Sunteți pe pagina 1din 5

2167

Vector Control of Trapezoidal Back-EMF PM


Machines Using Pseudo-Park Transformation
A. Lidozzi*, L. Solero*, F. Crescimbini
*
and R. Burgos
**
*
University ROMA TRE
Department of Mechanical and Industrial Engineering
Via della Vasca Navale 79, 00146 ROMA (Italy)
Phone: +39 06 55173522, e-mail: lidozzi@ing.uniroma3.it
**
Center for Power Electronics Systems (CPES)
The Bradley Department of Electrical and Computer Engineering
Virginia Polytechnic Institute and State University
Blacksburg, VA 24061 USA
AbstractThis paper deals with an extension of Parks
transformation to model permanent magnet (PM) non-
sinusoidal machines by means of the so-called pseudo-dq
axes reference frame. Speed and torque vector control can
then be implemented resulting in a motor drive
performance close to that of sinusoidal PM machines drives.
Similarly, the proposed control algorithm is derived by
decomposing the machine current into two components, one
being linked to the torque and the other one to its flux. The
paper discusses the benefits resulting from the use of vector
control in a brushless dc motor drive. Both simulation and
experimental results are used to validate the proposed
motor drive model and the related control strategy.
I. INTRODUCTION
Permanent magnet (PM) motor drives are often
divided into two main categories, namely brushless ac
motor and brushless dc motor drives. Brushless ac
motors, which have sinusoidal back electromotive force
(back-EMF), can take advantage of the well known
vector control scheme, where the electric machine can be
easily modeled by using Parks transformation. On the
other hand, brushless dc motor drives use a PM machine
with non-sinusoidal back-EMF; hence they are normally
controlled by using an abc-frame or scalar control
approach, as well as a two-phase dc current injection.
Within the class of non-sinusoidal back-EMF PM
machines, trapezoidal back-EMF motors are widely used
in automotive, aerospace, medical and industrial
applications. These machines are less expensive, much
easier to build, and posses a high torque density, facts
have made them suitable candidates for a number of high
speed applications. Compared to brushless ac motor
drives, the major drawbacks exhibited by PM motor
drives using a non-sinusoidal back-EMF machine are the
higher torque ripple and the relatively slow dynamic
response displayed when torque and speed are regulated
by means of a conventional control approach.
To overcome the above drawbacks, Parks
transformation can be readily extended to the modeling of
non-sinusoidal back-EMF machines on so-called pseudo-
dq axes. On this newly defined frame, speed and torque
vector control can be implemented for this type of PM
machines in a similar manner as the dq-axes vector
control of sinusoidal back-EMF PM motor drives.
Specifically, this paper discusses the implementation of a
linear vector control strategy for brushless dc motor
drives by using both simulation results and experiments
carried out with a 35 kW PM machine based DSP-
controlled motor drive. To this goal, a comprehensive
model for a PM machine with non-sinusoidal back-EMF
is described as basis of the proposed control algorithm.
Some of the results obtained from both simulation and
experiments are finally discussed in order to show the
advantages of the Pseudo-Park approach over the
classical two-phase dc current injection.
II. COMPARISON BETWEEN PMSM AND BLDC BASED
DRIVES
The main difference between sinusoidal back-EMF
PM synchronous motors (PMSM) and brushless dc
(BLDC) motors is found in the shape of the machine
back-EMF waveform. This usually forces the motor drive
designer to adopt two main different control strategies,
