Predictive Torque and Flux Control of an Induction

Machine fed by an Indirect Matrix Converter with

Reactive Power Minimization
J. Rodriguez*, J. Kolar**, J. Espinoza***, M. Rivera*, C. Rojas*
*
Departamento de Electrónica, Universidad Técnica Federico Santa María.
Avenida España 1680, Casilla 110-V Valparaíso, Chile.
E-mail: [email protected], [email protected].
**
Power Electronics System Laboratory, ETH Zurich, Zurich, Switzerland.
E-mail: [email protected]
***
Departamento de Ingeniería Eléctrica, Universidad de Concepción, Concepción, Chile.
E-mail: [email protected]

Abstract- A predictive torque and flux control method for an same performance that the DMC [2]-[4], such as four-
induction machine fed by an indirect matrix converter is quadrant operation, unit power factor, sinusoidal waveforms
presented in this paper. The control scheme selects the switching
state that minimizes the error in the torque and flux predictions
with variable frequency and amplitude during motoring and
according to their reference values and, at the same time, the regeneration operation, however, the control in this converter
control scheme is enhanced by including a reactive power is simpler and less complex compared to the DMC, allowing
minimization strategy with the goal to have unity power factor also the system to commutate securely [5][6] without
in the input side of the converter. The control objectives are particular sensing devices as required for the DMC [7]. A
accomplished by using a prediction horizon of one sample time
and a very intuitive control law. The proposed control scheme is further improvement compared to the DMC is that no
validated trough simulation results where is shown that the additional over-voltage protection circuit is required.
predictive approach can be implemented simply with a good Different works have been reported in the literature [8]-[11]
tracking of the torque and flux to their respective references and that implements the torque control in electrical machines
input currents with unity power factor. considering back-to-back and matrix converters with the use
of complex Pulse Width Modulation (PWM) schemes to
I. INTRODUCTION
achieve the goal of unity power factor and sinusoidal output
Nowadays, ac/ac converter systems with either a voltage current. However, since power converters have a discrete
or current dc-link have been widely studied and are mainly nature, the application of predictive control constitutes a
used in the industry. Due to the dc-link storage element promising and better suited approach as compared to standard
(capacitor or inductor), there is the advantage that both schemes that use mean values of the variables. Furthermore,
converter stages (rectifier and inverter) are to a large extend predictive control utilizes the system model to predict the
decoupled for control purposes, but the dc-link energy storage future behavior of variables to a predefined time horizon and
component has a relative large physical size and thus reduce selects the optimal action based on the minimization of a cost
the system lifetime. Moreover, there are also ac/ac converter function [12]-[14]. This structure is characterized because its
systems without any intermediate dc-link storage element, concepts are very intuitive and easy to understand, it can be
where different topologies have been reported in the applied to a wide variety of systems, may involve multiple
literature, which are classified into three main groups: the systems, dead time compensation, and non-linear constraints,
cycloconverter in a wide power variety, the direct matrix resulting in an easy controller to be implemented, being open
converter (DMC) and the indirect matrix converter (IMC), to modifications and extensions for specific applications as
both in low power range [1]. The cycloconverter is very reviewed in [15]-[21]. This paper considers the advantages of
common in high power applications such as cement kilns and predictive control and the indirect matrix converter, to control
ball mills in mineral processing, but has a major limitation in in an easy, an intuitive and a new manner, the torque and flux
terms of output frequency with respect to the input, because variables of an induction machine fed by the converter while
the presence of a high harmonic content due the ensuring a unitary power factor at the input system of the
commutations, which can not be filtered by the load converter. The predictive approach is based on the fact that
inductance. On the other hand, the DMC is based on bi- only a fixed numbers of feasible switching states can be
directional switches and carry out voltage and current generated by a static power converter and that the models of
conversion in only one stage, being a preferred choice when the system can be used to determine the behavior of the
the size and the need to remove the dc-link stage are variables for each commutation state. The work is presented
significant issues. But the biggest drawback of this as follows. Section II presents the fundaments of the indirect
technology is the great control complexity. IMC offers the matrix converter where is described its mathematical model

