of torque in dc commutator motors is based
new algorithms for the control of adjustable speed drives.
This paper introduces the concept of "instantaneous on the classical theory in which torque is proportional to
torque control" as an objective worth pursuing in the ap- the vector product of armature current and flux. Under
plication of such IC's to drive systems. Instantaneous sufficiently light load the commutator maintains the spatorque control would, in principle, provide the fastest tial orthogonality of these quantities so that the vector
possible response and would permit the elimination of product becomes a scalar product. Since both the flux and
torque ripple along with a wide range of instabilities and the current are normally constant, the instantaneous torque
even shaft resonances, by software techniques alone. It is is also constant. There is ideally no torque ripple, and the
even conceivable that some of these features could be torque is varied by changing the current or the flux or
both. In sine-wave ac machines the situation is similar.
automatic.
In most ac and dc drives the control algorithms are set The actual phase currents and flux linkages are not conup to control a time-averaged torque rather than the in- stant but can be made to appear constant by means of a
reference-frame transformation that has appropriately
works well for dc co- been
called the commutator transformation; (an example
is
Park's
transformation). The instantaneous torque of an
Manuscript received September 27, 1985; revised May 5, 1986. This
paper was presented at the Power Electronics Specialists Conference, Tou- ideal sine-wave ac machine is constant and is proportional
chines.
tiseapproach
spta contaneous toq.
louse, France, June 24-28, 1985, under the title, "Nonlinear Torque ConM. Ilic-Spong is with rthe Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, 1406 West Green
th
SR moo.sdul
aiet
ti
adrt
ncthSRmtr1dobysle, Shaero
construct a reference frame transformation that will elim0885-8993/87/0100-0055$ 1.00 C 1987 IEEE
56
inate the rotor shaft position from the voltage and torque
equations. It is therefore a good vehicle for exploring
whether the use of reference-frame transformation is likely
to be of use in generalizing the principles of instantaneous
torque control to a wider class of machines. In this paper
a Floquet transformation is constructed to express the
equations of the SR motor in a rotating reference frame.
Magnetic saturation is ignored, and the motor is characterized by a piecewise linearization of its inductance
curves. While these assumptions are very restrictive, the
resulting model is proposed as only one of a number of
possible approaches, and it is hoped that others may in
the future prove more powerful.
MAIN RESULTS
Mathematical models of physical systems are usually
represented for control purposes as sets of first-order nonlinear differential equations
dti
dtdt = fi (xI,
x2,
* ** ,
* ** ,
dX2
Um)
, Um)
dt
dx (t)
dt = A(t) x(t) + B(t) u(t)
(6)
Rp
d;
d
(8)
x2, * * * , xn; ul, u2, * * * , Um). (1) p = 1, 2, * * * , P where P is the number of phases. A
dt-= fn(xI,
crucial difference between the switched reluctance motor
If we define a state vector x = [x1, X2, ... , Xnj T, an (a doubly salient machine) and nonsalient machines lies
input (control) vector u = [U1, u2,
Um] T and an in the fact that the dependence of flux linkage on the curUl, T then (1) can be rent does not vary sinusoidally with the rotor position 0;
output vector y = [ Yl,
in general it could be any periodic function of 0 (with
represented in a vector form
higher order harmonics). Under the commonly made asdx = f (x, u).
(2) sumption of magnetic linearity (saturation neglected) exdt
perimentally measured magnetic curves (Fig. 1) can be
The output variables are variables that one can easily approximated as
measure in the system and are related to state and input
(9)
W (0, i) = L(0) i.
variables as
Here v is the vector flux linkages and i is the vector of
y = g(x, u).
(3) stator currents. L ( 0 ) is the inductance matrix.
A much
realistic into
of the
representation
magnetic
As a result, most of the dynamical systems (2) are ap curves
takesmore
saturation
account via
a piecewise
as cre ae
auainit
conto current
i
proximated by a particular subclass ofproximated
icws
systems known
by apriuaslinearization
technique
with respect
variables.
lnear time-invariant systems for which control
tech- All results presented
in this paper can be extended to tbis
niques are well-known and implementable in real time.
while
the
computational
complexity is slightly inNote: Real-time implementation off contromeans
control means that case,
We
will
these
extensions in future pubpresent
m r s t t vcreased.
if measurements of the output vector y areifperformed
lications.
every N time units, one can,- based on the previous meaFromon
. instant
.
~~~~~~~From(2) the vector of phase currents as a function of
surement, compute control 'at'
the next sampling
' .units.
