IEEE TRANSACTIONS ON POWER ELECTRONICS,VOL 11, NO 2, MARCH 1996 213

Loss Minimization in Scalar-Controlled Induction

Motor Drives with Search Controllers
Iordanis Kioskeridis and Nikos Margaris

Abstruct- Loss minimization in scalar-controlled induction

motor drives (IMD) with search controllers (SC) is investigated.
The problems arising when the input power is used as the
controlled variable are described. It is proved that better results
are achieved if the stator current is used as the controlled '.
'n

controIlcr
NOMENCLATURE
I I
Stator resistance.
Rotor resistance. Fig. 1. Optimal IMD.
Magnetizing reactance.
Magnetizing reactance at flux 1 pu. I. INTRODUCTION
Stator leakage reactance.
Rotor leakage reactance.
Magnetic loss resistance.
T HREE-PHASE induction motors consume 60% of indus-
trial electricity [l], and it takes considerable effort to
improve their el'ficiency. This effort focuses mainly on the
Stray loss resistance.
improvement of the materials, the design, and the construction
Supply frequency.
techniques [l], 121. However, in spite of the progress made
Slip.
in the fields mentioned above, there will always be signifi-
Slip frequency.
cant margins for efficiency improvement, especially when the
Speed.
motor operates at light load.
Base speed.
The efficiency and power factor can be improved by making
Air-gap flux.
the motor excitation a monotone increasing function of the
Stator voltage.
load. To achieve: this goal, the induction motor should either
Air-gap voltage.
be redesigned or fed through an inverter. In the first case, no
Stator current.
considerable efforts have been made yet, and even those that
Rotor current.
have been made are not widely accepted [3], [4]. The second
Magnetizing current.
case is very expensive except in the case of the induction motor
Electromagnetic torque.
drive (IMD) where feeding through an inverter is necessary.
Load torque.
A block diagram of the optimal IMD is shown in Fig. 1.
Total power losses. In the basic diagram, it is possible to embody one of the
Copper losses. two controllers that minimize the losses of the drive. The loss
Iron losses. model controller (LMC) measures the speed and stator current,
Stray losses. and through the motor loss model determines the optimal air-
Mechanical losses. gap flux. The search controller (SC) measures the input power
Eddy current and hysteresis coefficients. of the drive and searches for the optimal excitation value.
A review of the relevant literature [5]-[12] shows that
Stray loss coefficients.
the volume of mathematical operations and the difficulty in
Mechanical loss coefficient.
measuring the parameters of the loss model does not permit the
Magnetizing curve coefficients. implementation of an LMC. Consequently, loss minimization
has been mainly achieved by using lookup tables [5], [6] or
SC'S [7]-[ 111.
Manuscript received February 28, 1995; revised September 2, 1995. This A lookup table requires a number of costly and time-
work was supported by grants from the Ministry of Industry, Energy, and consuming measurements for each motor. Such measurements
Technology of Greece. are practically impossible on motors already in operation. The
The authors are with the Department of Electrical and Computer Engineer-
ing, Aristotle University of Thessaloniki, 54006 Thessaloniki, Greece. implementation of an SC does not require the knowledge
Publisher Item Identifier S 0885-8993(96)01937-0. of the motor loss model. However, the SC does require the
0885-8993/96$05.00 0 1996 IEEE
214 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 11, NO. 2, MARCH 1996

TABLE I
PARAMETERS OF THE MOTORSUSED IN THE ANALYSIS
AND THE EXPERIMENTS

Fig. 2. Per-unit induchon motor equivalent circuit.

The magnetizing and rotor currents are given, respectively, by

and

From (3) we obtain the air-gap flux

am = I i d ( r : / s a ) z +xi:.
The electromagnetic torque is given by

I 0.77 I 0.45 r:.

