Converter Load
power converters and drives are also addressed.
Load
PI
Load
Sopt
n
Minimization
Predictive J of
model cost function
J
n
Fig. 7. Block diagram of the FCSMPC control strategy for the AFE.
Fig. 6. Power circuit of the AFE.
800
700 vdc
Load
600
PI v?dc
vdc [V]
Sabc 500
n 400
vI;abc 300
Predictive PDPC
J 200
model PWMSVM
J
100
0
Fig. 9. Block diagram of the PDPC control strategy for the AFE.
0 0.2 0.4 0.5 0.6 0.8 1
Time (s)
(a)
with SVM modulation strategy (PDPC) can be considered as
q [kVAr], p [kW]
well established [14], [20][22]. 10 q
The block diagram of the FCSMPCDPC strategy is pre 5 q?
sented in Fig. 7. In this case, the model of the system is used to p
predict values of the instantaneous active and reactive power 0
over a prediction horizon N = 1, P p (k + 2), Qp (k + 2). 5
In [14], [20] a threephase twolevel AFE was controlled
10
adopting this strategy. The algorithm was developed in the
frame. Therefore, only the seven possible output vectors 0.45 0.47 0.49 0.51 0.53 0.55
were considered to perform the prediction, thus the number Time (s)
of switching states is J = 7. Once the seven output voltage (b)
predictions are calculated the cost function
400
vSa [V], iLa x10 [A]
300 vSa
2 2 200
g = (P (k+2) P p (k+2)) + (Q (k+2) Qp (k+2)) (2) 100 iLa x10
0
is minimized in order to find the inverter output vector that
100
should be applied in the next sampling period. 200
Fig. 8 presents experimental results obtained using this 300
strategy [14]. It should be noticed that predictions in instants 400
(k + 2) are used in order to compensate for the control action 0.45 0.47 0.49 0.51 0.53 0.55
delay of the digital implementation of the control strategy.
Time (s)
(c)
Another way to perform the predictive controller for the
AFE is the PDPC. The block diagram of the PDPC strategy Fig. 10. Experimental results of the PDPC for a threephase twolevel
AFE for a reactive power command step from 10 to 10 kVAr. (a) dclink
is presented in Fig. 9. Like GPC, PDPC strategy does not voltage; (b) Instantaneous active and reactive power; (c) Grid voltage and
take into account switching of power semiconductors therefore output current for phase a.
it just provides an exact solution to an approximated opti
mization problem. Besides, PDPC considers an unconstrained
MPC problem. Thus an explicit solution can be obtained It should be noticed that the outer control loop to regulate
providing the control action to be applied once the cost func the dclink capacitor voltage is usually solved using a conven
tion (2) is minimized. Therefore, an optimal switching vector tional PI controller. However, there exist some solutions that
sequence can be calculated. The control strategy provides the replace the PI control for a MPC strategy [14].
switching vectors and the switching times thus a PWMSVM
modulation strategy is necessary to generate the firing pulses.
Compared with FCSMPC, the PDPC algorithm uses an B. Control of an Active Filter
external modulator thus constant switching frequency is ob In its classical configuration, an active power filter (APF)
tained. This can be considered as an advantage specially in basically consists of a voltage source inverter (VSI) whose dc
the AFE application because for grid connected converters side is connected to a capacitors bank whereas its ac side is
exist high demanding codes that impose strict limits to the connected to the mains through a suited filter, usually formed
low order harmonics that can be injected into the grid. FCS by a set of series inductors as shown in Fig. 11 (referring
MPC presents variable switching frequency thus grid current to the most common 3wires configuration without neutral).
has a widespread harmonic spectrum. On the other hand P In such configuration, ideally the APF is able to operate as
DPC provides constant switching frequency thus grid current a controllable current generator, drawing from the mains any
harmonic spectrum is concentrated around the switching fre set of current waveforms having a null sum. Therefore, an
quency decreasing the cost of the output L filter. Fig. 10 shows APF is ideally able to compensate the unbalanced, reactive
experimental results obtained using the PDPC strategy for a and harmonic components of the currents drawn by any load,
STATCOM application when an instantaneous reactive power in such a way that the global equivalent load, as seen from
command step is imposed [22]. the grid, resembles a resistive balanced load drawing about the
5
iSa
iSb
iSc
ila ilb ilc
Nonlinear
load (a)
iSa
Sopt(k) iSb
iSc
Reference
design ila ilb ilc
Sopt
vS,abc(k)
iL,abc(k) J Minimization Nonlinear
Predictive of load
il,abc(k) model J
cost function
J
(b)
Fig. 12. Block diagram of the FCSMPC control strategy for the APF.
1 p 2
g = v (k + 2) vdc
(e (k + 2)) (3) rated values of active power and dc link voltage, 1 , 2 and
vdcR dc 3 are weighting factors that allow a proper balance among
2 2
+ (P (k + 2) PSp (k + 2)) deviations in voltage, active and reactive power, vedc (k + 2)
PSR S represents a filtered voltage reference with a certain prediction
3 2 horizon [23].