978-1-4244-6392-3/10/$26.00 ©2010 IEEE 3177

 Fig. 1. Indirect matrix converter topology.
and the relation between the input and output variables. C f v i  i s  i e , (6)
Section III describes the mathematical model of the induction where Lf involves the mains and filter inductances, and Rf the
machine, section IV depicts the control strategy proposed, mains and filter damping resistances. The rectifier stage can
validated with simulation results in Section V, and finally produce only positive dc-link voltage in each sampling time
conclusions are expressed in Section VI. (3 of 9 possible switching states accomplish this request), so
the number of valid switching states is 24. It should be noted
II. FUNDAMENTS OF THE INDIRECT MATRIX CONVERTER that the IMC topology includes as many switches as the
DMC, but the former features an extra freedom degree that
The converter topology is shown in Fig. 1, and it consists
alleviates the complexity of the commutation sequence [5][6].
of a rectifier connected to the inverter through a dc-link
without energy storage element. The converter synthesizes a
positive voltage in the dc-link by selecting a switching state III. MODEL OF THE INDUCTION MACHINE
in the rectifier that connects one phase to the point P and the To obtain the model of the system, it will be assumed that
other phase to the point N. In addition, the rectifier includes the three phase quantities of the converter are symmetrical
an LC filter in the input side which is needed to prevent over and therefore, can be represented by a two-dimensional space
voltages and to provide filtering of the high frequency vector. For example, the phase components xu, xv, and xw will
components of the input currents produced by the be described by the complex space vector:
commutations and the inductive nature of the load. The dc-
link voltage vdc is synthesized by the input voltages vi = [vir vis x  x  jx , (7)
vit]T and the switching states as follow, which is defined as:

vdc  S r1  S r 4 Sr3  Sr 6 
Sr5  Sr 2 vi ,
2 xu  xv  xw 
(1) 1
x 
where Sr1… Sr6 are the switching states of the rectifier stage. 3 ,
In addition, the input current ii = [iir iis iit]T, is doing by the
 (8)

switching state and the dc-link current idc, as: x 

1
 xv  x w  

3
 S r1  S r 4  This space vector is referred to a stationary reference frame
i i   S r 3  S r 6  i dc , (2) that will be considered as an -reference frame. The model
S  S  of the induction machine referred to stator is obtained as
 r5 r2  described in [17]. Hence, the induction machine model in
The dc-link current idc, is determined by the switching space vector representation is given by (9) and (10), where
states of the inverter stage Si1… Si6, and the output current io = the stator and rotor voltage equations in fixed coordinates are
[iou iov iow]T as follow, presented as:
v o  Rs i o  Ls ψ
idc  S i1 S i 3 S i 5 i o
s (9)
(3)
v r  Rr i r  Lr ψ r  jpψ s (10)
and the output voltage vo = [vou vov vow]T is determined by the
switching states of the inverter stage and the dc-link voltage where Rs, Rr and  correspond to the stator resistance, rotor
vdc as, resistance, and rotor angular frequency, p as the ac machine
number of pole pairs, in that order. The stator and rotor fluxes
 S i1  S i 4  are related with their respective currents trough the equations,
v o   S i 3  S i 6  v dc . (4)
ψ s  Ls i o + Lm i r and ψ r  Lr i r  Lm i o (11)
S  S 
 i5 i 2  where Ls, Lr, and Lm correspond to the self and mutual
The line side of the rectifier consists of a second order inductances respectively. Finally, the electric torque can be
system described by: expressed in current and flux terms such as,
L f i s  v s  v i  R f i s , (5)