namely the dq-axes vector control for the PMSM drives,
or the dc current injection for the BLDC drives. However,
it is widely recognized that the vector control approach is
the most effective and powerful dynamic control strategy,
as it allows the suitable regulation of the machine torque
and flux, as well as its speed and position. In this way the
drive system can work in the four torque-speed quadrants
in a very simple and robust manner. Hence, the use of a
vector control approach also in the case of BLDC motor
drives would be highly desirable, as it can be
demonstrated that superior motor drive and overall
performance could be achieved in comparison with
PMSM motor drives.
Trapezoidal back-EMF machines can be divided into
two categories with respect to the shape of the machine
back-EMF waveform, which can be characterized by the
angular intervalexpressed by electrical degreesover
which the back-EMF has an almost constant value.
According to that, in some PM machines the back-EMF
waveform is approximately flat over 120 electrical
degrees (in the following these machines are referred to
978-1-4244-1668-4/08/$25.00 2008 IEEE
2168
as 30-120-30 back-EMF waveform machines), whereas in
other types of PM machines the back-EMF waveform has
constant value over a 60 degree angular interval (in the
following these machines are referred to as 60-60-60
back-EMF waveform machines). By assuming idealized
waveforms of the current commanded in the machine
phases, Fig. 1 shows the injected currents when either the
classical 2-phase dc current control or the pseudo-Park
based vector control is used for a 30-120-30 back-EMF
waveform PM machine. As shown, whenever the rated
machine RMS current is used in both cases (i.e., a same
I
2
R power loss would be allowed in the machine), in the
case of current waveforms resulting from a nonlinear
Park transformation, the peak value of current waveform
could be reduced by 7.42 %. Such an advantage becomes
more evident whenever the vector control is applied to a
60-60-60 back-EMF waveform PM machine, as shown in
Fig. 2. In this case the peak value of the inverter output
current would be reduced by 22.54 %.
If the above comparison is performed with respect to
a PMSM motor drive, more benefits arise whenever the
pseudo-Park control is used in a BLDC motor drive. In
PMSM machines, the ratio between peak and rms current
values is the well know
2
factor. If the rms value of the
commanded current is still assumed to be the same for
both machines, the use of a 30-120-30 back-EMF
waveform machine in a BLDC motor drive with pseudo-
Park control would allow a 60.33 % reduction of the
inverter peak current compared to a PMSM motor drive.
Figure 1. Waveform of motor drive electrical quantities (top to
bottom): 30-120-30 machine back-EMF and injected currents for the
case of 2-phase dc current injection and 3-phase vector control,
respectively
Figure 2. Waveform of motor drive electrical quantities (top to
bottom): 60-60-60 machine back-EMF and injected currents for the case
of 2-phase dc current injection and 3-phase vector control, respectively
On the other hand, a comparison can also be established
by assuming the same power rating of the three-phase
inverter feeding the PM motor. As shown in [8], if the
currents are assumed to have same peak values, then the
ratio between the inverter output power when a
conventional BLDC motor drive, i.e., a motor drive using
both a 30-120-30 back-EMF waveform machine and dc 2-
phase current control, and a PMSM motor drive yields,
BLDC 2ph 2 2 BLDC 2ph
pk pk 3 3
PMSM PMSM
pk pk
BLDC 2ph BLDC 2ph
pk pk pk
PMSM
PMSM
pk
pk pk
3 E I
P
P E I
3
2 2
2E I I
4
1.33
3 3 I
E I
2


