Paper

Deadbeat

Control

State

Observer

Member Atsuo
Non-member

of Induction
with Adaptive

Kawamura

Lazhar

Motor

Current

Poles

using

Selection

(Yokohama National University)

Ben-Brahim

(Yokohama NationalUniversity)

Summary

For a high power induction motor drive, the switching frequency of inverter can not become
higher than several kHz, and such switching, frequency produces a large current ripple, which in
turn, produces

torque ripple. To minimize the current

ripple, a new method

based on deadbeat

control theory for current regulation is proposed. The PWM pattern is determined at every
sampling instant based on stator currents measurements, motor speed, current references and rotor
flux vector, which is predicted by a full-order observer with adaptive poles selection, so that the
stator currents
instant.

are controlled

to be exactly

equal to the reference

currents

at every sampling

The proposed method consists of two parts : ( 1) derivation of a deadbeat control ; (2) construction of a full-order observer which predicts the rotor flux and the stator currents in the next
sampling instant. This paper
various system conditions.

describes

key word : Induction Motor, Deadbeat

1.

a theoretical

and computer

simulations

under

Control, Observer
torque ripples. The deadbeat technique is a type of
control, by which the system exhibits a finite set

Introduction

The vector control technique, also known as field

tling time response, as already applied to the PWM

inverter in Ref. (2 ) ^S ( 4) In this paper the dead
beat control was extended to the induction motor

oriented control, has made it possible for induction

motor drives to be used in high performance appli
cation. However, the current controlled vector con
trol,

analysis

in general,

requires

the

following

drive to satisfy the above second condition"'. The

current of induction motor is controlled without

assump

tions(8)' ;

delay time by the proposed method as follows. The

(1)
An exact
vector is available.

information

on the rotor

flux

inverter with induction motor load is considered as

a plant of a closed loop digital feedback system.

(2)
The actual stator currents are adjusted
instantaneously
and precisely to the references.

the proposed

Unlike the conventional

The latter condition (2) is satisfied by a high

switching frequency. However, the switching fre

pattern at every sampling instant based on stator

currents, rotor flux, motor speed and current refer

quency remains

ences. As a result the stator currents are controlled

to be exactly equal to the reference currents at

as low as a few kHz for a high

power induction motor drives. As a result, large

current ripples are produced, which in turn, produce

every
590

bang-bang

current control,

control scheme determines

sampling

instant,

and the

T. IEE Japan,

the PWM

torque

ripples

Vol. 110-D, No. 5, '90

Descriptor

become

less

To

than

satisfy

rotor

flux

may

be

sensing

Thus,

more

desirable,

this
the

minimize

the

indirect

of

observers(")-(19),

works
tive

and

on

Ref.

estimated

flux

term

error

to

zero

are

to
in

the

Among

the

design
the
of

order

to

of

related
a

predic

converge

quickly.

the

minimum
The

order

flux

observer

using

corrective

meanings

was

these

reduced-order

designed

in

Ref.

(14)`(16).

=1-M2/L1L2

are

sensitive

to

the

measurements

rents(13)(17).

rotor

Furthermore

observers
this

of

do

not

problem,

observer

the
is

speed

around

estimate

the

poles

shifted

and
zero

flux

stator

of

adaptively

, R2

: stator

and

rotor

resistance,

L2

: stator

and

rotor

self

solve

full-order

depending

: stator/rotor

wr

: electrical

Vds,

on

Ids,

The

Combining

the

state

controller,

controlled

at

estimator

(observer)

the

stator

values

and

wide

source
ed

Iqs

: d-axis

digital
concludes

2.

angular

and
and

q-axis
q-axis

velocity,
stator
stator

volt-ages,
currents

Qr : d-axis
torque

equation

and

q-axis
is

rotor

fluxes

then

(3)

range

equation

of the mechanical

system

is

of

as follows,
deadbeat
inverter

controller
fed

in Section

observer

rotor

are

speeds.
The

inductance,

inductance,

the

currents

over

mutual

VQS : d-axis

and the state

any

follows,

the

speed.

deadbeat

as

these
To

dr,
motor

coefficient,

cur

speed,

well(17).

position

are

: leakage

noise
L1,

in

parameters

However
R

observers

of

prediction
o

error

the time derivative,

machine

and

importance
in

denotes

expen

proposed

evaluates

symbol"""

where,

of

variation

(13)
the

feedback

on
been

analysis

The

sensor,

and

method's'-i10'.

emphasizes

error

hall

methods

parameters

works

direct

viewpoint

dependent
have

control

interesting

by
the

sensing

are

(2)