+ (QS (k + 2) QpS (k + 2))
PSR According to the scheme of Fig. 11, the APF experimental
Where vdc , PS and QS are the reference values of dc prototype chosen to verify the effectiveness of the FCSMPC
p
link voltage, active and reactive power respectively; vdc , PSp control includes a standard 3leg IGBT based VSI inverter.
p
and QS are predicted values of dclink voltage, active and The implemented control algorithm operates at fixed sampling
reactive power respectively; PSR , and vdcR are respectively frequency fs = 50 kHz. To reproduce a distorted current in the
6
grid, a nonlinear load constituted by a 3phase diode bridge Input filter Direct Matrix Coverter
rectifier supplying a resistor having a rated power PSR = 5
kW was considered [23].
At full power the load draws the distorted current in Fig. 13a
where the vertical axis measures 5A/div while the horizontal
one 10ms/div; after APF compensation the mains currents
waveform is shown in Fig. 13b (5A/div) together with the
supply voltage (100V/div). The compensation action results
in a unity power factor operation and quasi sinusoidal current
with a superimposed highfrequency ripple due to inverter
commutation and the nature of the FCSMPC control action
itself. The achieved benefits and therefore the effectiveness of
the control action were also confirmed in spectral terms by
comparing the mains current spectrum and the load current Load
spectrum in Fig. 13c, resulting in a reduction of major low
order harmonics, which permits to achieve a THD< 5% Fig. 14. Power circuit of a threephase direct matrix converter.
starting from a THD> 29%, where the THD is calculated
including up to the 50th harmonic. DMC Load
Input Filter
3
IV. MPC FOR INVERTERS WITH RLL OAD
A. Control of a matrix converter
The direct matrix converter (DMC) is a type of static power
converter in which the load is directly connected to the mains 9
through a set of bidirectional switches. The power circuit of
a DMC is shown in Fig. 14. This topology does not require Reactive power 27 Minimization 27 Output current
prediction of cost function prediction
a dclink stage with energy storage as most power converters;
this reduces the weight and size of the converter, making it
suitable for applications that require high power density, such
as aerospace.
Control approaches based on FCSMPC have been exten
Fig. 15. Block diagram of the predictive current and reactive power control
sively tested for the DMC [24], [25], showing the effectiveness of a direct matrix converter.
and relative simplicity of the predictive methods over the
traditional ones such as spacevector modulation [26]. The
block diagram of the predictive control of load current and B. SHE and SHM for power converters
input power factor is presented in Fig. 15. The predictive 1) Selective harmonic elimination technique: The selective
controller relies on mathematical models for the prediction harmonic elimination (SHE) strategy is specially well suited
of both the input reactive power and the output current. Each for highpower applications where the power losses must be
prediction block depicted in Fig. 15 yields 27 predictions for kept below strict limits. The wellknown SHE technique is
the controlled variables, one for each of the different valid based on offline calculations and the generation of a pre
switching states of the DMC. A further stage of the algorithm programmed voltage waveforms with very low number of
then minimizes a cost function in order to determine the commutations and eliminating some low order harmonics [27].
optimal combination of gating signals to be applied to the Using the predictive control to implement the SHE method, the
converter at the next sampling period. MPCSHE control objective is to follow the voltage reference,
The cost function for the simultaneous control of input to eliminate low order harmonics and to reduce switching
reactive power Q and output current io is the following: losses [28]. These three objectives are included in the cost
g = io ipo  + Q Qp  , (4) function
vs* J
SDFT
5 + f0 Power Converter

.. Minimization vs
.. Sopt
ia [A]
of
0 . cost function
J SDFT
fK
5
J vsp Predictive
Model
10
0 0.02 0.04 0.06 0.08 0.10
(a)
Weighting
i Factors
400 Tuning
executed each 20ms
200
Fig. 17. Block diagram of the proposed MPCSHM technique.
va [V]
0 a b c
+ + +
HBridge 1a HBridge 1b HBridge 1c
200
  
400 + + +
0 0.02 0.04 0.06 0.08 0.10
HBridge 2a HBridge 2b HBridge 2c
(b)
  
10 400
vs
+ + +
5 200 HBridge 3a HBridge 3b HBridge 3c
is ia ib ic
  
vs [V]
is [A]
0 0
N
5 200 La Lb Lc
van vbn vcn
10 400 Ra Rb Rc
0 0.02 0.04 0.06 0.08 0.10
Time [s] n
(c)
Fig. 16. Predictive current control of a direct matrix converter. (a) Output
current. (b) Output voltage. (c) Input current and grid voltage (phase a). S1 S2
+ V dc
+
HBridge kj
the discrete Fourier Transform algorithm used to calculate a
 S1 S2
finite number of single frequency spectral components with
very low computational cost [29].
