You are on page 1of 6

LOSS MINIMIZATION CONTROL OF

AN INDUCTION MOTOR DRIVE

P. Famouri J. J. Cathey, Senior Member


Department of Electrical Engineering
University of Kentucky
Lexington, KY

Abstract - This paper presents a practical method for operation


of an inverter-fed induction motor drive at the point of maximum
efficiency while maintaining any particular torque-speed load point.
Design and operation is described of an economical,
microprocessor-based adaptive controller used to modify a com-
mercially available six-step inverter for minimum loss operation of
an induction motor. Experimental results are presented in graphi-
cal form giving comparison of overall drive system efficiencies
between loss minimization control and constant volts/hertz control
operation for a standard-efficiencyinduction motor of 10 HP rating.
The greatest potential for energy savings is found to exist for the
case of loads with nonlinear torque-speed characteristics, such as a
fan or pump load.
P e r c e n l c ~ c Speed

LIST OF PRINCIPAL SYMBOLS


Fig. 1 Torque and efficiency characteristics
frequency of primary currents
frequency of secondary currents The open-loop inverter-fed drives offer the lowest initial cost,
primary current and consequently, dominate the industrial and commercial applica-
secondary current (reflected) tion population. At any particular control (frequency) setting, the
shaft speed open-loop inverter presents an available torque-speed curve; the
speed at maximum efficiency curves labelled Tel and T,, of Fig. 1 exemplify control setting for
number of poles two frequencies f, and f2. The point of operation is determined by
total average power across airgap the intersection of the load torque-speed characteristic with the
rotational losses available motor torque-speed curve. The typical efficiency-speed
primary resistance curves of Fig. 1 are used to illustrate that at some point A the
secondary resistance (reflected) efficiency may have a maximum value (qA) while, at some other
core loss resistor (parallel) control setting where the simultaneous satisfaction of the motor
core loss resistor (series) and load torque curves is given by point B, it is observed that the
slip efficiency (qs) is a lower value than when operating at point A. It
total developed torque is seen that operation at maximum efficiency only occurs when the
load torque load and motor torque-speed characteristics intersect at the speed
corresponding to the inflection point of the efficiency-speed curve
primary leakage reactance
(Nqmm). Operation at maximum efficiency frequently takes place at
secondary leakage reactance (reflected)
near rated load torque; other load conditions result in reduced
magnetizing reactance (parallel)
efficiency operation.
magnetizing reactance (series)
Temperature variation also has a significant effect on motor
efficiency
efficiency. As the temperature of the rotor increases, R, increases
shaft speed
in value, the slope of the torque-speed curve decreases, and the
synchronous speed
operating point shifts such that it results in a lower value of
efficiency.
The closed-loop inverter-fed drives can take advantage of the
INTRODUCTION speed feedback signal for use in control of the inverter frequency
(0 to assure a constant rotor frequency (f,) according to the algo-
Present commercially available inverter-fed adjustable speed rithm
induction motor drives are of two classifications:
f = f,+&Nm
1. Open-loop. Frequency is set to give approximately a
desired speed while a constant voltshertz ratio is main- The rotor frequency chosen by setup procedure is normally
tained. the value at rated or base speed, voltage, and frequency. If this
2. Closed-loop. The speed loop is closed and speed feedback value of rated rotor frequency results in near maximum efficiency,
is also used to calculate a stator frequency command assur- this closed-loop, constant rotor frequency (CRF) inverter can offer
ing that rotor frequency is held to a constant value (usually a small efficiency improvement over the popular open-loop con-
the value at rated or base operation) subject to the con- stant voltshertz (V/Hz) inverter. Although potentially offering an
straint that a volts/henz ratio is not violated. efficiency improvement, the rotor frequency that yields maximum
89CH2792-0/89/0000-0226$01 .oO 0 1989 IEEE

