Вы находитесь на странице: 1из 10

Magnetic cross-coupling effects on the performance of the Synchronous

Reluctance Machine
Alberto Colotti, Konrad S. Stadler
ZURICH UNIVERSITY OF APPLIED SCIENCES
Institute of Mechatronic Systems
8401 Winterthur, Switzerland
Tel.: +41 / (0) 58 934.65.41.
Fax: +41 / (0) 58 935.65.41.
E-Mail: alberto.colotti@zhaw.ch
URL: http://www.ims.zhaw.com

Acknowledgements
The authors express their thanks to Jrg Nf, scientific assistant at the institute and to Andreas Heussi
and Fabian Schb, former students of the Zurich University of Applied Science for the installation and
measurements on the system.

Keywords
Reluctance Drive, Vector control, Modelling, Nonlinear Model identification

Abstract
Performance prediction and control of a synchronous reluctance machine is based on the knowledge of
the direct and the quadrature axis inductances. The properties of the machine depend on the saliency
ratio of these inductances and this is implemented by introducing internal flux barriers. In an idealized
linear magnetic topology the superposition principle would allow to derive both properties separately
by exiting a single axis. However, in a real case with saturation effects a cross coupling of the axes
appears, which alters the resulting values. Both methods are adopted in the numerical simulation and
in practical measurements. By comparing the inductance values, the effects of saturation and coupling
can be observed and interpreted. The ideal current trajectories in the operation regions or the resulting
curves for a variable speed operation alter when different properties of the machine are assumed. The
comparison to practical measurements show a certain match of the simplified methods. However, the
cross coupling are more pronounced in regions with low currents and therefore the differences
increase in the low flux areas at high speed.

Introduction
Since the early 90s the synchronous reluctance machine (SynRM) has been widely investigated ([1]),
The technology itself is well known and was mentioned already at the beginning of the last century
([2]). Compared to the induction motor it offers a nearly loss free rotor, with benefits on the thermal
behaviour and the efficiency of the motor. Due to the absence of the internal excitation the SynRM
does not reach the power density of the permanent magnet motor. The advantage lies in the absence of
the expensive hard magnetic materials and therefore in the absence of iron losses and of induced backemf in idling conditions. The trend towards energy saving and a growing concern for material usage
has revived this motor technology. Some suppliers (e.g. ABB, KSB) offer this product in their
portfolio.
In Fig. 1 the principle of the SynRM is shown. The rotor consists of lamination with strong reluctance
anisotropy. The d-axis is orientated along the path with the lowest reluctance. Here the flux passes
mainly through iron parts. In the q-axis flux barriers increase the magnetic resistance. The magnetic
anisotropy leads to a torque production, because the rotor tends to reduce the magnetic energy
([3],[4]) and follows therefore the flux vector created by the stator current.

Fig. 1: Configuration of a four pole SynRM with internal flux barriers. The d-axis is orientated along
the path with lowest magnetic resistance.

Analytical description and calculations


The equivalent circuit of the motor in the dq-reference frame is shown in Fig. 2. As known from the
Park-Theory the flux in the d-axis induces a voltage in the q-axis and vice versa. In stationary
operation no self-inductance term appears in the rotor reference frame description. Its worth
remembering, that with a permanent magnet excitation in the d-axis a further voltage term for the
back-emf appears in the q-axis. Permanent magnet materials have a low permeability. For standard
topologies this leads to a low inductance in the d- and a higher one in the q-axis what is inverse
compared to the SynRM case.

Fig. 2 Equivalent circuit of the SynRM in dq-reference frame. Stray inductances and core losses are
neglected. The dashed voltage source indicates the back-emf voltage when a magnetic excitation on
the rotor is present. On the right: Vector diagram without considering the rotor resistance RS for
stationary operation.
For further interpretations and calculation the vector diagram resulting from the equivalent circuit is
useful. The components of voltage, currents and fluxes are rearranged in the dq-frame to form the
resulting space vectors. The pole angle is known from synchronous machine theory and describes
the angle between the rotor d-axis and the resulting airgap flux S. The angle is the current angle
and the torque angle. Finally, the phase shift between stator voltage and current determines the
power factor when no harmonics are considered. When the stray inductance and the rotor resistance
are taken into account they lead to an internal and an external power factor.
The voltage equation of the SynRM in the rotor reference frame is given by

ud
id Ld id
Lqiq
R
=
u
i +  +

Ld id
q
q Lqiq

(1)

where is the supply angular frequency. Evaluating the input power leads to the torque produced by
the motor.