The first term (5) evaluates the error between the reference
and the predicted output voltage vector tuned to the funda 
mental frequency f1 . Fig. 18. Power circuit of a Hbridge multilevel converter (three Hbridges
The second term (6) is the sum of all those frequencies (up per phase) connected to a RL load.
to M th order) that need to be eliminated. The weighting factor
f is used to control the importance of this term in the cost
function, in this way the frequency elimination can be relaxed the low order harmonics but reducing the distortion below the
or strengthened in comparison to the fundamental frequency limits imposed by a grid code [30], [31].
tracking, depending on design considerations. The control objective of the MPCSHM is to follow the
Finally, the third term (7) is used to reduce the number of voltage reference, to control the harmonic distortion keeping
commutations introducing weighting factor sw to keep the it below the limits imposed in the grid code and to reduce
power losses below acceptable limits. the switching losses as much as possible [32]. These three
2) Selective harmonic mitigation technique: An evolution objectives are included in a cost function which is similar
of SHE is the selective harmonic mitigation (SHM) technique to that introduced for the MPCSHE method. In this case,
which is based on preprogrammed waveforms non eliminating the second term (6) has to be modified being the sum of
8
1
400
Phase Voltage vaN [V]
0.8 with the SHE method and this fact makes easier to find better
300
200
0.6 solutions leading to lower power losses.
0.4
100
0.2
0 0 C. Control of multilevel inverters
1.79 1.795 1.8 1.805 1.81 1.815 1.82 1.825 1.83 1.835 1.84 0 500 1000
400 1 The FCSMPC method has been applied to multilevel con
Line Voltage vab [V]
200 0.8 verters for multiple applications. Among the multilevel con
0
0.6
verter topologies, the neutralpointclamped converter (NPC),
0.4
200
0.2
the flyingcapacitor converter (FC) and the cascaded Hbridge
400 0 converter (CHB) are the ones with vast industrial success.
1.79 1.795 1.8 1.805 1.81 1.815 1.82 1.825 1.83 1.835 1.84 0 500 1000
300
These topologies are normally used for mediumvoltage high
1
power applications at the expense of a large number of power
Load Voltage van [V]
200
0.8
100
0.6 semiconductors and more complex control and modulation al
0
100
0.4 gorithms. The FCSMPC method for multilevel inverters has to
200
0.2
take into account the usual control objectives present in other
300 0
1.79 1.795 1.8 1.805 1.81 1.815 1.82 1.825 1.83 1.835 1.84 0 500 1000 converter topologies and applications but extra control targets
15
1 have to be included such as the balance of the floating dc
Phase Currents iabc [A]
10
5
0.8
voltages (if needed) and the reduction of the switching losses
0.6
0
0.4
(required because for highpower applications the effective
5
0.2 switching frequency and consequently the power losses have
10
15 0
0 500 1000
to be limited). Some examples are here addressed:
1.79 1.795 1.8 1.805 1.81 1.815 1.82 1.825 1.83 1.835 1.84
Time [s] Frequncy [Hz] 1) NPC inverter topology: The threelevel NPC converter
has the dclink bus divided in two parts that should be
Fig. 19. Predictive harmonic mitigation phase output voltage vaN , line balanced. So, this fact has to be included in the cost function.
voltage vab , load voltage van and load currents ia ,ib ,ic for a 7level CHB.
The dc voltage balance is achieved by the FCSMPC method
but at the expense of changing the switching state almost each
sampling time. So, the result is not satisfactory because it
the distortion of those harmonics that have to be kept below
leads to high switching losses. So, a limitation in the switching
acceptable limits imposed by the grid code from the 2nd to
frequency has to be included in the cost function as well. Thus
the K th harmonic. In fact, the mathematical expression of the
a possible cost function could be
MPCSHM cost function is similar to the MPCSHE one but
considering harmonics up to K th order where K is higher
than M . Another difference between both cost functions is g = i ip  +  i ip  +dc  vC1
p p
vC2  +n nc . (8)
that a weighting factor i is associated to each ith order
In the cost function, the first term is focused on the current
harmonic distortion that has to be reduced. This fact gives
tracking which is the application of this FCSMPC method.
the chance to relax the condition of eliminating the distortion
The second term is proportional to the absolute difference
of those harmonics. The proposed MPCSHM technique can
between the voltage predictions of both capacitors, so a switch
be summarized using the block diagram represented in Fig. 17
ing state that generates smaller differences will be preferred
where a weighting factors tuning block has been added to the
leading to a voltage balance situation. Finally, the third term
previously introduced MPCSHE method. For each harmonic,
is proportional to the number of commutations to get to the
the weighting factors tuning block relaxes the weighting
next switching state nc , so a switching state that implies fewer
factor i if the ith harmonic distortion is below the limit of
commutations of the power semiconductors will be preferred.
the grid code or make it higher in the opposite case. This
The weighting factors dc and n handle the relation between
dynamic weighting factors adjustment is made at fundamental
terms dedicated to current reference tracking, voltage balance
frequency.
and reduction of switching frequency [33].