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.
efficiency operation depends upon parameters of the induction
motor equivalent circuit which can vary with temperature or satura-
tion. Further, the value of rotor frequency that yields maximum
efficiency performance is a function of motor speed. Thus, with no
provision to adjust rotor frequency as either motor parameters or
speed change, operation usually occurs at less than maximum
efficiency.
During the past few years, several papers have appeared in the
literature offering control regimes attempting to maintain operation
of adjustable speed drives at the maximum efficiency [l-51. The
potential for energy savings is significant as documented by Refer-
ences 4 and 5. However, all of these proposed control schemes Fig. 2 Per phase equivalent circuit
require an accurate knowledge of the motor parameters for imple-
mentation. Maximum efficiency slip is more sensitive to the value
of rotor resistance .than any other motor parameter, yet rotor resis- The input impedance of the equivalent circuit of Fig. 2 can be
tance changes with load profile, and dynamic measurement of its written as
value is quite difficult, if possible.
Kusko and Galler [2] have offered mention of a testing con- Z,=[ R ~ + R ~ , ( f . s ) l + j [ x ~ + x , ( f . s ) l (7)
troller concept that can be utilized to vary rotor slip frequency in a where
manner to minimize input power while maintaining a particular
torque-speed load point. Kirschen, Novotny, and Lip0 [6] describe
R, (f, s) + s) = (Rms+ j
implementation of a testing or perturbing controller for minimum
loss operation of a field-oriented controlled induction motor. Using (8), the primary copper losses (PI) are given by
Significant energy savings are found for operation at light torque
loads for all speeds. For torque loads near nameplate rated value,
(9)
energy savings at low speed operation are reported.
In a recent thesis, Mehanna [7] presents an adaptive controller
Through use of current division to determine I,, the total average
modification of a constant rotor frequency inverter to operate at
optimal efficiency. The concept is similar to that used by Kirschen value of power crossing the induction motpr airgap (P,) is
et a1 [6] except that the extra complexity of the field-oriented con- expressed as follows:
trol is not required. Further, since the adaptive controller can be R, 3e+X:)I:R,/s
p =31~-=
realized by modification of a constant voltshertz inverter, it offers , s A
the potential of retrofit to inverters already in field service. The
initial results of a computer simulation of this model are promising -
3R
,(: + X2', ) V: R2/ s (10)
in that the controller has been demonstrated capable of seeking out A [(RI +R2, 1' + (XI + X, ),I
the maximum efficiency slip point. Obviously, the advantage of
where
this (perturbing rotor frequency or P E ) scheme is that no
knowledge of motor parameters is necessary; thus, parameter varia-
tions with temperature and saturation are of no consequence.
Implementation and test of this concept are the subject of this
A=
I :I2
R,,+L +(X,,,s+X2)2

Forming I,, the current through the equivalent series excitation


paper. branch of the equivalent circuit, by current division allows expres-
sion of the core losses (P,) as
THEORETICAL INVESTIGATION

A. Efficiency Determination

It is observed from Fig. 1 that for operation at any particular


frequency a pronounced inflection of the efficiency-speed curve of
an induction motor exists. The Steinmetz equivalent circuit for the
induction motor shown by Fig. 2 can be modified to account for
frequency effects on copper and core losses as follows [8]:
As a reasonable approximation, the rotational power losses (PR)are
dependent on the square of speed, or
PR = kN 0: (13)
Since
where RIO and R,, are the dc values of R I and R,, respectively. = (1 -s) 0, = (1 - s) 2 04xf (14)
Constants c1. q,c,, a,and p are determined from equivalent circuit P P
parameters evaluated at two or more frequencies. Further, R,, (and (13) can be expressed for a 4-pole motor as
G,) are the series equivalent elements for the parallel-connected
excitation branch of Fig. 2 given by
PR = kN [ X (1 -S) f2 (15)
n "2 Recalling that the power converted from electrical to mechanical
form is given by (l-s)Pg, the per unit efficiency of the induction
motor is formed as a function of s, f, and V,:
(1 - s) P, - PR
11= fl(s,f'v1)=
P, + PI + P,
221

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.
where PI, P,, P,, and PR are given by (9), (IO), (12) and (13), frequency to restore maximum efficiency operation. Thus,
respectively. However, VI depends upon the load torque T,. efficiency improvement is expected when using PRF control for the
Specifically, the developed electromechanical torque (T,) must case of reduced value constant load torque.
equal the load torque, or In order to substantiate the above predictions, two computer
programs were built up to model the motor performance for both
PRF and constant VMz operation. For both programs, the motor
was modelled by the approximate frequency-dependent parameters
Substitution of (lo), (13) and (14) into (17) and rearrangement documented in the Appendix and selected by guidance from Refer-
ence 8. The motor load toque was characterized as dependent on

