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

6, JUNE 2009 1875

Model Predictive Control of an Inverter With Output

LC Filter for UPS Applications
Patricio Cortés, Member, IEEE, Gabriel Ortiz, Juan I. Yuz, Member, IEEE, José Rodríguez, Senior Member, IEEE,
Sergio Vazquez, Member, IEEE, and Leopoldo G. Franquelo, Fellow, IEEE

Abstract—The use of an inverter with an output LC filter allows bank resonant filters [10], [11], and repetitive-based controllers
for generation of output sinusoidal voltages with low harmonic [12], [13]. Some tuning strategies have been presented, includ-
distortion, suitable for uninterruptible power supply systems. ing H-infinity control design [14]. In most of these schemes, the
However, the controller design becomes more difficult. This paper
presents a new and simple control scheme using predictive control output voltage and one of two currents are used by a cascaded
for a two-level converter. The controller uses the model of the control considering outer and inner control loops, with linear
system to predict, on each sampling interval, the behavior of the or nonlinear controllers, to generate the reference voltages and
output voltage for each possible switching state. Then, a cost by an external modulator to generate the firing pulses for the
function is used as a criterion for selecting the switching state power semiconductors.
that will be applied during the next sampling interval. In addi-
tion, an observer is used for load-current estimation, enhancing Predictive control appears as an attractive alternative for the
the behavior of the proposed controller without increasing the control of power converters due to its fast dynamic response
number of current sensors. Experimental results under linear and [15]. Several control algorithms have been presented under the
nonlinear load conditions, with a 5.5-kW prototype, are presented, name of predictive control, as presented in [16]. The most well-
verifying the feasibility and good performance of the proposed known scheme is deadbeat control, and it has been applied
control scheme.
to current control in inverters [17], [18], rectifiers [19], active
Index Terms—Power conversion, predictive control, uninter- filters [20], [21], power-factor preregulators [22], and UPSs [6],
ruptible power systems. [23]. When implemented in a digital system, deadbeat control
needs to be modified in order to improve robustness. Several
I. I NTRODUCTION modifications have been proposed in the last years [24]–[27]
considering adaptive schemes, neural networks, and other
T HE CONTROL of inverters with an output LC filter
has a special importance in applications where a high-
quality voltage is needed. Such applications include distributed
changes that make the controller robust but more complex.
Another approach is the model predictive control (MPC), also
known as receding-horizon control; it uses a model of the
generation, energy-storage systems, stand-alone applications
system to predict the behavior of the variables until a certain
based on renewable energy, and uninterruptible power supplies
horizon of time, and a cost function is used as criterion to select
(UPSs) [1], [2]. In these systems, it is desired, particularly for
the optimal future actions [28]–[31]. Several works have intro-
stand-alone applications and UPS systems [3], to achieve a
duced the models for power converters [32], [33], which can be
good output-voltage regulation with any kind of load, being
used to develop new model predictive controllers, improving
very important that the functionality of the system does not
the performance of the system. MPC is a very flexible control
deteriorate under nonlinear loads, such as diode rectifiers.
scheme that allows the easy inclusion of system constraints
The inclusion of an LC filter at the output of the inverter
and nonlinearities in the design stage of the controller. In
makes more difficult the controller design and controller pa-
MPC, different formulations of the cost function are possible,
rameters’ adjustment. Several control schemes have been pro-
considering different norms and including several variables and
posed for this converter, including deadbeat control [4]–[6],
weighting factors [15]. It is also possible to consider different
multiloop feedback control [7]–[9], adaptive control based on
prediction horizons, as shown in [34]. The inputs of the system
can be considered continuous, by using a modulator to apply
Manuscript received April 16, 2008; revised August 15, 2008 and the optimal voltages, as presented in [28] and [35]. In order
December 29, 2008. First published February 27, 2009; current version pub-
lished June 3, 2009. This work was supported in part by the Chilean National
to simplify the implementation of MPC, the converter can be
Fund of Scientific and Technological Development (FONDECYT) under Grant modeled as a system with a finite number of switching states,
1080443 and in part by the Universidad Técnica Federico Santa María. and only one time step horizon can be considered for the
P. Cortés, J. I. Yuz, and J. Rodríguez are with the Electronics Engineering
Department, Universidad Técnica Federico Santa María, Valparaíso 110-V,
optimization, as presented for the current control in a matrix
Chile (e-mail: patricio.cortes@usm.cl). converter in [36] and [37], a three-phase inverter in [38] and
G. Ortiz is with the Power Electronic Systems Laboratory, Eidgenoessische [39], an active front-end rectifier in [40], a multilevel inverter
Technische Hochschule Zürich, 8092 Zürich, Switzerland.
S. Vazquez and L. G. Franquelo are with the Department of Electronic in [41], and a flying capacitor converter in [42]. This way, all
Engineering, University of Seville, 41092 Seville, Spain (e-mail: svazquez@ possible switching states can be evaluated online; then, the one
zipi.us.es). that minimizes the cost function is selected.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. This paper proposes a new and simple MPC scheme for a
Digital Object Identifier 10.1109/TIE.2009.2015750 three-phase inverter with output LC filter. The controller uses