The MPCSHM method has been tested in a threephase
2) CHB inverter topology: The FCSMPC method can be
cascaded Hbridge multilevel converter (three Hbridges per
also applied to achieve current tracking control in a threephase
phase) connected to a RL load as depicted in Fig. 18. On
cascaded Hbridge converter with two Hbridges per phase
the other hand, Fig. 19 shows the converter phase output
[34]. In this case, the cost function only takes into account
voltage VaN , the lineline voltage Vab , the load voltage Van
the current tracking:
and the load currents ia , ib and ic . In addition, the respective
harmonic spectra are shown next to each waveform. A deep
g = i ip  +  i ip  . (9)
analysis of the data shows that the distortion of harmonics
considered in the cost function (up to harmonic 10th ) is always In the CHB inverter case, there are a high number of possi
below the limit imposed. In addition, the average switching ble switching combinations (125 for this topology where the
frequency of the MPCSHM method is lower to that obtained dc voltages of the Hbridges are fixed) so the computational
with the MPCSHE technique dealing with the same number cost of the FCSMPC method can become excessively high. In
of harmonics. This phenomenon appears because the SHM [34], this is solved eliminating the redundant switching states
relaxes the conditions of the harmonic distortions compared with higher commonmode voltages. In addition, only the last
9
vdc
Load
300 vc2a
200
vc1a
100
Fig. 21. Power circuit of a threephase inverter with output LC filter.
0
(a)
LC Filter
Inverter Output Voltage [V]
400
Load
300 Sopt
Minimization
200 of
Predictive
cost function
model J
100
van
Fig. 22. Block diagram of the FCSMPC control strategy for a threephase
0 inverter with output LC filter.
(b)
5
THDc=1.5%
4
Output Currents [A]
300
3
200
2
vC;abc [V]
100
1
0
0
100
1
200
2
300
3 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
4 ia ib ic
5 20
5 10 1 20 5 0
(c) Time [ms] 10
iO;abc [A]
0
Fig. 20. Experimental results for the FCSMPC applied to the FC with ratio
3:2:1: Capacitor voltages of phase a, inverter output voltage of phase a and 10
output currents (taken from [35]).
20
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Time [s]
applied switching state and the six states surrounding it in the
space vector diagram are taken into account. This fact reduces Fig. 23. Experimental results for the FCSMPC: output voltages and currents
to seven the number of possible combinations leading to the in steady state for a nonlinear load and a reference amplitude of 200V.
same computational cost of a conventional threephase two
level converter at the expense of solving a suboptimal problem
and loosing dynamic response. computational cost of the FCSMPC method can become
3) FC inverter topology: A similar FCSMPC strategy can excessively high. In [35], this is solved ignoring the interaction
be applied to the FC converter. In [35], it is presented a cost through the load neutral point in the prediction step. This
function to achieve current tracking and floating voltages con reduces the possible switching combinations to 24 leading to
trol of a threephase FC converter with two floating capacitors a high reduction in the computational cost at the expense of
per phase. In this case, the cost function includes the current limiting a control degree.
tracking term and the floating voltages control term as follows: In order to illustrate the good performance of FCSMPC
method for multilevel converters, the results for the FC inverter
g = ga + gb + gc (10) are represented in Fig. 20. As can be observed, the control
objectives, current tracking and control of the floating voltages,
where, ga , gb and gc follow the next expression (x a, b, c) are achieved.
p 2 p 2 2
gx = (ix ipx ) + dc1 (vc1x
vc1x
) + dc2 (vc2x vc2x ) .
(11) V. MPC FOR INVERTERS WITH OUTPUT LC FILTER
An interesting point to be highlighted is that, with the Inverters with output LC filter are used when it is necessary
threephase FC converter with two floating capacitors per to obtain a sinusoidal output voltage with very low harmonic
phase, there are 512 possible switching combinations so the content. This is the case of sensitive loads, or drives for
10
LC Filter
200 200
vCa vVCa
Load 150 150
aO
* *
vCa vCa
V
aO
100 *  v
100 * V
Sabc vCa Ca vVCa  vCa
aO aO
50 50
vCa [V]
vCa [V]
vI;abc 0 0
GPC
Predictive 50 50
model PWMSVM
100 100
150 150
Fig. 24. Block diagram of the GPC control strategy for a threephase inverter 200
0 10 20 30 40
200
0 10 20 30 40
with output LC filter. Time [ms] Time [ms]
(a) (b)
Fig. 25. System performance for the GPC strategy for different values of
machines in order to reduce the input voltage harmonics and N and : (a) Phase a voltage and its reference for N = 5 and = 0.0. (b)
increase its lifespan or avoid problems caused by high values Phase a voltage and its reference for N = 6 and = 1.05.
of dv/dt. These inverters are also employed when a voltage
source is needed like in FACTS devices as SSSC, DVR or
as a part of an UPFC or UPQC. However, its most important
application is as the main converter of uninterruptible power
supply systems (UPS).