T, = I,' 2 -.
sa
Substituting (3) into (3,we obtain
measurement of the input power. The performance of this
measurement with the desired accuracy is complex and costly
[13]. The minimization of the input power results in serious
limitations. For example, SC's cannot successfully find the
Usually, the induction motor operates with a small slip and
minimum of the input power function that is smooth and flat
the condition rC/s >> a x ; , holds. In this case, (4) and (6)
around the minimum [12]. Furthermore, SC in a real system
become, respectively
does not reach a steady state and causes oscillations in the air-
-1 TI
gap flux that results in undesirable torque disturbances [lo],
U 11. (7)
In the following, loss minimization in IMD by the use of an sa
T, 2 -a$ % @,I;. (8)
SC is examined. In Section 11, the relationships between the r c
induction motor variables are given. In Section ILI, the loss
Also, according to the analysis presented in [14]
model of the induction motor is presented. In Section IV, the
problems related to the minimization of the input power are I," = I; + CL12 (9)
discussed. In Section V, it is stated that the loss minimization
can be successfully achieved by minimizing the stator current where
instead of the input power. Experimental results are given in
Section VI.
CL =1 + 2-.Xlr (10)
Xm
Leakage reactance is not affected by motor core saturation,
while magnetizing reactance is highly affected [15]. The
11. RELATIONSHIPS BETWEENINDUCTION MOTORVARIABLES
magnetization curve is given by [I61
The per-phase equivalent circuit of the induction motor
shown in Fig. 2 and the motor equations are given in the I, = SI@" s2aL s3@% + (11) +
per-unit system. The per-unit frequency is
and the magnetizing reactance is
KIOSKElUDIS AND MARGARIS: LOSS MINIMIZATION IN SCALAR-CONTROLLED INDUCTION MOTOR. DRIVES 21s

"
0.2 0.4 0.8 0.8 1 1.2
*.gsph@u) r-rrsh(md -"n -)

(a) Fig. 4. Sensitivity of the input power with respect to air-gap flux at w = 1
pu and various load torques.

Stray Losses: These arise on the copper and iron of the

motor [17], [18] and are given by
r, = c,& + csqJ,2 + ce.I,". (15)
In the equivalent circuit of Fig. 2, the stray losses are repre-
sented by the equivalent resistance r,t, in the stator branch. As
it is known, the stray losses are mainly attributed to the rotor
current [3], [19], [20]. Since the rotor current in a squirrel cage
induction motor is not measurable, it is a common practice to
express the stray losses as a function of the stator current.
Another expression for the stray losses is
P, = CstrW 2 I,'2 . (16)
This expression holds for dc motors [12] and is also acceptable
82 0.4 0.8 0.8 1 1.2
*.os ilJx @U) for induction motors [3].
(b)
Fig. 3. (a) Power loss and (b) input power versus air-gap Rux at w =1 pu IV. INPUTPOWERMINIMIZATION
and various load torques. Commonly, the input power is used as the controlled
variable when SC's are used for the loss minimization of IMD.
111. LOSS MODELOF INDUCTION MOTOR Neglecting rotor iron losses and expressing stray losses by
(16), the total power losses are given by
The induction motor losses that can be minimized by
excitation adjustment are the following. PlOS,= ?-,I: + ?-;1;2+ (Icew2 + khW)@$
Copper losses: These are due to flow of the electric current
through the stator and rotor windings and are given by + CstrW21i2 + CfwW2 (17)

P,, = r,I: + .:I?. where c f w w 2 are the mechanical losses of the motor. Ne-
(13)
glecting rectifier and inverter losses, the input power is given
Zron losses: These are the losses due to eddy currents and by
hysteresis, given by
pin = 90,s + Pout =9oss + Tew (18)
Pfe= [ k e ( l + 2 ) a 2 + k h ( l + .).]@.E. (14)

where Pout= Tew is the drive output power.

The terms that contain the motor slip (3) correspond to rotor Considering (2), (8), (9), and (17), (18) becomes
iron losses. When the motor operates under normal conditions
( 3 << l), the rotor iron losses can be neglected. In the Pin =(cL,rS + r i + CstrW2)-T," + ( h e w 2 + I C ~ W ) @ ~
equivalent circuit of Fig. 2, iron losses are represented by the @L
equivalent resistance r, . + ?-,I: + q w w 2 + T,w. (19)
216 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 11, NO. 2, MARCH 1996

Oonoo-'s' - - --.. .. ..
25 0000 i' - 50 0000 s-
5 00 S / d l V

Fig 5 Loss minimization process in the 1-hp dnve when the controlled
variable is the input power

Taking into account the saturation effects from ( l l ) , (19)

becomes

As illustrated in Fig. 3, the input power Pi, is smoother

around the minimum than P~,,,, particularly in large efficient
motors. However, since the measurement of PI,,,is very
difficult and expensive, only the minimization of Pi,,can be
accomplished.
The sensitivity function of Pi, with respect to am at steady
state, is as follows:

*-es fiux @U)