0278-0046/$25.00 © 2009 IEEE

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
1876 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

Fig. 1. Three-phase inverter with output LC filter. Fig. 2. Possible voltage vectors generated by the inverter.

a model of the system to predict, on each sampling interval, the

behavior of the output voltage for each possible switching state,
and then, a cost function is used as a criterion for selecting the
switching state that will be applied during the next sampling
interval. There is no need of internal current-control loops and
no modulators; the gate-drive signals are generated directly by
the control. In addition, an observer is used for load-current Fig. 3. LC filter model.
estimation, enhancing the behavior of the proposed controller
without increasing the number of current sensors. eight voltage vectors are obtained. Note that v0 = v7 , resulting
in only seven different voltage vectors, as shown in Fig. 2.
Using modulation techniques like pulsewidth modulation,
II. C ONVERTER M ODEL the inverter can be modeled as a continuous system. Never-
The three-phase inverter with output LC filter considered in theless, in this paper, the inverter is considered as a nonlinear
this paper is shown in Fig. 1. The converter and filter models discrete system with only seven different voltage vectors as
are presented here, and the load is assumed unknown. possible outputs.
The switching states of the converter are determined by the Using vectorial notation, the filter current if , the output
gating signals Sa , Sb , and Sc as follows: voltage vc , and the output current io can be expressed as space
vectors and are defined as

1, if S1 on and S4 off 2
Sa = (1)
0, if S1 off and S4 on if = (if a + aif b + a2 if c ) (7)
 3
1, if S2 on and S5 off 2
Sb =
0, if S2 off and S5 on
(2) vc = (vca + avcb + a2 vcc ) (8)
3
 2
1, if S3 on and S6 off io = (ioa + aiob + a2 ioc ). (9)
Sc = (3) 3
0, if S3 off and S6 on

and can be expressed in vectorial form by The LC filter is modeled as shown in the block diagram in
Fig. 3. This model can be described by two equations, one that
2  describes the inductance dynamics and the other describing the
S= Sa + aSb + a2 Sc (4)
3 capacitor dynamics.
The equation of the filter inductance expressed in vectorial
where a = ej(2π/3) . form is
The output-voltage space vectors generated by the inverter
are defined by dif
L = vi − vc (10)
dt
2 
vi = vaN + avbN + a2 vcN (5) where L is the filter inductance.
3
The dynamic behavior of the output voltage can be expressed
where vaN , vbN , and vcN are the phase voltages of the inverter, by the following:
with respect to the negative terminal of the dc-link N (see
Fig. 1). Then, the load voltage vector vi can be related to the dvc
switching state vector S by C = if − io (11)
dt
vi = Vdc S (6) where C is the filter capacitance.
These equations can be rewritten as a state-space system as
where Vdc is the dc-link voltage.
Considering all the possible combinations of the gating sig- dx
= Ax + Bvi + Bd io (12)
nals Sa , Sb , and Sc , eight switching states and, consequently, dt

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
CORTÉS et al.: MPC OF AN INVERTER WITH OUTPUT LC FILTER FOR UPS APPLICATIONS 1877

where considerations consist of assuming a certain dynamic behavior

  of the load current.
if
x= (13) A simple consideration is to assume that the load current
vc
is changing very slowly, compared to the sampling frequency.
 
0 −1/L Lets assume that the load current can be approximated as a con-
A= (14) stant, so its behavior is described by the following differential
1/C 0
  equation:
1/L
B=
0 dio
  = 0. (22)
0 dt
Bd = . (15)
−1/C Then, including this load-current model in the filter model,
the system is described by the following state-space equations:
Variables if and vc are measured, while vi can be calculated
using (6), and io is considered as an unknown disturbance. In ⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ 1 ⎤
i 0 − L1 0 if
d ⎣ f⎦ ⎣1 L
this paper, the value of Vdc is assumed fixed and known. vc = C 0 − C1 ⎦ ⎣ vc ⎦ + ⎣ 0 ⎦ vi . (23)
The output of the system is the output voltage vc and written dt
io 0 0 0 io 0
as a state equation         
x A B
vc = [0 1]x. (16)
The output of this system are the two measured variables,
the filter current and the output voltage, and is defined by the
following equation:
A. Discrete-Time Model of the Filter
⎡ ⎤
A discrete-time model of the filter is obtained from (12) for   if
1 0 0 ⎣ ⎦
a sampling time Ts and is expressed as y= vc . (24)
0 1 0
   o i
x(k + 1) = Aq x(k) + Bq vi (k) + Bdq io (k) (17) C

