See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/287796890

Model Predictive Control: MPC's Role in the Evolution of Power

Electronics

Article in IEEE Industrial Electronics Magazine · December 2015

DOI: 10.1109/MIE.2015.2478920

CITATIONS READS
108 864

5 authors, including:

Samir Kouro Marcelo A Perez

Universidad Técnica Federico Santa María Universidad Técnica Federico Santa María
172 PUBLICATIONS 11,133 CITATIONS 118 PUBLICATIONS 6,463 CITATIONS

SEE PROFILE SEE PROFILE

Jose Rodriguez A.M. Llor

Universidad Técnica Federico Santa María Ecole Nationale Supérieure d’Electrotechnique, d’Electronique, d’Informatiq…
317 PUBLICATIONS 18,854 CITATIONS 44 PUBLICATIONS 656 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Inspection of transmission lines View project

High efficiency PV inverters View project

All content following this page was uploaded by Samir Kouro on 29 June 2016.

The user has requested enhancement of the downloaded file.

 © comstock

Model Predictive Control

MPC’s Role in the Evolution of Power Electronics

T
he evolution of power electronics and its control has been mainly
driven by industry applications and influenced by the development
achieved in several technologies, such as power semiconductors,
converter topologies, automatic control, and analog and digital
electronics. Digital signal processors (DSPs), in particular, have ex-
perienced an exponential development in processing power, which
SAMIR KOURO, MARCELO A. PEREZ,
until now has not been fully exploited for control purposes in power converters
JOSE RODRIGUEZ, ANA M. LLOR,
and drive applications. Presently, the control system technology finds itself in a
and HECTOR A. YOUNG
paradigm-changing tipping point, in which more demanding control goals, system
flexibility, and functionalities required by emerging applications are driving the
control system technology development, in addition to stabilization and robust-
ness, which was the main focus in the past. This article walks briefly through the
history of the mainstream power converter control scene, with an emphasis on
the more recent introduction of predictive control, and gives a glimpse on the
challenges and possibilities ahead. Special attention is given to finite control set

Digital Object Identifier 10.1109/MIE.2015.2478920

Date of publication: 21 December 2015

8 IEEE industrial electronics magazine  ■ december 2015 1932-4529/15©2015IEEE

(FCS)-model predictive control (MPC),
because of its simplicity, flexibility, in-
herent adaptation to power electronic

Becomes Standard in Hex Core 177.777 MIPS

Digital Implementation 2010: Microprocessor
circuits and their discrete nature, both
in the finite amount of switching states

Enters the Market

and the digital implementation with

Controlled Drive
microprocessors.

2011: First MPC

2010
The History of Power
Electronics and Its Control
A major breakthrough in power elec-
tronics, which started a revolution in

Power Electronics
the control of power, was the thyristor,

Current Control of
introduced by General Electric in 1957

2007: Predictive

2000
[1]. The introduction of this first con-
trolled semiconductor device marks
the beginning of modern power elec-

VSC
tronics as we know it, as shown in the

1996: Introduction
time line in Figure 1. Since then, sev-
eral devices have been introduced; cur-
rently, the metal–oxide–semiconductor

of IGCT
field-effect transistor and insulated-

1990
gate bipolar transistor (IGBT) domi-
nate low and medium power, while the
high-voltage IGBT, integrated gate-com-

1985: Introduction
1983: SVM Introduced

mutated thyristor, and thyristors domi-

1971: FOC Introduced 1985: DTC Introduced

nate high- power applications [2].

1985: First FPGA

The first control systems for power

of IGBT
converters were based on linear cas-
1980

caded control loops, assisted by coor-

dinate transformations and modulation
stages to linearize the power converter
1970: First Micro-
1978: MPC Introduced

as much as possible. This enabled the

Becomes Standard in Processor Patent
in Process Control

use of the well-known proportional-in-

1979: First DSP

tegral (PI) controllers for almost every

(Classic Linear)

converter and application [3].

1970

FIGURE 1 – The time line of control theory and control technology milestones.
The modulation stage generates
the commutation signals for the con-
verter based on a time-average oper-
Analog Electronics

ating principle to achieve the desired

Power Electronics

output. These are commonly based

1965: Hysteresis

on a firing angle delay control for

<<1950: Classic Linear 1965: Fuzzy Logic

Introduced

thyristor-based topologies working at

1960
Control
Introduced

fundamental switching frequency, and

on pulsewidth modulation (PWM) for
transistor-based topologies operating
1957: Introduction

at higher switching frequencies [4].

1950: Sliding Mode

The practical implementation for

Control Theory

(First Controlled

both the control and modulation meth-

Control Theory Is

of Thyristor
Introduced

od was performed using analog circuitry,

1950

Device)

which became the early standard control

Mature

platform for power electronics. Micro-

processors were introduced in the early
Con y
trol olog
The chn Te
1970s; nevertheless, they experienced ory trol
Con
an explosive development in computa-
tional power during the 1990s. During
this decade, digital microprocessors

december 2015 ■ IEEE industrial electronics magazine  9

were introduced to control power elec- give and take a few pros and cons. How- a transistor, which combined operation
tronics, which has continued since then ever, considering the recent introduc- connects the input voltage to the output
[5]. Currently, DSP and field-programma- tion of these technologies, compared terminals. The dc load current is con-
ble gate array (FPGA) platforms are the to classic control theory that has ma- trolled, changing the mean value of the
dominant technology in power electron- tured for more than a century, together load voltage by adjusting the conduc-
ics applications. with the fact that their development is tion time, or duty cycle, of the power
closely related to advances in comput- transistor. The gating signal is obtained
Control Methods Used in ing technology, unlike linear control- by comparing a voltage reference with
Power Electronics lers, suggests there is plenty room for a saw-tooth carrier. This carrier defines
The classic linear control theory and its improvement. This article analyzes this the switching frequency of the converter.
fundamental principles—linear system potential, particularly MPC, through an ­Figure 3(a) shows the typical waveforms
dynamics, feedback regulation, and overview of the present state-of-the-art of the carrier, modulating signal, switch-
stability—were developed in the late and the existing challenges, to provide ing pattern, and the time-average value of
1800s and was already a mature theo- a glimpse of what predictive control has the discontinuous output voltage.
ry in the 1950s. Despite the migration to offer and the role it can play in the
from analog to digital implementation, evolution of power electronics control. AC to DC Converter: Thyristor Rectifier
the fundamentals in control and modu- The time-average principle can be used
lation used in power electronics have The Principle Behind the Classic not only with fully controlled semicon-
remained the same [6], i.e., linear cas- Linear Control: Time-Average ductor, but also with semicontrolled
caded control loops with time-average- In power converters, the semiconduc- semiconductors such as thyristors.
based modulation schemes, as shown tors are operated only in the cutoff and Figure 3(b) shows the waveforms of the
in Figure 2. Nonlinear control tech- saturation regions, behaving essentially input voltage, output voltage, and its
niques were also developed for power as an electronic switch. By avoiding the reference of a three-phase thyristor rec-
electronics, some of which found prac- linear region, the efficiency of power tifier. In this converter topology, the cur-
tical application, although less com- converters is greatly increased, which rent delivered to the load is controlled
mon than their linear counterparts. A comes at the expense of discontinuous by changing the firing angle of the thyris-
successful example has been hystere- waveforms. Because of this switched tors, which modifies the waveform and
sis controllers shown in Figure 2, which behavior, linear control techniques mean value of the load voltage.
have been used in direct torque control cannot be directly applied to power
[7]. These nonlinear control techniques converters, requiring an additional DC to AC Converter: PWM Inverter
were also implemented originally with stage in which the continuous control The conversion from a dc power sup-
analog electronics and later digitalized signal is transformed into high-frequen- ply to an ac load is achieved using an
and programmed in microprocessors. cy binary gating signals that carry the inverter. In this case, the time-average
The revolutionary development of low frequency information as a time- values of the output voltage also fol-
microprocessing technology and the average value, which is then amplified low the continuous reference, but this
ever-increasing computational power by the power converter and filtered by average is valid only during one cycle
has triggered the exploration of new the load (and additional filters in some of the carrier signal. Therefore, a vari-
and more sophisticated approaches applications). This stage is known as able amplitude signal, which resem-
to the control of power electron- modulation and is an essential tool to bles a sinusoidal, can be obtained in
ics, particularly in the last decade. linearize the discontinuous behavior of the output voltage. Figure 3(c) shows
Among them, fuzzy logic control [8], the converter, enabling the analysis and the waveforms of triangular carrier,
sliding mode control [9], and model design of the controller using a continu- continuous reference, and output volt-
predictive control (MPC) [10] have at- ous converter-load model. The modula- age. In this figure, it can be seen that in
tracted increasing attention for power tion algorithm is closely related to the every period of the triangular carrier
electronics because of the nonlinear converter function (ac–ac, dc–dc, dc– signal, the output voltage has a mean
nature of the converters and motor ac, and ac–dc) and the specific power value equal to the reference voltage.
drives, as shown in Figure 2. converter topology. A few examples of
All these control methods have them are shown in Figure 3 and briefly AC to AC Converter: Cycloconverter
unique advantages and disadvantages. described in the next section. The cycloconverter is a type of direct
Hence, it is fair to say that there is no converter in which the output voltage
control technology that outperforms DC to DC Converter: is generated by two thyristor-based rec-
all others in every aspect. Some initial Step-Down Converter tifier connected in antiparallel. These
comparison works show that results The dc–dc converters are the best ex- rectifiers are controlled in such a way
obtained with these more advanced ample to show the relationship between to generate a sinusoidal output voltage.
controllers are comparable to classic the switching pattern and its time aver- Figure 3(d) shows the input voltage,
linear control performance [11], [12] age. These converters are composed by output voltage, and reference of a three-
in their present state of development, two power semiconductors, a diode and phase to single-phase cycloconverter.