T=

3
p(Ld Lq )id iq
2

(2)

The saliency ratio is defined by

= Ld L
q

(3)

For a positive torque both of the current components must be of the same sign. In a linear system with
constant coefficients the equation would have a maximal torque at a current angle of 45 degree.
Considering Fig. 2 the resulting air gap flux S is leading the d-axis of the motor by the pole angle .
The minimal magnetic energy would be achieved when the flux would be completely arranged on the
minimal reluctance. That is the magnetic explanation for the torque production. At a pole angle of 90
degree the d-current gets negative and the torque sign inverses.
The air including the air gap on the d-axis path and all of the iron present in the q-axis direction
also when used for mechanical reasons tends to reduce the saliency ratio and must be avoided as far
as possible to get a good machine performance. It is meanwhile an established knowledge that axially
laminated motors despite their higher saliency ratio have a disadvantage in respect to the transverse
laminated version. Therefore, only the latter one is treated in this paper ([4], [5]). An overview
regarding the dimensioning aspects of the SynRM is presented in [6].
When a given stator voltage is assumed, the above torque equation can be evolved to a function of the
voltage amplitude:
2
3 p 1
1 VS

T=
sin (2 ) .
2 2 Lq Ld

(4)

The Equations (2) and (4) are important for the control strategy. When the supply voltage offers a free
choice of the current angle a maximal torque per current strategy can be driven. At high speed, when
the voltage must be limited, a strategy with maximal torque per voltage must be chosen. Several
authors derived the control strategies and gave analytical expressions and interpretation methods for
the behaviour of the machine ([8], [9]).
The base for all this investigations is the knowledge of the motor inductances. Both are exposed to the
saturation by fluxes created the current components. Therefore, the inductances become nonlinear in
functions of two variables: Ld (id, iq) and Lq (id, iq) [5]. In the following, a simplified analysis and
measurement method is proposed. The inductances are identified by decoupling the excitation in dand q-axis. A comparison is given to evaluate the errors and the limitations of the method.

Test motor
For the comparison of the method a motor from SERVAX Drives / Landert Motoren AG
(Switzerland) was used. The test assembly was set up in the laboratory at the Institute for
Mechatronics Systems (IMS) of the Zurich University of Applied Science.
Type:
Nominal Power
Torque
Supply

MHP-S-90-04AB-01
8.5 kW@1500 rpm
54 Nm
400 V / 22.5 Arms

Fig. 3 Test motor MHP-S-90-04AB-01 from SERVAX Drives / Landert Motoren AG with
additional inertia for acceleration measurements

Identification by FEM methods


The internal barrier rotor must be constructed to offer a maximal air gap path in q-axis direction and a
minimal in d-axis. By locating the barriers in the internal region several ribs and iron wedges must be
introduced to get a mechanical stiff element for the required speed capability. In this complex
geometrical building no analytical calculation is possible. The FEM calculation is mandatory to
achieve the correct properties. Due to the small air gap the reason for that has already been explained
and the small wedges a very thin meshing is needed to get a good accuracy in the FEM results. The
example of a motor model with a mesh detail is shown in Fig. 4

Fig. 4 Motor model and mesh detail


For the further calculation process two different model variants have been chosen. For the
identification of the d-axis inductances the winding U has been laid exactly in this axis as shown in
Fig. 5 on the left. With this layout the flux in the d-axis coincides for every current situation with the
flux in the phase U. The inductance Ld is therefore simply the ratio between the linked flux and the
winding current. In the same manner the q-axis parameters are evaluated. The rotor has to be
positioned as shown in Fig. 5 on the right. Doing these placements all FEM calculations can be done
by only shifting the current distribution. No time consuming movement of the rotor is needed to obtain
all different values of current angle.