where A full-order observer for the system can be used to estimate

the state vector x. An observer is an open-loop model of the
Aq = eATs (18) system which includes a correcting term based on the measured
Ts output. This is
Bq = eAτ Bdτ (19) dx̂
0
= Ax̂ + Bvi + J(y − ŷ) (25)
dt
Ts
where ŷ = Cx̂ and J is the so-called observer gain [43].
Bdq = eAτ Bd dτ. (20) This equation can be rewritten as
0
⎡ ⎤
vi
These equations are used as the predictive model in the dx̂
= Aobs x̂ + [B J] ⎣ if ⎦ (26)
proposed predictive controller. dt
vc
In order to predict the output voltage using (17), the output
current io is needed, but usually, this current is not measured, where Aobs = A − JC. The output of the observer is the
and the load is unknown. A simple estimation of the load estimated load current
current can be calculated from filter-current and output-voltage
measurements using the following equation obtained from (11): îo = [0 0 1]x̂. (27)
C
io (k − 1) = if (k − 1) − (vc (k) − vc (k − 1)) . (21) Note that the observer can be understood just as a filter which
Ts
gives an estimate of the (unknown) load current îo , based on
However, this estimation is very sensitive to noise in the mea- measurements of the filter current if , the output voltage vc , and
surements, because it is based on the derivative of the output the inverter voltage vi .
voltage, so it will be preferred to use an observer such as the Matrix gain J will define the observer dynamics. As a design
one presented in the next section. parameter, it must take into account the tradeoff between band-
width and noise rejection. In fact, such gain can be chosen to be
optimal if we characterize the statistical properties of the noise,
III. L OAD -C URRENT O BSERVER
and then, the observer is designed as a steady-state Kalman
The load current depends on the load connected at the output filter [44]. A simpler alternative is to choose the observer gain
of the filter which is unknown. However, some considerations such that the poles of the observer give dynamics several times
can be taken in order to build an appropriate observer. These faster than the open-loop system dynamics.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
1878 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

Fig. 4. Block diagram of the predictive controller.

IV. P ROPOSED P REDICTIVE C ONTROLLER TABLE I

PARAMETERS OF THE EXPERIMENTAL SETUP
The use of MPC to control the power converter is proposed
in this paper. MPC presents several advantages that make it
suitable to control this kind of system: concepts are easy to
understand, constraints and nonlinearities of the system can
be easily included, and the desired behavior of the system
is formulated as a cost function to be minimized. This paper
takes into account an important restriction of the inverter, it can sidered in this function, such as current limitation, switching-
generate only seven different output-voltage vectors, and takes frequency reduction, and spectrum shaping.
advantage of this restriction, making it possible to solve online When implemented in a real system, the time needed for
the optimization problem of MPC. In these control schemes, performing all the calculations of the control algorithm intro-
an open-loop model is used for prediction and selection of the duces a one sampling time delay that has to be compensated as
optimal actuations, but the use of a receding horizon provides explained in [45].
the feedback to the control. This means that only the first
element of the optimal actuation sequence is applied, and all
the optimization is calculated again each sampling time. V. E XPERIMENTAL R ESULTS
The block diagram of the proposed predictive control for a The proposed predictive current-control strategy was tested
three-phase inverter with output LC filter is shown in Fig. 4. experimentally using a Danfoss VLT5008 5.5-kW three-
Here, measurements of the output voltage vc (k) and the filter phase inverter with an output LC filter. The dc-link is fed
current if (k) are used to predict, using (17), the value of the by a three-phase diode-bridge rectifier. The inverter is con-
output voltage at the next sampling instant vc (k + 1) for all the trolled externally through an interface and protection card. A
possible voltage vectors that the inverter generates. TMS320C6713 floating-point digital signal processor (DSP)
Note that in order to obtain the prediction vc (k + 1) from was used for the control. A field-programmable-gate-array-
(17), an estimate of the (unmeasured) output current io (k) is based daughter card handles the analog-to-digital and digital-
required. In fact, a comparison is presented in Section V with to-analog conversions and provides the digital outputs used
results obtained using the simple derivative approximation in as firing signals for power switches of the converter. The
(21) and the load-current observer presented in the previous parameters of the system are shown in Table I.
section. In the implementation of the predictive-control algorithm,
To choose the optimal voltage vector vi to be applied by including the load-current observer, a minimum sampling time
the inverter, the seven predictions obtained for vc (k + 1) are of Ts = 33 μs was achieved. However, further optimization
compared using a cost function g. The voltage vector vi that in the programming is possible, but it is not the subject of
minimizes this function is then chosen and applied at the next this paper.
sampling instant. The behavior of the proposed predictive controller in steady-
In this paper, we choose a cost function g expressed in state operation for a resistive load of 20 Ω is shown in Fig. 5.
orthogonal coordinates and defines the desired behavior of the The amplitude of the reference voltage is set to 200 V, and the
system: To minimize the error in the output voltage frequency is 50 Hz. It is shown in the figure that the output
∗
 ∗ 2 voltages are sinusoidal with low distortion. The output voltages
g = (vcα − vcα )2 + vcβ − vcβ (28) and currents with a different amplitude of the reference voltage
of 150 V are shown in Fig. 6.
∗ ∗
where vcα and vcβ are the real and imaginary parts of the The output voltage, output current, and filter current in one
output-voltage reference vector vc∗ , while vcα and vcβ are the phase are shown in Fig. 7. It is shown that, due to the resistive
real and imaginary parts of the predicted output-voltage vector load, the load current is proportional to the output voltage
vc (k + 1). while the filter current measured at the output of the converter
This cost function has been chosen in order to obtain the low- presents high-frequency harmonics which are attenuated by the
est voltage error. However, additional constraints can be con- filter.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
CORTÉS et al.: MPC OF AN INVERTER WITH OUTPUT LC FILTER FOR UPS APPLICATIONS 1879