[
leads to
speed squared with value equal to the rated value at 100% motor
I = s x f A [(RI+ Rz,,,1' + (XI + XZ,,,)'I[TL+k, (1 - s) x f I
3' Rz eS+
x&)
,]"20 8)
speed. The first program predicted PRF performance using the
limited range search approach described in the previous section.
The second program calculated constant V/Hz performance
In general, TL= TL(q,,) is speed dependent; thus, substitution of with V , / f maintained at the value determined by ratio of nameplate
(18) into (16) yields efficiency as a function of s, f, and N,: rated values. Implementation was carried out by substituting
VI = f ( V l / f ) m d and f = o , / ( l - s ) x into (17) and incrementing s
(19) from a small value until the equality of (17) was established.
Results of the theoretical study are. presented by Figs. 3-5.
where Fig. 3 shows that for speed less than approximately 65% with the
squared-law torque load, significant efficiency improvement is
f3 = X S (1 -S> f R, (&,+x~,)[TL(O,)+kNK(l -s) fl (20)
shown by PRF control over constant V/Hz control. Figs. 4 and 5
€4 = kN S R, [(I -S)f12 + es
(Cs ) (21) illustrate that for the lower speed operation the PRF control acts to

i
f, = [ T ~ ( O , ) + k ~ x ( 1 - - S ) f ] sR,&,+X&) +
lower voltage and increase frequency over constant VMz control at
each operating point to result in a reduction in core losses and,
subsequently, to reduce the total losses.
R,[(S~,+R~)~+S~(X~+X,)~I
+R,,,,,[R;+s2X;1

In principle, each occurrence of s in (19) could be replaced by


1 (22)
EXPERIMENTAL INVESTIGATION

s = ( x f - q , , ) / x f ('&pole case) to find q = f6(f,a,,,). Then, for any A. Controller Realization


particular value of a,,,, the value of f = f,,,,, that satisfies h / a f = 0
is the value of frequency to yield maximum efficiency operation. In order to verify the theoretical predictions, an adaptive con-
However, the resulting expressions do not allow closed-form solu- troller is built up by modification of a commercially available, six-
tion for f = f,,,,,. Since a numerical solution for f,,,,, is inevitable, step, 15 kVA, constant voltshertz inverter. An optional control
an alternate limited range search approach is utilized in that
s = ( x f - q , , ) / x f is substituted into (19), and with a,,,held con-
stant, 11 is evaluated as f is varied by small increments until
f = f,,,., is determined.

B. Theoretical Predictions

The plots of Fig. 1 are calculated using manufacturer supplied


typical rated frequency and voltage values of equivalent circuit
constants (See Appendix) for the standard-efficiency induction
motor under test. Study of the curves of Fig. 1 allows important
observations concerning the efficiency comparison of PRF and
VMz control. First, consider the two load toque characteristics
TLl (= k, N): and TLz(= k, ) that both result in maximum efficiency
operation at point A. If the motor frequency is reduced so that the
motor is characterized by curve 2, the constant torque load (TLZ) Fig. 3 Efficiency-speed characteristics
results in operation at point C where efficiency (q,) remains near
maximum possible value. However, if the motor load torque is 1 cc

described by T,,, operation at point B results with associated -


efficiency qe, which is less than the maximum possible value. If U
1 80- ~ -

the control regime is constant V/Hz, operation must remain with C


0
cr
3
efficiency qe. On the other hand, if PRF control is used, the per- f 60-
L
turbing controller will increase frequency and reduce voltage so
a,
that speed is unchanged while operation occurs at the maximum m
0 40-
efficiency point of a new curve slightly to the right of curve 2. C
W
U
Hence, it is predicted that PRF control offers efficiency improve-
2 20-
ment when the driven mechanical load presents a nonlinear
torque-speed characteristic.
0 1 , , , , I , , , ,

Secondly, if the constant load torque curve Tu of Fig. 1 were 0 IO 20 30 40 50 60 70 80 90 100

reduced in magnitude, points A and C would move to the right Percentage Speed

along Teland Td,respectively, to yield operation at less than max-


imum efficiency. PRF control can allow adjustment of voltage and Fig. 4 Frequency-speed characteristics

228

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.
(a,) are compared by the speed summer to yield the speed error
loo 1 / ” signal (vE). vE is processed by the PI controller to give the error
- PRF Control
signal vk used to set the value of bias signal Avds that adjusts the
V/Hz Control
volts/hertz ratio, thus changing the dc link voltage. If a, is greater
~

than Q,,, the dc link voltage is reduced. Conversely, if R, exceeds


CO, in value, the dc link voltage is increased, and thus, speed is
regulated.
; 40 The inverter frequency command (a)is formed by the fre-
0
W
quency summer as a linear combination of the speed command
Q 20 (Cl,,,) and the set rotor frequency command (Rr). If the perturbing
subcontroller is active, rotor frequency command ( C l , ) is altered by
0 7 - 1 I I
-- r .
-I ~ 1 ~
signal 6w,.
0 10 20 30 40 50 60 70 50 9c ‘Li