(b)
Fig. 6 . Stator current versus air-gap flux at w = '1 pu and various load
torques. (a) 1-hp motor. (b) 100-hp motor.
From Fig. 4, it is concluded that the sensitivity around the
minimum is low, particularly in large efficient motors. Further-
more, from (21) it is concluded that the sensitivity depends on reach to the conclusion that the minimization of Pi, or i d c can
speed. successfully be applied only on less efficient motors, such as
Even though the measurement of Pi, is easier than the mea- small motors.
surement of Pl,,,,its performance with the desired accuracy Fig. 5 presents the loss minimization process in a 1-hp drive
remains complex and costly. For this reason, instead of Pi,,, using SC when the input power is used as the controlled
the current of the inverter dc link ( i d c ) can be minimized [7]. variable. At steady state, the air-gap flux exhibits oscillations
Neglecting the rectifier losses, the input power of the drive due to low sensitivity. Fig. 5 refers to light loads where the SC
is given by performance is the best possible one. At higher loads the air-
gap flux is more oscillatory because the minimum of the input
power is less distinct. This behavior also has been observed
in [lo] and [lll.
Since the voltage 'ZLdc is practically constant, Pi, variations
follow i d c variations. Even though the minimization of i d c V. STATORCURRENTMINIMIZATION
reduces the cost, the method still suffers from low sensitivity. A variable more sensitive to the air-gap flux variations
Also, taking into account the noise that attends real drives, we than the input power is the stator current I,. Considering the
 KIOSKERIDIS AND MARGARIS: LOSS MINIMIZATIONIN SCALAR-CONTROLLED INDUCTION MOTOR. DRIVES 217

0.5 0.6 0.7 0.8 O.@ I 1.1 12 1.3 1.4 1.5

0.5 0.6 0.7 0.8 0.9 1 1.1

*gap Iln (real MLrJoplhun value)
(b)
1.2 1.3 1.4

Fig. 7. Sensitivity of the stator current with respect to air-gap flux for various
1.5
-tor -
(b)
(pl)

Fig. 8. Input power versus stator current at w = 1 pu and various load

load torques. (a) 1-hp motor. (b) 100-hp motor. torques. (a) 1-hp motor. (b) 100-hp motor.

nonlinear behavior of the motor, (9), due to (8) and (ll), is to the conclusion that the minimum of I , is more distinct than
rewritten the minimum of Pi,.
The sensitivity function of the stator current with respect
to the air-gap flux is given by (24), shown at the bottom of
the page. Fig. '7 demonstrates the sensitivity function of the
stator current with respect to air-gap flux. Comparing Fig. 7
= /(SI@- + SZ@% + s ~ @ L+) -.CLT:
~
%7
(23)
with Fig. 4, we see that the stator current sensitivity with
respect to air-gap flux is greater than the respective sensitivity
of the input power. Additionally, the stator current sensitivity
Fig. 6 illustrates the variation of the stator current with function is independent of the speed and does not significantly
respect to air-gap flux. Comparing Fig. 6 with Fig. 3, we reach depend on the motor efficiency.
218 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 11, NO. 2, MARCH 1996

1FJ

,./... ..... ........ ......... .........,.

.
. . ........... .............. .................
i :.- .......... i

I.UA

-0 00000 s- - I O 0000 s - 20-0000 s+

. .
0 00000 s 25 0000 5 50 0000 s
2 00 s/dlv 5 00 s / d i v

Fig 9 Loss minimization process in the 1-hp dnve when the controlled
vanable is the stator current Fig. 11. SC response to load torque variations (U = 1 pu,
T = 0.2 t 0.8 t 0.2 pu).