.
flux linkage vector can be expressed as
~~~~~~~~~~the
within these N time
Linear time-invariant systems have a form
i = L-1(0) AV(0, i),
(L0)
()
= Ax(t) + Bu(t)
(4) and (8) and (10) combined in a vector form obtain
Y2=
yU,]
notue:areawell-timenand
implementat
y(t) = Cx(t)
th.
(5)
dt =
R(0)V~ + v
(11)
57
e constant
becomes
L_I(-)
+
dH
-TAT-' _2aT,
-_o T 1
(12)
(20)
C(W)
where C(w) is not a function of a rotor position 0; ob(13) viously, if dI/dt =r = constant, the right-hand side of
A(M) -RL71(0).
Then the change of flux linkages with rotor position (20) is time invariant.
can be shown that the dynamic equation (15) exasIt
3), as
change iS defined, using (1 1) and (Idefind,usng(1)an(13,>
pressed in terms of transformed flux variables XN via T
takes on the form
dv
=
A(0)A
dt
(14)
'
,
-R2/L2(6), * -Rp/Lp(0)) (16)
and B = diag ( 1, 1, * * * ). When w = co = constant,
matrix A (0) becomes a function of time only, since
0 = 00 + Wrt
(17)
and (15) yields
d~~~~~~~~~XN
(WC(W)
dt
-d6
dt dw
and
(22)
VN = Tv.
In particular, for constant rotor velocity d2 /dt2 = 0,
which gives
d
XN = -COC(W) XN + VN.
(23)
dt
Now (23) has constant linear coefficients and a time-varying control VN and clearly belongs to the class of dynamical systems given by (4), (5).
Based on the Floquet theory and its application to rotating machinery as presented in [2], it is known that a
reference frame transformation T exists even for doubly
salient machines. A constructive procedure for obtaining
such a reference frame transformation is hard in general
if A is a full matrix. At this point we take advantage of
the fact that the model of the switched reluctance drive is
accurate enough if the matrix A is considered to be diagonal. In this case a transformation T can be obtained by
deriving it for separate decoupled phase flux linkages.
Condition (20) for any phase is in a scalar, rather than
vector, form:
+
(18) A solution to (24) is
dtd(t) A(t) x(t) Bu(t)
T =vhO/cr(5
which is clearly the same as (6).
Next, let us introduce change of variables via a trans- with
_
.dh (6) K R 8(6
formation T so that a new state vector XN iS related to the
L0
-d
original state vector xas
XN = Tx
(19) and
a = __
where T = T(6, cor) and possesses the following proper27r Jo L(0)
ties.
(24)
(27)
58
Wr
Torque Loop
Nq,
Tdes
v CoCordinatetFlu
TransformEsiaoSR
N~~~~~
Commutator
e~~~~~W
dfxNNi = -axNI + VN
VNi
dt
alNi
(28)
with
(29)
VNi e(h( )/')vi,
and finally, (28) written with respect to rotor position o
(rather than time t) is
dXNi _ N_ + VNi
(30)
XNi +
dO Note: In the case when the saturation is not neglected,
L(e) and the transformation Tdepend on current, making
ai a function of current in (29) and (30).
INSTANTANEOUS TORQUE FOLLOWING ALGORITHM
After introducing the change of state variables from x
into XN and change of controls from v into VN via the transformation T for the switched reluctance drive, we pose
the instantaneous torque following problem as a problem
of finding a control VNfin (30) which causes the resulting
closed-loop system to follow a desired torque trajectory.
A particular case of a torque following trajectory is to
follow a constant torque Tdes. A control scheme for this is
proposed in Fig. 2 where the error in the inner loop is
defined as the difference between the flux linkages seen in
the new reference frame, the one corresponding to the desired torque Tdes XN and the actual flux linkage at a given
sampling instant, i.e.,
e (0) = xN (0)- XN (09).*
(31 )
The outer loop corresponds to the control needed to obtain
a constant speed orqThe state of the art in the motor control area is such that these controls are designed indepen-
dently.
Because of decoupling in (16), one can write a scalar
version of (31) corresponding to a single phase:
e(0) = x4 (0) - XN (0)*
(32)
Finding control
VN
(33)
59
merical integration is involved. At today's level of technology, a real-time implementation of this control is possible. Typically, m(6) is computed in the block named
"commutator" in Fig. 2. For the case when saturation is
neglected,
we
discuss
(0)
Phase A L (a)
in detail.