PI Inverter
The power circuit of an inverter with output LC filter + Induction Machine
connected to a generic load is shown in Fig. 21. As shown, the  Minimization Sopt
of
main system variables are the output inductor current iL,abc , cost function
output capacitor voltage vC,abc , output capacitor current iC,abc
and output load current iO,abc . The main objective of the
control strategy is to calculate the output inverter voltage vI,abc J J
ia
in order to track an output capacitor voltage reference vC,abc Predictive
Estimate
for any connected load. model is ib
There are several alternatives for designing the control
algorithm for an inverter with output LC filter [36]. MPC
is a very interesting option for this application, because high
Fig. 26. Block diagram of the predictive torque control of an induction
performance of the overall system can be achieved with a very machine.
simple algorithm. Application of FCSMPC and GPC for an
UPS system can also be found in the literature [12], [37].
The block diagram of the FCSMPC strategy is presented a GPC strategy to calculate the control action. The GPC uses
in Fig. 22. In this case, the model of the system is used to a CARIMA model to predict the system variables with long
predict the output capacitor voltage over a prediction horizon prediction horizon values [11]. The block diagram of the GPC
p
N = 1, vC (k + 1). In [37] this strategy was adopted and the strategy is presented in Fig. 24. In this case, it is considered
controller was developed for a threephase twolevel inverter an unconstrained MPC problem. Thus an explicit solution can
in the frame. Therefore, only the seven possible output be obtained providing the control action to be applied once
vectors were considered to perform the prediction, thus the the cost function
number of switching states is J = 7. Once the seven output
voltage predictions are calculated the cost function N
X
2
p
g=
vC,abc (t + j) vC,abc (t + j)
u(t)2 (13)
2 2
p p
g = vC, vC, + vC, vC, (12) j=1
is minimized.
is minimized in order to find the inverter output vector that
Compared with FCSMPC, the GPC algorithm requires an
should be applied in the next sampling period.
external modulator but this provides the benefit of presenting
Fig. 23 shows experimental results obtained using this
constant switching frequency making the design of the output
strategy when a non linear load is connected to the inverter
LC filter easier. Fig. 25 shows experimental results obtained
[37]. It should be noticed that iO,abc has been considered as
using the GPC strategy when a linear load is connected to
a perturbation. In this way, iO,abc was not measured but a
the inverter for different values of N and [12]. It should be
observer was used to enhance the performance of the system.
noticed that tuning correctly the prediction horizon N and the
Prediction horizons N higher than 1 can provide in some
weighting factor can increase significantly the performance
cases better performance than using N = 1. This issue has
of the system.
been investigated for a UPS system using the FCSMPC
strategy [38]. The main problem is that computation burden
increases exponentially with N . Therefore it is difficult the VI. MPC FOR HIGH  PERFORMANCE DRIVES
practical implementation of this approach. For the control of highperformance drives there exist two
Another way to increase the prediction horizon N is to use wellestablished methods: fieldoriented control (FOC) and
11
150
control objectives. Following the operation principle of FCS
100 MPC, the switching state applied to the converter at each
50 sampling period corresponds to the one that minimizes the
[rad/s]
0
cost function. The block diagram of PTC is presented in
Fig. 26. An estimator is required in order to compute the
50
stator and rotor fluxes. Then, a mathematical model is used
100 to predict the future behavior of the torque and stator flux
150
using the flux estimations and the measurements of stator
0 1 2 3 4 5 current and mechanical angular speed of the machine. In [40]
(a) this technique was used together with threephase twolevel
inverter. In this way, the number of switching states is J = 7.
20 Therefore, the predictions associated to the seven different
voltage vectors are evaluated in an optimization stage in order
10 to select the optimum switching state S to be applied to the
[N]
0
inverter. For the generation of the electrical torque reference,
an external control loop with a PI controller was used.
,
0
VII. PAST, PRESENT AND FUTURE CHALLENGES OF MPC
10 FOR POWER CONVERTERS AND DRIVES
20 MPC has been used as a good solution for industrial appli
0 1 2 3 4 5 cations since decades [5], [41]. However, once the simplicity
Time [s] and the good performance of the MPC controller in the power
(c) electronics field has been demonstrated, the question to be
Fig. 27. Predictive torque control of an induction machine. (a) Angular solved is why it is not already extensively used in the industry.
speed. (b) Electrical torque. (c) Stator current (phase a). As a major challenge, the MPC needs an accurate model of
the system and this is not usually a simple task in highly
dynamic systems. However, in the last years the modeling
direct torque control (DTC) [39]. The first one is able to of complex electrical systems has been greatly improved and
perform a decoupled control of torque and flux by controlling this challenge can be solved. Although still is necessary more
separately the quadrature and direct components of the stator research, now it is possible to find applications of MPC
current in a rotating coordinate frame which is synchronous to power converters where Luenberger and extended state
with the rotor flux. observers are used to avoid the effects of systems parameters
On the other hand, the DTC approach is also able to perform uncertainties [37], [42], [43].
