You are on page 1of 6

A R O T O R FLUX OBSERVER FOR THE V E C T O R

CONTROL O F THE INDUCTION MOTOR DR

D e p a r t m e n t o f Electrical E n g i n e e r i n
Faculty o f Engineering
O S A K A City U n i v e r s i t y *

ABSTRACT 2. Minimum-order ux observer. (')


An improved rotor flux observer was developed. This
observer can be used effectively in vector control
(Field oriented control) systems o f the current source (a,b axis) is d
inverter drive induction motor. The observer is com-
posed of a 16 bit micro-computer which calculates
the observer equation and A / D converters which mea- The state vector are fluxes of the primary and
sure currents and voltages o f the motor. As to the secondary clrcui
observer, the stability and correctness of the expected
flux vector must be considered as most important things.
Then, in our observer, the first-order hold and The state vector
aexcellent pole-placement design are adopted. In this supplied voltage Vs.
paper, simulated results o f estimated rotor flux and
V S = [ v a s , vbs] (2)
experimental results of the tested motor are presented.
I f the state vector o f the output si
1 . Introduction primary current o f I s= [ias, l b s l the output value
is shown as follows.
Recently, some interesting work has been published
concerning the application o f the rotor flux observer.
For these application, the values of the rotor flux
ri
is expected to a accurate and a time lagging-less
vector, then the observer is required to have char-
acteristics of the quick and accurate calculation.
In the case that the observer is composed in where, A = Ls
a micro-computer, the delay time of the estimated
value is delayed cause o f both sampling observer
signals and calculating the observer equation.
Then,it is the most important problem for the rotor
flux observer to make these delay times as short as
possible. When a stable and accurate observer is com-
posed on a micro-computer, the essential problem is
how to arrange the poles o f the observer's equation.
In this paper, a rot lux observer system for
the practical vector cont o f the induction motor
the induction m
I S discussed. The observ s composed of the first
-order hold to reduce th elay time which i s made
by sampling the observer s I s , and a method o f the
observer poles arrangement i s proposed.
Finally, we show experimental results o f a vector
controlled induction motor which i s driven by a
controlled current type inverter.

A1 1= -{RsLr /A + RrM'/LrA}* I
KEY WORDS: AC motor control: state observer; current A12=M/fl{Rr/Lr* I -w r-JI
source inverter; induction motor, A21=MRr/Lr+ I
A22=-Rr/Lr. I +or.J
Bi=Lr/A- I
*Address f o r correspondence:
3-138 SUGIMOTO 3 , SUMIYOSHI-KU
OSAKA 558 JAPAN
Phone 06-605-2586

IECON '88 I 4 7 2
The state equation ( 4 ) can be rewrite as follows. Te=(3/2)*(P/Z>*(Lr/M)
X(A areibs- A br*ias) (11)
hr=A22*hr+A21* i s (6) In the case of using the vector control in the
ir-All. is-Bl-Vs=A12*Ar (7) induction motor, the orthogonal currents i ds and
From eq.(6) of the state equation we can understand i qs are controlled independently. In this case,
that h r is the state vector, and that is i s the these currents should be converted to the primary
vector of the input signal and eq.(7) is the output current vector of i as and i bs.
equation in which (p. i s-All. i s-Bl.Vs) is the vector In this conversion, we should have the angle L h r
of the output signal. between an axis of the static coordinate and the rotor
Then, we can show the minimum order observer as the flux vector A r axis. In the vector control system,
equations (6) and ( 7 ) . the angle L A r is obtained by expecting the values
in this observer, the expected value.of the rotor of the rotor flux observer.
flux A r is obtained from eq.(6). and the error in
the flux between the command value and the estimated 3 . Design of the rotor flux observer.
one of eq.(7) is feedbacked to the system by the gain
G shown in the following equation. 3-1. Observer equation using the first-order hold.
h
When we calculate the equation of the rotor flux
h r=A22*x,r+A21* i s
+
G { i s-AllLi s-Bl-Vs-A12*T r > observer by using the micro-computer, the detected in-
=(A22-G.A12)Ar put signals i s(t) and Vs(t) should be converted to
+(A21-G*All) i s - G * B l * V s + G * i s (8) the discrete-time input signals of i s(k) and Vsfk).
In the eq.(8), we must detect the value of the And the observer equation has to be converted from
time-derivative of the primary current. Where, the differential equation of the continuous-time
introduce a new variable F and express the equation system to the difference equation of the discrete-time
of the A r as follows. system.
A As the poles of the observer equation can be
Ar=E+G.is (9)
determined arbitrarily by a designer, eigenvalues of
* =
E=fir-G*is the observer equation can be given in advance. The
=(A22- G eA12)Tr general solution of the observer equation can be ob-
+(A21-A11*G) i s - G*Bl*Vs
tained and the differential equation of the observer
In this manner, the lime derivative of i s can be converted easily to the difference equation.
ignored. The eq.(9) is the minimum order state equa- However, as for the input signals, they have
tion of the rotor flux observer. delay-time between continuous-time values and the dis-
The gain matrix G can be calculated as in the crete-time ones. We discussed the eliminating method
following eq.(lO) in order to make the pole of the of the delay-time of the observer signals, and we
*
flux observer - d j p . adopted the first-order hold instead of the ordinary
zero-order hold.
G=gl* I + g2.J (10) By adopting the first-order hold, the delay-time be-
tween the input signal of the motor f (t) and the
input signal of the observer g (t) can be eliminated.
At the sampling time KTs , the expected rotor flux
A w r *a -(Rr/Lr) p
g2=-
M (Rr/LrI2 + wr2 by using the first-order hold is shown as follows.