Fig. 5 Winding distribution to get coincidence of d- and q-axis with the winding U.
In a first calculation the inductance Ld(id, iq) is derived. The results are reported in Fig. 6 on the left.
For the parameter iq = 0 A the behaviour with single axis excitation the d-axis is shown. Between 1 and
10 A of id a nearly constant inductivity can be observed. By increasing iq more regions are saturated
and therefore slightly reduced inductances Ld emerge. Above 10 A in the region of the nominal
current iS,N even the direct flux path increasingly gets saturated. The cross-coupling from the d-current
on the q-axis becomes marginal. At low id and iq currents the value of Ld reflects the permeability
properties of the rotor iron. The desaturation of the ribs with small currents has no effect on Ld. The
increase of Ld at high iq current however, is not due to the iron properties. The curve reflects the fact,
that the high q-current generates fluxes in the d inductance because of cross coupling.
The same procedure is applied to the Lq axis (see Fig. 6 right). The intention of the design is to realize
a low value of inductance in the q-axis. It is clearly visible how the saturation affects the results. With
low current values Lq has practically the same value as Ld. In the case with id = 0 A a low iq is not able
to put the iron rips into saturation. In a first stage, the increasing current leads to a clear rise of the
inductance, corresponding to the initially increasing permeability of iron materials. The desaturation of

the ribs leads to an inductance, which


w
approaches the same value as for Ld. No saliency is offered from
the rotor when the ribs are not yeet saturated.

Fig. 6 Inductances Ld (id) with iq as parameter and Lq (iq) with id as parameter up to the nominal
current iS,N.
In the region of the nominal curreent iS,N = 32 A the cross coupling has very low im
mpact on the
inductances. At low currents the iron characteristics is sensitive to the current loadding. So a more
pronounced variation of the induuctance due to the perpendicular current can be obbserved.
As presented in Equation (2) the torque production is a function of the interaction between inductance
and current. In the generalized caase functions Ld(id, iq) and Lq(id,iq) must be used too reflect a specific
linkage situation. For the sake off convenience a reduction to Ld(id, iq = 0 A) and Lq(id = 0 A, iq) would
be of interest.
To investigate the consequences the torque was calculated in three different ways::
A) Evaluation of the torque withhin the FE programm with variation of the pole anngle for different
amplitudes of the stator curreent and shifting of the current angle.
B) Neglection of the cross couplling and calculation of the torque by Eq. (2) usingg the inductances
derived by means of FE withh single axis id or iq current.
C) Calculation of torque with Eqq. (2) and the inductances resulting from a combiined current situation
with variable current angle .
In all three cases the effects of haarmonics or torque ripples is not considered in thiis investigation. The
scope is to establish a simplified calculation method. Considering such effects woould increase the
FEM calculation efforts considerrably. The results are shown in Fig. 7.

Fig. 7 Comparison of calculatioon methods.

Measurements
The same simplification method as used in the calculation has also been applied too the measurements.
Instead of evaluating the inductannces as function of both the current components Ld,q=f(id,iq) the
orthogonal components have beeen applied separately. As further simplifications thhe measurements
were conducted without couplingg the test motor to a braking machine as shown inn Fig. 3. Such a set-up
is easily feasible and there is no need
n
for a second control system to ensure a controlled braking
torque.
No torque is produced when singgle axis d- or q-currents are applied. By measuringg the stator voltage,
the d- and q-components of the innductance can be derived by

Ld , q =

(v

(RiS )
id , q

2
S

(5)