3178
 Fig. 2. Proposed control strategy scheme.

2 the valid switching state that produces the minimum value of

Te  p (ψ s  i o ) . (12)
the cost function is selected for the next sampling period. As
3
mentioned before, the method is represented in Fig. 2, where
it is possible to appreciate that the control algorithm is easy to
IV. PROPOSED SCHEME: PREDICTIVE TORQUE AND FLUX understand and implement which demonstrates that the use of
CONTROL OF AN INDUCTION MACHINE predictive control can avoid the use of complex modulation
techniques.
A. Proposed Control Strategy
Predictive control considers the advantage of the discrete B. Steps 1-2 Measurements, References and Speed Control
nature of power converters, which have a finite number of As mentioned before, to implement the control algorithm it
valid commutation states because they are based on discrete is necessary to obtain the measurements and references values
switches, which have only two states: ON and OFF. The of the interesting variables. The speed is controlled using an
proposed predictive control scheme is represented in Fig. 2. external controller which generates the torque reference used
to make the gating patterns as mentioned before. The
The approach pursues the selection of the switching state of
controller is a PI because the integral part is required in order
the converter that leads the torque and flux closest their
to achieve zero steady state error, due to the fact that the
respective references at the end of the sampling instant. At predictive torque control fast dynamic can be represented just
the same time, the line side of the rectifier must deliver active as a unity gain between the reference and the controlled
power and finally the dc-link voltage must be always positive. variables.
These before mentioned conditions are fulfilled by the
predictive controller in five steps as follow: C. Steps 3-4 Flux Estimations and Calculation of Predicted
1. The input voltages vsk, vik, input currents isk, stator currents Values
iok and speed k of the induction machine are measured. With the assumption that it is possible to define a first
2. The speed *k, flux ψs*k and reactive power qs*k references order approximation for the derivatives due to the first order
correspond at given values, although the torque Te*k reference nature of the state equations that describes the induction
is established by a PI linear controller which have as input the machine model (9)-(12) as,
error between the reference and measured speed values of the x k+1  x k
x  (13)
induction machine as explained in section IV.B. Ts
3. The stator and rotor flux estimations, are obtained by a where Ts is the sampling period, so, the stator and rotor fluxes
flux estimator described in section IV.C. can be estimated from (9)-(10) resulting in,
4. The model of the system is used to predict in the next
ψ sk  ψ sk -1  v ok Ts  Rs i ok Ts (14)
sampling time the value of the torque Tek+1, flux ψek+1 and
reactive power qsk+1, for each of the valid switching states Lr k 1  LL 
according exposed in section IV.C. ψ rk 
ψ s   Lm  s r  i ok (15)
Lm  Lm 
5. As a final point, the predicted values are used to evaluate
a cost function which deals with the torque, flux and the input Thus, according to [18] it is possible to obtain a stator flux
power factor errors as indicated in section IV.D. After that, prediction as,

3179
 150

100

wmec [rad/s]
50

0
(a)
-50

-100

-150
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
60

40
Teref - Te [Nm]

20

(b) -20

-40

-60

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Time [s]

1.2

1
Flux [Wb]

0.8

0.6
(c)
0.4

0.2

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
20

10
Io [A]

0
(d)
-10

-20
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
20

10
Is [A]

0
(e)
-10

-20
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
4
x 10
2

1.5
Reactive Power [VAR]

0.5

0
(f)
-0.5

-1

-1.5

-2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
Fig. 3. Simulation results without reactive power minimization:
(a) mec: mechanic speed, (b) Te: electric torque and Te*: reference; (c) s: stator flux and s*: reference;
(d) io : load current; (e) isr: source current and vsr/50: source voltage; (f) qs: source reactive power.
u

ψ sk+1  ψ sk  v ok 1Ts  Rs i ok 1Ts , (16) where,

k+1
where vo is given by (4) and the stator current prediction rσ  Rs  Rr kr2 ,   1  kr ks , r  Lr Rr , kr  Lm Lr ,
equation [18] is, ks  Lm Ls (18)
 rT 
i ok 1   1   s  i ok Ts  s   v ok 1   r k r  jkr   ψ rk (17)
T The predicted electrical torque, for the next sample time, is
  Ls   Ls deduced from (12) and (16)-(17) as,

3180
 150

100

wmec [rad/s]
50

0
(a) -50

-100

-150
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
60

40
Teref - Te [Nm]

20

(b) -20

-40

-60

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Time [s]

1.2

1
Flux [Wb]

0.8

0.6
(c)
0.4

0.2

0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]

20

10
Io [A]

0
(d)
-10

-20
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
20

10
Is [A]

(e)
-10

-20
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
4
x 10
2

1.5
Reactive Power [VAR]

0.5

0
(f)
-0.5

-1

-1.5

-2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Time [s]
Fig. 4. Simulation results with reactive power minimization:
(a) mec: mechanic speed, (b) Te: electric torque and Te*: reference; (c) s: stator flux and s*: reference;
(d) iou: load current; (e) isr: source current and vsr/50: source voltage; (f) qs: source reactive power.

