IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 47, NO.
1, FEBRUARY 2000
Loss Minimization in Surface Permanent-Magnet
Synchronous Motor Drives
Christos Mademlis, Jannis Xypteras, and Nikos Margaris
Abstract—The loss minimization in surface permanent—magnet
synchronous motor drives is investigated. Based on theoretical
analysis, a loss model controller is introduced to specify the
optimal air-gap flux that minimizes losses. Theoretical results are
verified experimentally. The proposed loss model controller is
simple and does not affect adversely the cost and complexity of the
drive. Implementation of the loss model controller does not require
knowledge of the loss model. The suggested loss minimization
method can be applied both in V/f- or current-controlled schemes.
Index Terms—Adjustable-speed drives, efficiency optimization,
loss minimization, permanent magnet, synchronous motor.
NOMENCLATURE
Stator resistance.
Magnetizing reactance.
Stator leakage reactance.
Supply frequency.
Stator voltage.
Stator current.
Magnetizing current.
Air-gap flux.
Excitation current.
Electromagnetic torque.
Total power losses.
Copper losses.
Iron losses.
Stray losses.
Mechanical losses.
Iron loss coefficient.
Stray loss coefficient.
Mechanical loss coefficient.
I. INTRODUCTION
E LECTRIC machines are employed in almost every indus-
trial and manufacturing process. The need to save energy
has drawn the industrial attention to the losses and efficiency of
motors. The main efforts for higher efficiency are focused on
improvement of materials and optimization of design strategies
[1]. However, efficiency can also be improved by intervening in
the operational principle of motors. Such methods can be im-
plemented on adjustable-speed drives fed through an inverter.
Several simple and effective control methods have been
proposed in order to minimize the losses of dc, induction,
Manuscript received July 16, 1998; revised May 20, 1999. Abstract published
on the Internet November 11, 1999.
The authors are with the Department of Electrical and Computer Engineering,
Aristotle University of Thessaloniki, GR-54006 Thessaloniki, Greece.
Publisher Item Identifier S 0278-0046(00)01323-X.
0278–0046/00$10.00 © 2000 IEEE
115
and synchronous motor drives [2]–[8]. All these methods are
based on the air-gap flux weakening and attempt to make the
air-gap flux an increasing function of the load torque. This can
be achieved by using mainly two control methods: one based
on search controllers (SC’s) and another based on loss model
controllers (LMC’s).
In recent years, permanent-magnet (PM) synchronous motors
have gained an increasing popularity in a variety of industrial
applications. The advantages of high torque density, high ef-
ficiency, and high power factor make PM motors competitive
over induction motors [9]. The availability of low-cost power
electronic devices and the improvement of PM characteristics
enable the use of PM motors in more demanding applications
[10].
Adjustable-speed PM motor drives are rapidly introduced in
the area of electric vehicles, computer applications, medical in-
struments, and motion control devices [11]. These drives are
often powered by a battery source. Due to restrictions in energy
supply, the improvement of motor efficiency is a most important
priority.
Attempts to minimize the loss on PM synchronous motor
drives are made in [12]–[16]. An effort to specify the loss min-
imization condition for surface and interior PM synchronous
motor drives is presented in [16]. However, the proposed con-
dition is complicated and its implementation is based on the
knowledge of the machine parameters. Consequently, the sug-
gested method is implemented by using a lookup table where a
number of costly and time-consuming measurements for each
motor are required. Such measurements are practically impos-
sible on motors already in operation and, therefore, the method
is not widely applicable.
In this paper, the loss minimization problem in surface PM
synchronous motor drives is investigated in detail. The proposed
method is based on the air-gap flux weakening and its imple-
mentation does not require knowledge of the machine param-
eters. The proposed method is compared with the conventional
“ = 0 control” method, where the direct axis component of the
stator current is always kept at zero and demagnetization of the
permanent magnet is prevented [17], [18].
A block diagram of optimal PM synchronous motor drive is
shown in Fig. 1. In this diagram, it is possible to embody one
of the two controllers that minimize the losses on the drive. The
SC measures the input power to the drive and adjusts the stator
voltage, while it searches the minimum input power. The LMC
measures the speed and armature current and, through the motor
loss model, it specifies the optimal value of stator voltage. Ac-
cording to relevant literature [7], [8], and [12], SC’s have sev-
eral disadvantages and, consequently, LMC’s offer a superior
116 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 47, NO. 1, FEBRUARY 2000

Fig. 1. Optimal surface PM synchronous motor.