10 IEEE industrial electronics magazine  ■ december 2015

 Linear Control Hysteresis Control Fuzzy Control Sliding Mode Control MPC

x i Ts
Surface
ia* δ p
i (t ) i1
S1 p
Reaching i2
S2
Very Low Medium High Very p
Low S3 i3
High

Fuzzy Value
Sa
i *(t ) Sn p
Closed-Loop in
Real Value Sliding
Root Locus
t
y tk tk+1

ia* Sa i *(tk )
+ S (tk )
– Cost Function
vd* va* ia Fuzzy i* Minimization
va*
id* Rules Sa

d,q
+ Sa Sa
– vb* ib* Sb S n
vb* Calc.
id Sb + i*

PWM
vq* vc* Sb Sb

PWM
– + vc* i
–

a,b,c
iq* + Sc ib Sc

Fuzzyfication
– Sc i p(tk + 1) i (tk )

Defuzzyfication
iq i Predictive
ic* Sc Model
+
–
ic

Known Bandwidth by Design Nonlinear Controller, Nonlinear Controller Directly Applicable to Nonlinear Controller
Use of Modulator Makes Very Robust Simple Analogy to Human Switched Converters No Modulation and
Simple to Extend to Different No Modulator Required Response Robust Against Model Coordinate Transformation
Topologies Very Fast Dynamic Linguistic Variables Uncertainties and Can Include Nonlinearities
Fixed Switching Frequency Performance Disturbances and Requirements and
Well Established (Mature, Simple Design Heuristic Design Constraints
Used in Commercial Drives) Well Established (Mature, Problems with Very High Control Activity, i.e., Simple Design Based on
Used in Commercial Drives) Defuzzyfication “Chattering” Prediction Model and Cost
Robustness for Nonlinear Function
Systems Digital Implementation
Not Easy to Adapt for Special Requires High Sample Rate Variable Switching Frequency
Requirements (Constraints, Variable Switching Frequency High Computational
Nonlinearities, etc.) Resonance Issues Requirements
Requires Modulator (Slower Not Easy to Extend to Parameter Uncertainty
Dynamics) Different Converter Heuristic Design of Weighting
Usually Requires Coordinate Topologies Factors
Transformation
FIGURE 2 – The control methods used in power electronics.

december 2015 ■ IEEE industrial electronics magazine  11

 vdc vo* vo ∧ v0 voref
V

t
0 t 0
ton Ts ton Ts α = 20° α = 60°

dc ac

(a) dc dc (b)

dc ac

ac ac
vdc vaN varef voa voaref
∧
2 V

0 t
0
vdc
–
2 Ts ∧
ton ton Ts –V α = 28° α = 148°

(c) (d)

FIGURE 3 – The classic linear control using time-average models for different power converter families. (a) Step-down dc-dc converter. (b) Three-
phase thyristor rectifier. (c) Three-phase PWM inverter. (d) Cycloconverter.

The output voltage waveform in this are several conditions present in many and transform electrical energy using
converter is obtained, similarly to the applications, such as nonlinear plants, nonlinear power semiconductors.
three-phase thyristor, by controlling where linear control presents several
the firing angles appropriately. restrictions in terms of dynamic re- Digital Microprocessors
The four power conversion types sponse, handling of constrains, and Early digital processors appearing at the
operate following the exact same un- even stability when parameters change. beginning of the 1970s had very restrict-
derlying principle, in which a non- Control theory has shown great advanc- ed features regarding processing power
linear waveform is generated with a es during the last years, proposing new and calculation speed. However, the fast
time-average value equal to the de- control techniques with higher capabil- developments of semiconductor tech-
sired reference provided by the linear ities, such as robust control, H-infinite, nology increase their processing capabil-
controller. Although the platforms in fuzzy logic, generalized control, passiv- ity [measured in million instructions per
which these methods are implemented ity based control, etc. These advanced second (MIPS)] at an exponential rate
have changed from analog electronics control strategies outnumbered linear shown in Figure 4. Other features such
to microprocessors, it has become a control in applications with complex, as number of bits representation, float-
mature and reliable technology widely nonlinear, and heavily constrained dy- ing point arithmetic, fast memory op-
adopted across most power electronic namical systems. erations, and the number of processing
applications over the last decades. Among these advanced control cores have been continuously enhanced
strategies, MPC has been found to over time. On the other hand, the price
New Possibilities of Control be the most relevant. After years of of the digital processor has been con-
Theory and Microprocessors development, predictive control has tinuously reduced, as shown in Figure 4.
Technology found increasing acceptance, par- Although early generations of micro-
ticularly in the process industry [13], processors did not have enough capa-
Modern Control Theory being considered as one of the major bilities for control purposes, dedicated
Linear control is the most commonly advances during the last two decades hardware such as DSPs and FPGAs ap-
used control strategy, employed in in the field of control theory [14]. Now, pear from the early 1980s [15]. The
almost every industrial application the scientific community is observing capabilities of such processors also
because of its well-known design and an increasing application of predictive includes analog to digital signal conver-
simple implementation. However, there control to power electronics, to control sion, sequential and parallel processing

12 IEEE industrial electronics magazine  ■ december 2015

capability, digital filter implementation,
and an ever increasing number of math- 106 101
ematical and logical functions. FX-8150
105 10–1
In industrial power electronics con- Athlon 64 Core i7
verters and digital controllers have Core 2
104 Athlon 10–3
steadily been used from since the 1990s

USD/Transistor/MHz
Pentium 4
[16]. They are usually in charge of tasks 103 Pentium 10–5
Pentium III
such as handling signal acquisition, car-

MIPS
102 10–7
rier generation and modulation, filtering, ADSP2100
and control algorithm implementation TMS32010
101 Intel 486 10–9
[6]. The control algorithm includes a PI Motorola
68000 Intel 386
controller, observers, and estimators, 100 10–11
Intel 286
which are usually digital implementa- Intel 8080
tions of their linear counterpart [17], [18]. 10–1 10–13
Intel 4004
The increasing of processing ca-
10–2 10–15
pabilities allows intense calculation 1970 1980 1990 2000 2010 2020
tasks, such as audio, image, and video Year
processing. However, the requirements
for standard control applications have FIGURE 4 – The evolution of microprocessor processing capabilities and costs.
not been increased; hence, there exists
a large gap between calculation require- to have an accurate prediction. These such as commutation losses, common-
ments of power electronics applica- models depend directly on the applica- mode voltages, switching frequency
tions and the calculation capability of tion and the power converter used. For and others, can be also included. Fur-
state-of-the-art microprocessors. This example, in drives, they are basically thermore, nonlinear operation, such
gap has pushed the research, devel- electrical machines whose physical as restrictions and constraints, can be
opment, and implementation of more models have been widely studied such easily added greatly increasing the flex-
complex controller strategies, which as induction, synchronous, permanent ibility and reach of this control strategy.
requires higher processing capabilities, magnets, switched reluctance, etc. For Today, it is possible to find very
such as fuzzy logic, genetic algorithms, other applications, the loads, sources, powerful microprocessors in the mar-
and MPC, as previously discussed. and filters can be modeled as combina- ket, which can perform a large amount
Particularly, in MPC, the increase tions of resistive, inductive, and capac- of calculations at a reduced cost. These
in processing capabilities will allow itive components. Mechanical systems calculations can be fast enough to pre-
its application to tasks that require a and their interaction with electrical dict the behavior of variables like elec-
high amount of calculations such as systems can also be modeled and in- trical currents or voltages in real time
multilevel converters and extended cluded in the MPC algorithm. These without negatively affecting the perfor-
horizon predictions. elements have well-known mathemati- mance of the system under control, en-
cal models established in the theory of abling the use of predictive control in
MPC in Power Electronics electrical networks and machines. power electronics applications.
MPC has found successful application Power converters are composed by
in the chemical and process industry semiconductors operating in only two Operational Principle of
for several decades, mainly because states: cutoff and saturation. Therefore, Predictive Control
of its complex models and slow speed there always exist a finite number of The working principle of MPC is based
dynamics [19]. Power electronics con- possible combinations of the switching on the prediction of the system be-
verters also have complex models and states in any power converter. This fea- havior using a mathematical model
faster dynamics than chemical pro- ture greatly simplifies the application of of it and the optimization of the cost
cesses; therefore, improvements to predictive control because, instead of a function defined using the predicted
microprocessor capabilities and dis- time consuming continuous optimiza- values to fulfill the control objectives.
crete modeling were necessary to en- tion algorithm, direct evaluation of the When applied to power converters
able the use of MPC in such converters switching states can be performed. this working principle can be separated
[20], [21]. On the other hand, one of the The control objectives in power in the following three steps:
most attractive features of predictive electronics are usually to follow ref- ■■ calculation of the predicted variables
control is its intuitive and logical pro- erences of currents, voltages, power, of the system using a discrete model.
cedure to set out the control problem, torque, flux, etc. These objectives are ■■ evaluation of the cost function for
which makes it easy to understand as represented in MPC by a cost function, each one of the switching states of
a concept and simple to implement. which evaluates the errors between the converter.
One of the key elements required in these references and the actual vari- ■■ selection—no optimization and appli-
MPC is a precise model of the system ables. Additional control objectives, cation of the optimal switching state.