2 v ik 1  11v ik  12i sk   11v sk   12i ik . (21)

p  ψ sk+1  i ok 1 
Tek+1  (19)
3 The real coefficients ij and ij are defined so that the
The prediction of the input current and capacitor voltages obtained values for the predicted currents correspond to those
are computed from a first order differential equation, as is of the continuous time system after one sampling time, as
described follows, exposed in [23].
i sk 1  21v ik  22i sk   21v sk   22i ik , (20)

3181
D. Step 5. Cost Function (Switching State Selector) power is included in the cost function g, i.e. C ≠ 0, the input
Different control criteria will be expressed in different current (Fig. 4.e) is in phase with its respective phase voltage
quality function but in this work, the absolute error is except when the reversal manoeuvring is applied, as
considered for computational simplicity. So, the error expected. In the second case, a good behaviour of the reactive
between the predicted electrical torque and its reference can power is appreciated in Fig. 4.f, because it is maintained
be expressed as follow, always closes to zero.
Tek 1  Te*  Tek 1 , (22) The simulation results were done with the exact parameters
of the machine, Table I. A future research is the study of the
where Tek+1 denotes the electrical torque for k+1 sample time,
control sensibility with changes in the system parameters.
and Te* its respective reference. Also, the error between the
predicted values of the stator flux and its reference can be
represented as, VI. CONCLUSIONS
 sk 1   s*   sk 1 , (23) A very simple and effective predictive torque and flux
furthermore, the error between the reference and the predicted control method with reactive power minimization applied in
value of the instantaneous reactive power is given by, an indirect matrix converter has been presented in this
qsk 1  0  (vsk1isk1  vsk1isk1 ) . (24) publication. The strategy offers the possibility to control both
the input currents, to maintain unity power factor, and the
Finally, the equations (22)-(24) are combined into a single output voltage or machine’s variables at the same time. The
so-called quality function: control scheme uses a discrete model of the converter,
g k 1  ATek 1  B sk 1  Cqink 1 . (25) induction machine and input filter to predict the behaviour of
where A, B, and C are the weight coefficients which denotes torque, flux and input current of the system and to obtain the
the priority in the control. The proposed predictive scheme best-suited converter switching state considering the torque,
operates as follows: at each sampling time, all possible flux and reactive power error by the evaluation of the 24
switching states are used to calculate the eq. 25. The possible combinations of the topology. The ideal minimum of
switching state that produces the minimum value of g is the cost function is zero and represents the perfect regulation
selected to be applied for the next one sampling period [16]- of the controlled variables, this is unity power factor and a
[23]. given machine torque and flux. With this control strategy, no
modulators are required and also can be considered, in a very
convenient form, the discrete nature of the switching state of
V. SIMULATION RESULTS
the converter and the function of the microprocessor used to
The predictive strategy is simulated in order to validate the obtain the predictions.
proposed control scheme. The simulation parameters are
indicated in Table I and the sampling period of the control
APPENDIX
algorithm was set in Ts = 10s. Two cases are considered; the
first one includes the control of torque and flux, but does not The simulation parameters are detailed in Table I.
includes the minimization of the reactive power (C = 0) and TABLE I
the second one minimizes at the same time, the torque and SIMULATION PARAMETERS
flux of the induction machine and reactive power in the input
Variables Description Simulation
system. Both test considers the starting of the induction Vs RMS supply phase voltage 540(V)
machine at t = 0.05s without a load torque, applying a speed fs Supply frequency 50(Hz)
reference change from 0 to 1430 RPM (nominal speed of the Input Filter
Lf Input filter inductance 400(H)
induction machine); during the starting, the torque of the
Cf Input filter capacitor 21(F)
machine is limited at its nominal value 51 Nm. In the instant t Rf Input filter resistance 0.5()
= 0.3s is applied a load torque equal to 40 Nm and finally is Load
introduced a reversing in t = 0.5s, changing the speed Pn Nominal power 4(kW)
In Nominal current 12(A)
reference from 1430 RPM to -1430 RPM (-149.7 rad/s). The
n Nominal speed 1430(RPM)
torque reference generated by the speed controller is different Tn Nominal torque 51(Nm)
from zero during the transients and load torque steps and can Rs Stator resistance 1.35()
be appreciate a good tracking of the speed (Fig. 3.a and Fig. Ls Stator inductance 0.2861(mH)
Rr Rotor resistance 7.2037()
4.a), torque (Fig. 3.b and Fig. 4.b) and stator flux (Fig. 3.c Lr Rotor inductance 0.2861(mH)
and Fig. 4.c) to their references. Fig. 3.d and Fig. 4.d Lm Magnetization inductance 0.2822(mH)
indicates the behaviour of the output current where is P Pole Number 2
Control Method -
visualized the frequency changes depending of the motor Ts Sample time 10(s)
operation, but is verified that this output current presents in qs* Reactive power reference 0(VA)
all the time sinusoidal waveforms. In Fig. 3.e and Fig. 3.f, A weighting factor 63
results shown the chaotic behaviour of the input current and B weighting factor 14000
C weighting factor 1
reactive power, but when the term that minimizes the reactive