Fig. 5. Experimental results: output voltages and currents in steady state for a Fig. 8. Experimental results: output voltage and output current for a resistive
reference amplitude of 200 V with 20-Ω load. Voltage THD: 2.65%. load step from no load to full load.

Fig. 6. Experimental results: output voltages and currents in steady state for a Fig. 9. Experimental results: output voltage and output current for a
reference amplitude of 150 V with 20-Ω load. Voltage THD: 2.82%. resistive–inductive load step.

Fig. 10. Diode-bridge rectifier used as nonlinear load. C = 3000 μF,

R = 60 Ω.

The load, with values R = 20 Ω and 10 mH, is connected at

time 0.1 s. The amplitude of the reference voltage is 200 V for
the resistive load and 150 V for the resistive–inductive load.
The diode-bridge rectifier shown in Fig. 10 was used as
nonlinear load for the results shown in Fig. 11. Here, the output
Fig. 7. Experimental results: output voltage, output current, and filter current voltage presents a small distortion, but it is still sinusoidal de-
in steady state. spite the highly distorted load currents. A noticeable unbalance
in the load currents is present in this result due to unbalanced
The transient behavior of the system for a load step from voltages when a nonlinear load is connected. This result could
no load to full load is shown in Fig. 8. Here, a 20-Ω load is be improved by using a higher sampling frequency, but this
connected at time of 0.05 s. It can be seen in this result that solution is difficult to implement due to hardware restrictions.
the output voltage is not affected by this change in the load. Alternative solutions to improve the quality of the control for
A similar test is shown in Fig. 9 for a resistive–inductive load. nonlinear loads are under consideration.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
1880 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

Fig. 11. Experimental results: output voltages and currents in steady state for
a nonlinear load and a reference amplitude of 200 V. Voltage THD: 4.60%.
Fig. 13. Experimental results. (a) Output voltages using derivative-based
load-current estimation. THD = 7.8%. (b) Output voltages using observer-
based load-current estimation. THD = 3.8%.

Fig. 14. Experimental results: observer-based load-current estimation for the

nonlinear load. (An offset has been added to the measured signals.)

Fig. 12. Experimental results: derivative-based and observer-based load-

current estimations.

Two estimation methods were implemented experimentally,

the derivative-based estimation (21) and the observer-based es-
timation. All the results previously shown were obtained using
the load-current observer. In Fig. 12, the performances of both
estimators are shown. It can be noted that the derivative-based
estimation is very noisy while the observer-based estimation
is much cleaner due to the filtering action of the observer.
The output voltages obtained using the proposed predictive
controller with both estimation methods are shown in Fig. 13
for the same operating conditions. The use of a noisy estimation
introduces errors in the calculation of the output-voltage predic-
tion producing distorted voltages, as shown in Fig. 13(a). When Fig. 15. Experimental results: load current and estimated load current using
the observer for a load step. (Top) Using a fast observer. (Bottom) Using a slow
comparing this result with the one shown in Fig. 13(b) using observer.
the observer, it is clear that the use of the observer enhances the
behavior of the control, achieving a lower THD, without the use The effect of two different pole placement for the load-
of additional current measurements. Results from Fig. 13 where current observer is shown in Figs. 15 and 16. A plot of the
obtained using a smaller filter, compared to the results shown in pole placement for the fast and slow observers is shown in
Fig. 5, and consequently present a higher THD value. When Fig. 17. The load current in one phase of the load is shown
the nonlinear load is connected, the error introduced by the in Fig. 15 when a resistive load step is applied at time 0.1 s. A
derivative-based estimator makes the controller fail to control fast observer is used in the upper figure where it is shown how
the output voltage, and the UPS does not work. However, the the estimated value is equal to the measured value in a fraction
load-current observer delivers a good estimation, as shown in of fundamental cycle. On the other hand, if a slow observer
Fig. 14, allowing for a good control of the output voltages. is used, the estimated value reaches the real value after a full