Percentage Speed
The perturbing subcontroller senses the inverter input power
(P) and then either increases or decreases signal So,. If a change in
Fig. 5 Voltage-speed characteristics 60, reduces the value of P, the perturbing subcontroller then
changes 60, another increment in the previous direction. If inverter
card offered by the inverter manufacturer is installed in the power P increases due to a change in &or, then the subcontroller
inverter. The card provides the normal inverter adjustments for begins to alter 60,in the opposite direction. Hence, the inverter
setting acceleration, deceleration, and volts/hertz ratio. However, output frequency is adjusted until a point at which minimum input
two features of the card are of particular interest in the design of power to maintain the mechanical load point is reached.
the controller that is the focus of the described research. First, the Although the controller concept has been presented by analog
slip signal input can be bidirectional signal to dynamically alter the block diagram, the actual realization is better suited to a digital
volts/hertz control ratio, allowing control of the dc link voltage of implementation. An 8-bit microprocessor-based implementation of
the inverter independent of inverter output frequency. Second, the the controller is used. Introduction of 12-bit A/D and D/A convert-
card accepts a square-wave input reference signal to directly ers are necessary to obtain the performance accuracy required by
control the inverter output frequency. In-house circuit designs are the problem. An abbreviated flow chart of the adaptive controller
implemented to allow operation in either of two modes: (1) con- program is displayed by Fig. 7 to give an understanding of the con-
stant rotor frequency control, in which external command of rotor troller logic.
frequency is set and the controller acts to adjust dc link voltage
magnitude to maintain the set speed; (2) perturbing control, in B. Test Descriptions
which rotor frequency command is perturbed in a manner to search
for minimum input power while dc link voltage is adjusted to Losses of components under test are measured by the princi-
maintain the set speed. ple of air calorimetry [9] to circumvent the error that can be intro-
A block diagram of the adaptive perturbing controller is duced by electrical instrument with inadequate frequency response.
presented by Fig. 6. Speed command (a,) and actual motor speed
3 - PHASE
I N PUT

I
TRANS-
DUCER

1
PERTURBING
SUB-
CONTROLLER
I
a,
ROTOR FREP.

,a,T,-r
COMM.
.--

FW-1
FREO
SUMMER

P I CONTROLLER
1“avdc

’ v;
,
FREP C O M M .

NONLINEAR
FEED B A C K
8
SPEED S U M M E R I
Fig. 6 Block diagram of adaptive controller
229

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.
Fig. 9 Rotor frequency (upper trace) and
input power (lower trace)

traces of input power and rotor frequency command for the case of
50% speed- 114 torque, illustrating the dynamic nature of the PRF
control. The rotor frequency has decreased by 23 steps for .05
Hdstep from 1.33 Hz to .19 Hz; meanwhile input power has
decreased by 27.5% for standard-efficiency induction motor.
The PRF control is successful in finding an operating point to
improve motor efficiency and, consequently, the overall system
efficiency over the V/Hz control. Figs. 10 and 11 show percentage
efficiency vs. speed for 114 and 414 torque, respectively, comparing
overall system efficiency for V/Hz control and PRF control. As
torque increases, the efficiency improvement of PRF over V/Hz
decreases to an insignificant value, thus substantiating the earlier
prediction that significant efficiency improvement only for reduced
torque loads is to be expected.
The specific impact of the previous conclusion that the non-
linear torque load type is best suited for PRF control can be seen
quantitatively from the plot of the square-law torque load data in
Fig. 12. Then the torque demand at 75% and 50% speed would be
56.25% (9/16) and 25% (114) of rated torque, respectively. The
Fig. 7 Flow chart of controller program

The calorimetric loss determination technique was used to measllre


converter (cascaded ac-dc converter and inverter) losses (Pc,) and
motor losses (P,J according to the test setup indicated by Fig. 8.
Two power conditioner-motor arrangements were tested: (1)
Standard-efficiency motor fed by six-step inverter using constant
voltshenz control; (2) Standard-efficiency motor fed by six-step
inverter using perturbing rotor frequency control. Tests were con-
ducted for motor speed (%) of 25%, 50%, 75%, and 100% of E ]
; 20
1750 rpm. At each speed point, mechanical load was adjusted to
require a shaft torque of 114, 214, 314, and 4/4 of 30 ft-lbs, all
i
under thermal steady-state condition. In addition, data were
recorded for both power conditioner-motor arrangements at 75%
speed with 9/16 of 30 ft-lbs. Thus, data were recorded for a total
of seventeen torque-speed points for each of the two power Fig. 10 Efficiency-speed data for 1/4 torque
conditioner-motor arrangements.