2) A step variation Au in the value of the control variable

is applied on.
3) The new value of the controlled variable y Ay,that +
corresponds to the new value of the control variable
U AIL, is measured. +
4) If Ay < O(Ay > 0), a new step variation Au(-Au)is
forced on the control variable.
In case of the loss minimization of IMD, some limitations
are set in the implementation of SC. These limitations are the
following:
The value of the controlled variable y should be measured
after the steady state has been established. At the time the
process reaches the minimum, because of the air-gap flux
weakening, the settling time of the drive increases resulting
in the increment of the SC sampling period. The sampling
period is larger as the load of the motor, and consequently the
air-gap flux, is lighter.
A large positive step variation of the control variable is
acceptable and should be set, when the operation conditions
of the drive demand such an increment (torque demand).
On the contrary, the amplitude of the negative step variations
I
must be limited. Large decrement of the air-gap flux causes
j
ow.
._ __- 25.m 50.00 a
(I
increment of the rotor current and consequently increment
5.00 .Ibr of the losses and undesirable disturbances in the flux and
(b) the torque. The decrement of the air-gap flux should be
Fig 10 Loss minimizaaon process in the 1-hp dnve when the controlled
accomplished in small steps.
variable IS (a) the input power and (b) the stator current The decrement of the amplithde of the step variation Au of
the control variable when reaches the minimum, that is the use
Fig. 8 is a plot of P,, versus I,, which results from (20) of an adaptive step, is not a practical solution for the following
and (23) after the elimination of @ .,. As illustrated in this two reasons: 1) The variation Ay of the controlled variable is
figure, the stator current I , and the input power P,, are not measurable, and 2) the noise level that accompanies the
minimized almost simultaneously. This conclusion has been measurement of the controlled variable always sets a minimum
confirmed experimentally in several small motors and by on the value of the step Au.
computer simulation in several large motors. Therefore, in Fig. 9 shows the loss minimization process in the 1-hp
practice the stator current can be used as the controlled variable drive when the stator current is the controlled variable. In
in the loss minimization procedure. this case, the SC output attains a steady state. Fig. 10(a)
and (b) shows the loss minimization process in the 1-hp
VI. EXPERIMENTAL RESULTS
drive when (the input power (stator current) is the controlled
In general, the performance of the SC is the following: variable). The control variable are the input power and the
1) The value of the controlled variable y (input power, stator current, respectively. Comparing Fig. 1O(a) with 10(b)
stator current), which corresponds to the actual value we notice that the stator current minimization leads to even
U of the control variable (air-gap flux), is measured. more loss reduction, due to the absence of oscillations in the
KIOSKERIDIS AND MARGARIS: LOSS MINIMIZATION IN SCALAR-CONTROLLED INDUCTION MOTOR DRIVES 219

dynamic behavior of the drive, if an adaptive PI controller is

excluded, the nominal flux level must be provided whenever
a torque demand occurs. This implies that during transients
I l l
the motor operates at nominal flux level. In Fig. 11 the SC
response to an abrupt torque demand is shown. The air-gap
flux takes its nominal value immediately. After equilibrium is
reestablished, the air-gap flux reduces slowly to reach its new
optimal value. As soon as the torque demand is withdrawn,
the air-gap flux tends to its former optimal value.
Finally, Fig. 1%shows the experimental loss reduction and
the ratio of the optimal efficiency to the nominal one in the
1-hp drive for various speeds and load torques achieved with
the suggested stator current minimization.

VII. CONCLUSION
I l l In this paper, the loss minimization of IMD using SCs
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 has been examined. The weak points of this method, when
the input power or the dc link current are used as controlled
variables, are presented. The minimum of the input power
is not distinct, especially in large efficient motors, resulting
in its approximation not being successful. In addition, the
oscillation of the air-gap flux around its optimal value causes
undesirable torque disturbances. Then, the possibility of mini-
mizing the drive losses through the stator current minimization
is proved. The minimum of the stator current can be more
easily detected than the minimum of the input power. The
loss minimization, through the stator current, exceeds the loss
minimization through the input power, especially in highly
efficient motors.

REFERENCES
[I] R. Fei, E. F. Fuchs, and H. Huang, Comparison of two optimization
techniques as applied to three-phase induction motor design, IEEE
Trans. Energy Conversion, vol. 4, no. 4, pp. 651-660, Dec. 1989.
[2] J. C . Andreas, Energy-efJicient Electric Moiors. New York Marcel
Dekker, 1982.
[3] S. D. Umans and H. L. Hess, Modelling and analysis of the Wan-
lass three-phase induction motor configuration, ZEEE Trans. Power
Apparai., vol. PAS-102, no. 9, pp. 2912-2926, Sept. 1983.
[4] J. B. Medarametla, M. D. Cox, and Y . Baghzouz, Calculations and
measurements of the unity plus three phase induction motor, IEEE
Trans. Energy Conversion, vol. 7, no. 4, Dec. 1992.
[5] H. G. Kim, S. K. Sul, and M. H. Park, Optimal efficiency drive of a
current source inverter fed induction motor by flux control, IEEE Trans.
Ind. Applicai., vol. IA-20, no. 6, pp. 1453-1459, Nov./Dec. 1984.
[6] S. K. SUI and M. H. Park, A novel technique for optimal efficiency
control of a current-source inverter-fed induction motor, ZEEE Trans.
air-gap flux and to the minimization of the inverter losses. Power Electron., vol. 3, no. 2, pp. 192-199, Apr. 1988.
The inverter losses are proportional to the stator current [21]. [7] J. C. Moreira, T. A. Lipo, and V. Blasko, Simple efficiency maximizer
for an adjustable frequency induction motor drive, IEEE Trans. Znd.
Therefore, the stator current minimization leads to the losses Applicat., vol. 27, no. 5, pp. 940-946, Sept./Oct. 1991.
minimization of the inverter. [8] P. Famouri and I. J. Cathey, Loss minimization control of an induction
Thus, we reach the conclusion that the minimization of the motor drive, IEEE Trans. Ind. Applicai.,
_. vol. 27, no. 1, pp.
.~32-37,
Jan./Feb. 1991.
stator current has advantages over the minimization of the __
191 D. S . Kirschen. D. W. Novotny, and W. Suwanwisoot. Minimizing
input power. These advantages are not clear enough because induction motor losses by excitation control in variable frequenc;
drives, IEEE Trans. Ind. Applicai., vol. IA-20, no. 5, pp. 1244-1250,
the above conclusions are extracted from a low power drive. Sept./Oct. 1984.
As mentioned above, the advantages of the suggested method [lo] D. S. Kirschen, D. W. Novotny, and T. A. Lipo, On-line efficiency
will become more clear in the case of larger motors. optimization of a variable frequency induction motor drive, IEEE
Trans. Znd. Applicai., vol. IA-21, no. 4, pp. 610-615, May/June 1985.
Even though field weakening does not cause stability prob- [I I] D. S. Kirschm, Optimal efficiency control of induction machines,
lems, it alters the dynamics of the drive. To maintain the Ph.D. dissertation, Univ. Wisconsin-Madison, 1985.
220 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 11, NO. 2, MARCH 1996