december 2015 ■ IEEE industrial electronics magazine  13

 These stages are present in all pre- sequentially or for FPGA, which work Cost Function
dictive control schemes, since it is in parallel. The cost function is a representation
possible to modify each one indepen- of the control objectives which are
dently to adjust the controller scheme Models for Predictive Control usually related to make the variables
to any configuration. For example: if The majority of the models in power follow the references. Hence, the qua-
the load changes, only the discrete electronics converters are based on dratic value of the error, or its absolute
model must be modified; if the con- physical phenomena with well-known value, is commonly employed to find
verter changes, only the switching mathematical models. For example, the minimum value of the cost function.
states in which the model is evaluated dynamical equations of currents, When two or more objectives must be
must be adjusted; and if the control voltages, charges, or fluxes are used achieved simultaneously, they can be
objective changes, only the cost func- to model the most common electri- combined in a sum of error terms. If
tion must be adapted accordingly. cal variables. The instantaneous ac- these variables have the same nature,
The MPC steps can be described tive and reactive power can be also such as direct and quadrature compo-
in the algorithm shown in Figure 5. modeled based on mathematical nent of the current or the active and
The algorithm starts with the vari- arrangements of voltages and cur- reactive power they can be added di-
able measurements at the beginning rents. Mechanical models of torque rectly. If the variables have different na-
of the sampling time. Once all of the and speed, which can be obtained tures, such as magnetic flux and torque
variables are available, the model is from the analysis of electrical ma- or voltages and currents, they must
evaluated for the first switching state chines, can be also useful. Further- be added including a normalization. In
(k = 1, with k being a counter) ob- more, models not directly related to both cases, a weighting factor can be
taining the predicted variables, which physical equations, such as switch- multiplied by these terms to give rela-
are used in the cost function. Depend- ing frequency, commutation losses, tive importance to one or another.
ing on the result, the switching state and even costs can be included in the In power electronics, the most
is selected or discarded and the loop predictive control. common objective is to control the
is repeated. Once the switching states These mathematical models must currents; hence, the cost function is
were evaluated (k = N, where N is be written in discrete time; usually, a composed by the quadratic sum of the
the number of switching states), the first-order model for the next sample current errors
selected switching state is applied time, but more sophisticated models,
to the converter. The previous algo- can be also used depending on the g = (i *s - i s (k + 1)) 2, (1)
rithm can be implemented, adapting system, as will be shown in the follow-
it for microprocessors which work ing sections. where the current and its reference
are expressed as a vector.
Another objective is to control
the magnetic flux and torque of a ma-
Measurements chine connected to an inverter. This
alternative, namely, predictive torque
control, employs a cost function com-
Delay Compensation
posed by the error in torque and the
error in flux amplitude multiplied by a
weighting factor
k = 1 to N
g = (Te (k + 1) - T *e) 2
+ m } (} s (k + 1) - } *s ) 2 . (2)
Variable Predictions

A typical topology of power convert-

Cost Function Evaluation ers correspond to multilevel convert-
ers with floating capacitors. In these
converters, the control of such capaci-
Minimization
tors can be easily included in the cost
function adding the voltage error
k = N?
g = (Te (k + 1) - T *e) 2 + m } (} s (k + 1)
- } *s ) 2 + m v (Tv (k + 1)) 2 . (3)

Apply Optimal State
One of the aspects where MPC
FIGURE 5 – The MPC algorithm. shows its flexibility is the inclusion of

14 IEEE industrial electronics magazine  ■ december 2015

nonlinear elements in model. For ex- connected in parallel with a nonlinear the system and their predicted values.
ample, constraints can be added sim- load through an inductive filter. The idea Since all of the components of the cost
ply by using a logical function where is to control the filter currents to pro- function have the same units and rel-
the value is zero if the variable lies vide the reactive and harmonic current evance within the control scheme, no
inside the allowed region and a very components demanded by the nonlinear weighting factors are required.
high value if it is outside that region load. In this way, the grid currents can be The simulation results for the con-
sinusoidal and balanced, just like in the trol of the active filter are illustrat-
g = (i *s - i s (k + 1)) 2 + m s ( ~ > ~ lim) .  case of feeding a linear balanced load. In ed in Figure 6. A step change in the
(4) this section, we present simulation re- load resistance is applied at instant
sults of a predictive controller in the cur- t = 0.06 [s] to show the transient be-
Additional terms that improve par- rent control of a shunt active filter, where havior of the controlled system. The
ticular features in each converter such the references are computed according dc-link voltage is controlled to a ref-
as to minimize the common-mode volt- to the generalized instant reactive power erence of 1,000 [V]. The nonlinear
age, to reduce the switching frequen- theory. The current reference of the filter load current as well as the grid cur-
cy, to minimize the losses or to reduce has not only reactive and harmonic con- rent is shown together for a phase
total harmonic distortion (THD) can tent but also an active part to control the only. The difference between both
be also included in the cost function. dc-link voltage. This reference is gener- currents is the filter current, which
This feature gives high flexibility to ated by means of a PI controller. is shown along with its correspond-
predictive control, which is very difficult For this simulated example, the ing reference. In the zoom, it can be
to achieve with linear controllers that filter was characterized by a 5-mH observed that, despite the very high
require additional and complex control- inductor with an equivalent series harmonic content of the current ref-
lers or modifications to the modulation. resistance of 0.5 Ω. The predictive al- erence, the tracking provided by the
gorithm was implemented considering predictive controller is very exact.
An Example of MPC: Active Filter a sampling time of 100 μs and a cost With an adequate reference genera-
The current control in an active filter is function given by tion and current control schemes for
a good example of an application that the active filter, the grid currents be-
requires extraordinary fast transient g = i *a - i a (k + 1) + come sinusoidal, as can be seen in the
responses, difficult to achieve through i *b - i b (k + 1) + i c* - i c (k + 1) . waveforms. This example illustrates
a linear control scheme with modula- one of the most important aspects
tion. The power circuit of this system This cost function takes into account of finite-control-set MPC, namely, the
is depicted in Figure 6, which shows the errors between the current refer- excellent dynamic performance of the
a two-level voltage–source converter ences for each of the three phases of reference tracking while maintaining
Grid and Load current (A)

20
is il
10
0
−10
if
−20
0.02 0.04 0.06 0.08 0.1 0.12 0.14
Time (s)
is il
Nonlinear
if Load
Grid

20 1,005
Filter dc Voltage (V)

Vdc
Filter Current (A)

if Vdc
10 1,000
Active
0
Filter
995
−10
−20 990
0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Time (s) Time (s)

FIGURE 6 – The MPC of an active filter: system configuration and results.

december 2015 ■ IEEE industrial electronics magazine  15

a simple structure both from the con- and controllable power supplies. Fur- control power electronics. However,
ceptual and implementation point of thermore, each category is also classi- this control system is far from being
view. A detailed comparison between fied by converter topology. It is worth mature and it has plenty of room for
FCS-MPC and classic linear control noting that several converter topolo- improvements. In this section, some of
methods has been presented in [22]. gies are repeated among different ap- the new advances in predictive control
plications since the voltage source are presented and discussed.
Applications of Predictive inverter used in all four. In addition, the
Control in Power Electronics references for each converter topology Advanced Plant Modeling
Over the last decade, MPC has been ap- are classified by the control objectives. The modeling of the plant is one of the
plied to a wide variety of applications, A recent survey has been fully dedicat- key elements of predictive control. In the
power converters, and control objec- ed to discuss applications of MPC [21]. linear control approach, when a nonlin-
tives. A classification of these applica- ear model is found, a linearization in a
tions is summarized in Figure 7. The Recent Advances given operating point is performed to ob-
references shown in the table are classi- and Future Trends tain a linear model. In predictive control,
fied by application in power quality, ma- As shown in previous sections, MPC has nonlinear models such as the coupled
chine drive, grid-connected converters, proved to be a promising alternative to dynamics of machines and rectifiers can

Application Converter Topology Control Objective References

Current Control, Active and Reactive Power, Voltage
Cascaded H-Bridge [23], [24]
Balance
Current Control, Voltage Control, Voltage Balance,
Flying Capacitor [25], [26]
Power Switching Frequency
Quality Voltage Control, Current Control, Voltage Balance,
NPC Inverter [27]–[30]
Switching Frequency
Voltage Source Inverter
Current Control, Voltage Control [31], [32]
(Single Phase)
Torque and Flux, Input Reactive Power, Common Mode
Direct Matrix Converter [33]–[36]
Voltages, Switching Losses

Indirect Matrix Converter Torque and Flux, Input Reactive Power [37]
Current control, Torque and Flux, Switching Losses,
NPC Inverter [38]–[42]
Voltage Balance
Machine Current Control, Switching Frequency Speed, Current
Drive Voltage Source Inverter Magnitude Minimization, Torque Ripple Torque and Flux, [22], [43]–[54]
Current Limitation
Voltage Source Inverter
Current Control [55]–[61]
(5–6 Ph)
Voltage Source Inverter Hybrid Torque, Current Magnitude Minimization [62], [63]

Current Control, Input Reactive Power, Grid

Indirect Matrix Converter [64], [65]
Synchronization

Modular Multilevel Converter Current Control, Voltage Balance [66]–[72]

Grid-
Connected Active and Reactive Power, Capacitor Balance Current
Converters NPC Rectifier Control, Voltage Control, Resonance Damping, Switching [73]–[75]
Frequency

Active and Reactive Power, Switching Frequency, Current

Voltage Source Rectifier [76]–[86]
Control, Current Ripple, dc Voltage,

Cascaded H-Bridge Current Control, Reduced Number of States [87], [88]

Cascaded H-bridge Asymmetric Current Control, Voltage Control, Voltage Balance [89]
Controllable
Power Flying Capacitor Current Control, Voltage Control, Capacitor Balance [90]–[92]
Supply
Voltage Source Inverter Current Control, Switching Minimization [93]–[96]

DC-DC Current Control, Voltage Control, Power Flow [97]–[103]

FIGURE 7 – The applications, converters, and control objectives of predictive control in literature.