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
CORTÉS et al.: MPC OF AN INVERTER WITH OUTPUT LC FILTER FOR UPS APPLICATIONS 1881

trolled variables. The gate-drive signals are generated directly

by the controller, so a modulator is not needed.
The output voltage is directly controlled, without using a
cascaded control structure, with an inner current-control loop.
This allows for a fast dynamic response of the voltage control.
It has been shown that the use of an observer allows for a
good estimation of the unknown load current and improves the
overall behavior of the system without the need of additional
current measurements.
Predictive control presents a different approach for the con-
trol of power converters, taking into account the discrete nature
of the converters and the microprocessors used for the control.
In addition, the high calculation power of the current exist-
ing DSPs makes this method very attractive to control power
converters.

Fig. 16. Experimental results: output voltages for a load step. (Top) Using a R EFERENCES
fast observer. (Bottom) Using a slow observer.
[1] J. M. Carrasco, L. G. Franquelo, J. T. Bialasiewicz, E. Galvan,
R. C. PortilloGuisado, M. A. M. Prats, J. I. Leon, and N. Moreno-Alfonso,
“Power-electronic systems for the grid integration of renewable energy
sources: A survey,” IEEE Trans. Ind. Electron., vol. 53, no. 4, pp. 1002–
1016, Jun. 2006.
[2] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview
of control and grid synchronization for distributed power generation
systems,” IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1398–1409,
Oct. 2006.
[3] J. M. Guerrero, L. Garcia de Vicuna, and J. Uceda, “Uninterruptible power
supply systems provide protection,” IEEE Ind. Electron. Mag., vol. 1,
no. 1, pp. 28–38, 2007.
[4] O. Kukrer, “Deadbeat control of a three-phase inverter with an output
LC filter,” IEEE Trans. Power Electron., vol. 11, no. 1, pp. 16–23,
Jan. 1996.
[5] M. Kojima, K. Hirabayashi, Y. Kawabata, E. C. Ejiogu, and T. Kawabata,
“Novel vector control system using deadbeat-controlled PWM inverter
with output LC filter,” IEEE Trans. Ind. Appl., vol. 40, no. 1, pp. 162–
169, Jan./Feb. 2004.
[6] P. Mattavelli, “An improved deadbeat control for UPS using disturbance
observers,” IEEE Trans. Ind. Electron., vol. 52, no. 1, pp. 206–212,
Feb. 2005.
[7] P. C. Loh, M. J. Newman, D. N. Zmood, and D. G. Holmes, “A com-
parative analysis of multiloop voltage regulation strategies for single and
three-phase UPS systems,” IEEE Trans. Power Electron., vol. 18, no. 5,
pp. 1176–1185, Sep. 2003.
[8] P. C. Loh and D. G. Holmes, “Analysis of multiloop strategies for
Fig. 17. Pole placement for fast and slow observers. LC/CL/LCL-filtered voltage-source and current-source inverters,” IEEE
Trans. Ind. Appl., vol. 41, no. 2, pp. 644–654, Mar./Apr. 2005.
[9] S. Buso, S. Fasolo, and P. Mattavelli, “Uninterruptible power sup-
cycle. The effect of these differences in the estimation over the ply multiloop control employing digital predictive voltage and current
output voltage is shown in Fig. 16. Here, a resistive load step regulators,” IEEE Trans. Ind. Appl., vol. 37, no. 6, pp. 1846–1854,
Nov./Dec. 2001.
is applied at time 0.05 s. While with the fast observer the effect [10] A. Kulka, T. Undeland, S. Vazquez, and L. G. Franquelo, “Stationary
of the load step is a very short drop in the voltages, with the frame voltage harmonic controller for standalone power generation,” in
slow observer, there is a noticeable reduction of the voltage Proc. Eur. Conf. Power Electron. Appl., Sep. 2007, pp. 1–10.
[11] M. N. Marwali and A. Keyhani, “Control of distributed generation
amplitude, with respect to the reference amplitude (dashed systems—Part I: Voltages and currents control,” IEEE Trans. Power Elec-
lines), between times 0.05 and 0.06 s. tron., vol. 19, no. 6, pp. 1541–1550, Nov. 2004.
[12] G. Escobar, A. A. Valdes, J. Leyva-Ramos, and P. Mattavelli, “Repetitive-
based controller for a UPS inverter to compensate unbalance and har-
monic distortion,” IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 504–510,
VI. C ONCLUSION Feb. 2007.
[13] G. Escobar, P. Mattavelli, A. M. Stankovic, A. A. Valdez, and
In this paper, a new and simple control scheme was presented J. Leyva-Ramos, “An adaptive control for UPS to compensate unbalance
for a three-phase inverter with output LC filter. The feasibility and harmonic distortion using a combined capacitor/load current sensing,”
of the proposed predictive controller has been demonstrated by IEEE Trans. Ind. Electron., vol. 54, no. 2, pp. 839–847, Apr. 2007.
[14] G. Willmann, D. F. Coutinho, L. F. A. Pereira, and F. B. Libano,
implementing it in a laboratory prototype. Results show that the “Multiple-loop H-infinity control design for uninterruptible power sup-
proposed scheme achieves a good voltage regulation with linear plies,” IEEE Trans. Ind. Electron., vol. 54, no. 3, pp. 1591–1602,
loads as well as with nonlinear loads. Jun. 2007.
[15] P. Cortes, M. P. Kazmierkowski, R. M. Kennel, D. E. Quevedo, and
The proposed controller has no parameters to adjust; it needs J. Rodriguez, “Predictive control in power electronics and drives,” IEEE
a model of the system for calculating predictions of the con- Trans. Ind. Electron., vol. 55, no. 12, pp. 4312–4324, Dec. 2008.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
1882 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 56, NO. 6, JUNE 2009