Fig. 8 Inductances Ld (id) and Lq (iq) from single axis excitation measurements inn straight lines.
Dashed lines indicate the correspponding values from FEM simulation with single excitation
calculation, dotted lines with low
w cross coupling component in the calculation.
In the region of the nominal curreent iSN, the values based on the measurements andd the values derived
from simulation match well as shhown in Fig. 8. When approaching lower currentss the differences
increase and the values derived from
fr
the measurements approach each others. Noo effect of saliency
remains effective in this mode. The
T reduction of permeability which is shown in usual
u
iron curves
appears only in the simulated currve. An interesting aspect appears when the Ld / Lq curves with a low
amount of cross coupling are addded (dotted lines in Fig. 8). At low currents the caalculation with low
cross coupling (dotted lines) is clloser to the measurement than the simulation withh single axis
excitation. We assume that this reesults from the measurements not being preciselyy aligned to the d-q
axis.
In Fig. 9 the results of two differrent methods to determine the torque with inductaance values
originating from the measuremennts are presented and compared to curve A in Fig.. 7.
These methods are:
D) The torque is determined by calculating the acceleration when a combined exxcitation current with
amplitude iS and current anglle is applied.
E) The torque is determined by evaluating Eq. (2) with inductances derived by Eq.
E (4) when a single
excitation id or iq is applied.
Method D is used as the referencce because it results from practical measurements.. The FEM
simulation (A) provides a better fit
f for the optimal current angle. The simplified calculation based on
measurements (E), has a comparaably large difference and a clear error in the optim
mal current angle.
Therefore, using method E is nott suggested.

Fig. 9 Comparison of measurem


ment methods.

Control aspects
Current trajectories
When driving a motor the scope is to maximise the performance within given connstraints. E.g. for the
stator current

i S = i d2 + iq2 I S , max

(6)

and for the stator voltage

v S ( , id , iq ) =

(L i ) + (L i )

d d

q q

VS ,max

(7)

Three different modes of operatioons can be observed:


MTPA region: at low speed the motor voltage is below the maxim
mal value (vS<VS,max),
therefore the motor currrent can be at the maximal value (iS=IS,max). In the case of constant
inductances the maximal torque per ampre (MTPA) value is reached att the current angle of
MTPA = 45 degree accordding to Eq. (2). The torque is constant in this reggion. Solving Eq. (7)
for at vS=VS,max gives thhe maximal speed which can be reached with IS,maax:

MTPA,max =

VS ,max
I S ,max

2
L + L2q

2
d

(8)

MTPVA region: exeedinng MTPA,max a further increase of speed must bee obtained by shifting
the current angles towardds a lower inductance, still the maximal current can
c be impressed, but
the torque is now reduceed in this maximal torque per volt and ampre (M
MTPVA) region. The
operation point moves on
o the circle with constant current iS=IS,max. This shifting makes sense
as long as the torque cannnot be reached with less current at a lower currennt angle. This limit at
= MTPVA,max introduces the next operation region.
MTPV region: the operaation is limited by the available voltage. The optim
mal current angle can
be deduced by inserting Eq.
E (7) in Eq. (2) and solving thus for the torque as a function of id :

V
3
T (id ) = p(Ld Lq )id S , max
I S , max
2

VS , max

2id2
L
q

(9)

The derivative dM/did = 0 leads to the currents which give the maximal torque.

id =

VS ,max
2Ld

, iq =

VS ,max
2Lq

and for the ratio

iq
id

Ld
=
Lq

For a given stator current id becomes

id =

1
1+ 2

iS

(10)

With iS = IS,max the maximal speed in the MTPVA region results to be

MTPVA,max =

VS ,max
2 I S ,max

1
1
+ 2
2
Lq Ld

(11)

In the MTPVA region the current angle is constant at

MTPVA = tan 1 ( )

30

30

25

25

20

20
q

i /A

i /A

(12)
In Fig. 10 the trajectory in the idealized case of constant inductances is presented. As a reference the
locus of constant current (circles in the id /iq frame), constant torque (hyperbolas) and constant voltage
(ellipses) are drawn. As results from Eq. (7) the locus of a constant voltage will form an elliptic shape.
With increasing speed the ellipse becomes smaller. The saliency ratio of the machine is visible in the
ratio between the main axes of the ellipse. The current curves for constant torques form hyperbolas.
The minimal current is reached on the MTPA locus at a current angle of 45 degree.

15

L ,L const.