16 IEEE industrial electronics magazine  ■ december 2015

be used directly, because no linear mod- manipulate the inputs to improve sta- [118]. In this category, electrical vari-
el is required [46], [31], [40]. bility [109], it still remains as one of ables with slow dynamics, such as dc
The degree of complexity of the the main concerns of MPC. links voltage and currents in rectifiers
plant model represents a compromise MPC does not need the exact val- or inverters, can be considered [79].
between accuracy of the estimated ues of the plant parameters because it Some alternatives can be considered
variables and the calculation burden uses the information of the derivative in these cases. Either the model can
at each sampling time. Higher-order of the controlled variables. However, be calculated in each sample time ap-
models such as LC filters are usu- when there are parameter mismatches, plying restrictions in the evaluation of
ally controlled to work far from the MPC can still control the plant, but the the cost function or the model consid-
resonant frequency; however, in MPC, dynamical performance is degraded ering the mechanical variables can be
including models of higher order in- [45]. The performance degradation de- calculated with a slower sample time,
creases its controllability [78], [80]. pends on several factors such as cost so it must be modeled separately.
MPC can be also used with non- function, delay compensation, and the
minimum phase plants, which usu- nominal operating point [22]. Nonconventional Control Objectives
ally appear in rectifiers connected to Several methods to deal with this One of the most important advantages
the grid, avoiding linear controllers, parameter mismatch have been pro- of predictive control is the flexibility to
which, in these cases, have a very re- posed in literature [110], the most include nonconventional control objec-
stricted performance [76], [79], [83]. common solution being the use of a tives such as losses [119], THD [120],
parameter estimation algorithm in- temperature [34], and even costs,
Model Discretization cluded into MPC to obtain a more ad- among others. These objectives can be
There are several alternatives to obtain justed model and improve dynamical modeled using derivations of the state
a discrete model of the plant, even if it is response [111]. This parameter esti- variables such as in the THD or using
linear or nonlinear. Usually, a first-order mation can be integrated completely additional models such as thermal dy-
forward Euler approximation is enough into the MPC algorithm using the same namics to calculate the temperature.
to obtain the discrete model [36], [44]. information that MPC uses [112]. Fu-
However, when the sampling frequency ture trends can look for including pa- Cost Function Design
is low, the order of the model is higher rameter supervision and online model and Weighting Factors Calculation
than one or there are high pass filters actualization to improve the stability, The cost function must be construct-
in the plant, first-order Euler approxi- robustness, and dynamics of the con- ed according to the control objec-
mation cannot be used. In those cases, trolled system [113]. tives using the predicted variables
bilinear discretization [40], [45] or ex- from models and its references. The
act discretization [104] must be used. Frequency Domain error between these variables is usu-
An interesting advantage that MPC of- ally evaluated using a convex function
Delay Compensation fers is the direct use of information in such as absolute or quadratic values.
The well-known problem of compensa- the frequency domain. This informa- The selection between these func-
tion of the calculation delay is not par- tion is obtained using discrete models tions depends on the required track-
ticularly related to predictive control in frequency instead of time. There are ing error. If a close tracking of the
but to any digitally implemented control several alternatives to provide these error is required, a quadratic function
technique. When the calculation time is models, such as a complete fast Fou- is better than absolute value [20] for
significant compared to the sampling rier transform implementation [93] terms of the cost function related to
time, there will be a significant delay or directly implementing filters to ob- the tracking.
between the instant where the variables tain a specific frequency range of the The use of constrains in MPC is
are measured and the instant of the ap- modeled variable [114]. Usually, these very intuitive. To apply a constraint,
plication of the next switching state. methods require information stored the variable is simply compared with
Some compensation methods have been from past sample times. the limit value using a logic function.
proposed [105], [106], [107], the most The result of this function is multi-
simple compensation methods consider Mechanical/Electrical Variables plied by a very high weighting factor.
the calculation time in the evaluation Predictive control can be expanded Therefore, if the variable is inside the
of the cost function for the next period, to control not only electrical vari- allowed region, this term has a zero
and apply the selected switching state ables, but also mechanical variables, value, if it is outside the allowed re-
after the next sampling instant based on such as position and speed in ma- gion the term has a very high value
a prediction model in (k + 2) instant. chine drive applications. The main [118]. A continuous function based on
challenge to combine these control exponential or an arctan function can
Stability, Robustness, objectives is the difference in time be also used.
and Parameters Estimation responses, which usually are in milli- It is also necessary to use weighing
Although several efforts have been seconds for current, but they can even factors to combine variables of differ-
developed to prove stability [108] and be seconds for mechanical speed [43], ent natures, such as currents, power, or

december 2015 ■ IEEE industrial electronics magazine  17

speed, to compensate the offset of val- advances in calculation capacity of Biographies
ues or to define priorities. In these cases, digital control systems could allow con- Samir Kouro received his M.S. and
these weighting factors can be obtained siderably longer prediction horizons, Ph.D. degrees in electronics engineer-
using an established procedure [47], but it seems that it is not always neces- ing from the Universidad Tecnica Fed-
[91], [122]. Additionally, it is possible to sary in many electrical systems where erico Santa Maria, Valparaíso, Chile, in
avoid the use of weighting factors replac- controlling fast dynamic transients is 2004 and 2008, respectively, where he
ing the optimization by a ranking-based very important, and simple and fast al- currently serves as an associate profes-
search [123], or including only variables gorithms are more appropriate. sor. From 2009 to 2011, he was a post-
of the same nature in the cost function Several MPC algorithms are focused doctoral fellow in the Department of
and evaluating some of the variables on the application of fixed frequency Electrical and Computer Engineering,
through a state machine [124], [125]. methods to improve harmonic content Ryerson University, Toronto, Canada.
of controlled signals [132], [133]. This He is principal investigator of the So-
Optimization of the Optimization leads to new possibilities in control lar Energy Research Center (Chile)
Although the processor capability has in- strategies, keeping the advantage of not and titular researcher of the Advanced
creased over time, there are some cases using a modulator, but applying the op- Center of Electrical and Electronics En-
where there is still a lack of processor timal switching state for a fixed period. gineering AC3E. He received the IEEE
capability. The best examples are multi- These arguments confirm the fact Industrial Electronics Society J. David
level and multiphase converters where, that MPC has an open future in the Irwin Early Career Award in 2015, the
during each sample time, there can be a coming years. Improvements in con- IEEE Power Electronics Society Rich-
high number of switching states, which trol algorithms are often motivated ard M. Bass Outstanding Young Power
in most of the cases represent redun- by industrial applications or new re- Electronics Engineer Award in 2012, the
dant states, which must be evaluated in search topics, so the evolution of this IEEE Industry Applications Magazine
the model. Several strategies have been control technique will probably come first prize paper award in 2012, the IEEE
developed to find the optimal state with from many of these areas in the future. Transactions on Industrial Electronics
minimum processing requirements. For Best Paper Award of 2011, and the IEEE
example, the search space can be re- Conclusions Industrial Electronics Magazine Best
duced to use the nearest switching states During the last decades, the advances Paper Award of 2008. He is a Member
in terms of number of commutations or in modern control theory and micro- of the IEEE.
distance in the real-imaginary plane [17], processors have made it possible to Marcelo A. Perez received his engi-
[87], [89]. This technique not only re- apply MPC in power electronics in a neer degree in electronic engineering,
duces the number of states that must be natural and simple way. This method his M.S. degree in electrical engineer-
evaluated, but also reduces the dynamic shows a similar performance to the ing, and his D.Sc. degree in electrical
response of the converter. To avoid that, well-established PWM with linear con- engineering from the University of
information of the nearest switching trollers. Moreover, MPC offers high Concepcion, Chile, in 2000, 2003, and
states can be combined with information flexibility to control different convert- 2006, respectively. From 2006 to 2009,
from selected switching states located er topologies and to manage several he held a postdoctoral position at the
far from the actual point to increase the control objectives, without adding sig- Universidad Tecnica Federico Santa
dynamical response [126], [127]. nificant complexity. Maria, Valparaiso, Chile, conducting
MPC has proved to be a modern, research in the area of power convert-
Looking Forward attractive, and competitive alternative ers. Since 2009, he has been an asso-
To improve the steady-state response, for the control of electrical energy us- ciate researcher at the same institu-
in particular, the steady-state error, it ing power semiconductors. MPC can tion. His main interests are control of
is possible to perform the prediction of dramatically change the way electrical power converters, multilevel convert-
the switching states using an extended energy is controlled using power semi- ers, and high-voltage direct current
prediction horizon [128], [129]. Also, conductors, and it offers challenging systems. He is the IEEE-IES Region 9
this extended prediction combined opportunities for research in power Chapter coordinator and president of
with a low-switching frequency can electronics and electrical drives. the IEEE-Industrial Electronics Soci-
be used to improve current THD [95]. ety Chapter Chile. He is a Senior Mem-
Some MPC techniques dealing with Acknowledgment ber of the IEEE.
long prediction horizons and keeping The authors acknowledge the support Jose Rodriguez received his engi-
reasonable levels of computational ef- provided by the Chilean Scientific and neer degree in electrical engineering
fort are present in bibliography, taking Technologic Fund (FONDECYT 1150829), from the Universidad Federico Santa Ma-
into account the compromise between by the Advanced Center for Electrical ria, Valparaiso, Chile, in 1977, and his Dr.-
the exponential increase of the com- and Electronic Engineering(CONICYT/ Ing. degree in electrical engineering from
putational complexity for long predic- FB0008), and by the Solar Energy Re- the University of Erlangen, G ­ ermany, in
tion horizons and the accuracy of the search Center Chile (CONICYT/FONDAP/ 1985. He has been president of Universi-
control method [130], [131]. Recent 15110019). dad Andres Bello, S­ antiago, Chile, since