an independent control of torque and flux, by employing non A drawback of the MPC strategies is the exponential in
linear hysteresis controllers and a logic based on a lookup crease of the computational burden if the prediction horizon
table in order to select the appropriate switching states of the (N ) is longer than 1 and in the case of FCSMPC is the
converter. number of switching states to be studied (J) is high. This
A third alternative, predictive torque control (PTC), has fact was critical in the past but nowadays the highspeed
been proposed recently [40]. In the PTC algorithm, a cost microprocessors can carry out complex iterative calculations
function that gathers the tracking error of torque and stator and the FCSMPC methods can be executed with sampling
flux magnitude is employed: times around several decades of microseconds [44]. Besides,
MPC techniques like GPC can deal with long prediction
g = Te Tep  + s  s p  , (14) horizons without increasing significantly the computational
burden [12]. In addition, some authors have developed FCS
where Te , s  , Tep and s p are the reference and predicted MPC techniques that evaluate a reduced set of switching states
values of electrical torque and the magnitude of the stator in cases where the possible switching states are high. For
flux, respectively. The parameter is a weighting factor instance, in [34] a threephase cascaded Hbridge multilevel
that allows giving more or less relevance to each of the converter has been considered with N = 1. This converter has
12
125 possible voltage vectors, but the proposed method just [4] A. Linder, R. Kanchan, R. Kennel, and P. Stolze, ModelBased Predictive
calculates the cost function for the 7 vectors located around Control of Electric Drives. Cuvillier Verlag Gottingen, 2010.
[5] J. Rodriguez and P. Cortes, Predictive control of power converters and
the last voltage vector applied to the converter. In spite of this, electrical drives, 1st ed. WileyIEEE Press, 2012.
finding computational efficient MPC control algorithms is an [6] J. Rodriguez, M. P. Kazmierkowski, J. R. Espinoza, P. Zanchetta,
open issue. H. AbuRub, H. A. Young, and C. A. Rojas, State of the Art of
Finite Control Set Model Predictive Control in Power Electronics, IEEE
Usually considered an advantage, the FCSMPC method Transactions on Industrial Informatics, vol. 9, no. 2, pp. 10031016,
avoids using a modulation stage. However, this usually leads to May 2013.
spread harmonic spectra of the output waveforms. This can be [7] http://ieeexplore.ieee.org/xplore/home.jsp. [Online]. Available: http:
//ieeexplore.ieee.org/Xplore/home.jsp
solved taking it into account in the cost function [45] or using [8] E. F. Camacho and C. Bordons, Model Predictive Control. NewYork:
a modulation stage and applying the FCSMPC considering SpringerVerlag, 1999.
all the possible combinations of the switching states of the [9] P. Cortes, M. P. Kazmierkowski, R. M. Kennel, D. E. Quevedo, and
converter [46]. J. Rodriguez, Predictive control in power electronics and drives,
Industrial Electronics, IEEE Transactions on, vol. 55, no. 12, pp. 4312
Another MPC concern is about the design of an efficient 4324, 2008.
cost function and the tuning of the weighting factors. In this [10] S. Mariethoz and M. Morari, Explicit modelpredictive control of a
case, it can be affirmed that a systematic way to design the cost pwm inverter with an lcl filter, Industrial Electronics, IEEE Transac
tions on, vol. 56, no. 2, pp. 389399, 2009.
function with the best weighting factors tuning is still missing. [11] D. W. Clarke, C. Mohtadi, and P. S. Tuffs, Generalized predictive
However, some works have introduced a first approach to solve controlpart i. the basic algorithm, Automatica, vol. 23, no. 2, pp.
the problem facilitating the electrical engineers design work 137148, Mar. 1987. [Online]. Available: http://dx.doi.org/10.1016/
00051098(87)900872
[47]. [12] S. Vazquez, C. Montero, C. Bordons, and L. G. Franquelo, Model
Finally, it should be noticed that there is a lack of analytical predictive control of a vsi with long prediction horizon, in Industrial
tools to evaluate performance of MPC for power converters Electronics (ISIE), 2011 IEEE International Symposium on, 2011, pp.
18051810.
and drives without having to carry out extensive simulations [13] P. Cortes, J. Rodriguez, C. Silva, and A. Flores, Delay Compensation
or experiments. Therefore, it is expected that another area of in Model Predictive Current Control of a ThreePhase Inverter, IEEE
future research would be the development of such tools. Transactions on Industrial Electronics, vol. 59, no. 2, pp. 13231325,
2012.
[14] D. E. Quevedo, R. P. Aguilera, M. A. Perez, P. Cortes, and R. Lizana,
VIII. C ONCLUSIONS Model predictive control of an afe rectifier with dynamic references,
Model predictive control is a wellknown technique to Power Electronics, IEEE Transactions on, vol. 27, no. 7, pp. 31283136,
2012.
achieve a high performance operation in a wide applications [15] J. M. Carrasco, L. G. Franquelo, J. T. Bialasiewicz, E. Galvan, R. C. P.
range. Since decades it has been successfully applied to Guisado, M. A. M. Prats, J. I. Leon, and N. MorenoAlfonso, Power
chemical processes with low sampling requirements. However, electronic systems for the grid integration of renewable energy sources:
A survey, Industrial Electronics, IEEE Transactions on, vol. 53, no. 4,
in the last decade the academia has demonstrated that MPC can pp. 10021016, 2006.