3182
 ACKNOWLEDGMENT Transactions on Power Electronics, Vol. 23, Issue 2, March 2008, pp.
612 – 618.
The authors wish to thank the financial support from the [20] M. Rivera, R. Vargas, J. Espinoza, J. Rodriguez, P. Wheeler, C. Silva,
“Current Control in Matrix Converters Connected to Polluted AC
Chilean Fund for Scientific and Technological Development Voltage Supplies”, in Conf. Rec. PESC’08, Rhodes, Greece, Jun. 15 –
(FONDECYT) through project 1100404. 19, 2008.
[21] R. Vargas, M. Rivera, J. Rodriguez, J. Espinoza, P. Wheeler,
REFERENCES “Predictive Torque Control with Input PF Correction Applied to an
Induction Machine fed by a Matrix Converter”, in Conf. Rec. PESC’08,
[1] P.W. Wheeler, J. Rodriguez, J.C. Clare, L. Empringham, A. Weinstein, Rhodes, Greece, Jun. 15 – 19, 2008.
“Matrix Converters: A Technology Review”, IEEE Transactions on [22] M. Rivera, P. Correa, J. Rodriguez, I. Lizama, J. Espinoza, “Predictive
Industrial Electronics, Vol. 49, Issue 2, Apr 2002 pp. 276 – 288. Control of the Indirect Matrix Converter with Active Damping”, in
[2] S. Round, F. Schafmeister, M. Heldweind, E. Pereira, L. Serpa and proceedings of 6th International Power Electronics and Motion
J.W. Kolar, “Comparison of Performance and Realization Effort of a Control Conference IPEMC 2009, Wuhan China.
Very Sparse Matrix Converters to a Voltage dc-link PWM Inverter with [23] P. Correa, M. Rivera, J. Rodriguez, J. Espinoza, “Predictive Control of
Active-Front-End”, IEEJ Transactions of the Institute of Electrical an Indirect Matrix Converter”, IEEE Transactions on Industrial
Engineers of Japan, Vol. 126-D, no 5, pp.578 – 588, May 2006. Electronics, Vol.56, Nº 6, pag. 1847-1853. June 2009.
[3] T. Friedli, M.L. Heldwein, F. Giezendanner, J.W. Kolar, “A High [24] T. Friedli, S.Round, S. J. W. Kolar, “A 100 kHz SiC Sparse Matrix
Efficiency Indirect Matrix Converter Utilizing RB-IGBTs”, in Converter;” in proceedings of the Power Electronics Specialists
proceedings of the 37th IEEE Power Electronics Specialists Conference, PESC 2007, IEEE 17-21 June 2007, pp.2148 – 2154.
Conference, PESC '06, 18-22 June 2006, pp. 1 – 7.
[4] J. Schonberger, T Friedli, S.D.Round, J.W.;Kolar, “An Ultra Sparse
Matrix Converter with a Novel Active Clamp Circuit”, in proceedings
of the Power Conversion Conference, Nagoya, PCC '07, 2-5 April
2007, pp. 784 – 791.
[5] J.W. Kolar, F Schafmeister, S.D. Round, H.Ertl, “Novel Three-Phase
AC–AC Sparse Matrix Converters”, IEEE Transactions on Power
Electronics, Vol. 22, Issue 5, Sept. 2007, pp.1649 – 1661.
[6] L. Wei, T.A. Lipo, “A Novel Matrix Converter Topology with Simple
Commutation”, Record of the IEEE Industry Applications Society
Annual Meeting, Chicago, USA, pp. 1749 – 1754, 2001.