18 IEEE industrial electronics magazine  ■ december 2015

March 2015. He has coauthored more [5] J. Petrovcic and S. Strmcnik, “A microcomput- [23] C. Townsend, T. Summers, and R. Betz, “Mul-
er-based speed controller for lift drives,” IEEE tigoal heuristic model predictive control
than 350 journal and conference papers. Trans. Ind. Appl., vol. 24, no. 3, pp. 487–498, technique applied to a cascaded H-bridge stat-
His main research interests include mul- May/June 1988. com,” IEEE Trans. Power Electron., vol. 27, no. 3,
[6] R. Gabriel, W. Leonhard, and C. J. Nordby, pp. 1191–1200, Mar. 2012.
tilevel inverters, new converter topolo-
“Field-oriented control of a standard AC motor [24] P. Karamanakos, K. Pavlou, and S. Manias,
gies, control of power converters, and using microprocessors,” IEEE Trans. Ind. Appl., “An enumeration-based model predictive
adjustable-speed drives. He is an associ- vol. IA-16, no. 2, pp. 186–192, Mar. 1980. control strategy for the cascaded H-bridge
[7] M. Depenbrock, “Direct self-control (DSC) of multilevel rectifier,” IEEE Trans. Ind. Elec-
ate editor of IEEE Transactions on Power inverter-fed induction machine,” IEEE Trans. tron., vol. 61, no. 7, pp. 3480–3489, July 2014.
Electronics, Industrial Electronics, and Power Electron., vol. 3, no. 4, pp. 420–429, Oct. [25] F. Defay, A. M. Llor, and M. Fadel, “Direct
1988. control strategy for a four- level three-phase
Industrial Informatics. He has received a [8] G. C. D. Sousa and B. K. Bose, “A fuzzy set flying-capacitor inverter,” IEEE Trans. Ind. Elec-
number of best paper awards from dif- theory based control of a phase-controlled tron., vol. 57, no. 7, pp. 2240–2248, July 2010.
converter DC machine drive,” in Proc. IEEE In- [26] F. Defay, A. M. Llor, and M. Fadel, “A predictive
ferent IEEE journals. He is a fellow of the dustry Applications Society Annual Meeting, IAS control with flying capacitor balancing of a
Chilean Academy of Engineering and 1991, pp. 854–861. multicell active power filter,” IEEE Trans. Ind.
[9] A. Savanovic, R. Benitez, H. Hashimoto, and F. Electron., vol. 55, no. 9, pp. 3212–3220, 2008.
a Fellow of the IEEE. He was the recipi- Harashima, “VSS approach to DC drives con- [27] J. D. Barros and J. F. Silva, “Multilevel optimal
ent of the National Applied Sciences and trol,” in Proc. IEEE Power Electron. Specialists predictive dynamic voltage restorer,” IEEE
Conf., PESC 1988, 11–14 Apr. 1988, vol.1, pp. Trans. Ind. Electron., vol. 57, no. 8, pp. 2747–
Technology Award of Chile 2014. 235–242. 2760, Aug. 2010.
Ana M. Llor received her M.S. [10] P. Cortes, M. P. Kazmierkowski, R. M. Kennel, [28] J. D. Barros and J. F. Silva, “Optimal predictive
D. E. Quevedo, and J. Rodriguez, “Predictive control of three-phase NPC multilevel convert-
degree from the Escuela Politécnica control in power electronics and drives,” IEEE er for power quality applications,” IEEE Trans.
Superior de la Universidad Carlos III, Trans. Ind. Electron., vol. 55, no. 12, pp. 4312– Ind. Electron., vol. 55, no. 10, pp. 3670–3681,
4324, Dec. 2008. Oct. 2008.
Madrid, Spain, in 1998, and her Ph.D. [11] M. Preindl and S. Bolognani, “Comparison of [29] M. Chaves, E. Margato, J. F. Silva, S. F. Pinto,
degree from the Escuela Politécnica direct and PWM model predictive control for and J. Santana, “Fast optimum-predictive con-
power electronic and drive systems,” in Proc. trol and capacitor voltage balancing strategy
Superior de la Universidad Carlos III, 28th IEEE Applied Power Electronics Conf. Expo- for bipolar back-to-back NPC converters in
Madrid, Spain, and the Institut Na- sition (APEC 2013), pp. 2526–2533. high-voltage direct current transmission sys-
[12] T. Geyer, “A comparison of control and modu- tems,” IET Generation, Trans. Distrib., vol. 5, no.
tional des Sciences Appliquées, Lyon, lation schemes for medium-voltage drives: 3, pp. 368–375, 2011.
France, in 2003. She is currently with Emerging predictive control concepts versus [30] P. Acuna, L. Moran, M. Rivera, R. Aguilera,
PWM-based schemes,” IEEE Trans. Ind. Appl., R. Burgos, and V. G. Agelidis, “A single-objective
the Laboratoire de Plasma et Conver- vol. 47, no. 3, pp. 1380–1389, Mar. 2011. predictive control method for a multivariable
sion d’Energie in Toulouse, France, as [13] M. Morari and J. H. Lee, “Model predictive con- single-phase three-level NPC converter-based
trol: Past, present and future,” Comp. Chem. active power filter,” IEEE Trans. Ind. Electron.,
assistant professor, since 2005. Her Eng., vol. 23, no. 4–5, pp. 667–682, May 1999. vol. 62, no. 7, pp. 4598–4607, July 2015.
research interests include parallel and [14] J. H. Lee, “Model predictive control: Review [31] E. Oyarbide, J. Galarza, S. Aurtenechea, and
of the three decades of development,” Int. J. M. A. Rodriguez, “Second-order predictive di-
series multilevel converters for high- rect control of a voltage source inverter cou-
Control Autom. Syst., vol. 9, no. 3, pp. 415–424,
performance applications and predic- 2011. pled to an LC filter,” IET Power Electron., vol. 1,
[15] E. Monmasson, L. Idkhajine, M. Cirstea, I. Bahri, no. 1, pp. 38–49, 2008.
tive control of power converters.
A. Tisan, and M. Naouar, “FPGAs in industrial [32] C. Kumar and M. K. Mishra, “Predictive voltage
Hector A. Young received his B.Eng. control applications,” IEEE Trans. Ind. Inf., control of transformerless dynamic voltage re-
and M.S. degrees in electronics engi- vol. 7, no. 2, pp. 224–243, May 2011. storer,” IEEE Trans. Ind. Electron., vol. 62, no. 5,
[16] C. Buccella, C. Cecati, and H. Latafat, “Digital pp. 2693–2697, May 2015.
neering in 2009 from the Universidad de control of power converters—A survey,” IEEE [33] R. Vargas, U. Ammann, B. Hudoffsky, J. Rodri-
la Frontera, Temuco, Chile. He received Trans. Ind. Inf., vol. 8, no. 3, pp. 437–447, Aug. guez, and P. W. Wheeler, “Predictive torque
2012. control of an induction machine fed by a ma-
his Ph.D. degree in power electronics [17] T. Atalik, M. Deniz, E. Koc, C. Gercek, trix converter with reactive input power con-
from the Universidad Tecnica Federico B. Gultekin, M. Ermis, and I. Cadirci, “Multi-dsp trol,” IEEE Trans. Power Electron., vol. 25, no. 6,
and -fpga based fully-digital control system for pp. 1426–1438, 2010.
Santa Maria, Valparaiso, Chile, in 2014. cascaded multilevel converters used in facts [34] R. Vargas, U. Ammann, and J. Rodriguez, “Pre-
Currently, he is with the Universidad de applications,” IEEE Trans. Ind. Inf., vol. 8, no. 3, dictive approach to increase efficiency and re-
pp. 511–527, Aug. 2012. duce switching losses on matrix converters,”
La Frontera as a lecturer in the area of [18] M. Kazmierkowski, M. Jasinski, and G. Wrona, IEEE Trans. Power Electron., vol. 24, no. 4, pp.
power electronics. His research inter- “Dsp-based control of grid-connected power 894–902, 2009.
converters operating under grid distortions,” [35] R. Vargas, U. Ammann, J. Rodriguez, and
ests include model-based predictive IEEE Trans. Ind. Inf., vol. 7, no. 2, pp. 204–211, J. Pontt, “Predictive strategy to control com-
control of power converters and drives. May 2011. mon-mode voltage in loads fed by matrix con-
[19] H. Han and J. Qiao, “Nonlinear model-pre- verters,” IEEE Trans. Ind. Electron., vol. 55, no.
He is a member of the IEEE Industrial dictive control for industrial processes: An 12, pp. 4372–4380, 2008.
Electronics Society. application to wastewater treatment pro- [36] R. Vargas, J. Rodriguez, U. Ammann, and
cess,” IEEE Trans. Ind. Electron., vol. 61, no. 4, P. W. Wheeler, “Predictive current control of an
pp. 1970–1982, Apr. 2014. induction machine fed by a matrix converter
References [20] S. Kouro, P. Cortes, R. Vargas, U. Ammann, with reactive power control,” IEEE Trans. Ind.
[1] B. J. Baliga, “Power semiconductor devices for and J. Rodriguez, “Model predictive control-a Electron., vol. 55, no. 12, pp. 4362–4371, 2008.
variable-frequency drives,” Proc. IEEE, vol. 82, simple and powerful method to control power [37] M. Lopez, J. Rodriguez, C. Silva, and M. Rivera,
no. 8, pp. 1112–1122, Aug. 1994. converters,” IEEE Trans. Ind. Electron., vol. 56, “Predictive torque control of a multidrive sys-
[2] F. Filsecker, R. Alvarez, and S. Bernet, “Com- no. 6, pp. 1826–1838, June 2009. tem fed by a dual indirect matrix converter,”
parison of 4.5-kV press-pack IGBTs and IGCTs [21] S. Vazquez, J. I. Leon, L. G. Franquelo, J. Rodri- IEEE Trans. Ind. Electron., vol. 62, no. 5, pp.
for medium-voltage converters,” IEEE Trans. guez, H. A. Young, A. Marquez, and P. Zanchetta, 2731–2741, May 2015.
Ind. Electron., vol. 60, no. 2, pp. 440–449, Feb. “Model predictive control: A review of its ap- [38] T. Geyer, G. Papafotiou, and M. Morari, “Model
2013. plications in power electronics,” IEEE Ind. Elec- predictive direct torque control – Part I: Con-
[3] E. P. Cornell and T. A. Lipo, “Modeling and tron. Mag., vol. 8, no. 1, pp. 16–31, Mar. 2014. cept, algorithm, and analysis,” IEEE Trans. Ind.
design of controlled current induction mo- [22] H. A. Young, M. A. Perez, J. Rodriguez, and Electron., vol. 56, no. 6, pp. 1894–1905, June 2009.
tor drive systems,” IEEE Trans. Ind. Appl., H. Abu-Rub, “Assessing finite-control-set [39] G. Papafotiou, J. Kley, K. G. Papadopoulos,
vol. IA-13, no. 4, pp. 321–330, July 1977. model predictive control: A comparison with P. Bohren, and M. Morari, “Model predictive di-
[4] J. Holtz, “Pulsewidth modulation-a sur- a linear current controller in two-level volt- rect torque control - Part II: Implementation and
vey,” IEEE Trans. Ind. Electron., vol. 39, no. 5, age source inverters,” IEEE Ind. Electron. Mag., experimental evaluation, “ IEEE Trans. Ind. Elec-
pp. 410–420, Oct. 1992. vol. 8, no. 1, pp. 44–52, Mar. 2014. tron., vol. 56, no. 6, pp. 1906–1915, June 2009.