[16] R. Kennel and A. Linder, “Predictive control of inverter supplied electrical [40] P. Cortes, J. Rodriguez, P. Antoniewicz, and M. Kazmierkowski, “Direct
drives,” in Proc. IEEE PESC, Galway, Ireland, Jun. 2000, pp. 761–766. power control of an AFE using predictive control,” IEEE Trans. Power
CD-ROM. Electron., vol. 23, no. 5, pp. 2516–2523, Sep. 2008.
[17] G. Bode, P. C. Loh, M. J. Newman, and D. G. Holmes, “An improved [41] M. A. Perez, P. Cortes, and J. Rodriguez, “Predictive control algorithm
robust predictive current regulation algorithm,” IEEE Trans. Ind. Appl., technique for multilevel asymmetric cascaded h-bridge inverters,” IEEE
vol. 41, no. 6, pp. 1720–1733, Nov./Dec. 2005. Trans. Ind. Electron., vol. 55, no. 12, pp. 4354–4361, Dec. 2008.
[18] S.-M. Yang and C.-H. Lee, “A deadbeat current controller for field ori- [42] E. I. Silva, B. P. McGrath, D. E. Quevedo, and G. C. Goodwin, “Predictive
ented induction motor drives,” IEEE Trans. Power Electron., vol. 17, control of a flying capacitor converter,” in Proc. Amer. Control Conf.,
no. 5, pp. 772–778, Sep. 2002. New York, Jul. 2007, pp. 3763–3768.
[19] Y. Nishida, O. Miyashita, T. Haneyoshi, H. Tomita, and A. Maeda, “A [43] H. Kwakernaak and R. Sivan, Linear Optimal Control Systems. New
predictive instantaneous-current PWM controlled rectifier with AC-side York: Wiley, 1972.
harmonic current reduction,” IEEE Trans. Ind. Electron., vol. 44, no. 3, [44] B. D. O. Anderson and J. Moore, Optimal Filtering. Englewood Cliffs,
pp. 337–343, Jun. 1997. NJ: Prentice-Hall, 1979.
[20] L. Malesani, P. Mattavelli, and S. Buso, “Robust dead-beat current control [45] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. Silva, “Predictive current
for PWM rectifier and active filters,” IEEE Trans. Ind. Appl., vol. 35, no. 3, control strategy with imposed load current spectrum,” IEEE Trans. Power
pp. 613–620, May/Jun. 1999. Electron., vol. 23, no. 2, pp. 612–618, Mar. 2008.
[21] J. Mossoba and P. W. Lehn, “A controller architecture for high bandwidth
active power filters,” IEEE Trans. Power Electron., vol. 18, no. 1, pt. 2,
pp. 317–325, Jan. 2003.
[22] P. Mattavelli, G. Spiazzi, and P. Tenti, “Predictive digital control of
power factor preregulators with input voltage estimation using disturbance
observers,” IEEE Trans. Power Electron., vol. 20, no. 1, pp. 140–147,
Jan. 2005.
[23] A. Nasiri, “Digital control of three-phase series–parallel uninterruptible Patricio Cortés (S’05–M’08) received the B.Eng.
power supply systems,” IEEE Trans. Power Electron., vol. 22, no. 4, and M.Sc. degrees in electronics engineering in 2004
pp. 1116–1127, Jul. 2007. and the Ph.D. degree in 2008 from the Universidad
[24] Y. A.-R. I. Mohamed and E. F. El-Saadany, “An improved deadbeat Técnica Federico Santa María (UTFSM), Valparaíso,
current control scheme with a novel adaptive self-tuning load model for Chile.
a three-phase PWM voltage-source inverter,” IEEE Trans. Ind. Electron., Since 2003, he has been with the Electronics
vol. 54, no. 2, pp. 747–759, Apr. 2007. Engineering Department, UTFSM, where he is cur-
[25] Y. A.-R. I. Mohamed and E. F. El-Saadany, “Robust high bandwidth rently a Research Associate. In 2007, he visited
discrete-time predictive current control with predictive internal model—A the Institute of Control and Industrial Electronics,
unified approach for voltage-source PWM converters,” IEEE Trans. Warsaw University of Technology, Warsaw, Poland.
Power Electron., vol. 23, no. 1, pp. 126–136, Jan. 2008. His main research interests include power electron-
[26] Y. A.-R. I. Mohamed and E. F. El-Saadany, “Adaptive discrete- ics, adjustable-speed drives, and predictive control. He received the Best
time grid-voltage sensorless interfacing scheme for grid-connected Paper Award from the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