[7] P. W. Wheeler, J.C. Clare, L. Empringham, M. Bland, M. Apap, “Gate
Drive Level Intelligence and Current Sensing for Matrix Converter
Current Commutation”, IEEE Transactions on Industrial Electronics,
Vol. 49, no 2, pp. 382 – 389, 2002.
[8] K. Lee, F. Blaabjerg, “Sensorless DTC-SVM for Induction Motor
Driven by a Matrix Converter Using a Parameter Estimation Strategy”,
IEEE Transactions on Industrial Electronics, Vol. 55, Issue 2, Feb.
2008, pp. 512 – 521.
[9] J. W. Kolar, T. Friedli, F. Krismer, S. D. Round, “The Essence of
Three-Phase AC/AC Converter Systems” in proceedings of 13th Power
Electronics and Motion Control Conference EPE-PEMC 08, Sept. 2008
pp. 27-42.
[10] H. Cui, C. Chen, W. Jiang, L. Hou, “Direct Torque Control Method for
Induction Machines using Three-Level Space Vector Modulation”, in
proceedings of 6th International Power Electronics and Motion
Control Conference IPEMC 2009, Wuhan China.
[11] Z. Yingehao, Z. Zhengming, L. Ting, L. Long, “Direct Power and
Torque Control for Three-Level NPC based PWM AC/DC/AC
Converter”, in proceedings of 6th International Power Electronics and
Motion Control Conference IPEMC 2009, Wuhan China.
[12] M. Morari, J. H Lee “Model Predictive Control: Past, Present and
Future”, Computers and Chemical Engineering, Vol 23, no 4-5, pp.
667- 682, 1999.
[13] A. Linder, R. Kennel, “Model Predictive Control for Electrical Drives”,
in proceedings of the 36th IEEE 2005 Power Electronics Specialists
Conference, PESC '05, pp.1793 – 1799.
[14] R. Kennel, A. Linder, “Direct Model Predictive Control - a New Direct
Predictive Control Strategy for Electrical Drives”, in proceedings of
the, 2005 European Conference on Power Electronics and
Applications, 11-14 Sept. 2005, pp. 10.
[15] S. Muller; U. Ammann, S. Rees; “New Time-Discrete Modulation
Scheme for Matrix Converters”, IEEE Transactions on Industrial
Electronics, Vol. 52, Issue 6, Dec. 2005, pp. 1607 – 1615.
[16] J. Rodriguez, J. Pontt, C.A. Silva, P. Correa, P. Lezana, P. Cortes, U.
Ammann, “Predictive Current Control of a Voltage Source Inverter”
IEEE Transactions on Industrial Electronics, Vol. 54, Issue 1, Feb.
2007 Page(s):495 – 503.
[17] P. Correa, M. Pacas, J. Rodriguez, “Predictive Torque Control for
Inverter-Fed Induction Machine”, IEEE Transactions on Industrial
Electronics, Vol. 54, Issue 2, April. 2007, pp. 1073 – 1079.
[18] M. Rivera, J. Espinoza, R. Vargas, J. Rodriguez, “Behavior of the
Predictive DTC based Matrix Converter under Unbalanced AC Supply”
in Conf. Rec. IAS’07, New Orleans, LA, USA, Sept. 23 – 27, 2007.
[19] P. Cortes, J. Rodriguez, D. E. Quevedo, C. Silva, “Predictive Current
Control Strategy with Imposed Load Current Spectrum”, IEEE

3183