, the

unreliable

methods

effects

vector

recent

is
flux

they

as

From

method

Several

method.
(1)

such

theory.

indirect
but

parameters.

conventional
condition

method
in

use,

sive.

in

former

preferable

practical

flux

those

the

2 of
is

this

Deadbeat

induction
this

paper.

described

computer

applied

in

motor
The

drive
design

Section

simulation

for

3.
results,

where,
Te : electromagnetic torque,
TL : load torque Jm : moment of inertia,
D : damping factor
2. 2 Deadbeat control theory
Fig. 1 is the proposed digital control system of the
induction motor drive. Assuming that the inverter is
for high power application, 1.8 kHz sampling fre

is present
of

Section
and

(4)

voltage

full-order
4

gives

Section

paper.

control

law

2. 1 Induction motor state equation

The basic equation of an induction motor with the

quency is chosen. To derive a sampled-data model

of Eqs. (1) and (2 ), inverter output voltages Vab,
Vb, and Vcaare chosen as one or two voltage pulses
of magnitude +E or -E, satisfying the relation of
Vab+ Vbc+Vca=0 at any moment as shown in Fig. 2.
JTab(k), ZTbjk), JTca(k) are the pulse widths of
Vab,Vb, and Via respectively in the k-th sampling
interval. The two phase pulse widths ZTd(k) and
QTQ(k) are given by Eqs. (5) and (6 ).

d-q coordinates fixed in the stator are expressed as,

(1)
The output

equation

D,1105,

is
2

591

Fig. 1.

Proposed

induction

motor

drive

system.

Fig. 3.

Timing
beat

chart

of the proposed

dead

control.

4T(k)=(/JTd(k), 4Tq(k))r. F, H, Is(k), A,(k) and

4T(k) are the values at the sampling instant of kT.
Thus, using Eq. (7) and replacing 18(k+l) with
the reference Isref(k+1), the deadbeat control law
is given as follows",

Fig. 2.

Pulse pattern selection

(8).

(example).

This control law forces the output currents to be

exactly equal to the reference signal at the (k+1)th
sampling instant. For computation of Eq. (8), the
values of F,1, F12and Hi, which are derived from the
actual motor speed and plant parameters, have to be
calculated or searched from look-up table, and 18(k)
and .ir(k) must be provided at every sampling
instant. The references are one sampling ahead
preview values of stator currents, which are pro
vided by the vector control method(5).If IS(k) is
detected at kT from which Ar(k) is estimated, fol
lowed by computation of Eq. (8 ), then the pulse
width zT(k) in Eq. (8) cannot be as large as the
sampling interval T due to the finite calculation
time. To get rid of this, 12(k) and ar(k) are predicted
at the (k-l)th sampling instant, and Eq. (8) is
computed before k-th sampling instant, which
enables the maximum pulse width to be equal to T,
as shown in Fig. 3.

(5)

(6)
As long as these pulses are symmetrical
center of the sampling
Vq, the sampled-data
becomes",

at the

interval and so are Vds and

model of Eqs. (1)

where, F,,, F,z, Fu and F22 are corresponding

and (2 )

(2 x 2)

2. 3

matrices of exp (AT), and H, and H2 are corre

sponding (2x2) matrices of exp (AT/2)BE,
and

Pulse

pattern

selection

The three phase pulse widths zTab(k), JTbr(k),

and JTca(k) are calculated in Eq. (5) from JTd(k)
592

T. IEE Japan,

Vol. 110-D, No. 5, '90

Descriptor

and Tq(k)
for

obtained

selection

(1).
in

An
Fig.

of

in

actual

example

Eq.

(8

pulse

of pulse

).

is

The

the

method

same

pattern

used

as

in

Ref.

selection

is shown

corrective

predic

2.

3.

Full-order

3.1

Design

of

tion

observer

error

Recently,

are

a full-order

flux
deal

of no

change

with

low

in

approaches
was

However,

applied

none

problem

even

to

of

these

the

case

with

cor

in

parameters.

observer

prediction

published

theory

speed

motor

full-order

rective

several
observer

estimation(17)`(19).

papers

The

with

feedback

there

in which
rotor

observer

state

error

equation

feedback

becomes

as

Fig. 4.

Torque,
waveforms.

current

and

voltage

fol

lows(21),

law Eq. (8 ), digital simulations

were carried

out,

assuming that the flux vector could be measured.

