IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO.

6, JUNE 2009 1847

Predictive Control of an Indirect Matrix Converter

Pablo Correa, Member, IEEE, José Rodríguez, Senior Member, IEEE,
Marco Rivera, Jose R. Espinoza, Member, IEEE, and Johann W. Kolar, Senior Member, IEEE

Abstract—This paper presents the implementation of a predic- path from input to output produces higher power losses. This
tive control scheme for an indirect matrix converter. The control can be partly mitigated by the use of semiconductors such as
scheme selects the switching state that minimizes the reactive reverse blocking insulated-gate bipolar transistors (RB-IGBTs)
power and the error in the output currents according to their ref-
erence values. This is accomplished by using a prediction horizon [4], which have already been used in conventional matrix
of one sample time and a very intuitive control law. Experimental converters [5], [6]. Additionally, some studies have focused
results with a 6.8-kVA indirect matrix converter prototype are on sparse matrix topologies, which can improve power density
provided in order to validate the proposed control scheme. The by reducing the number of semiconductors at the expense of
converter uses standard digital signal processor operating at a functionality. This is the case of the sparse matrix [7], [8],
sampling frequency of 20 µs. It is shown that the idea of con-
trolling this converter topology with a predictive approach can be the very sparse matrix [9], and the ultrasparse matrix converter
implemented simply and input currents with unity power factor [10], which feature 15, 12, and 9 switches, respectively.
and a total harmonic distortion lower than 5% can be obtained. Indirect matrix converters use complex pulsewidth modula-
Index Terms—AC motor drives, current control, predictive tion (PWM) schemes to achieve the goal of unity power factor
control. and sinusoidal output current. However, since power converters
have a discrete nature, the application of predictive control
I. I NTRODUCTION constitutes a promising and better suited approach as compared
with standard schemes that use mean values of the variables

M ATRIX converters feature several advantages compared

with standard two-level converters, such as a bidirec-
tional power flow and the reduced size of the reactive compo-
[11]. Furthermore, predictive control utilizes a very intuitive
control law that can easily deal with multivariable cases and the
treatment of constraints and compensate for the dead time [12].
nents [1]. Different kinds of converters without dc link have Early practical applications of predictive control can be
been presented in the technical literature. They are classified found in the 1980s for the case of two-level inverters [13].
into three main groups: 1) the cycloconverters in the high-power Nowadays, it is possible to find applications in machine con-
range; 2) the standard matrix converters; and 3) indirect matrix trol [14], [15], active rectifiers [16], [17], matrix converters
converters [2], all of them in the low-power range, as it is shown [18], [19], and even multilevel converters [20], [21]. Among
in Fig. 1. The conventional indirect matrix converter is similar the broad family of predictive control methods, those with a
to a back-to-back inverter but includes bidirectional switches in
prediction horizon of one sample time are particularly attractive
the rectifier and has no dc-link capacitor, as shown in Fig. 2.
due to the simplicity of implementation. More advanced control
References to this topology can be traced back to 20 years ago,
techniques with a prediction horizon greater than one have
when the concept was first introduced using a converter with
been proposed for simple converter topologies such as dc–dc
gate turnoff (GTO) thyristors [3].
converters and two-level inverters [22], [23]. These types of
Over the last few years, research on indirect matrix convert-
schemes offer some advantages in terms of faster dynamics
ers has benefited from advances in semiconductor technologies,
and reduced harmonic distortion at the expense of an increased
which have mainly contributed toward enhancing efficiency.
calculation burden. In order to solve this problem, some meth-
Compared with the standard matrix converter, this topology can
ods have been proposed to move part of the calculations’
use a simpler modulation scheme and does not need an addi-
effort offline [23]. Unfortunately, matrix converters operate at
tional overvoltage protection system. Conversely, the current
a very fast sampling rate and feature a high number of possible
switching states.
Manuscript received April 15, 2008; revised January 5, 2009. First published This paper presents an application of predictive control in
February 3, 2009; current version published June 3, 2009. This work was
supported in part by the Chilean Research Fund CONICYT under Grant
indirect matrix converters. The switching state is selected by
1080059, in part by the Industrial Electronics and Mechatronics Millennium minimizing a quality function that considers the instantaneous
Science Nucleus, and in part by the Universidad Técnica Federico Santa María. reactive power in the input, the current error in the output, and
P. Correa, J. Rodríguez, and M. Rivera are with the Departamento de
Electrónica, Universidad Técnica Federico Santa María, Valparaíso 110-V,
the generation of a positive voltage in the dc link. Feasibility,
Chile (e-mail: [email protected]; [email protected]; [email protected]). implementation details, and advantages and disadvantages are
J. R. Espinoza is with the Departamento de Ingenieria Eléctrica, Universidad also discussed. Due to the limited capacity of the controller,
de Concepción, Concepción 160-C, Chile (e-mail: [email protected]).
J. W. Kolar is with the Power Electronic Systems Laboratory (PES), Swiss the use of a prediction horizon of one sample time will be a
Federal Institute of Technology (ETH) Zurich, CH-8092 Zurich, Switzerland design criterion. This paper is organized in the following man-
(e-mail: [email protected]). ner. Section II presents the converter topology. In Sections III
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. and IV, the control scheme is explained in detail. Section V
Digital Object Identifier 10.1109/TIE.2009.2013686 describes the commutation sequence for the rectifier and the