be applied to control other systems such as electrical machines [16] S. Vazquez, S. M. Lukic, E. Galvan, L. G. Franquelo, and J. M. Carrasco,
and drives. Critical challenges as the accuracy of the models, Energy storage systems for transport and grid applications, Industrial
Electronics, IEEE Transactions on, vol. 57, no. 12, pp. 38813895, 2010.
high sampling rates and high computational cost have been [17] J. Dixon, L. Moran, J. Rodriguez, and R. Domke, Reactive power
overcome due to continuous evolution of the microprocessors compensation technologies: Stateoftheart review, Proceedings of the
technology and the effort of the researchers. The last step IEEE, vol. 93, no. 12, pp. 21442164, 2005.
of the MPC to become mature is currently being done and [18] S. Vazquez, J. A. Sanchez, J. M. Carrasco, J. I. Leon, and E. Galvan,
A modelbased direct power control for threephase power converters,
some companies have been attracted by this control method. Industrial Electronics, IEEE Transactions on, vol. 55, no. 4, pp. 1647
Hopefully just one step ahead, the MPC will be extensively 1657, 2008.
applied to control complex electrical systems. [19] M. Reyes, P. Rodriguez, S. Vazquez, A. Luna, R. Teodorescu, and
J. M. Carrasco, Enhanced decoupled double synchronous reference
frame current controller for unbalanced gridvoltage conditions, Power
ACKNOWLEDGMENT Electronics, IEEE Transactions on, vol. 27, no. 9, pp. 39343943, 2012.
[20] P. Cortes, J. Rodriguez, P. Antoniewicz, and M. Kazmierkowski, Direct
The authors gratefully acknowledge the financial support power control of an afe using predictive control, Power Electronics,
provided by the Andalusian Government and the European IEEE Transactions on, vol. 23, no. 5, pp. 25162523, 2008.
Commission (ERDF) under the project P11TIC7070 and [21] S. A. Larrinaga, M. A. R. Vidal, E. Oyarbide, and J. R. T. Apraiz,
the Ministerio de Economia y Competitividad of the Spanish Predictive control strategy for dc/ac converters based on direct power
control, Industrial Electronics, IEEE Transactions on, vol. 54, no. 3,
Government and the European Commission (ERDF) under the pp. 12611271, 2007.
project ENE201236897. [22] R. Aguilera, S. Vazquez, D. Quevedo, and L. G. Franquelo, Generalized
predictive direct power control for ac/dc converters, in ECCE 2013 
IEEE ECCE Asia Downunder, 2013.
R EFERENCES [23] P. Zanchetta, P. Cortes, M. Perez, J. Rodriguez, and C. Silva, Finite
[1] J. H. Lee, Model Predictive Control: Review of the Three Decades states model predictive control for shunt active filters, in IECON 2011
of Development, International Journal of Control, Automation, and  37th Annual Conference on IEEE Industrial Electronics Society, 2011,
Systems, vol. 9, no. 3, pp. 415424, 2011. pp. 581586.
[2] M. Morari and J. H. Lee, Model predictive control: past, present and [24] R. Vargas, J. Rodrguez, U. Ammann, and P. W. Wheeler, Predictive
future, Computers and Chemical Engineering, vol. 23, pp. 667682, Current Control of an Induction Machine Fed by a Matrix Converter
1999. With Reactive Power Control, IEEE Transactions on Industrial Elec
[3] S. Kouro, P. Cortes, R. Vargas, U. Ammann, and J. Rodriguez, Model tronics, vol. 55, no. 12, pp. 43624371, 2008.
Predictive ControlA Simple and Powerful Method to Control Power [25] F. Villarroel, J. R. Espinoza, C. A. Rojas, J. Rodriguez, M. Rivera, and
Converters, IEEE Transactions on Industrial Electronics, vol. 56, no. 6, D. Sbarbaro, Multiobjective Switching State Selector for FiniteStates
pp. 18261838, 2009. Model Predictive Control Based on Fuzzy Decision Making in a Matrix
13
Converter, IEEE Transactions on Industrial Electronics, vol. 60, no. 2, [37] P. Cortes, G. Ortiz, J. I. Yuz, J. Rodriguez, S. Vazquez, and L. G.
pp. 589599, Feb. 2013. Franquelo, Model predictive control of an inverter with output lc filter
[26] M. Rivera, A. Wilson, C. A. Rojas, J. Rodriguez, J. R. Espinoza, P. W. for ups applications, Industrial Electronics, IEEE Transactions on,
Wheeler, and L. Empringham, A Comparative Assessment of Model vol. 56, no. 6, pp. 18751883, 2009.
Predictive Current Control and Space Vector Modulation in a Direct [38] P. Cortes, J. Rodriguez, S. Vazquez, and L. G. Franquelo, Predictive
Matrix Converter, IEEE Transactions on Industrial Electronics, vol. 60, control of a threephase ups inverter using two steps prediction horizon,
no. 2, pp. 578588, Feb. 2013. in Industrial Technology (ICIT), 2010 IEEE International Conference on,
[27] H. S. Patel and R. G. Hoft, Generalized techniques of harmonic 2010, pp. 12831288.
elimination and voltage control in thyristor inverters: Part iharmonic [39] M. P. Kazmierkowski, L. G. Franquelo, J. Rodriguez, M. Perez, and
elimination, Industry Applications, IEEE Transactions on, vol. IA9, J. I. Leon, Highperformance motor drives, Industrial Electronics
no. 3, pp. 310317, 1973. Magazine, IEEE, vol. 5, no. 3, pp. 626, 2011.