(12)
In the induction motor, the current component of
the torque o r the rotor flux is shown in the primary
current. And the torque equation can be shown in the
same way as in the direct current motor.
There, the reference flame is set on the rotor flux
A r axis, and the primary current is divided into two
components. One is the exciting current i ds, and the
other is the current o f the torque component i qs.
As well known, the generate torque of the induction
motor can be calculated by eq.(ll). In the above equations, conversion matrixes @ (Ts).

IECON '88 I 4 7 3
t7 (Ts) and U ( T s ) are not affected by the rotating real-axis. When the poles of the system are set
speed of the motor, and they are constant values. on these straight line, the flux observer is set
When the input signal in the observer is hold by in the critical damping condition. And the maximum
the first-order hold, the delay-time o f the expected driving frequency less than 1/2 of the
output slgnal can be cancelled. However, the sampling sampling frequency.
time Ts of the observer increases because of the When we adopt th
calculating time of the first-order hold in the ob- ment, i t Is desira t the poles of the rotor
server. In practice, the percentage of the cal- flux observer should be placed on the
culating time in the first-order hold is small In in the complex-plane.
comparison with the sampling time of the observer.
The construction of the above the observer system

pT
is shown in Fig.1.

Y dt) 1 t( t)

1s \ S t a t e Observer 1% \

..._.... ...~..~.[ .k .- -.
l l
... ........................................... ___..
Bd=-C.Bl
Cd=AZI-AII- G -A22.C + G * A OG larger according to the di n the pole po-
Fig.1. Farst-Oder Hold Flux Observer. sition - R r / L r 2 j o
the f l u x observer
3-2, Stabilization and pole-position o f the observer. rotor flux observ
In the equation o f an observer, the real-part o f
the pole -
a is related to the settling time of the damping conditi
observer output signals. When we make - a smaller, In that situation,
that is, the poles of the observer are placed on the
far left side on the complex-plane, the output signal
of the observer is settled i n the more short time.
On the other hand, the imaginary part P o f the the observer s y
pole affects the angular ncy of the pulsation of disturbances.
the output signal. I f the of P is made a larger To keep the ga nstant independent of the
one, the observer is put in an unstable condition. rotating speed, bserver must
However, the value of 0 is made too small, the step
-response of the observer becomes vague.
T o obtain the excellent output signals from t h e ob- should be placed i
server, i t is not perfect that the pole should be set tlon of the open-1
in the left side on the complex-plane. In the
closed loop system, the stable boundary of the pole
position i s limited the slant line shown in Fig.2. the observer mu
In our observer system, the pole position in the real axis
open-loop i s the point of -Rr / L r + j o r , then, the
real-part - a must be set smaller than -Rr / L r .
And the pole position of the stable system must be
the area In the inner one between 2 4 5 ' line to the rotor flux correspond to the fundamental component.