0278-0046/$25.00 © 2009 IEEE

1848 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

Fig. 1. Classification of power converters.

Fig. 2. Simplified scheme of the indirect matrix converter.

inverter. Section VI describes the selection of the weight factors represented by the well-known 2-D space vector. For example,
for the predictive controller. Finally, the experimental setup and the phase current components isu , isv , and isw will be described
results are explained, showing the feasibility of the proposed by the complex space vector
method in Section VII.
is = isα + j · isβ (1)
II. P OWER C IRCUIT OF THE I NDIRECT which is defined as
M ATRIX C ONVERTER
isα = 31 (2isu − isv − isw )

The indirect matrix converter consists of an array of power . (2)
semiconductors that is very similar to the ac/dc/ac back-to- isβ = √13 (isv − isw )
back converter, as shown in Fig. 2. The converter synthesizes
This space vector is referred to a stationary reference frame that
a positive voltage in the dc link by selecting a switching state in
will be denoted as an αβ reference frame.
the rectifier that connects one phase to point P and the other
Predictive control aims to select the converter switching state
phase to point N . Additionally, the rectifier includes an LC
that leads the controlled variables closest to their respective
filter in the input which is needed to provide a path for the phase
references at the end of the sampling period. In order to
current which is momentarily not connected to the dc link.
meet this requirement, the load current and input voltages are
It should be noted that the indirect matrix converter topology
measured, and predicted values of the input and output currents
includes as many switches as the standard matrix converter, but
are generated for each possible switching state.
the former features an extra freedom degree that alleviates the
Three main conditions must be fulfilled for the converter
complexity of the commutation sequence.
to properly operate. First, the line side of the rectifier must
An indirect matrix converter with RB-IGBTs is considered in
deliver active power. Second, the load current must follow the
this paper. The conduction losses of this kind of semiconductor
reference with high accuracy, and third, the dc-link voltage
are lower than a bidirectional switch consisting of a standard
must be positive. This last condition differentiates the proposed
IGBT and a diode connected in series [7], [8].
control law from the method presented in [18]. The first condi-
tion is accomplished by minimizing the predicted value of the
III. P REDICTIVE C ONTROL M ETHOD instantaneous reactive power
In the following, it will be assumed that the three phase 
k+1 k+1

q k+1 = uk+1 k+1 
sα · isβ − usβ · isα  . (3)

quantities of the converter are symmetrical and, hence, can be
CORREA et al.: PREDICTIVE CONTROL OF AN INDIRECT MATRIX CONVERTER 1849

On the other hand, the second condition requires a minimum and expressing the system in its discrete form as
error between the predicted load currents and its references
ue k+1 ue k
    k
us
  =Φ +Γ

∆ik+1 = ilα − ik+1  + i∗lβ − ik+1 
∗
(4) (12)
l lα lβ is k+1 is k ie k

where ilα and ilβ denote the load current in αβ coordinates and where
i∗lα and i∗lβ denote their respective references.
The third condition necessary for the operation of the rectifier
Φ = eA·Ts Γ = A−1 (Φ − I)B (13)
is to discard those states that produce negative dc-link voltages.
In order to fulfill this condition, the following function is
defined: with

uk+1
    
k+1 0, dc > 0 0 1/Cf 0 −1/Cf
h = (5) A= B= . (14)
M, uk+1 dc ≤ 0
−1/Lf Rf 1/Lf 0

where M is the maximum positive number that can be gener- The load model is obtained similarly. Assuming an
ated by the arithmetic unit of the controller. Equations (3)–(5) inductive–resistive load as shown in Fig. 2, after representing
are merged into a single so-called quality function the three-phase system in α−β coordinates, the following
equation describes the behavior of the load:
g k+1 = q k+1 + A · ∆ik+1
l + hk+1 . (6)