december 2015 ■ IEEE industrial electronics magazine  19

[40] T. Laczynski and A. Mertens, “Predictive [55] F. Barrero, M. R. Arahal, R. Gregor, S. Toral, [71] S. Mariethoz, A. Fuchs, and M. Morari, “A VSC-
stator current control for medium voltage and M. J. Duran, “One step modulation pre- HVDC decentralized model predictive control
drives with LC filters,” IEEE Trans. Power dictive current control method for the asym- scheme for fast power tracking,” IEEE Trans.
Electron., vol. 24, no. 11, pp. 2427–2435, Nov. metrical dual three-phase induction machine,” Power Deliv., vol. 29, no. 1, pp. 462–471, Feb.
2009. IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 2014.
[41] V. Yaramasu, W. Bin, S. Alepuz, and S. Kouro, 1974–1983, June 2009. [72] A. Fuchs, M. Imhof, T. Demiray, and M. Morari,
“Predictive control for low-voltage ride- [56] R. Gregor, F. Barrero, S. L. Toral, M. J. Duran, M. R. “Stabilization of large power systems us-
through enhancement of three-level-boost Arahal, J. Prieto, and J. L. Mora, “Predictive-space ing VSC–HVDC and model predictive con-
and NPC-converter-based PMSG wind tur- vector PWM current control method for asymmet- trol,” IEEE Trans. Power Deliv., vol. 29, no. 1,
bine,” IEEE Trans. Ind. Electron., vol. 61, no. 12, rical dual three-phase induction motor drives,” IET pp. 480–488, Feb. 2014.
pp. 6832–6843, Dec. 2014. Electr. Power Appl., vol. 4, no. 1, pp. 26–34, 2010. [73] V. Yaramasu and B. Wu, “Model predictive de-
[42] V. Yaramasu, W. Bin, M. Rivera, and J. Rodri- [57] M. J. Duran, J. Prieto, F. Barrero, and S. Toral, coupled active and reactive power control for
guez, “A new power conversion system for “Predictive current control of dual three- high-power grid-connected four-level diode-
megawatt PMSG wind turbines using four-level phase drives using restrained search tech- clamped inverters,” IEEE Trans. Ind. Electron.,
converters and a simple control scheme based niques,” IEEE Trans. Ind. Electron., vol. 58, vol. 61, no. 7, pp. 3407–3416, July 2014.
on two-step model predictive strategy— no. 8, pp. 3253–3263, Aug. 2011. [74] A. Calle-Prado, S. Alepuz, J. Bordonau, J. Nico-
Part II: Simulation and experimental analysis,” [58] F. Barrero, M. R. Arahal, R. Gregor, S. Toral, las-Apruzzese, P. Cortes, and J. Rodriguez,
IEEE J. Emerg. Sel. Top. Power Electron., vol. 2, and M. J. Duran,”A proof of concept study of “Model predictive current control of grid-
no. 1, pp. 14–25, Mar. 2014. predictive current control for vsi-driven asym- connected neutral-point-clamped converters
[43] E. J. Fuentes, A. Silva, and J. I. Yuz, “Predictive metrical dual three-phase ac machines,” IEEE to meet low-voltage ride-through require-
speed control of a two-mass system driven Trans. Ind. Electron., vol. 56, no. 6, pp. 1937– ments,” IEEE Trans. Ind. Electron., vol. 62, no. 3,
by a permanent magnet synchronous mo- 1954, June 2009. pp. 1503–1514, Mar. 2015.
tor,” IEEE Trans. Ind. Electron., vol. 59, no. 7, [59] M. Arahal, F. Barrero, S. Toral, M. Duran, and [75] V. Yaramasu, B. Wu, and J. Chen, “Model-pre-
pp. 2840–2848, July 2012. R. Gregor, “Multiphase current control using fi- dictive control of grid-tied four-level diode-
[44] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, nite-state model-predictive control,” Control Eng. clamped inverters for high-power wind energy
P. Lezana, P. Cortes, and U. Ammann, “Predic- Practice, vol. 17, no. 5, pp. 579–587, May 2009. conversion systems,” IEEE Trans. Power Elec-
tive current control of a voltage source in- [60] F. Barrero, J. Prieto, E. Levi, R. Gregor, S. Toral, tron., vol. 29, no. 6, pp. 2861–2873, June 2014.
verter,” IEEE Trans. Ind. Electron., vol. 54, no. 1, and M. J. Duran, “An enhanced predictive [76] P. Cortes, J. Rodriguez, P. Antoniewicz, and
pp. 495–503, Jan. 2007. current control method for asymmetrical six- M. Kazmierkowski, “Direct power control of an
[45] J. Rodriguez, R. Kennel, J. R. Espinoza, M. phase motor drives,” IEEE Trans. Ind. Electron., AFE using predictive control,” IEEE Trans. Pow-
Trincado, C. A. Silva, and C. A. Rojas, “High vol. 58, no. 8, pp. 3242–3252, Aug. 2011. er Electron., vol. 23, no. 5, pp. 2516–2523, 2008.
performance control strategies for electrical [61] H. Guzman, M. J. Duran, F. Barrero, B. Bogado, [77] S. Aurtenechea, M. A. Rodriguez, E. Oyarbide,
drives: An experimental assessment,” IEEE and S. Toral, “Speed control of five-phase induc- and J. R. Torrealday, “Predictive control strat-
Trans. Ind. Electron., vol. 59, no. 2, pp. 812–820, tion motors with integrated open-phase fault egy for DC/AC converters based on direct pow-
Feb. 2012. operation using model-based predictive current er control,” IEEE Trans. Ind. Electron., vol. 54,
[46] H. Miranda, P. Cortes, J. I. Yuz, and J. Rodri- control techniques,” IEEE Trans. Ind. Electron., no. 3, pp. 1261–1271, 2007.
guez, “Predictive torque control of induction vol. 61, no. 9, pp. 4474–4484, Sept. 2014. [78] K.-H. Ahmed, A. M. Massoud, S. J. Finney, and
machines based on state-space models,” IEEE [62] D. Zhou, J. Zhao, and Y. Liu, “Predictive torque B. W. Williams, “A modified stationary refer-
Trans. Ind. Electron., vol. 56, no. 6, pp. 1916– control scheme for three-phase four-switch in- ence frame-based predictive current control
1924, 2009. verter-fed induction motor drives with DC-link with zero steady-state error for LCL coupled
[47] S. Davari, D. A. Khaburi, and R. Kennel, “An im- voltages offset suppression,” IEEE Trans. Power inverter-based distributed generation sys-
proved FCS-MPC algorithm for induction mo- Electron., vol. 30, no. 6, pp. 3309–3318, June tems,” IEEE Trans. Ind. Electron., vol. 58, no. 4,