IECON '88 I474


In Fig.J(a), the relation between components gl, g2 vector hr. And the error of the expected flux vector
of the gain matrix and the rotating speeds of the can be divided into two parts such as the amplitude
motor are shown. And Fig.3(b) shows the same relation and the phase angle component.
as in the case of the Fig.3(a), but these poles of the
observer are fixed on constant values of -400 k J 4 0 0 . Error of the amp1 i tude 6= I xr I / I Ar I
In the case * o f Fig.3(a), the variable Kaln de-
creases at lower rotating speeds of the motor. Error of the Phase-angle E =Lxr -LAr
From these results, we concluded that by adopting
the pole placement method mentioned above, we can de- 4. Characteristics of the rotor flux observer.
sign a stable rotor flux observer even at the lower
rotating speeds of the motor. In this chapter, the experimental results of the
rotor flux observer are glven.
Fig.4 shows the construction of the vector control
system of the induction motor for the experiment.
In this rotating speed control system, the switch-
ing devices of the current source inverter are GTO's
and the inverter is supplied by the current controlled
thyristor converter.
(b) Fixed Pole n Ld n
(-4002 1400)

U (11r( IOHz/d iv)

- (a) Variable Pole I t


wr
16 b i t CPU

Fig.4. Induction M o t o r Drive System


Fig.3. Values of Gain Matrix G. with Current Sorce Inverter.

3-3. Error of the estimated flux. A 100 V. 4 poles, 0.4 KW three-phase induction
An induction motor is supplied by the primary motor is directly connected to the torque transducer,
current whose amplitude is Is and whose angular the loads of the eddy current brake and the Prony
frequency is o e . The voltage vector Vs of the -brake. A rotary-encoder, whose output signal is 1024
primary voltage vs and the flux vector A r of the PPR, was connected to the other end of the tested
secondary flux h r can be calculated as follows. motor shaft. And these output signals are converted
A r =(A21 ' I s)/( j we- A22') to the 4096 PPR signal by the multiplying circuit,
then we obtained 1 . 8 RPM/bit accurately.
V s = ( j o e * I s-All'* I s-A12'*Ar)/B1' (13)

Where, each variable is a complex value. And ( I )


values of eq.(13) denote the practical values of the
equivalent circuit or the values which are calculated Control
-
from the equivalent circuit parameters.
On the other hand, the expected values of the sec-
/-
ondary flux vector Ar is obtained from the observer
equation. These result are explained in eq.(14). 1 r-
"nr={(A21- G - A l l + J o e * G ) I s- G*BI*Vs} Fig.5. Construction of Vector Control
/ { j me-(A22- G.Al2)I (14) System using Flux Obserber.
The observer error can be obtained by the subtrac- The thyristor converter is controlled by the phase
tion. that is, to subtract the practical value of the control signals from the 8 bit CPU(Z-80A.4MH). The
st-condary flux vector A r from the expected flux value of the DC reactor is about 250 mH. Gate slgnals

IECON '88 I475


of the GTO are supplied by a 8 bit CPUfZ-80A,4MH) the rotating speeds o
which is controlled by the 16 bit personal-computer
(8086CPU.8MH). And the wave forms o f the controlled
current inverter are the 120 rectangular or the PWM motor. In this
wave, in which the switching frequency is unsynchro- and the DC link cu
nized to the fundamental frequency. The filtering of the f s= & 1.
capacitors of 1 fi F are connected to the output results, at the
terminals of the inverter. in order to detect primary values o f the
currents and voltages o f the inverter, 12 bit A / D measured ones, bu
converter system is adopted, which has the function tion the expect
of sampling in the six-channels A/D converter at the decrease s 1i gh t 1 y .
same time, and the conversion time is about 50 usec..
The 1 6 bit personal-computer (8086CPU,8MH) was used
in calculating the data o f the rotor flux observer.
The sampling time of a tested system is determined braking in comparison with the power running condition
as 3 . 2 msec which includes the caiculating time I t of the observer
o f the vector control for the induction motor.
In this system, the upper limit of the inverter
frequency is about 30 Hz, in which case the observer
samples the data about 10 times t o one cycle o f these
signals. (a) Measured
Value
The functional block diagram of the induction
motor drive with the rotor flux observer i s given in
Fig. 5.
i
Expected wave forms of the rotor flux by the
observer system are shown in Fig.6.
E (b) First-Order
i Observer