The control method operates as follows. At each sampling dil

Ll = ul − Rl · il (15)
time, all possible switching states are used to calculate (6). dt
The switching state that produces the minimum value of g is
selected to be applied for one sampling period. which is discretized as follows:

ik+1
l = d1 ukl + d2 ikl (16)
IV. C ALCULATION OF P REDICTED V ALUES
A mathematical model of the input filter and the load pro- where d1 and d2 depend on L1 , R1 , and the sample time.
vides the basis for the prediction of the values of the input
and output currents, which are needed for evaluating the quality
function. V. D EAD -T IME C OMPENSATION
The line side of the rectifier consists of a second-order system
The previously explained method does not deal with the dead
described by
time generated by the calculations. In fact, the inverter and
dis the rectifier can generate eight and nine valid switching states,
Lf = us − ue − Rf · is (7) respectively, which, altogether, produce 72 possible switching
dt
combinations that must be taken into account in the quality
due
Cf = is − ie (8) function g. A simplification is possible if h is a priori evaluated
dt using each of the nine possible rectifier states and the present
where Lf comprises the mains and filter inductances and Rf voltage values of the capacitor filter. In this way, three of
represents the mains and filter damping resistances. The pre- the nine possible rectifier states are selected beforehand, and
diction of the input current and capacitor voltages are computed just 24 switching combinations have to be taken into account
from a first-order difference equation, as is described in [18] for the evaluation of g. The dead time associated to these
calculations cannot be neglected if a standard low-cost digital
ik+1
s = c1i · uks + c2i · uke + c3i · iks + c4i · ike (9) signal processor (DSP) controller is used.
A simple way to consider the dead time is by calculating
uk+1
s = c1u · uks + c2u · uke + c3u · iks + c4u · ike . (10) the quality function at the end of the next sampling period,
i.e., g k+2 . Thus, the selected switching state can be applied at
The real coefficients c1 , c2 , c3 , and c4 are defined so that the tk+1 , and a period of time equivalent to one sampling period
obtained values for the predicted currents correspond to those is available for calculations. The aforementioned compensation
of the continuous-time system after one sampling time. They requires the calculation of uk+1 , uk+1 , ik+1 , and ik+1 in order
s e s l
can be calculated by representing (7) and (8) by a state-space to have the basis for the calculation of g k+2
. These terms are
system with state variables is and ue obtained out of (9), (10), (16), and the current switching state
  
u̇e
  S k . Considering that the change in the input voltage is small
0 1/Cf ue
= in one sampling time, uk+1 s will be considered equal to uks .
i̇s −1/Lf Rf is
   A block diagram of the control scheme and the sequence of
0 −1/Cf us events for the dead-time compensation are shown in Fig. 3(a)
+ (11)
1/Lf 0 ie and (b), respectively.
1850 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

Fig. 3. Predictive controller. (a) Block diagram. (b) Time diagram of the algorithm.