tor with imposed optimized weighting factor,” 2015. pp. 1359–1370, Apr. 2011.
IEEE Trans. Power Electron., vol. 27, no. 3, pp. [63] H. Guzman, F. Barrero, and M. J. Duran, “IGBT- [79] D. E. Quevedo, R. P. Aguilera, M. A. Perez,
1540–1551, 2012. gating failure effect on a fault-tolerant predic- P. Cortes, and R. Lizana, “Model predictive
[48] Y. Cho, K.-B. Lee, J.-H. Song, and Y. I. Lee, tive current-controlled five-phase induction control of an AFE rectifier with dynamic ref-
“Torque-ripple minimization and fast dy- motor drive,” IEEE Trans. Ind. Electron., vol. 62, erences,” IEEE Trans. Power Electron., vol. 27,
namic scheme for torque predictive control no. 1, pp. 15–20, Jan. 2015. no. 7, pp. 3128–3136, July 2012.
of permanent-magnet synchronous motors,” [64] J. L. Elizondo, A. Olloqui, M. Rivera, M. E. Macias, [80] J. Scoltock, T. Geyer, and U. K. Madawala,
IEEE Trans. Power Electron., vol. 30, no. 4, pp. O. Probst, O. M. Micheloud, and J. Rodriguez, “A model predictive direct current control
2182–2190, Apr. 2015. “Model-based predictive rotor current control strategy with predictive references for MV
[49] Y. Zhang and H. Yang, “Model predictive for grid synchronization of a DFIG driven by an grid-connected converters with LCL-filters,”
torque control of induction motor drives indirect matrix converter,” IEEE J. Emerg. Sel. IEEE Trans. Power Electron., vol. 30, no. 10,
with optimal duty cycle control,” IEEE Trans. Top. Power Electron., vol. 2, no. 4, pp. 715–726, pp. 5926–5937, Oct. 2015.
Power Electron., vol. 29, no. 12, pp. 6593–6603, Dec. 2014. [81] Z. Song, W. Chen, and C. Xia, “Predictive direct
Dec. 2014. [65] M. Rivera, J. Rodriguez, B. Wu, J. R. Espinoza, power control for three-phase grid-connected
[50] S. Kwak and S.-k. Mun, “Model predictive con- and C. A. Rojas, “Current control for an indi- converters without sector information and volt-
trol methods to reduce common-mode voltage rect matrix converter with filter resonance age vector selection,” IEEE Trans. Power Elec-
for three-phase voltage source inverters,” IEEE mitigation,” IEEE Trans. Ind. Electron., vol. 59, tron., vol. 29, no. 10, pp. 5518–5531, Oct. 2014.
Trans. Power Electron., vol. 30, no. 9, pp. 5019– no. 1, pp. 71–79, Jan. 2012. [82] Y. Zhang and C. Qu, “Model predictive direct
5035, Sept. 2015. [66] M. A. Perez, J. Rodriguez, E. J. Fuentes, and power control of PWM rectifiers under unbal-
[51] L. Tong, X. Zou, S. Feng, Y. Chen, Y. Kang, Q. F. Kammerer, “Predictive control of AC-AC anced network conditions,” IEEE Trans. Ind.
Huang, and Y. Huang, “An SRF-PLL-based modular multilevel converters,” IEEE Trans. Electron., vol. 62, no. 7, pp. 4011–4022, July 2015.
sensorless vector control using the predic- Ind. Electron., vol. 59, no. 7, p. 2839, 2012. [83] S. Kwak, U.-C. Moon, and J.-C. Park, “Predic-
tive deadbeat algorithm for the direct-driven [67] B. S. Riar, T. Geyer, and U. K. Madawala, “Model tive-control-based direct power control with
permanent magnet synchronous generator,” predictive direct current control of modular an adaptive parameter identification tech-
IEEE Trans. Power Electron., vol. 29, no. 6, pp. multilevel converters: Modeling, analysis, and nique for improved AFE performance,” IEEE
2837–2849, June 2014. experimental evaluation,” IEEE Trans. Power Trans. Power Electron., vol. 29, no. 11, pp. 6178–
[52] C. A. Rojas, J. I. Yuz, C. A. Silva, and J. Rodri- Electron., vol. 30, no. 1, pp. 431–439, Jan. 2015. 6187, Nov. 2014.
guez, “Comments on: Predictive torque con- [68] Z. Wei, Y. Yuan, X. Lei, H. Wang, G. Sun, and [84] J. R. Fischer, S. A. Gonzalez, I. Carugati, M. A. Her-
trol of induction machines based on state- Y. Sun, “Direct-current predictive control ran, M. G. Judewicz, and D. O. Carrica, “Robust
space models,” IEEE Trans. Ind. Electron., vol. strategy for inhibiting commutation failure predictive control of grid-tied converters based
61, no. 3, pp. 1635–1638, Mar. 2014. in HVDC converter,” IEEE Trans. Power Syst., on direct power control,” IEEE Trans. Power Elec-
[53] C.-K. Lin, T.-H. Liu, J.-t. Yu, L.-C. Fu, and vol. 29, no. 5, pp. 2409–2417, Sept. 2014. tron., vol. 29, no. 10, pp. 5634–5643, Oct. 2014.
C.-F. Hsiao, “Model-free predictive current [69] J.-W. Moon, J.-S. Gwon, J.-W. Park, D.-W. Kang, [85] S. Vazquez, A. Marquez, R. Aguilera, D. Que-
control for interior permanent-magnet syn- and J.-M. Kim, “Model predictive control with vedo, J. I. Leon, and L. G. Franquelo, “Pre-
chronous motor drives based on current a reduced number of considered states in a dictive optimal switching sequence direct
difference detection technique,” IEEE Trans. modular multilevel converter for HVDC sys- power control for grid-connected power con-
Ind. Electron., vol. 61, no. 2, pp. 667–681, Feb. tem,” IEEE Trans. Power Deliv., vol. 30, no. 2, verters,” IEEE Trans. Ind. Electron., vol. 62,
2014. pp. 608–617, Apr. 2015. no. 4, pp. 2010–2020, Apr. 2015.
[54] P. Alkorta, O. Barambones, J. A. Cortajarena, [70] J. Bocker, B. Freudenberg, A. The, and S. Di- [86] S. Kwak and J.-C. Park, “Model-predictive di-
and A. Zubizarrreta, “Efficient multivariable eckerhoff, “Experimental comparison of model rect power control with vector preselection
generalized predictive control for sensorless predictive control and cascaded control of the technique for highly efficient active rectifiers,”
induction motor drives,” IEEE Trans. Ind. Elec- modular multilevel converter,” IEEE Trans. Pow- IEEE Trans. Ind. Inf., vol. 11, no. 1, pp. 44–52,
tron., vol. 61, no. 9, pp. 5126–5134, Sept. 2014. er Electron., vol. 30, no. 1, pp. 422–430, Jan. 2015. Feb. 2015.