Fig. 3. Approximate phase diagram of surface PM synchronous motor.

and

(3)

Furthermore, the - and -axis components of magnetizing cur-

rent are given, respectively, by
Fig. 2. Per-unit equivalent circuit of surface PM synchronous motor.
(4)

performance. For example, SC’s cannot successfully find the and

minimum of the input power function that is smooth and flat
around the minimum. An SC in a real system does not reach a (5)
steady state and causes oscillations in the air-gap flux that re- Finally, the electromagnetic torque of the motor is given by
sults in undesirable torque disturbances.
The contents of this paper are organized as follows. In Sec- (6)
tion II, the relationships between the PM synchronous motor
variables are given. In SectionIII, the motor loss model is pre-
sented. The loss minimization condition is derived in Section IV III. LOSS MODEL
and its implementation with SC and LMC in controlled The main losses on the PM synchronous motor are the fol-
scheme is described in Section V. The experimental results are lowing.
presented in Section VI. Finally, in Section VII, the implemen- 1) Copper Losses: These are due to the flow of load current
tation of the LMC in a current-controlled scheme is briefly pre- through the stator windings and are given by
sented.
(7)
II. RELATIONSHIPS BETWEEN PM SYNCHRONOUS MOTOR 2) Iron Losses: These are due to hysteresis and eddy cur-
VARIABLES rents and are given by the following empirical formula [9], [20]:
In Fig. 2, the per-phase equivalent circuit of the PM syn-
(8)
chronous motor is given in the p.u. system. In this circuit, the
effects of iron and stray losses are ignored. The phasor diagram where
in synchronously rotating - reference frame is illustrated in 3) Stray Losses: These arise on the copper and iron of the
Fig. 3. Since the resistance of the stator winding is very small motor and are given by [11], [19]
[9], the respective voltage drop could be neglected. The stator
current is resolved into the respective and -axis compo- (9)
nents. The excitation current is aligned with the axis. The
4) Mechanical Losses: These are due to friction and
and -axis components of stator voltage are given, respec-
windage losses and are proportional to the square of rotating
tively, by
speed [20]
(1) (10)

5) Harmonic Losses: These are additional losses due to

(2) nonsinusoidal stator voltage that supplies the synchronous
MADEMLIS et al.: LOSS MINIMIZATION IN SURFACE PM SYNCHRONOUS MOTOR DRIVES 117

Fig. 4. Power loss versus stator voltage in a 1.25-kW drive.

motor. The presence of harmonic currents increases the stator Since the electromagnetic torque is constant, it is deduced that
copper losses. The presence of harmonic voltages increases the
iron losses [9]. (17)
Since mechanical losses are independent from the electrical
variables, they are not controlled by flux weakening. Addition-
From (6), we conclude that (17) is satisfied when
ally, harmonic losses are not directly controlled by flux weak-
ening. However, these losses are indirectly controlled by the de-
(18)
creasing of the harmonic voltages, because of flux weakening.
The losses that can be minimized by flux weakening are
From (16) and (18) the following loss minimization condition
(11) is deduced:

where (19)
(12)
Equation (19) gives the optimal -axis component of stator
and current. This equation can be applied in current-controlled
schemes. It should be noted that is independent of load
(13) conditions; however, it depends on the frequency and loss
model parameters. A similar condition is presented in [16].
Finally, (10) can be expressed in – axes components as
However, stray losses are not included and the condition is
follows:
implemented using a lookup table.
(14) Loss minimization condition for -controlled schemes is
obtained by substituting (19) in (1). Combining (1), (2) and (3),
the optimal ratio is given by
IV. LOSS MINIMIZATION CONDITION
Generally, the synchronous motor is a double-fed machine
and its inputs are the stator voltage and excitation current
In the PM motor, the field excitation is provided from the
PM’s, thus, the excitation current is constant. From (14), the (20)
loss minimization condition at steady state ( and constant)
with respect to is given by The existence of the loss minimum is experimentally verified.
Fig. 4 provides the loss reduction versus stator voltage measured
in an 1.25-kW PM synchronous motor drive. Note that the loss
(15)
curves are smooth and flat around the minimum.
In Fig. 4, the points on the dotted line correspond to “ =
Because of (4) and (5), condition (15) is satisfied when 0 control.” In constant-speed operation and “ = 0 control,”
stator voltage increases as load torque increases, hence, power
(16) losses increase, as per [16].
118 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 47, NO. 1, FEBRUARY 2000

where

(23)

(24)

(25)

and

(26)
Fig. 5. Performance of optimal PM motor drive with SC (! = 1 p.u. and
T = 0:4 p.u.). The excitation current in the PM machine is constant, hence,
the gain in (24) is constant. is easily derived from the
open-circuit test when the machine is operated as a generator.
The measured value of the open-circuit stator voltage provides
the value of From (22), it is implied that the is com-
posed of two terms

(27)

and

(28)

The former depends on the load condition. The latter is propor-

Fig. 6. SC response to load torque variations (! = 1 p.u. and T = 0:4 p.u. tional to the excitation electromotive force and depends on the
→ 0.9 p.u. → 0.4 p.u.). speed. The square-root term at the right side of (28) is the am-
plitude of the frequency response of a first-order low-pass filter
V. IMPLEMENTATION OF THE LOSS MINIMIZATION CONDITION with transfer function
WITH -CONTROLLED SCHEME
The use of the SC is the obvious method for implementing
the loss minimization condition (20). As shown in Fig. 1, the (29)
SC measures the input power to the drive and adjusts the stator
voltage so that the input power is minimized. However, ex- The filter pole depends on the motor speed. However, as proved
periments prove that the performance of the optimal PM syn- in [3] and [8], the filter can be approximated by a first-order
chronous motor drive with the SC is not satisfactory. As shown low-pass filter with a constant pole provided that the cutoff fre-
in Fig. 5, the drive does not reach a steady state. The transient quency is the same.
response of SC in load torque disturbances is not satisfactory, Fig. 7(a) and (b) gives and for various speeds
as shown in Fig. 6. Generally, the SC approach has several dis- and load torques. Since and are approximately con-
advantages and such performance is expected from the relevant stant at low speed, the gains and affect LMC param-
literature [7], [8], and [12]. eters response at low speed. As the speed increases, is re-
Substituting (12) and (13) in (20), the optimal ratio duced. Therefore, at medium speed, the LMC paremeters are
which is a practical measure of the air-gap flux, is given by affected by Finally, reduction beyond a certain speed
is constrained by the filter corner frequency . Thus, af-
fects LMC parameters response at high speeds. Consequently,
the LMC parameters are practically decoupled with respect to
speed and they can be easily tuned.
In order to calculate the optimal ratio from (22), the
(21) values of the measurement of the stator current and motor speed
are required. Since the loss minimization occurs at steady state,
From (21), we have the equation of LMC the reference speed can be used instead of the actual speed.
Since the LMC parameters are decoupled from each other with
(22) respect to speed, they can be adjusted experimentally, as fol-
lows.
MADEMLIS et al.: LOSS MINIMIZATION IN SURFACE PM SYNCHRONOUS MOTOR DRIVES 119

TABLE I
1.25-kW MOTOR PARAMETERS (p.u.)

TABLE II
LMC PARAMETERS

(a)

(b)

Fig. 7. Variation of (a) K and K and (b) optimal ratio V =! for various
speeds and load torques.