The motor parameters used for simulation are listed
in the appendix. Fig. 4 shows the waveforms of
stator current Ids, electromagnetic
torque Te and

(9)
The

symbol"-"

where,

denotes

G=kG1 G2l

The

error

is

the

the

gain

dynamics

predicted

quantities.

stator voltage Vds. Fig. 5 shows the comparison of

torque and currents ripple between the proposed
and the conventional hysteresis comparator method

matrix.

of

the

observer

is,

in which the phase currents are forced to track the

reference generated by the vector controller. The

(10)
where,

switching frequency of the proposed pulse pattern in

Fig. 2 is about 2/3 of the sampling frequency, thus

(11)
The

gain

matrix

of Eq.

(10)

take

placement
first

poles

choice
Eq.

the

technique

transformed

then

of
(12),

it

specified

of

the

is

desired
is

such

in

a controllable

the

observer
gain

possible

conjugate

that

values.

used,

into

observer
is

designed

to

Thus
which,

the
Eq.

are

pole

(7)

place

the

pairs

as

With
shown

poles
in

the

trol method is about 40% less than that of hysteresis

method.

is

form,

assigned.
G

the average of the switching frequency of the hyster

esis comparator method is adjusted to be equal to it.
From Fig. 5, the torque ripple of the proposed con

poles

canonical

matrix

complex

the

at
s

4. 2
in

Simulation
poles

of

state

observer

with

fixed

placement

The dynamic of the observer is determined by the

selection of poles position. Simulation results are

any

doma

shown in Fig. 6, using the gain matrix given by Eq.

in(12)`(17).

(12) and setting the poles of the observer

domain, for example, as follows,

(12)

in s

a,}jB1=-2,000}j2,000
a2}jB2=-1,000}j1,000

4.

Fig. 6 shows

Simulations

that the stator

flux are estimated

4. 1

Deadbeat

In order

control

to verify

D,1105,

simulations

the proposed
2

deadbeat

control
593

low speed

where

with

real

the

without

the predicted
one.

The

currents

any error,

and the rotor

except

at very

flux does not coincide

observer

poles

should

be

Fig. 5.

chosen
all

in such

state

of fixed

a way

variables

prediction

error
poles

allocations,
ably small
To

Torque

an adaptive

ripples

Fig. 6.

Reverse

that

the observer

at any

speed

as fast as possible.
were

simulated

but fixed
prediction

overcome

and current

this
poles

poles
errors

problem,
placement,

for

did not

and

speed

operation

reduce

kinds

the

poles

best

provide

method,

is presented

with

4.3

the

Different

with

1.2 kHz

fixed

switching

frequency.

poles.

following section.

can predict

in the low speed

a new

at 60 Hz operation

Simulation

of the observer

with adaptive

poles placement
The induction motor poles as a function of rotor
speed are computed as shown in Fig. 7. Poles in this
figure are eigenvalues of the controllable canonical

reason
range.
called

form of the matrix

in the

they
594

are

open

A in Eq. (1).

loop

system

T. IEE Japan,

In other words,

poles.

Thus

if the

Vol. 110-D, No. 5, '90

Descriptor

Fig. 7.

observer

poles
these

which

deteriorates

adaptive

are

function

of

estimated
reverse
,

speed

positions

of the

Fig. 8.

Reverse

speed

fixed

by

become
the

shifted

when

motor

which

the

was

averaged

and

the

selected,

as

with

proposed

the

poles
move

observer
the

poles

speed.

placement.

the

Since the whole sampling interval (0.555 ms) is

available
for necessary calculation,
the actual

the

implementation

as

through

adaptive

of rotor

tem, and Fig. 8 shows the reverse speed operation

using the proposed controller.

of

of

as a function

predicted, deadbeat controller with observer under

vector control (DBOV) is implemented in the sys

speed,

error

provided

motor

observer
low

the

poles

Then,

variables

were

In

technique,

operation,

error,

at

induction

operation

the

high

prediction.

speed.

a2}jB2,

adjusting
very

placement

rotor

state

and

average

gains

poles

observers

}jB1

are

gains,

Poles

poles

can be made by a digital

processor, which can execute the required computa

tion using look-up tables for coefficients of deadbeat
control and adaptive observer poles. The look-up

a,

minimum

follows,

table for poles should be prepared using Eq. (13).

However, the speed step has to be chosen carefully

(13)
where, y, and y2 are the real parts of the motor poles

taking into account the desired performance

in Eq. (1) .
This choice provides

as the hardware constraints.