[28] S. Kouro, B. La Rocca, P. Cortes, S. Alepuz, B. Wu, and J. Rodriguez, [40] H. Miranda, P. Cortes, J. I. Yuz, and J. Rodriguez, Predictive Torque
Predictive control based selective harmonic elimination with low Control of Induction Machines Based on StateSpace Models, IEEE
switching frequency for multilevel converters, in Energy Conversion Transactions on Industrial Electronics, vol. 56, no. 6, pp. 19161924,
Congress and Exposition, 2009. ECCE 2009. IEEE, 2009, pp. 3130 2009.
3136. [41] S. Qin and T. Badgwell, A survey of industrial model predictive
[29] E. Jacobsen and R. Lyons, The sliding dft, Signal Processing Maga control technology, Control Engineering Practice, vol. 11, no. 7,
zine, IEEE, vol. 20, no. 2, pp. 7480, 2003. pp. 733764, Jul. 2003. [Online]. Available: http://dx.doi.org/10.1016/
[30] L. G. Franquelo, J. Napoles, R. C. P. Guisado, J. I. Leon, and M. A. s09670661(02)001867
Aguirre, A flexible selective harmonic mitigation technique to meet [42] C. Xia, M. Wang, Z. Song, and T. Liu, Robust model predictive
grid codes in threelevel pwm converters, Industrial Electronics, IEEE current control of threephase voltage source pwm rectifier with online
Transactions on, vol. 54, no. 6, pp. 30223029, 2007. disturbance observation, Industrial Informatics, IEEE Transactions on,
[31] J. Napoles, J. I. Leon, R. Portillo, L. G. Franquelo, and M. A. Aguirre, vol. 8, no. 3, pp. 459471, 2012.
Selective harmonic mitigation technique for highpower converters, [43] Z. Song, C. Xia, and T. Liu, Predictive current control of threephase
Industrial Electronics, IEEE Transactions on, vol. 57, no. 7, pp. 2315 gridconnected converters with constant switching frequency for wind
2323, 2010. energy systems, Industrial Electronics, IEEE Transactions on, vol. 60,
[32] H. Aggrawal, J. I. Leon, L. G. Franquelo, S. Kouro, P. Garg, and J. Ro no. 6, pp. 24512464, 2013.
driguez, Model predictive control based selective harmonic mitigation [44] R. K. Cavin, P. Lugli, and V. V. Zhirnov, Science and engineering
technique for multilevel cascaded hbridge converters, in IECON 2011 beyond moores law, Proceedings of the IEEE, vol. 100, no. Special
 37th Annual Conference on IEEE Industrial Electronics Society, 2011, Centennial Issue, pp. 17201749, 2012.
pp. 44274432. [45] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. Silva, Predictive current
[33] R. Vargas, P. Cortes, U. Ammann, J. Rodriguez, and J. Pontt, Predic control strategy with imposed load current spectrum, Power Electronics,
tive control of a threephase neutralpointclamped inverter, Industrial IEEE Transactions on, vol. 23, no. 2, pp. 612618, 2008.
Electronics, IEEE Transactions on, vol. 54, no. 5, pp. 26972705, 2007.
[34] P. Cortes, A. Wilson, S. Kouro, J. Rodriguez, and H. AbuRub, Model [46] S. Vazquez, J. I. Leon, L. G. Franquelo, J. M. Carrasco, O. Martinez,
predictive control of multilevel cascaded hbridge inverters, Industrial J. Rodriguez, P. Cortes, and S. Kouro, Model predictive control with
Electronics, IEEE Transactions on, vol. 57, no. 8, pp. 26912699, 2010. constant switching frequency using a discrete space vector modulation
[35] P. Lezana, R. Aguilera, and D. E. Quevedo, Model predictive control of with virtual state vectors, in Industrial Technology, 2009. ICIT 2009.
an asymmetric flying capacitor converter, Industrial Electronics, IEEE IEEE International Conference on, 2009, pp. 16.
Transactions on, vol. 56, no. 6, pp. 18391846, 2009. [47] P. Cortes, S. Kouro, B. La Rocca, R. Vargas, J. Rodriguez, J. I. Leon,
[36] A. Kulka, T. Undeland, S. Vazquez, and L. G. Franquelo, Stationary S. Vazquez, and L. G. Franquelo, Guidelines for weighting factors
frame voltage harmonic controller for standalone power generation, design in model predictive control of power converters and drives, in
in Power Electronics and Applications, 2007 European Conference on, Industrial Technology, 2009. ICIT 2009. IEEE International Conference
2007, pp. 110. on, 2009, pp. 17.