(e) Zero-Order

-order hold, s a little

is maintained to

I lagging in the
0 aos om 01 0 005 aro hold can improve
1 I.e( 38c) Tlr(JPI;)

Fig.6. Flux Wave Forms obtained from


R o t o r Flux Observer.
In figure 6,(a) and (b) show the expected flux wave
forms on the fixed pole position (-125 +j125) of the
observer. And the tested motor i s driven under
steady-state conditlon in which the values of the DC flux observe the measured values, the
link current and slip-f ncy of the motor are kept length of lag d on the sampling-time o f
as constant values. Fig.6 (c) and (d) show the ex- the observer us problem. In which case
pected flux wave form under the condition in which the length of g time Is proportional to the
the poles o f the observer change in accordance with lagging angle o f the primary current. When the

IECON '88 I476


sampling time is set 3.2 msec., the lagging angle tating speed of the motor is about 600 rpm.
based on this sampling time becomes about 23 at Fig.9. shows the experimental results of decelera-
the 20 Hz of the supply frequency and 35 at the 30 tion and acceleration when the tested motor is oper-
Hz. This 1s a serious problem in the vector con- ated by the drive system shown In Fig.4 and the
trol system, therefore; we should make use of a state vector control system using the rotor flux observer
estimator or the same one which acts as a compensator shown in Fig.5.
of the lagging angle. Where,we propose a compensating
method of the lagging angle in the flux observer.
In this method, the absolute value of the expected
flux I A r I is used to compensate the lagging angle
with the expected rotor flux. That is, the error
value between the absolute value of the expected flux
I A r 1 and the command flux A r* is feedbacked to J:Homent of Inertia
the phase command value of the primary current.
z
0)

0
1.0 - Rotat lng Assembly
I-..
When the command current of the torque is a posi- 8':
b.5°;
tive value, and the phase angle of the expected flux 20 30 40 50
9
is lagging from the current command, the value of the -1.0 - Time(SEC1

exciting current becomes larger, and the value of the


rotor flux comes to be a larger than the command value.
Conversely, in the case that the phase angle of the In this experiment, the command values of the
expected flux is led to the command one. The values rotor flux maintained 0.15 Wb. And in this case, the
of the exciting current component becomes small than zero-order hold and a state estimator ware adopted in
that o f the command one, and the values of the ex- order to save the calculating time. The command value
pected rotor flux are reduced in these of the flux of the rotating speed changes on the step from -600
command. In the case of the inverse polarity of the to +600 rpm. The value of the DC link current i s set
torque command signal, this relation is reversed. under 3 . 0 A.
In the experiment, the absolute value of the ex-
pected flux I A r I was used for the compensation of 6 . Conclusion
the phase angle in the primary current. Therefore,
the value of I A r I can not be applied for compen- The rotor flux observer by the micro-computer was
sating the amplitude of the flux. applied to the servosystem of the induction motor.
However, we actually use the above system, the In the rotor flux observer, the first order hold
command value of the exciting current i ds is kept is adopted instead of the zero-order hold in the
constant. We suppose that this matter is not a ordinary observer system. With this method, the
serious problem in a practical application. lagging time which causes the sampling time of the
observer is eliminated . And in our observer, the
pole positlon of the observer equation changes in
accordance with the motor speed and, quick and
accurate response of the expected value of the rotor
flux is obtained.
When our proposed observer is applied to the
vector control system of the induction motor, the
phase angle of the primary current is compensated b y
the expected flux.
We suppose that our control method of the induc-
Fig.8. Transient Response of Flux Observer. tion motor can be practically used in the current
source inverter drive system.
Fig.8. shows the step response of the estimated
flux in the our experimental system. The flux command Reference
A r* is maintained as constant value of 0.15 Wb, and (I)Y.Hori,V.Cotter and Y.Kaya.Control Theoretical Con
the current command of the torque component iqs* siderations Relating to an lnduction Mashine Flux
Obsever. Trans.J1EE.106-B,11,1001-1008,(1986.N0V.~
changes on a step from 1.5 A to 0.3 A. And the ro-
IECON '88I 477