1) A three-phase wattmeter is used for measuring the total

(a)
input power of the drive.
2) The motor rotates at low speed, about 15%–25% of its
nominal speed. Under this condition (22) is
approximated by

(30)

The stator voltage is adjusted so that the minimum

wattmeter indication is accomplished. Two measure-
ments for two different load torque give an equation
system with unknowns the gains and The solu-
tion of the equation system gives the values of and

3) The speed is increased up to 40%–55% of its nominal

(b)
value. Under this condition, (22) is reduced to
Fig. 8. Performance of optimal PM motor drive with air-gap flux LMC (! =
1 p.u. and T = 0:4 p.u.). (a) Ratio V
=! and input power. (b) Ratio V =!
(31) and stator current.

and the minimum indication on the wattmeter is now ac-
complished by adjusting parameter
4) The speed is increased to its maximum value and the min-
imum indication is achieved by adjusting parameter
5) Steps 2)–4) are repeated until the desired accuracy is
reached.
During Steps 2)–5), the load torque is maintained steady and
equal to 0.3–0.5 p.u. to provide a more accurate location of the
minimum. The LMC parameters depend on the magnetizing
reactance and their values are affected by saturation. In the
PM drives, the saturation level is not significantly modified
because the strong excitation field of the permanent magnets
120 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 47, NO. 1, FEBRUARY 2000

(a)
(a)

(b)

Fig. 9. Air-gap flux LMC response to load torque variations (! = 1 p.u. and
T = 0:4 p.u. → 0.9 p.u. → 0.4 p.u.). (b)

Fig. 10. Comparison of optimal and “I = 0 control” PM motor drive. (a)