DG-inverters based on neural-network identification and deadbeat current for 2007.
regulation,” IEEE Trans. Power Electron., vol. 23, no. 1, pp. 308–321,
Jan. 2008.
[27] Q. Zeng and L. Chang, “An advanced SVPWM-based predictive cur-
rent controller for three-phase inverters in distributed generation sys-
tems,” IEEE Trans. Ind. Electron., vol. 55, no. 3, pp. 1235–1246,
Mar. 2008.
[28] A. Linder and R. Kennel, “Model predictive control for electrical drives,”
in Proc. IEEE PESC, Recife, Brazil, Jun. 12–16, 2005, pp. 1793–1799. Gabriel Ortiz was born in Chuquicamata, Chile, on
[29] E. F. Camacho and C. Bordons, Model Predictive Control. New York: September 13, 1984. He received the M.Sc. degree
Springer-Verlag, 2007. in 2008. His thesis concerned reconfiguration of re-
[30] G. C. Goodwin, M. M. Serón, and J. D. Doná, Constrained Control and generative and nonregenerative cascaded multilevel
Estimation. New York: Springer-Verlag, 2004. converters under fault condition. He has been work-
[31] J. M. Maciejowski, Predictive Control With Constraints. Englewood ing toward the Ph.D. degree in the Power Electronic
Cliffs, NJ: Prentice-Hall, 2001. Systems Laboratory, Eidgenoessische Technische
[32] V. Blasko and V. Kaura, “A new mathematical model and control of a Hochschule Zürich, Zürich, Switzerland, since 2009.
three-phase AC–DC voltage source converter,” IEEE Trans. Power Elec- Since early 2007, he has also been with the
tron., vol. 12, no. 1, pp. 116–123, Jan. 1997. power electronics group of the Electronics Engi-
[33] R. C. Portillo, M. M. Prats, J. I. Leon, J. A. Sanchez, J. M. Carrasco, neering Department, Universidad Técnica Federico
E. Galvan, and L. G. Franquelo, “Modeling strategy for back-to-back Santa María, Valparaíso, Chile.
three-level converters applied to high-power wind turbines,” IEEE Trans.
Ind. Electron., vol. 53, no. 5, pp. 1483–1491, Oct. 2006.
[34] J. Kley, G. Papafotiou, K. Papadopoulos, P. Bohren, and M. Morari,
“Performance evaluation of model predictive direct torque control,” in
Proc. IEEE PESC, Jun. 2008, pp. 4737–4744.
[35] M. Veenstra and A. Rufer, “Control of a hybrid asymmetric multilevel
inverter for competitive medium-voltage industrial drives,” IEEE Trans.
Ind. Appl., vol. 41, no. 2, pp. 655–664, Mar./Apr. 2005. Juan I. Yuz (S’01–M’06) was born in Valparaíso,
[36] S. Muller, U. Ammann, and S. Rees, “New time-discrete modulation Chile, in 1975. He received the Ingeniero Civil Elec-
scheme for matrix converters,” IEEE Trans. Ind. Electron., vol. 52, no. 6, trónico and M.S. degrees in electronics engineering
pp. 1607–1615, Dec. 2005. from the Universidad Técnica Federico Santa María
[37] R. Vargas, J. Rodriguez, U. Ammann, and P. W. Wheeler, “Predictive (UTFSM), Valparaíso, in 2001, and the Ph.D. de-
current control of an induction machine fed by a matrix converter with gree in electrical engineering from The University of
reactive power control,” IEEE Trans. Ind. Electron., vol. 55, no. 12, Newcastle, Newcastle, Australia, in 2006.
pp. 4362–4371, Dec. 2008. He currently holds a research position with the
[38] J. Rodríguez, J. Pontt, C. Silva, P. Correa, P. Lezana, P. Cortés, and Automatic Control Group in the Electronics Engi-
U. Ammann, “Predictive current control of a voltage source inverter,” neering Department, UTFSM. His research areas
IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 495–503, Feb. 2007. are in control and identification of sampled-data
[39] J. Rodríguez, J. Pontt, C. Silva, M. Salgado, S. Rees, U. Ammann, systems.
P. Lezana, R. Huerta, and P. Cortés, “Predictive control of a three-phase Dr. Yuz was the recipient of the Best Electronics Engineering Student Award
inverter,” Electron. Lett., vol. 40, no. 9, pp. 561–562, Apr. 29, 2004. from UTFSM in 2001.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.
CORTÉS et al.: MPC OF AN INVERTER WITH OUTPUT LC FILTER FOR UPS APPLICATIONS 1883