10

10

d q

sing. ex. calc


sing. ex. meas
comb. ex calc
comb. ex meas

5
0
0

15

10

15
i /A
d

20

25

30

0
0

10

15

20

25

30

i /A
d

Fig. 10 Left: Current trajectories in a case of constant inductances with MTPA, MTPAV and MTPA
modes ([7], [8]). The locus for constant torque at different current amplitudes, constant voltage at
different speeds and current are shown as indicated by the corresponding arrows. The speeds which
denote the transition to the next region are labelled. Right: Comparison of different current
trajectories. Blue: constant inductances; red: single axis excitation; green: combined excitation. Thick
lines are used for calculation and thin lines for measurement results. Star points from measurements
(see Fig. 11).
In the case of inductance variations due to cross coupling or saturation effects no analytical solution is
possible. Therefore, to determine the optimal current vectors a numerical evaluation of the operation
point was implemented. Three approaches for the inductances used to derive the voltages and torques
were adopted:
1) constant inductances with the values of the nominal operation point of the motor (Is,max = 30 A,
Lq=10 mH, Ld=32 mH, MTPA = 45 deg, MTPV = 73 deg, see Fig. 10 right, blue curve)
2) inductances from single axis excitation (see Fig. 10 right, red curves)
3) inductance with combined excitation (see Fig. 10 right, green curves)
Due to the higher saliency at lower saturation level the current angle has to be shifted to higher
values in the MTPA region. This effect appears more pronounced in the case of identification with the
single axis excitation. In the combined excitation the increase of saliency is less pronounced. So the
current angle is reduced either. At low currents all the inductances are lower than at the nominal point.
Therefore the MTPVA,max speed is higher and again the current angle is increased.
The MTPV region shows very little difference for the two identification methods.

Speed curves
When plotting power, torque and current as functions of the speed, the behaviour in a variable speed
drive arrangement becomes visible. In Fig. 11 the speed characteristics are shown for the following
methods of inductance identification:
I) voltage measurements (vd, vq) at single axis excitation with id or iq (see Fig. 8)
II) FE flux calculation with combined exitation with is and (see Fig. 6 and curve C in Fig. 7)
III) torque and voltage measurements (vd, vq) when applying combined excitation is and (see
curve D in Fig. 9)

Torque T /Nm

As reference additional measurements have been carried out on the test bench with a break motor as
shown in Fig. 11. In the measurements the angle currents were adjusted to meet the ideal operation
points within the operation modes. The resulting curve (drawn with star points in Fig. 11) is therefore
the reference for the evaluation of the speed curves.
60
40
20

Current i S /A

500

1000 1500 2000 2500 3000 3500 4000 4500

30
20
I) sing ex. meas
II) comb. ex. calc
III) comb. ex. meas

10
0

500

1000 1500 2000 2500 3000 3500 4000 4500

500

1000 1500 2000 2500 3000 3500 4000 4500

Angle /deg

90
80
70
60
50

Speed n /rpm

Fig. 11 Variable speed curves based on different inductance identification methods and measurements
points resulting from test bench.
In Table I the resulting values of the different methods are compared. The method II shows the best fit
to the measured reference curve. The main effect of the cross coupling as shown in Fig. 6 is a
reduction of the inductances. Therefore, all speed calculation (nMTPA, max with Eq. (8) and nMTPAV, max
with Eq. (11)) based on single axis excitation (method I) are too low.

Table I: Comparison of operating points

nMTPA, max
nMTPAV, max
@ nMTPA, max
@ nMTPAV, max
Tmax, MTPA
Tmax,MTPAV

I Measurement II Calculation III Measurement


with combined with combined
with single
axis excitation axis excitation axes excitation
1300
1900
1800
2800
4000
3000
70
65
57
85
84
78
56
54
50
20
20
29

* Measured
reference
curve
1861
4000
63
82
47
17

rpm
rpm
deg
deg
Nm
Nm

Control scheme outlook


The control of the SynRM with the strong nonlinearity of the voltage and torque properties requires a
control scheme implementation with look-up tables (LUT) ([9], [10]). A reliable knowledge of the
nonlinear properties of the equivalent circuit is used in the following two fields:
Implementation of a feedforward compensation with LUT in the current control loop. The
relations Ld (id, iq) and Lq (id, iq) are used in the -term in Eq. (1) to decouple the voltage
components and to obtain a faster response in the control loop.
Generation of the ideal current angle within the three operation modes with a LUT which
incorporates the functions id,C (T, ) and iq,C (T, ).
Parameter variations or identification errors will lead to control errors in a pure LUT based open loop
control structure. Investigations on these effects cannot be included in this publication and are matter
of further works.