1121 A Kusko and D Galler, Control means for minimization of losses in Iordanis Kioskeridis was born on January 29,
ac and dc motor dnves, IEEE Trans Ind Applicat , vol IA-19, no 4, 1965, in Thessalonilu, Greece He received the
pp 561-570, JulylAug 1983 Diploma degree in electncal engineenng and the
[13] S Mukherjee, R G Hoft, and J A McCormick, Digital measurement Ph D degree in industnal electronics from the Ans-
of the efficiency of inverter-induction machines, ZEEE Trans Ind totle University of Thessalonih, Greece, in 1990
Applicat , vol 26, no 5, pp 872-879, Sept /Oct. 1990 and 1994, respectively
[14] A Abhondanti, Method of flux control in induction motors dnven
by vanable frequency, vanable voltage supplies, in IEEE Press Ad-
justable Speed ac Drive Systems, B K Bose, Ed
1151 B W Williams and T C Green, Steady-state control of an induction
motor by estimation of stator flux magnitude, IEE Proc.-B, vol. 138,
no. 2, pp. 69-74, Mar. 1991.
R. J. Kerkman, Steady-state and transient analyses of an induction
machine with saturation of the magnetizing branch, IEEE Trans. Ind.
Applicat., vol. IA-21, no. 1, pp. 226-234, Jan./Feb. 1985.
P. L. Alger, G. Angst, and E. J. Davies, Stray-load losses in polyphase Nikos Margaris was bom on February 10, 1949,
induction machines, AIEE Trans. Power Apparat. Syst., vol. 78, pt. in Athens, Greece He received the Diploma degree
111-A, pp. 349-357, June 1959. in physics, the postgraduate degree in electronics,
P. K. Sen and H. A. Landa, Derating of induction motors due to and the Ph.D. degree in automatic control from
waveform distortion, IEEE Trans. Ind. Applicat., vol. 26, no. 6 , pp. the Aristotle University of Thessalonilu, Greece, in
1102-1107, Nov./Dec. 1990. 1972, 1975, and 1982, respectively.
P. J. Tsivitse and E. A. Klingshirn, Optimal voltage and frequency for Since 1977 he has been with the Electrical and
polyphase induction motors operating with variable frequency power Computer Engineering Department, Anstotle Uni-
supplies, IEEE Trans. Ind. Gen. Applicat., vol. IGA-7, no. 4, pp. versity of Thessalonilu, teaching graduate and post-
480487, July/Aug. 1971. graduate courses in electronics, automatic control,
P. G. Cummings, W. D. Bowers, and W. J. Martiny, Induction motor power electronics, and circuit theory. He is Asso-
efficiency test methods, IEEE Trans. Ind. Applicat., vol. IA-17, no. 3, ciate Professor and Vice President of the Electrical and Computer Engineering
pp. 253-265, May/June 1981. Department. From 1992 to 1994, he was Director of the Electronics and
S. C. Peak and J. L. Oldenkamp, A study of system losses in a Computer Division in the same department His current research interests
transistorized inverter-induction motor drive system, IEEE Trans. Ind. include the loss rmnimization in variable and constant speed drives and the
Applicat., vol. IA-21, no. 1, pp. 248-257, Jan./Feb. 1985. study of nonlinear oscillations