A four-phase switched reluctance drive and its associated power conditioner developed in the Power Electronics Laboratory, General Electric Company, is studied. The
information flow necessary for closed-loop control is
shown in Fig. 2.
In the linearized case, (3) is
=
(6, i)/L(6),
2e(8(5)2)
m()
i2 dL(O) _ ~2 dL(O)
T(t) = 2 dO - 2LM2() dc
(34)
Phase D
45 60
135 50
90 105
1X!
MB(3e5_3)B
E_i
( (
m(8)
MB
-
) = .(37)
90)2)
where
m (H
8 (degrees)
(a)
Let the superposition strategy for phase excitation be formally defined so that a torque per phase is
T(O) = Tdes * m(H)
(36)
180
|MD(e) /Xm(e-eP)=
Imc(O)
(b)
(39)
(~--XN+VN).
(45)
60
Tdes
15.0
a*
(46)
L )j (4)
dOL~Tdes~ dL (O)/IdO L(0)l
dO df
ds [
(47)
XN = XN - e().
Combining (42)-(47), we obtain
de(0)
d
dT(O, COr)
+ - (XN*
TO, Cr)
x* +
dO
- e
(0))
dx*
75
5.0I
2.5
60
90
75
&
I05
(48)
VN
15.0_
de(0) =
dO
12.5
= -eo)
dx*75
FdT(O9
+
'* + T(O, Or) dO
xdT(O)
d
L dO
de (O)
de(0)
dS =
5.0
(49)
VN]
X
or
Fig.
a
(50)
u1(0, x) -Me(a),
T(O,
Wr)ddO + -XN
'r
VN
(52)
dO
dO
(53)
5.
~~
5~
~~
90 7560~~~~~~4
105
v
150
-75
-150 _
dO450759
dT(or &.'r)
dx*
VN = d( 'r) X* + T(6, Wr)
+-
~~~~~~
or
10.0
045
dO
or
or
12.5
75
90IV
&
Equation (53) explicitly defines the voltage control necessary to make the developed torque follow a desired
torque Tdes. The same control seen from a stationary reference frame is obtained using the inverse transformation
CONCLUSION
T' (0@, (Or), i.e., the control required in (11) is
The concept of instantaneous torque control for motor
V- ( s XrJ VN- ~case of the switched reluctance motor, which is chosen as
Typical curves for the required current, voltage, and flux an example because of the simplicity of its elementary
linkage are illustrated in Figs. 4-6. These are given again torque production theory and because it is a good vehicle
-T-
(0,
')
(54)
drives
and explored
ILlIt-SPONG
61
for exploring the possibility of generalizing the idea of ing, Cornell University, Ithaca, NY. Currently she is an Assistant Profesand the Cosor in the Science
Department
of Electrical
and Computer
at Urbana-Champaign,
of IllinoisEngineering
University
ordinated
Laboratory,
by the voltage and current capabilities of the power converter and the switching frequencies of which it is capable.
ACKNOWLEDGMENT
IL. Her research interests are in the areas of control and network theory
applications to large-scale power systems and, recently, in the power elec-
electricofmachinery.
tronics Ilic-Spong
based control
AAAS and Sigma Xi.
is aofmember
lmaDr.
Timothy J. E. from
Millerthe(M'74-SM'82)
received the
University of Glasgow,
B.Sc. degree
the
Ph.D. degree from the
Glasgow, Scotland, and
University of Leeds, Leeds, England, in 1970 and
1977, respectively, both in electrical engineering.
Electrc Company Corprch staff of the General
velopment, in 1979. He has worked on the design
switched reluctance motor drives. His technical work includes the analysis
of the dynamic performance of permanent-magnetic ac motors during starting and when operating from solid-state power supplies. He has designed
the world's largest permanent-magnetic generator for industrial applications. His work on switched reluctance systems includes the development
of
novel typesManager
of powerofelectronic
withControl
associated
analysis.
He
Power
Program,
Electronics
the Powerconverters
is presently
Electronics Laboratory. He is also an Instructor in the Reactive Power Con-
Dr. Miller is a Chartered Engineer in the United Kingdom and an Associate Fellow of the Institute of Mathematics and its Applications.
REFERENCES
_
_
_ _ _ respectively.
In 1981-1982 she was with the Department of
Electrical and Computer Engineering, Drexel
University, Philadelphia, PA. From 1982 to 1984
she was with the School of Electrical Engineer-