VI. C OMMUTATION S EQUENCE OF THE

P OWER S EMICONDUCTORS
The converter requires a commutation sequence that allows
a safe change of the rectifier switching state. Basically, this
problem can be addressed by synchronizing the state changes
in the rectifier with the application of a zero-voltage space
vector in the inverter stage. Under this condition, no current
circulates through the dc link, and the rectifier state can be
changed without help of auxiliary commutation circuits [3].
An exact determination of the switching frequency with the
proposed scheme is not possible, considering the nondetermin-
istic generation of switching pulses. In order to estimate the
mean switching frequency of the switches of the inverter, the
following assumptions were considered.
1) The rectifier stage changes at each sampling time.
2) If the inverter stage changes between two active vectors,
i.e., those inverter stages that deliver to load a voltage
different than zero, one extra switching in one phase is
needed to provide the safe commutation of the rectifier.
This is the case in the time instant tk+3 in Fig. 4(a).
3) If the inverter stage has a zero space vector state or the last
state was a zero vector, no extra switchings are necessary
for the safe commutation of the rectifier. This is the case
Fig. 4. Switching behavior of the inverter stage. (a) Commutation sequence
in the time instants tk+1 ytk+2 . of the inverter and extra switching transitions. (b) Mean switching frequency of
The characteristic curve shown in Fig. 4(b) depicts the mean one of the IGBTs of the inverter stage for a sampling time of 20 µs.
switching frequency f¯igbt_o of one of the IGBTs of the inverter
stage, obtained by counting the number of switching transitions to the control of the output currents. This factor was adjusted
in several periods of the fundamental frequency. Other switches empirically based on [24]. First, A is set to a high value in
present the same behavior, which is why the curve correspond- order to prioritize the control of the output current. As such, the
ing to one semiconductor is shown. It can be observed that the inverter can control the output currents while the input currents
mean switching frequency is lower than the sampling frequency will be highly distorted. After that, factor A is slowly decreased,
and it decreases at a higher load current. thereby lending more importance to the control of the reactive
power. The output currents will not follow the reference if the
factor A is too low, as shown in Fig. 5(a). Therefore, A is
VII. S ELECTION OF THE W EIGHT F ACTORS
selected as the minimum value such that the output current
The weighting factor A of the quality function (6) decides if has no noticeable deviations with respect to the reference, as
the priority will be given to the control of the power factor or it is shown in Fig. 5(b). There is no specific design criterion
CORREA et al.: PREDICTIVE CONTROL OF AN INDIRECT MATRIX CONVERTER 1851

Fig. 6. Photograph of the indirect matrix converter.

Fig. 5. Selection of the factor A. (a) Output current A = 0.15, 95% of the
maximum output current. (b) Output current A = 0.45, 95% of the maximum
output current. (c) Factor A versus the load current.

to define the maximum allowable error in the output current,

except that the maximum discrepancy must be below 5% of the
maximum value of the reference. Since the adjustment of the
parameter A depends on the operating point, different values
are obtained using this empiric method, as shown in Fig. 5(c).

VIII. R ESULTS
A conventional indirect matrix converter, built by the Power
Electronics Systems Laboratory of the Eidgenössische Technis-
che Hochschule (ETH) Zürich, was used for the experimental
evaluation. The converter features reverse blocking IGBTs of Fig. 7. Voltages and currents of the converter. (a) Input voltage usU and input
type IXRH40N120 for the bidirectional switch, standard IGBTs current isU . (b) Output current with its reference.
with antiparallel diodes of type FII50-12E for the inverter
stage. A picture of the converter is shown in Fig. 6. The time could be considered high for a low-power converter, better
control scheme was implemented in a 160-MIPS fixed-point results can be expected from future generations of IGBTs.
ADSP21991 DSP board which is also shown in Fig. 6. The Fig. 7 shows the measured input current and voltage and
processor board is connected to additional stacked boards that output currents of phase U according to the parameters of the
include a field-programmable gate array for the commutation Appendix. As expected, the input current fulfils the condition
sequence generation and the signal conditioning for the mea- of unitary power factor and presents an approximated total
surement of voltages and currents. The input filter is integrated harmonic distortion (THD) of 3.5%. As it is shown in Fig. 7,
in the converter board, as well as the current and voltage the input current shows a ripple corresponding to the resonance
transducers. The original setup includes filter parameters of frequency of the input filter. On the other hand, the output
Lf = 130 µH and Cf = 10 µF. In this case, however, the ca- currents follow the reference accurately, with a THD lower than
pacity of the filter was increased to Cf = 40 µF to improve the 1%. For the THD calculations, harmonics up to the sampling
stability of the input current control. frequency were considered.
The sampling period of the control algorithm was set at Ts = It is not possible to directly compare the predictive control
20 µs. It should be noted that the dead time necessary for a safe scheme’s performance with other methods that use a constant
switching transition of the rectifier is approximately 10% of the switching frequency. Results in [9] for a PWM-based control
aforementioned sampling period. Despite the fact that this dead scheme operating at 25 kHz indicate that the proposed method
1852 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