= =
= = ~
(1)
which means that for a given size of the motor drive
inverter a BLDC machine having 33 % higher power than
a PMSM motor could be used. On the other hand,
whenever a comparison is performed between a vector
controlled BLDC motor drive and the other conventional
PM machine motor drive arrangements it yields,
BLDC pPark
BLDC pPark
pk pk
PMSM PMSM
pk pk
2 BLDC pPark BLDC pPark
pk pk pk
PMSM
PMSM
pk
pk pk
3a E a I
P
P E I
3
2 2
3a E I I
2.33
1.55
3 3 I
E I
2
2



= =
= ~ ~
(2)
BLDC pPark
BLDC pPark
pk pk
BLDC 2ph BLDC 2ph 2 2
pk pk 3 3
2 BLDC pPark BLDC pPark
pk pk pk
BLDC 2ph BLDC 2ph
pk pk pk
3a E a I
P
P 3 E I
3a E I 2.33 I
1.165
2 E I 2 I





= =

= = ~

(3)
with a 0.8819 ~ . These two equations show the benefits
in terms of inverter size when a BLDC motor using
vector control is used with respect to either a vector
controlled PMSM drive or a conventional BLDC motor
drive. For a given size of the converter the use of vector
control in a BLDC motor drive would allow 55 % higher
power than in a PMSM motor drive. Moreover, an
advantage is found also in the case of using pseudo-Park
based control together with a BLDC machine, as 16.5 %
higher power would be allowed.
III. TRAPEZOIDAL BACK-EMF MACHINE MODEL ON
DQ-AXES
The equations describing the machine operation in
the abc reference frame can be written in the following
form [2],
| |
a a a a
b s b b b
c c c c
a a a
b e b e b
e
c c c
V i i e
d
V R i L i e
dt
V i i e
e '
d
e '
d
e '
| | | | | | | |
| | | |
= + +
| | | |
| | | |
\ . \ . \ . \ .
u u | | | | | |
| | |
= e u = e u
| | |
0
| | |
u u
\ . \ . \ .
(4)
where, assuming x = a, b, c; e
x
, V
x
, are the back-EMF
and line-to-neutral phase voltage respectively; i
x
, is the
2169
machine current; u
x
is the flux induced in the stator phase
and u'
x
is its spatial derivative. In order to obtain the dq0-
axes equations of the machine, two transformations have
to be applied. Firstly, the classical transformation from
abc to o|0 can be obtained by using the stationary
reference frame transformation given in (5); secondly, the
conventional Park transformation has to be applied in
order to describe the model in the rotor reference frame.
1 1
2 2
3 3
2 2 abc 0
1 1 1
2 2 2
1
2
T 0
3
o|
(
(
( =
(
(

(5)
However, in case of non-sinusoidal back-EMF machines,
u
d
is not zero and u
0
may be or not equal to zero.
Therefore, in order to obtain a model of the trapezoidal
machine similar to the sinusoidal case, a new
transformation from o|0 to dq0 was proposed and
presented in [5]. With this transformation, u
d
is forced
to zero in a star-connected constant air-gap motor with
floating neutral. This implies that the zero sequence
current i
0
= 0, and that the electromagnetic torque
expression can be simplified to,
( )
( )
e 0 0
d d q q 0 0 q q
3
T P ' i ' i ' i
2
3 3
P ' i ' i ' i P ' i
2 2
o o | |
= u + u + u =
= u + u + u = u
(6)
This equation is the same as that of sinusoidal
machines; however u
q
is a periodic function dependent
on the rotor position 0
e
and the actual back-EMF shape.
The number of pole pairs is indicated as P. In order to
force u
d
to zero, a new matrix as shown in (7) is used,
where
e
corresponds to the displacement angle due to
the non-sinusoidal back-EMF shape [5]. Clearly,
e
is
zero in sinusoidal machines. The trigonometric functions
in (7) can then be evaluated as shown in (8).
d e e e e
e e e e q
0 0
dq0 0 e e q
0
' ' cos( ) sin( ) 0
' sin( ) cos( ) 0 '
0 0 1
' '
0
T ( ) '
'
o
|
o|
| | | | u u 0 + 0 + (
| |
(
u = 0 + 0 + u =
| |
(
| |
(
u u \ . \ .
| |
|
( = 0 + u
|
|
u
\ .
(7)
( ) ( )
e e
2 q 2
q
e e
q
'
sin( )
'
and ' ' '
'
cos( )
'
o
o |
|
u
0 + =