Conclusion
Due to the magnetic properties of a lamination with internal flux barriers as used in SynRM motors the
d- and q-axis are not completely decoupled in their magnetic behaviour. Therefore, two-dimensional
arrays must be used to describe the inductance properties of such a motor. The identification of these
parameters bears a high computational effort in FE-methods. A reduced and simplified function of the
inductances can be obtained by applying a single axis excitation in the dq-axes frame. This method,
which neglects cross-coupling effects, gives faster results when properties of a design and the
reachable operation characteristics are calculated. Also in measurement based identification the
application of a combined axes excitation implies a quadratically higher number of measurements.
The proposed simplified method is evaluated by comparing the results in the current trajectory plane
and also in the resulting speed curves.
At very low currents the flux leakage due to the special rotor design of a SynRM leads to
considerably large differences. At nominal points the steel properties become more stable and the
cross coupling effects reduce. Due to the errors produced generally in the single axis excitation
methods the deduced current angles are not ideal. This can be seen in the resulting current trajectories
and torque curves for variable speed operation.

References
[1] T. A. Lipo, Novel Reluctance Machine Concepts for Variable Speed Drives,Proc. 6th Mediterranen Elec.
Conf., pp 34-43, 1991.
[2] J.K. Kostko, Polyphase Reaction Synchronous Motor, Journal American Inst. Elec. Eng., Vol 42, pp
1162-1168, 1923.
[3] A. Vagati, A. Fratta, G. Franceschini and P. Rosso, AC motors for high-performance drives: a design
based comparison, IEEE Trans. Ind. Appl., Vol. 32, No. 5, pp 1211-1219, 1996.
[4] T. J. E. Miller, A. Hutton, C. Cossar and D. A. Staton, Design of Synchronous Reluctance Motor Drives,
IEEE Trans. on Industrial Applications, Vol. 27, pp 741-749, Jul./Aug. 1991.
[5] M. J. Kamper and A. F. Volschenk, Effect on rotor dimensions and cross magnetisation on Ld and Lq
inductances of reluctance synchronous machine with cageless flux barrier rotor, IEEE Proc. on Elec. Power
Appl., Vol. 141, No. 4, pp 213-220, 1994.
[6] A. Vagati, A. Canova, M. Chiampi, M. Pastorelli and M. Repetti, Design Refinement of Synchronous
Reluctance Motors Through Finite-Element Analysis IEEE Trans. on Industrial Applications, Vol. 36, No. 4,
pp 1094-1102, Jul./Aug. 2000.
[7] R. R. Moghaddam, F. Magnussen and Ch. Sadarangani, Theoretical and Experimental Reevaluation of
Synchronous Reluctance Machine IEEE Trans. on Industrial Electronics, Vol. 57, No. 1, 2010.
[8] W. L. Song and T. J. E. Miller, Field-weakening performance of brushless synchronous AC motor drives
IEE Proc. Electr. Power Appl., Vol. 141, No. 6, pp 331-340, 1994.
[9] S. Morimoto, Y. Takeda, T. Hirasa and K. Taniguchi, Expansion of operation limit for permanent magnet
motor by current vector control considering inverter capability, IEEE Trans. Ind. Appl., Vol. 26, No. 5, pp
866-871, 1990.
[10] W. Lu, L. Liang, Research on Saturation Compensation Control of Interior Permanent Magnetic (IPM)
Motor with Wide Speed Operation,Chinese Control and Decision Conference, pp 2467-2472, 2011.

Вам также может понравиться