performs similarly in terms of harmonic distortion. Better re- [6] C. J. Kaufman, private communication, May 1995.
sults can be expected by optimizing the input filter and adding [7] T. Friedli, S. Round, and S. J. W. Kolar, “A 100 kHz SiC sparse matrix
converter,” in Proc. IEEE PESC, Jun. 17–21, 2007, pp. 2148–2154.
active damping. [8] M. Young, The Technical Writer’s Handbook. Mill Valley, CA: Univ.
It must be acknowledged that the main advantage of the pro- Sci., 1989.
posed control method is the simplicity in the implementation, [9] F. Schafmeister, S. Herold, and J. W. Kolar, “Evaluation of
1200 V-Si-IGBTs and 1300 V-SiC-JFETs for application in three-
since the controller does not need a complex modulation unit. phase very sparse matrix AC–AC converter systems,” in Proc. 18th Annu.
This can reduce the overall cost of the complete system. The IEEE APEC, Feb. 9–13, 2003, vol. 1, pp. 241–255.
method also presents drawbacks as the quality function (6) is [10] J. Schonberger, T. Friedli, S. D. Round, and J. W. Kolar, “An ultra
sparse matrix converter with a novel active clamp circuit,” in Proc. PCC,
explicitly solved for each switching state. This can be a problem Nagoya, Japan, Apr. 2–5, 2007, pp. 784–791.
if a slow controller is used, as a higher sampling time could [11] P. Cortes, M. Kazmierkowski, R. Kennel, D. Quevedo, and J. Rodriguez,
increase the harmonic distortion in the currents. “Predictive control in power electronics and drives,” IEEE Trans. Ind.
Electron., vol. 55, no. 12, pp. 4312–4324, Dec. 2008.
[12] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, P. Lezana, P. Cortes, and
U. Ammann, “Predictive current control of a voltage source inverter,”
IX. C ONCLUSION IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 495–503, Feb. 2007.
[13] R. Würslin, “Pulsumrichtergespeiste Synchronmaschinen Antrieb mit ho-
As a result of the advances in power semiconductors and her Taktfrequenz und sehr großem Feldschwächbereich,” Ph.D. disserta-
processors, indirect matrix converter and predictive control tion, Universität Stuttgart, Stuttgart, Germany, 1984.
schemes have recently emerged as feasible approaches. This [14] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. Silva, “Predictive current
control strategy with imposed load current spectrum,” IEEE Trans. Power
paper takes advantage of these advances and proposes a predic- Electron., vol. 23, no. 2, pp. 612–661, Mar. 2008.
tive control scheme for an indirect matrix converter. The control [15] G. Abad, M. A. Rodriguez, and J. Poza, “Two-level VSC based predictive
scheme uses the predicted values of the input and output cur- direct torque control of the doubly fed induction machine with reduced
torque and flux ripples at low constant switching frequency,” IEEE Trans.
rents to evaluate the best suited converter state considering the Power Electron., vol. 23, no. 3, pp. 1050–1061, May 2008.
output current error, the input power factor, and the generation [16] S. A Larrinaga, M. A. R. Vidal, E. Oyarbide, and J. R. T. Apraiz, “Predic-
of a positive dc-link voltage. Experimental results at a sampling tive control strategy for DC/AC converters based on direct power control,”
IEEE Trans. Ind. Electron., vol. 54, no. 3, pp. 1261–1271, Jun. 2007.
time of Ts = 20 µs show that the unity power factor is obtained [17] P. Antoniewicz, M. P. Kazmierkowski, S. Aurtenechea, and
with input currents with THD of 3.5%. The output current M. A. Rodriguez, “Comparative study of two predictive direct power
follows the reference very accurately. The main advantage of control algorithms for three-phase AC/DC converters,” in Proc. Eur.
Conf. Power Electron. Appl., Sep. 2–5, 2007, pp. 1–10.
the method is that, with a very simple control approach, results [18] S. Muller, U. Ammann, and S. Rees, “New time-discrete modulation
comparable with other PWM-based schemes can be obtained. scheme for matrix converters,” IEEE Trans. Ind. Electron., vol. 52, no. 6,
pp. 1607–1615, Dec. 2005.
[19] R. Vargas, U. Ammann, J. Rodriguez, and J. Pontt, “Predictive strategy to
A PPENDIX control common-mode voltage in loads fed by matrix converters,” IEEE
Trans. Ind. Electron., vol. 55, no. 12, pp. 4372–4380, Dec. 2008.
The parameters of the input filter and the load are as follows. [20] P. Correa, M. Pacas, and J. Rodriguez, “Predictive torque control for
inverter-fed induction machines,” IEEE Trans. Ind. Electron., vol. 54,
no. 2, pp. 1073–1079, Apr. 2007.
Mains Filter: Load: [21] R. Vargas, P. Cortes, U. Ammann, J. Rodriguez, and J. Pontt, “Predictive
Lf = 130 µH R1 = 20 Ω control of a three-phase neutral-point-clamped inverter,” IEEE Trans. Ind.
Electron., vol. 54, no. 5, pp. 2697–2705, Oct. 2007.
Cf = 40 µF L1 = 10 mH [22] R. Kennel and A. Linder, “Model predictive control for electrical drives,”
Rf = 0.2 Ω in Proc. 36th IEEE Power Electron. Spec. Conf., 2005, pp. 1793–1799.
[23] R. Kennel and A. Linder, “Direct model predictive control—A new direct
predictive control strategy for electrical drives,” in Proc. Eur. Conf. Power
Electron. Appl., Sep. 11–14, 2005, p. 10.
ACKNOWLEDGMENT [24] S. Müller, “Neuartiges Steuerverfahren Für Einen Matrixconverter
Mit Sehr Kleinem Netzfilter,” Ph.D. dissertation, Universität Stuttgart,
The authors would like to thank Prof. J. Kolar for providing Stuttgart, Germany, 2003.
the laboratory facilities at the ETH Zürich and Dr. S. Round for
the interesting technical discussions.