constrains the values of the optimal air-gap flux between 0.7–1 Power saving. (b) Efficiency improvement.
p.u. [Fig. 7(b)].
The excitation current and the armature resistance vary
motor are given in Table I. The control unit of the drive performs
depending on the temperature. The and therefore the gain
either as a loss model controller or as a search controller. The
decreases as temperature increases [9], [17]. On the con-
LMC parameters given in Table II have been adjusted experi-
trary, increases as temperature increases. From (25), (26), and
mentally according to the rules of Section V.
(29), it is deduced that, as temperature increases, the gain of the
The performance of the suggested LMC is shown in Fig. 8.
LMC filter increases as well. According to (28), the increment
Although the optimal ratio is known a priori from (22),
of the filter gain is partially compensated by the decrement of
the stator voltage command decreases at a low rate. The strategy
and, consequently, the whole variation of is very narrow.
is based on the fact that an abrupt decrease in the air-gap flux
Furthermore, successful approximation of the loss minimum is
causes strong current in stator, loss increment, demagnetiza-
possible since the loss curves shown in Fig. 4 are smooth and
tion, and electromagnetic torque disturbances. Comparing Fig. 8
flat around the minimum. Therefore, the minimum is reached
to Fig. 5, we deduce that the optimal PM synchronous motor
for a wide range of stator voltage values and the influence of
with the LMC performs better than the optimal PM synchronous
the parameter variations on the LMC performance is very low.
motor with SC.
The conclusion is that, in practice, knowledge of the loss
Fig. 9(a) and (b) provides the LMC response to an abrupt
model is not required. The LMC minimizes not only the PM
torque demand. It can be seen that the LMC compensates
synchronous motor losses, but the whole drive losses as well.
the torque demand almost immediately. After equilibrium is
Due to the experimental measurement of the LMC parameters,
reestablished, air-gap flux reaches its new optimal value. As
the losses in all drive stages are implicitly included.
soon as the torque demand is withdrawn, air-gap flux tends
to its former optimal value. During the above procedure, the
speed remains constant (Fig. 9(b), lower waveform).
VI. EXPERIMENTAL RESULTS
Fig. 10(a) and (b) gives the power saving and the ratio of the
The effectiveness of the suggested LMC operation has been optimal efficiency of the suggested LMC method to that of the
experimentally tested in a 1.25-kW drive. The parameters of the “ = 0 control” method for various speeds and load torques.
MADEMLIS et al.: LOSS MINIMIZATION IN SURFACE PM SYNCHRONOUS MOTOR DRIVES
(a)
(b)
Fig. 11. Optimal values of (a) power factor and (b) load angle for various
speeds and load torques.
The efficiency improvement is remarkable at heavy load. The
above is consistent with the curves of Fig. 4.
Fig. 11(a) and (b) provides the variation of the optimal power
factor and load angle for various speeds and load torques, re-
spectively. It is noted that, the optimal power factor is increased
as the load torque increases and the optimal load angle is almost
independent from the speed.
VII. IMPLEMENTATION OF THE LOSS MINIMIZATION
CONDITION WITH CURRENT-CONTROLLED SCHEME
Substituting (12) and (13) in (19), the optimal current is
given by
(32)
From (32) and using (23)–(26), we get the equation of the LMC
(33)
where
(34)
121
Fig. 12. Variation of optimal current I versus speed.
The optimal current is independent of load conditions. To
calculate the optimal from (33), only the measurement of
motor speed is required. Fig. 12 shows the optimal versus
speed.
The LMC parameters can be easily adjusted, by performing
two measurements of the current for two different speed
values. The current is adjusted so that minimum power input
is accomplished. This two measurements give an equation
system with unknowns the parameters and The solution
of the equation system gives the values of and The
measurements are repeated for various couple of speeds, until
the desired accuracy is reached.
VIII. CONCLUSIONS
In this paper, the loss minimization problem in surface PM
synchronous motor drives at steady state has been investigated.
LMC’s for determining the optimal ratio in con-
trolled schemes and the optimal current in current controlled
schemes have been presented. Although the conception of the
suggested method is based on the motor loss model, its imple-
mentation does not require knowledge of the loss model. The
adjustment of LMC’s parameters is an easy experimental task
that can be performed on motors already in operation. The pro-
posed LMC’s do not affect significantly the dynamics of the
drive. The suggested loss minimization method can be used in
both open and closed speed loop drives.
REFERENCES
[1] J. C. Andreas, Energy-Efficient Electric Motors. New York: Marcel
Dekker, 1992.
[2] A. Kusko and D. Galler, “Control means for minimization of losses in