4. 4 Parameters
sensitivity

the best

results

through

simulations. The flux is well predicted even at low

speed. Now that all state variable are precisely
D,1105,

signal

The coefficients in Eqs. (8) and

dependent
on motor parameters.
595

as well

(9) are all

Since these

Fig. 9.

parameters

may

to

temperature

to

investigate

vary
or

to

changes

resistance

R2.

deadbeat

control

range

of

R2

complete

Fig.
at

the

wr=1rpm
slip
At

10%,

low

the

difference

flux

is

R2

causes

tude

between

about

and

R2
actual

+4

degrees

variation

of

Note

(Manuscript

that

flux
is

and

in

do

not

(1)

increased,

actual

+10%

change

and

+8

degrees.

method

has

beyond

this

5.

to

be

R2

rotor

variation

rotor

flux

deviation

in

phase,

causes

respectively

An

additional

considered

(2)

of
magni
while

for

+9%

compensating
these

cases,

but

it

is

(4)

paper.

Conclusions

new

ory,
selection,

using

(5)

technique
a
was

state
proposed

based

on

observer
to

poles selection

received Jan. 9, 1989,

revised May 16, 1989)
References

(3)
+20%

which, in turn,

rotor speeds.

exceeds

rotor

variation

example,

+6%

R2

method

variation

predicted

For

the

(DBOV).

R2.

enabled low current rip

showed the deadbeat current control to be effective

around zero rotor speed, and therefore contributed
to performance improvement over a wide range of

the

of

the follow-

rpm.

resistance

increased.

of
as

controller

ple with lower switching frequency,

resulted in low torque ripple.
The state observer with adaptive

practical

proposed
of

whenever
the

the

The deadbeat

the

of }5%

the

100

that

system

variation

and

(1)

variation

is

most
rotor

sensitivity

drive

+5%

the

is the

increased

under

(ws)

rotor

the

change

system

speed,

Ref.

retical analysis and digital simulations,

ing conclusions were made.

system

of

by

that

R2
A

predicted
As

in

complete

frequency

very

coincide.
the

Note

with

One

stator currents under vector control. From the theo

due

important

complete

affected

decreased.

is

parameters

not

to

the

the

shown

is

affect

9 shows

the

motor

was

change.

speed
not

of

it

changes.
in

system

motor
does

sensitivity

It

operation

effects,

parameters

significant

on-line

saturation

the

(DBOV)

during

Very low speed with 5% change of Rz.

deadbeat
with

control

control
adaptive
induction

the

(6)

poles
motor

596

L. Ben-Brahim and A. Kawamura : "Simulation Study of

Digital Control of Induction Motor Current with Deadbeat
Response", Technical Meeting of IEE of Japan on Semi
conductor Power Conversion, SPC-88-64 (1988)
K. P. Gokhale, A. Kawamura & R. G. Hoft : "Deadbeat
Microprocessor Control of PWM Inverter for Sinusoidal
Output Waveform Synthesis", IEEE Trans. Industr Ap
plic, IA-23,901 (1987)
A. Kawamura,
T. Haneyoshi & R. G. Hoft : "Deadbeat
Controlled PWM Inverter with Parameter
Estimation
using Only Voltage sensor", IEEE Trans. Power Elec
tronics, 3, 118 (1988)
A. Kawamura & K. Ishihara : "Real Time Digital Feed
back Control of Three Phase PWM Inverter with Quick
Transient Response Suitable for Uninterruptible
Power
Supply", IEEE. IAS Annual Meeting, p. 728 (1988)
A. Kawamura & R. G. Heft : "An Analysis of Induction
Motor Field Oriented or Vector Control", IEEE Pow. Elec.
Spec. Conf. Rec., p.91 (1983)
F. Alashhab, A. Kawamura & R. G. Heft : "Matrix Trans
formation Analysis to Derive Unique Features of FOC and
FAM Induction Motor Control",
IEEE IAS. Conf. Rec.

T. IEE Japan,

Vol. 110-D, No. 5, '90

Descriptor

p.586 (1985)
T. A. Lipo & E. P. Cornell : "State-Variable
Steady-State
Analysis of a Controlled Current Induction Motor", IEEE
Trans. Indtatr. Applic., IA-11, 704 (1975)
M. Koyama, M. Yano, I. Kamiyama & S. Yano, et al. :
"Microprocessor-Based
Vector Control System for Induc

(7)

(8)

(9)

(10)

(11)
(12)
(13)

(14)

(15)

(16)

(17)

(18)

(19)

(20)

Atsuo Kawamura

M. S. E. E., and Ph. D. degrees

Electrical

Engineering

parameters
60 Hz,

used

2 poles,

National University,
where he is presently

Damping

of

inertia

factor

R1=0.0793,
L1=31.25mH,

(Jm=0.1
(D=0.01

award in 1988.
Dr. Kawamura is a member of the IEE of Japan,
Robotics Society of Japan, the Institute of Elec
tronics, Information and Communication Engineers,
the Society of Instrument and Control Engineering,
and IEEE.