u = u + u

0 + =

(8)
In order to apply the Park-like transformation
described in (7), the controller needs to know the spatial
derivative of the machine fluxes in abc-coordinates at
every control cycle. However, by applying the pseudo-
Park transformation (7) and (8) to the normalized back-
EMF shown in Fig. 3, the dq-axes components can be
obtained in a very simple way. In effect, Fig. 5 shows that
the d-axis component is indeed forced to zero, and that
the 0-axis component is also zero for this machine. As
shown in Fig. 5 then, the flux is no more constant as in
PMSM machines, being directly related to both the back-
EMF harmonic contents shown in Fig. 4 and the electrical
rotor position.
Figure 3. Machine line-to-neutral back-EMF
Figure 4. Harmonic content for the machine back-EMF shown in Fig. 3
Figure 5. dq0 components for the machine back-EMF of Fig. 3
The dynamical model of a star connected brushless
dc machine is fully explained in [5] and repeated here in
(9). These equations could be coded in a simulation
environment, such as Matlab, in order to implement the
machine model and obtain simple simulation results. In
this model, L
s
stands for the per-phase self-inductance,
and M
s
for the mutual inductance between windings.
d e e
d s d e q
e
q ' e e
q s q e d e q e
e
'
0 e 0 e
'
e q e q
s s
di d ( )
V R i (1 )i
dt d
di d ( )
V R i (1 )i ( )
dt d
V ( )
3
T P ( )i
2
(L M )
0
= + A e A +

0
= + A + e A + + e u 0

= e u 0

= u 0

A =

(9)
With the goal of developing a suitable motor drive
control algorithm, the whole inverter-motor system was
modeled including speed controller, torque controller,
space vector modulator, and the actual trapezoidal back-
EMF machine as in [7-9]. The Matlab model built was
developed using the back-EMF waveforms shown in Fig.
3. These were measured on a trapezoidal back-EMF PM
2170
machine using a Tektronix TDS-7054 scope, measuring
the voltage across its terminals while it spun driven as a
generator with open terminals.
In order to control a PM machine, its phase currents
should be synchronized with the back-EMF. Since these
voltages are determined by the relative position between
the rotor and the stator, under the proposed control
scheme the rotor position must be detected first. Also, the
controller needs requires the values of the normalized
back-EMF u
abc
and the displacement angle
e
at every
control cycle. The latter is required in order to implement
the modified transformation matrix and derive the
appropriate decoupling terms. Unlike it was discussed in
[5], the control algorithm proposed in this paper utilizes
the information on the displacement angle
e
in the
current controller for the decoupling of the voltage terms
in the d and q axes. This is clearly shown below.
e e
d dec e s s q
e
e
s s e q
' e e
q dec e s s d e q e
e
' e
s s e d e q e
d ( )
V (L M )(1 )i
d
d (t)
(L M )( )i
dt
d ( )
V (L M )(1 )i ( )
d
d (t)
(L M )( )i ( )
dt