RESULTS OF TESTS

Data were recorded according to the description of test set


forth in previous section. Fig. 9 displays photographs of recorded

I -!

Ib 20 io 40 510 $0 A Bo 90 ,bo
Percentage Speed

Fig. 8 Block diagram of test arrangement Fig. 11 Efficiency-speeddata for 414 torque
230

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.
loo 1 Frequency Dependent Model:
I t perturbing c o n t r o i
Rio = 0.2151 R cI = 0.8868 x lo4
R20 = 0.1231 R c2 = 1.236x
C, = 2.2133~10-~ CL = 1.75
p = 1.45

ACKNOWLEDGEMENT

This work was supported by contract no. TCRD01-87-U011


, ! U I , , I I - - I
from the Tennessee Center for Research and Development through
0 10 20 30 10 50 60 70 80 90 100
the Power Electronics Applications Center, Knoxville, TN.
Percentage Speed

Fig. 12 Efficiency-speed data for square-law torque

experimentally obtained efficiency at these points are 78.6% and REFERENCES


65.27%, respectively, using V/Hz control while the efficiency
would be improved by 1.4% and 7.46%, respectively, by use of [l] D. Galler, "Energy Efficient Control of AC Induction Motor
PRF control. Driven Vehicles", Proceedings of 1980 IEEE-IAS Annual
Meeting, pp. 301-308.
CONCLUSIONS
[2] A. Kusko and D. Galler, "Control Means for Minimization of
Losses in AC and DC Motor Drives", IEEE Trans., vol. IA-
Comparison of test results leads to conclusions similar to
19, no. 4, July-Aug. 1983, pp. 561-570.
those of Reference 6 that reported work conducted with the more
sophisticated field-oriented control:
[3] M. H. Park and S. K. SUI, "Microprocessor-Based Optimal-
1. Significant efficiency improvement over constant volts/hertz Efficiency Drive of an Induction Motor";IEEE Trans., vol.
operation is offered by the loss minimization control at E-31, no. 1, Feb. 1984, pp. 69-73.
reduced torque loads for all speeds.
2. Small efficiency improvement is found for near rated torque [4] J.M.D. Murphy and V. B. Honsinger, "Efficiency Optimiza-
loads at low speed operation tion of Inverter-Fed Induction Motor Drives", Proceedings of
I982 IAS Annual Meeting, pp. 544-552.
It is concluded that the greatest efficiency improvement is
found for application of the loss minimization control when driving [ 5 ] D. S. Kirschen, D. W. Novotny, and W. Suwanwisoot,
loads with nonlinear torque-speed characteristics, such as fan or "Minimizing Induction Motor Losses by Excitation Control in
pump loads. Such loads are already the primary application of Variable Frequency Drives", IEEE Trans., vol. IA-20, no. 5,
constant volts/hertz inverter-fed induction motor drives. Further, Sept.-Oct. 1984, pp. 1244-1250.
since the described controller is suitable for economical retrofit of
existing constant volts/hertz inverter drives, the practical interest [6] D. S. Kirschen, D. W. Novotny, and T. A. Lipo, "Optimal
should be significant. Efficiency Control of an Induction Motor Drive", IEEE
Trans., vol. EC-2, no. 1, March 1987, pp. 70-76.

-
APPENDIX MACHINE PARAMETERS [7] C. Y. Mehanna, "Perturbing Controller for Minimum Loss
Operation of Induction Machines", M.S. Thesis, University of
Nameplate Data: Kentucky, August 1985.
10 hp, 2301460 V, 27113.5 A [8] A. Bellini, R. Miglio, U. Reggiani, and C. Tassoni, "Voltage-
1755 rpm, 60 Hz, Design B Frequency Law in Frequency-Controlled Induction Motor
1.15 SF Drives", Proceedings of I978 IEEE-IAS Annuat Meeting, pp.
690-693.
Equivalent Circuit Parameters (60 Hz):
RI = 0.2264 R R2 = 0.1256R [9] J. J. Cathey, M. Rabiee, and S. H. Jan, "Application of
XI = 0.5842 R X, = 0.7292 R Calorimetry in Efficiency Determination of Power Condi-
R,,,%= 0.8382 R X,, = 10.367 R tioned Drives", Electric Machines and Power Systems, vol.
14, no. 1, 1988, pp. 33-44.
230 V connection

23 1

Authorized licensed use limited to: Kyungsung University. Downloaded on March 13,2010 at 06:04:21 EST from IEEE Xplore. Restrictions apply.

You might also like