Lazhar Ben- Brahim (Non-member)

He was born in Chaalil, Tunisia.
He received the Engineer and M.S.
(DEA) degrees in Electrical Engi
neering from L'Ecole Nationale
D'Ingenieurs de Tunis (ENIT), Tunisia, in 1985 and
1986, respectively. From 1986 to 1987, he served as
an electrical engineer at the Tunisian Telecom
Electric and as a teaching assistant in the Depart
Engineering

of L'ENIT.

He is

currently working on his Ph. D. degree in AC motor

control at the Department of Electrical and Com

induction

puter engineering of Yokohama National Univer

sity, Yokohama, Japan. His interests include appli
cation of modern and adaptive control, robotics, and

NEm2)

power electronics.
He is a member of IEEE Industrial Applications,
Industrial Electronics, and Power Electronics Soci

NEm2/s)

R2=0.0785
L2=31.19mH

eties.
2

Yokohama, Japan, in 1986,

an Associate Professor in the

control, ultrasonic actuators, robotics and so on. He

is a recipient of IEEE IAS transaction prize paper

M=30.50mH

D,1105,

in
of

Department of Electrical and Computer Engineer

ing. His interests are in power electronics, digital

motor.
Moment

University

and was an Assistant Professor there from 1983

through 1986. He joined the faculty of Yokohama

for simulations
440 V,

the

Electrical and Computer Engineering at the Univer

sity of Missouri-Columbia as a Postdoctoral Fellow,

ment of Electrical
motor

from

Tokyo, Tokyo, Japan, in 1976, 1978, and 1981,

respectively. In 1981, he joined the Department of

Appendix

3 phases,

Pre

fecture, Japan on December 17,

1953. He received the B. S. E. E.,

tion Motor
Drives
with
Rotor
Time
Constant
Identification Function". IEEE Ind. Appl. Soc. Conf. Rec.,
p.564 (1985)
R. Krishnan & F. C. Doran : "Study of Parameter Sensitiv
ity in High Performance
Inverter-Fed
Induction Motor
Drive Systems", IEEE Trans. Industr. Applic., IA-23, No.
4 (1987)
T. Ohtani : "A New Method of Torque Control Free From
Motor Parameter Variation in Induction Motor Drives",
IEEE Ind. Appl. Soc. Conf. Rec., p. 203 (1986)
W. Leonhard : Control
of Electrical
Drives
(1985)
Springer-Verglas
B. K. Bose : "Power Electronics
and AC Drives (1986)
Englewood, NJ, Prentice-Hall
G. Verghese & S. Sanders : "Observers for Fast Flux
Estimation in Induction Machines", IEEE Pow. Elec. Spec.
Conf. Rec., p.751 (1985)
V. Hori, V. Cotter & Y. Kaya : "Considerations
of an
Induction Motor Flux Observer in a Vector Control Con
text : Control Theoretical
Connection and Simulation",
JSST International Conference, p. 448 (1986)
V. Cotter, Y. Hori & Y. Kaya : "Design Method for
Robust Induction Motor Flux Observer", Technical Meet
ing IEE of Japan on Semiconductor Power Conversion,
SPC-86-43
(1986)
Y. Hori, T. Sasaki & T. Uchida : "Implementation
of
Second Order Discrete-Time Flux Observer of Induction
Motor Using DSP". Technical Meeting IEE of Japan on
Industrial Electrical Application (in Japanese),
IEA-8819 (1988)
H. Kobayashi, M. Koizumi, H. Hashimoto, S. Kondo & F.
Harashima : "A New Controller
for Induction Motors
Using Flux Observer", IEEE Pow. Elec. Spec. Conf., p.
1063 (1988)
II. Kubota, K. Matsuse & T. Fukao : "Improvement
of
Vector Controlled CSI-Fed Induction Motor Drives using
State Observer" (in Japanese), Trans. IEE Japan, 108-D,
658 (1988)
H, Kubota, K. Matsuse & T. Fukao : "Indirect Field
Oriented Control of CSI-Fed Induction Motor with State
Observer", IEEE IAS. Conf. Rec., p. 138 (1987)
Y. Dote : "Stabilization
of Controlled Current Induction
Motor Drive System
Via New
Nonlinear
State
Observers", IEEE
Trans. Industr. Electronics
Control
Inshvmn. IEIC-27. 77 (1980)

Induction

(Member)

He was born in Yamaguchi

(21)D.
G.Luenberger
: "Observers
forAC-11,
Multivariable
systems",
IEEE
Trans. Automatic
Control,
190 (1966)

60 hp,

597