R EFERENCES
[1] P. W. Wheeler, J. Rodriguez, J. C. Clare, L. Empringham, and
A. Weinstein, “Matrix converters: A technology review,” IEEE Trans. Ind.
Electron., vol. 49, no. 2, pp. 276–288, Apr. 2002. Pablo Correa (M’07) received the Electrical Engi-
[2] J. W. Kolar, F. Schafmeister, S. D. Round, and H. Ertl, “Novel three-phase neering degree and the M.Sc. degree in electrical
AC–AC sparse matrix converters,” IEEE Trans. Power Electron., vol. 22, engineering from the Universidad Técnica Federico
no. 5, pp. 1649–1661, Sep. 2007. Santa María, Valparaíso, Chile, in 2001, and the
[3] J. Holtz and U. Boelkens, “Direct frequency convertor with sinusoidal Dr.Ing. degree from the University of Siegen, Siegen,
line currents for speed-variable AC motors,” IEEE Trans. Ind. Electron., Germany, with a scholarship awarded in 2002 by the
vol. 36, no. 4, pp. 475–479, Nov. 1989. German Academic Exchange Service for doctoral
[4] T. Friedli, M. L. Heldwein, F. Giezendanner, and J. W. Kolar, “A high studies, in 2006.
efficiency indirect matrix converter utilizing RB-IGBTs,” in Proc. 37th He is currently with the Power Electronics Group,
IEEE PESC, Jun. 18–22, 2006, pp. 1–7. Departamento de Electrónica, Universidad Técnica
[5] K. Sun, D. Zhou, H. Lipei, K. Matsuse, and K. Sasagawa, “A novel Federico Santa María. His research interests in-
commutation method of matrix converter fed induction motor drive us- clude digital control for high-power drives and renewable energy converters
ing RB-IGBT,” IEEE Trans. Ind. Appl., vol. 43, no. 3, pp. 777–786, and the development of high-performance control platforms based on field-
May/Jun. 2007. programmable gate arrays.
CORREA et al.: PREDICTIVE CONTROL OF AN INDIRECT MATRIX CONVERTER 1853