0
= e + =
0

= e +
0
= e + + e u 0 =
0

= e + + e u 0
(10)
The displacement angle can be derived directly from
(5) using inverse trigonometric functions and subtracting
the electrical rotor position 0
e
; however, this method is
complicated due to the periodicity of the sine and cosine
functions and the operations with the position 0
e
. The
critical part of the decoupling terms is related to the
calculation of the angle
e
. This can be avoided using the
following mathematical manipulation:
| |
e e e e e
e
d (t) (t) d (t) d (t) d (t)
dt dt dt dt
0 + 0 | | | |
e + = + =
| |
\ . \ .
(11)
By means of (11), the electrical rotor position and the
displacement angle can be considered as a new angle that
can be directly obtained as shown below.
| | | |
e e e
e
d (t) (t) d (t)
dt dt
(t) a tan 2( ' , ' )
o |
0 + o
=
o = u u
(12)
The atan2 operator above is commonly present in
DSP built-in functions and is defined as the four quadrant
arctangent from t to t. Once the o
e
angle is evaluated,
its time derivative can be computed in discrete form using
the algorithm sample time.
IV. SIMULATION AND EXPERIMENTAL RESULTS
The complete motor drive system was simulated and
tested on a 35 kW, 75 Nm, trapezoidal machine having
the 60-60-60 back-EMF waveform shown in Fig. 3. The
machine parameters are the following ones: stator
resistance 3 mO, synchronous inductance 180 H, and
peak magnet flux 0.067 V
p
/rad/s. The controller
bandwidths were set at 900 Hz for the torque and flux
controllers and 1 Hz for the speed controller. The inverter
switching frequency was 20 kHz, and the same frequency
value was used for the control sampling time.
The electromagnetic torque ripple depicted in Fig. 6
and Fig. 7 was sampled at high frequency in order to
show the switching harmonics as well as the harmonic
ripple. As seen, when the decoupling terms are included
in the control algorithm, the torque ripple can be
significantly reduced. In fact, from the comparison of
experimental results shown in these figures, it can be
observed that without decoupling terms the torque ripple
is more than 4 Nm, whereas only 2 Nm when the
decoupling terms are included in the control algorithm.
Figure 6. Electromagnetic torque, reference torque and load torque
(with decoupling terms)
Figure 7. Electromagnetic torque, reference torque and load torque
(without decoupling terms)
The motor drive controller was verified experimentally
and the corresponding results are shown in Fig. 8 and Fig.
9, which show the machine simulated and measured
currents respectively. Note that phase currents closely
resemble the machine back-EMF waveforms shown in
Fig. 3. This is due to the usage of the pseudo-Park
transformation that injects the back-EMF harmonics into
the inverter terminal voltages.
Figure 8. Simulation results: machine currents under linear control
2171
Figure 9. Experimental results: phase a current and electrical angle
under linear control
CONCLUSIONS
The use of Parks extended transformation has
allowed the development of the dq-axes model for PM
machines with arbitrary flux pattern distributions. In
consequence, high performance controls like vector
control could be implemented to reduce the torque ripple
and improve the overall performance of a trapezoidal
back-EMF machine. Moreover, the use of the pseudo-dq
machine model enabled the dynamic decoupling of the dq-
axes channels, which significantly reduce the machine
torque ripple by means of an appropriate cancellation of
the machine non-sinusoidal back-EMF shape. From power
electronics sizing point of view, trapezoidal machine
based drives show some benefits allowing the drive of
higher power brushless dc machines with respect to PM
sinusoidal machine drives.
REFERENCES
[1] P. Pillay, R. Krishnan, Modeling of Permanent Magnet Motor
Drives, IEEE Transactions on Industrial Electronics, vol. 35, no.
4, November 1988
[2] P. Pillay, R. Krishnan, Modeling, Simulation, and Analysis of
Permanent-Magnet Motor Drives, Part II: The Brushless dc Motor
Drive, IEEE Transactions On Industry Applications, vol. 25, no.
2, March/April 1989
[3] D. Grenier, R. Mende, J. P. Louis, Comparison of Several
Control Strategies for dc Brushless Drives, IEEE IECON, vol.
1, pages 5-9, Sept. 1994
[4] D. Grenier, L. A. Dessaint, O. Akhrif, J. P. Louis, A Park-like
Transformation for the Study and the Control of a Non-Sinusoidal
Brushless dc Motor, IEEE 1995
[5] D. Grenier, J. P. Louis, Use of an Extension of the Parks
Transformation to determine Control Laws Applied to a Non-
Sinusoidal Permanent Magnet Synchronous Motor, IEEE - EPE
1993
[6] J. P. Johnson, M. Ehsani, Y. Guzelgunier, Review of Sensorless
Methods for Brushless dc, IEEE 1999
[7] B. H. Bae and S. K. Sul, J. H. Kwon, and J. S. Byeon,
Implementation of Sensorless Vector Control for Super-High-
Speed PMSM for Turbo-Compressor, IEEE Trans. Ind. Applicat.,
Vol. 39, No. 3, pp. 811-818, 2005
[8] P. Pillay, R. Krishnan, Application Characteristics of Permanent
Magnet Synchronous and Brushless dc Motors for Servo Drives,
IEEE Trans. Ind. Applicat., Vol. 25, No. 5, September/October
1991

S-ar putea să vă placă și