José Rodríguez (M’81–SM’94) received the B.Eng. Leopoldo G. Franquelo (M’84–SM’96–F’05) was
degree in electrical engineering from the Universidad born in Málaga, Spain. He received the M.Sc. and
Técnica Federico Santa María (UTFSM), Valparaíso, Ph.D. degrees in electrical engineering from the Uni-
Chile, in 1977, and the Dr.-Ing. degree in electri- versity of Seville (US), Seville, Spain, in 1977 and
cal engineering from the University of Erlangen, 1980, respectively.
Erlangen, Germany, in 1985. Since 1978, he has been with the US, where he
Since 1977, he has been with the Electronics was first a Research Assistant, then an Associate
Engineering Department, UTFSM, where from 2001 Professor in 1982, the Director of the Electronics
to 2004, he was the Director and, currently, is a Engineering Department from 1998 to 2005, and has
Professor. From 2004 to 2005, he was the Vice- been a Professor in the same department since 1986.
Rector of Academic Affairs, and since 2005, has He is leading a large research and teaching team in
been the Rector at UTFSM. During his sabbatical leave in 1996, he was respon- Spain. In the last five years, his group activity can be summarized as follows:
sible for the Mining Division, Siemens Corporation, Santiago, Chile. He has 45 publications in international journals, 170 in international conference pro-
extensive consulting experience in the mining industry, particularly in the ceedings, ten patents, advisor for ten Ph.D. dissertations, and 96 R&D projects.
application of large drives such as cycloconverter-fed synchronous motors for His current research interests include modulation techniques for multilevel
SAG mills, high-power conveyors, and controlled ac drives for shovels and inverters and applications to power electronic systems for renewable energy
power-quality issues. He has directed more than 40 R&D projects in the field of systems.
industrial electronics. He has coauthored more than 250 journal and conference Dr. Franquelo was the Vice President of the IEEE Industrial Electronics So-
proceedings papers and contributed one book chapter. His research group has ciety (IES) Spanish Chapter from 2002 to 2003 and a Member-at-Large of the
been recognized as one of the two Centers of Excellence in Engineering in Chile IES AdCom from 2002 to 2003. He was the Vice President for Conferences of
from 2005 to 2008. His main research interests include multilevel inverters, new the IES from 2004 to 2007, of which he has also been a Distinguished Lecturer
converter topologies, and adjustable-speed drives. since 2006. He has been an Associate Editor for the IEEE TRANSACTIONS ON
Prof. Rodríguez has been an active Associate Editor of the IEEE INDUSTRIAL ELECTRONICS since 2007. Since January 2008, he has been the
TRANSACTIONS ON POWER ELECTRONICS and IEEE TRANSACTIONS ON President Elect of the IES.
INDUSTRIAL ELECTRONICS, since 2002. He has served as Guest Editor for
the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS in five instances
[Special Sections on matrix converters (2002), multilevel inverters (2002),
modern rectifiers (2005), high-power drives (2007), and predictive control of
power converters and drives (2008)]. He received the Best Paper Award from
the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS for 2007.

Sergio Vazquez (S’04–M’08) was born in Seville,

Spain, in 1974. He received the B.S. and M.S. de-
grees in industrial engineering from the University of
Seville (US), Seville, in 2003 and 2006, respectively.
In 2002, he was with the Power Electronics Group,
University of Seville, working on R&D projects. He
is currently an Assistant Professor with the Depart-
ment of Electronic Engineering, US. His research
interests include electronic power systems, model-
ing, modulation and control of power electronic con-
verters, and power quality in renewable generation
plants.

Authorized licensed use limited to: Universidad Tecnica Federico Santa Maria. Downloaded on July 28, 2009 at 11:04 from IEEE Xplore. Restrictions apply.