José Rodríguez (M’81–SM’94) received the Engi- Jose R. Espinoza (S’92–M’97) was born in
neer degree in electrical engineering from the Uni- Concepción, Chile, in 1965. He received the Eng. de-
versidad Técnica Federico Santa María, Valparaíso, gree in electronic engineering and the M.Sc. degree
Chile, in 1977, and the Dr. Ing. degree in electri- in electrical engineering from the Universidad de
cal engineering from the University of Erlangen, Concepción, Concepción, in 1989 and 1992, respec-
Erlangen, Germany, in 1985. tively, and the Ph.D. degree in electrical engineering
Since 1977, he has been with the Universidad from Concordia University, Montreal, QC, Canada,
Técnica Federico Santa María, where he is currently in 1997.
the President and a Professor with the Departa- Since January 2006, he has been a Professor with
mento de Electrónica. During his sabbatical leave the Departamento de Ingenieria Eléctrica, Universi-
in 1996, he was responsible for the mining division dad de Concepción, where he is engaged in teaching
of the Siemens Corporation, Chile. He has extensive consulting experience and research in the areas of automatic control and power electronics. He
in the mining industry, particularly in the application of large drives, such has authored or coauthored more than 100 refereed journal and conference
as cycloconverter-fed synchronous motors for semiautogenous grinding mills, papers and contributed to one chapter in the Power Electronics Handbook
high-power conveyors, controlled drives for shovels, and power quality issues. (Academic, 2007).
His research interests include power electronics and electrical drives. In the
last several years, his main research interests have included multilevel inverters
and new converter topologies. He has authored or coauthored more than Johann W. Kolar (M’89–SM’04) received the Ph.D.
250 refereed journal and conference papers and contributed to one chapter in degree (summa cum laude/promotio sub auspiciis
the Power Electronics Handbook (Academic, 2006). praesidentis rei publicae) in industrial engineering
Dr. Rodríguez is an Associate Editor of the IEEE TRANSACTIONS from the University of Technology, Vienna, Austria,
ON I NDUSTRIAL E LECTRONICS and IEEE T RANSACTIONS ON P OWER in 1998.
ELECTRONICS. He received the Best Paper Award in 2007 from the IEEE Since 1984, he has been an Independent Inter-
TRANSACTIONS ON INDUSTRIAL ELECTRONICS. national Consultant in close collaboration with the
University of Technology, in the fields of power elec-
tronics, industrial electronics, and high-performance
drives. He has proposed numerous novel pulsewidth
modulation (PWM) converter topologies, and mod-
ulation and control concepts, e.g., the VIENNA rectifier and the three-phase
ac–ac sparse matrix converter. On February 1, 2001, he was appointed Professor
and Head of the Power Electronic Systems Laboratory, Swiss Federal Institute
of Technology Eidgenössische Technische Hochschule (ETH) Zurich, Zurich,
Switzerland. During 2003, he was the Erskine Fellow of the University of
Canterbury, New Zealand. He is the author or coauthor of more than 300 sci-
entific papers published in international journals and conference proceedings.
He has filed 75 patents. His current research interests include ac–ac and ac–dc
converter topologies with low effects on the mains, e.g., for power supply
of telecommunication systems, more-electric-aircraft, and distributed power
systems in connection with fuel cells, realization of ultracompact intelligent
converter modules employing latest power semiconductor technology (SiC),
novel concepts for cooling and electromagnetic interference (EMI) filtering,
multidomain/multiscale modeling and simulation, pulsed power, bearingless
motors, and power microelectromechanical systems (MEMS).
Marco Rivera received the Electronics Engineering Dr. Kolar is a member of the Institute of Electrical Engineers of Japan (IEEJ)
degree and the M.Sc. degree in electrical engineering and Technical Program Committees of numerous international conferences
from the Universidad de Concepción, Concepción, (e.g., Director of the Power Quality Branch of the International Conference
Chile, in 2007 and 2008, respectively. He is currently on Power Conversion and Intelligent Motion). He also received an Erskine Fel-
working toward the Ph.D. degree in the Departa- lowship from the University of Canterbury, New Zealand, in 2003. In 2006, the
mento de Electrónica, Universidad Técnica Federico European Power Supplies Manufacturers Association (EPSMA) awarded the
Santa María, Valparaíso, Chile, with a scholarship Power Electronics Systems Laboratory of ETH Zurich as the leading academic
from the Chilean Research Fund CONICYT. research institution in Europe. He was the recipient of the Best Transactions
His research interests include direct and indirect Paper Award of the IEEE Industrial Electronics Society in 2005. From 1997
matrix converters, predictive and digital controls through 2000, he was an Associate Editor of the IEEE TRANSACTIONS ON
for high-power drives, and development of high- INDUSTRIAL ELECTRONICS, and since 2001, as an Associate Editor of the
performance control platforms based on field-programmable gate arrays. IEEE TRANSACTIONS ON POWER ELECTRONICS. Since 2002, he has been an
Mr. Rivera was awarded a scholarship from the Marie Curie Host Fellow- Associate Editor of the Journal of Power Electronics of the Korean Institute of
ships for Early Stage Research Training in Electrical Energy Conversion and Power Electronics and a member of the Editorial Advisory Board of the IEEJ
Conditioning Technology at University College Cork, Cork, Ireland. Transactions on Electrical and Electronic Engineering.