ac and dc motor drives,” IEEE Trans. Ind. Applicat., vol. IA-19, pp.
561–570, July/Aug. 1983.
[3] N. Margaris, T. Goutas, Z. Doulgeri, and A. Paschali, “Loss minimiza-
tion in dc drives,” IEEE Trans. Ind. Electron., vol. 38, pp. 328–336, Oct.
1991.
[4] D. S. Kirschen, D. W. Novotny, and W. Suwanwissot, “Minimizing in-
duction motor losses by excitation control in variable frequency drives,”
IEEE Trans. Ind. Applicat., vol. IA-20, pp. 1244–1250, Sept./Oct. 1984.
[5] P. Famouri and J. J. Cathey, “Loss minimization control of an induction
motor drive,” IEEE Trans. Ind. Applicat., vol. 27, pp. 32–37, Jan./Feb.
1991.
[6] “Method and device for determining the optimal excitation value which
minimizes the electromagnetic losses of electric machines,” U.S. Patent
5 075 612, Dec. 24, 1991.
122 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 47, NO. 1, FEBRUARY 2000
[7] I. Kioskeridis and N. Margaris, “Loss minimization in induction motor
adjustable-speed drives,” IEEE Trans. Ind. Electron., vol. 43, pp.
226–231, Feb. 1996.
[8] C. Mademlis, J. Xypteras, and N. Margaris, “Loss minimization in
wound-field cylindrical rotor synchronous motor drives,” IEEE Trans.
Power Electron., vol. 13, pp. 288–296, Mar. 1998.
[9] J. K. Gieras and M. Wing, Permanent Magnet Motor Technology: Design
and Applications. New York: Marcel Dekker, 1997.
[10] P. Vas and W. Drury, “Future of electrical machines and drives,” in Proc.
ICEM, Vigo, Spain, 1996, pp. 491–496.
[11] S. A. Nasar, Electric Machines and Power Systems. New York: Mc-
Graw-Hill, 1995.
[12] R. S. Colby and D. W. Novotny, “Efficient operation of surface-mounted
PM synchronous motors,” IEEE Trans. Ind. Applicat., vol. IA-23, pp.
1048–1054, Nov./Dec. 1987.
[13] , “Efficiency-optimizing permanent-magnet synchronous motor
drive,” IEEE Trans. Ind. Applicat., vol. IA-24, pp. 462–469, May/June
1988.
[14] B. K. Bose, “A high-performance inverter-fed drive system of an interior
permanent magnet synchronous machine,” IEEE Trans. Ind. Applicat.,
vol. 24, pp. 987–997, Nov./Dec. 1988.
[15] Y. Nakamura, T. Kudo, F. Ishibashi, and S. Hibino, “High-efficiency
drive due to power factor control of a permanent magnet synchronous
motor,” IEEE Trans. Power Electron., vol. 10, pp. 247–253, Mar. 1995.
[16] S. Morimoto, Y. Tong, Y. Takeda, and T. Hirasa, “Loss minimization
control of permanent magnet synchronous motor drives,” IEEE Trans.
Ind. Electron., vol. 41, pp. 511–517, Oct. 1994.
[17] J. R. Hendershot, Jr. and T. J. E. Miller, Design of Brushless Permanent-
Magnet Motors. Oxford, U.K.: Magna Physics and Clarendon, 1994.
[18] S. Morimoto, Y. Takeda, and T. Hirasa, “Current phase control methods
for permanent magnet synchronous motors,” IEEE Trans. Power Elec-
tron., vol. 5, pp. 133–138, Apr. 1990.
[19] I. J. Nagrath and D. P. Kothari, Electric Machines. New Delhi, India:
Tata McGraw-Hill, 1985.
[20] M. Kostenko and L. Piotrovsky, Electrical Machines. Moscow,
U.S.S.R.: Mir, 1974, ch. 7.
Christos Mademlis was born in Arnea Chalkidikis,
Greece, in 1964. He received the Diploma degree in
electrical engineering and the Ph.D degree in elec-
trical machines from Aristotle University of Thessa-
loniki, Thessaloniki, Greece, in 1987 and 1997, re-
spectively.
Since 1990, he has been with the Electrical
Machines Laboratory, Electrical and Computer
Engineering Department, Aristotle University of
Thessaloniki, as a Research Assistant.
Jannis Xypteras was born in Plomari Lesvos,
Greece, in 1937. He received the Diploma degree
in electrical engineering from the Fakultät für Elec-
trotechik, RWTH Aachen, Aachen, Germany, and
the Ph.D. in eddy currents from Aristotle University
of Thessaloniki, Thessaloniki, Greece, in 1965 and
1978, respectively.
From 1965 to 1980, he was with German and
Greek factories of electrical machines and, for
several years, he worked as the Chief of development
departments. From 1984 until his death in 1998,
he was with the Electrical and Computer Engineering Department, Aristotle
University of Thessaloniki, as an Associate Professor and Director of the
Electrical Machines Laboratory. From 1990 to 1995, he was also the Director
of the Electrical Energy Division.
Nikos Margaris was born in Athens, Greece, in
1949. He received the Diploma in physics, the Post-
graduate degree in electronics, and the Ph.D. degree
in automatic control from Aristotle University of
Thessaloniki, Thessaloniki, Greece, in 1972, 1975,
and 1982, respectively.
Since 1977, he has been with the Electrical
and Computer Engineering Department, Aristotle
University of Thessaloniki, teaching graduate and
postgraduate courses in electronics, automatic
control, power electronics, and circuit theory. From
1992 to 1994, he was the Director of the Electronics and Computer Division
and, from 1993 to 1995, the Vice President of the Electrical and Computer En-
gineering Department. His current research interests include loss minimization
in variable- and constant- speed drives and the study of nonlinear oscillations.