20 IEEE industrial electronics magazine  ■ december 2015

[87] P. Cortes, A. Wilson, S. Kouro, J. Rodriguez, [103] V. Smidl, S. Janous, and Z. Peroutka, “Im- [119] D. Andler, M. A. Perez, J. Rodriguez, and S.
and H. Abu-Rub, “Model predictive control proved stability of DC catenary fed traction Bernet, “Predictive control of three-level
of multilevel cascaded H-bridge invert- drives using two-stage predictive control,” active NPC converter with evenly energy
ers,” IEEE Trans. Ind. Electron., vol. 57, no. 8, IEEE Trans. Ind. Electron., vol. 62, no. 5, pp. losses distribution,” in Proc. IEEE Int. Power
pp. 2691–2699, Aug. 2010. 3192–3201, May 2015. Electronics Conf.—ECCE ASIA, June 2010, pp.
[88] L. Tarisciotti, P. Zanchetta, A. Watson, [104] M. Rivera, C. A. Rojas, J. Rodriguez, P. Wheel- 754–759.
S. Bifaretti, and J. C. Clare, “Modulated model er, B. Wu, and J. R. Espinoza, “Predictive [120] M. Rivera, J. Rodriguez, J. R. Espinoza, and
predictive control for a seven-level cascaded current control with input filter resonance H. Abu-Rub, “Instantaneous reactive power
H-bridge back-to-back converter,” IEEE Trans. mitigation for a direct matrix converter,” minimization and current control for an indirect
Ind. Electron., vol. 61, no.10, pp. 5375–5383, IEEE Trans. Power Electron., vol. 26, no. 10, matrix converter under a distorted AC-supply,”
Oct. 2014. pp. 2794–2803, Oct. 2011. IEEE Trans. Ind. Inf., vol. 8, no. 3, pp. 482–490,
[89] M. A. Perez, P. Cortes, and J. Rodriguez, “Pre- [105] P. Cortes, J. Rodriguez, C. A. Silva, and Mar. 2012.
dictive control algorithm technique for mul- A. Flores, “Delay compensation in model [121] D.-K. Choi and K.-B. Lee, “Dynamic perfor-
tilevel asymmetric cascaded H-bridge invert- predictive current control of a three-phase mance improvement of AC/DC converter us-
ers,” IEEE Trans. Ind. Electron., vol. 55, no. 12, inverter,” IEEE Trans. Ind. Electron., vol. 59, ing model predictive direct power control
pp. 4354–4361, 2008. no. 2, pp. 1323–1325, Feb. 2012. with finite control set,” IEEE Trans. Ind. Elec-
[90] P. Lezana, R. P. Aguilera, and D. E. Quevedo, [106] P. F. C. Goncalves and S. M. A. Cruz, “A tron., vol. 62, no. 2, pp. 757–767, Feb. 2015.
“Model predictive control of an asymmetric new delay compensation method for finite [122] P. Cortes, S. Kouro, B. La Rocca, R. Vargas,
flying capacitor converter,” IEEE Trans. Ind. control set model predictive control of in- J. Rodriguez, J. I. Leon, S. Vazquez, and L. G.
Electron., vol. 56, no. 6, pp. 1839–1846, 2009. duction motor drives,” in Proc. 2014 16th Franquelo, “Guidelines for weighting factors
[91] S. Thielemans, T. J. Vyncke, and J. A. A. European Conf. Power Electronics and Appli- design in model predictive control of power
Melkebeek, “Weight factor selection for mod- cations (EPE’14-ECCE Europe), 26–28 Aug. converters and drives,” in Proc. IEEE Int. Conf.
el-based predictive control of a four-level fly- 2014, pp. 1–10. Industrial Technology (ICIT 2009), pp. 1–7.
ing capacitor inverter,” IET Power Electron., [107] J. Hu, “Improved dead-beat predictive DPC [123] C. A. Rojas, J. Rodriguez, F. Villarroel, J. R. Es-
vol. 5, no. 3, pp. 323–333, 2012. strategy of grid-connected DC–AC convert- pinoza, C. A. Silva, and M. Trincado, “Predic-
[92] J. Druant, T. Vyncke, F. De Belie, P. Sergeant, and ers with switching loss minimization and de- tive torque and flux control without weight-
J. Melkebeek, “Adding inverter fault detection lay compensations,” IEEE Trans. Ind. Inf., vol. ing factors,” IEEE Trans. Ind. Electron., vol. 60,
to model-based predictive control for flying- 9, no. 2, pp. 728–738, May 2013. no. 2, pp. 681–690, Feb. 2013.
capacitor inverters,” IEEE Trans. Ind. Electron., [108] M. N. Zeilinger, M. Morari, and C. N. Jones, [124] E. Solano, A. Llor, G. Gateau, T. Meynard, and
vol. 62, no. 4, pp. 2054–2063, Apr. 2015. “Soft constrained model predictive control M. Fadel, “Control strategy with variable
[93] T. M. Wolbank, R. Stumberger, A. Lechner, with robust stability guarantees,” IEEE Trans. commutation instants for MPC based on two
and J. Machl, “Novel approach of constant Autom. Control, vol. 59, no. 5, pp. 1190–1202, flying capacitors connected in parallel,” in
switching-frequency inverter control with May 2014. Proc. 2013 15th European Conf. Power Elec-
optimum current transient response,” in Proc. [109] D. Bernardini and A. Bemporad, “Stabilizing tronics and Applications (EPE), 2–6 Sept. 2013,
Int. Power Electron. Elect. Drives Automatic Mo- model predictive control of stochastic con- pp.1–9.
tion (SPEEDAM) Symp., 2010, pp. 803–808. strained linear systems,” IEEE Trans. Autom. [125] F. Villarroel, J. R. Espinoza, C. A. Rojas,
[94] H. Abu-Rub, J. Guzinski, Z. Krzeminski, and Control, vol. 57, no. 6, pp. 1468–1480, June J. Rodriguez, M. Rivera, and D. Sbarbaro,
H. A. Toliyat, “Predictive current control of 2012. “Multiobjective switching state selector for
voltage-source inverters,” IEEE Trans. Ind. [110] R. P. Aguilera and D. E. Quevedo, “Predictive finite-states model predictive control based
Electron., vol. 51, no. 3, pp. 585–593, 2004. control of power converters: Designs with on fuzzy decision making in a matrix convert-
[95] R. P. Aguilera, P. Lezana, and D. E. Quevedo, guaranteed performance,” IEEE Trans. Ind. er,” IEEE Trans. Ind. Electron., vol. 60, no. 2,
“Finite-control-set model predictive control Inf., vol. 11, no. 1, pp. 53–63, Feb. 2015. pp. 589–599, Feb. 2013.
with improved steady-state performance,” [111] P. Antoniewicz and M. P. Kazmierkowski, [126] A. Wilson, P. Cortes, S. Kouro, J. Rodriguez,
IEEE Trans. Ind. Inf., vol. 9, no. 2, pp. 658–667, “Virtual-flux-based predictive direct power and H. Abu-Rub, “Model predictive control
May 2013. control of AC/DC converters with online for cascaded H-bridge multilevel inverters
[96] Z. Zhang, H. Xu, M. Xue, Z. Chen, T. Sun, R. Ken- inductance estimation,” IEEE Trans. Ind. with even power distribution,” in Proc. IEEE
nel, and C. M. Hackl, “Predictive control with Electron., vol. 55, no. 12, pp. 4381–4390, Dec. Int. Conf. Industrial Technology (ICIT 2010), pp.
novel virtual-flux estimation for back-to-back 2008. 1271–1276.
power converters,” IEEE Trans. Ind. Electron., [112] T. Orlowska-Kowalska and M. Kaminski, [127] C. Xia, T. Liu, T. Shi, and Z. Song, “A simplified
vol. 62, no. 5, pp. 2823–2834, May 2015. “Fpga implementation of the multilayer neu- finite-control-set model-predictive control
[97] S. Mariethoz, S. Almer, M. Baja, A. G. Beccuti, ral network for the speed estimation of the for power converters,” IEEE Trans. Ind. Inf.,
D. Patino, A. Wernrud, J. Buisson, H. Cormer- two-mass drive system,” IEEE Trans. Ind. Inf., vol. 10, no. 2, pp. 991–1002, May 2014.
ais, T. Geyer, H. Fujioka, U. T. Jonsson, C.-Y. vol. 7, no. 3, pp. 436–445, Aug. 2011. [128] T. Geyer and D. E. Quevedo, “Multistep finite
Kao, M. Morari, G. Papafotiou, A. Rantzer, and [113] S. Davari, D. A. Khaburi, F. Wang, and R. Kennel, control set model predictive control for pow-
P. Riedinger, “Comparison of hybrid control “Using full order and reduced order observers er electronics,” IEEE Trans. Power Electron.,
techniques for buck and boost DC-DC convert- for robust sensorless predictive torque con- vol. 29, no. 12, pp. 6836–6846, Dec. 2014.
ers,” IEEE Trans. Control Syst. Tech., vol. 18, trol of induction motors,” IEEE Trans. Power [129] P. Karamanakos, T. Geyer, N. Oikonomou,
no. 5, pp. 1126–1145, May 2010. Electron., vol. 27, no. 7, pp. 3424–3433, July F. D. Kieferndorf, and S. Manias, “Direct mod-
[98] P. Karamanakos, T. Geyer, and S. Manias, 2012. el predictive control: A review of strategies
“Direct voltage control of DC–DC boost con- [114] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. that achieve long prediction intervals for
verters using enumeration-based model pre- A. Silva, “Predictive current control strategy power electronics,” IEEE Ind. Electron. Mag.,
dictive control,” IEEE Trans. Power Electron., with imposed load current spectrum,” IEEE vol. 8, no. 1, pp. 32–43, Mar. 2014.
vol. 29, no. 2, pp. 968–978, Feb. 2014. Trans. Power Electron., vol. 23, no. 2, pp. 612– [130] T. Geyer and D. E. Quevedo, “Performance of
[99] B. Hredzak, V. G. Agelidis, and M. Jang, “A 618, Feb. 2008. multistep finite control set model predictive
model predictive control system for a hy- [115] S. Kwak and J.-C. Park, “Switching strategy control for power electronics,” IEEE Trans.
brid battery-ultracapacitor power source,” based on model predictive control of VSI to Power Electron., vol. 30, no. 3, pp. 1633–1644,
IEEE Trans. Power Electron., vol. 29, no. 3, obtain high efficiency and balanced loss dis- Mar. 2015.
pp. 1469–1479, Mar. 2014. tribution,” IEEE Trans. Power Electron., vol. [131] R. Cagienard, P. Grieder, E. C. Kerrigan, and
[100] Q. Zhang, R. Min, Q. Tong, X. Zou, Z. Liu, 29, no. 9, pp. 4551–4567, Sept. 2014. M. Morari, “Move blocking strategies in re-
and A. Shen, “Sensorless predictive current [116] R. O. Ramirez, J. R. Espinoza, F. Villarroel, E. ceding horizon control,” in Proc. 43rd IEEE
controlled DC–DC converter with a self- Maurelia, and M. E. Reyes, “A novel hybrid Conf. Decision and Control, CDC 2004, 14–17
correction differential current observer,” finite control set model predictive control Dec. 2004, vol.2, pp. 2023–2028.
IEEE Trans. Ind. Electron., vol. 61, no. 12, scheme with reduced switching,” IEEE Trans. [132] M. Tomlinson, T. Mouton, R. Kennel, and
pp. 6747–6757, Dec. 2014. Ind. Electron., vol. 61, no. 11, pp. 5912–5920, P. Stolze, “A generic approach to implement-
[101] J. P. Torreglosa, P. Garcia, L. M. Fernandez, Nov. 2014. ing finite-set model predictive control with a
and F. Jurado, “Predictive control for the en- [117] Z. Ma, S. Saeidi, and R. Kennel, “FPGA imple- fixed switching frequency,” in Proc. 2014 IEEE
ergy management of a fuel-cell–battery–su- mentation of model predictive control with 23rd Int. Symp. Industrial Electronics (ISIE),
percapacitor tramway,” IEEE Trans. Ind. Inf., constant switching frequency for PMSM 1–4 June 2014, pp. 330–335.
vol. 10, no. 1, pp. 276–285, Feb. 2014. drives,” IEEE Trans. Ind. Inf., vol. 10, no. 4, pp. [133] R. Mikail, I. Husain, Y. Sozer, M. S. Islam, and
[102] Amin, R. T. Bambang, A. S. Rohman, C. J. 2055–2063, Nov. 2014. T. Sebastian, “A fixed switching frequency
Dronkers, R. Ortega, and A. Sasongko, “En- [118] E. J. Fuentes, C. A. Silva, D. E. Quevedo, and predictive current control method for
ergy management of fuel cell/battery/super- E. I. Silva, “Predictive speed control of a syn- switched reluctance machines,” IEEE Trans.
capacitor hybrid power sources using model chronous permanent magnet motor,” in Proc. Ind. Appl., vol. 50, no. 6, pp. 3717–3726, Nov.–
predictive control,” IEEE Trans. Ind. Inf., IEEE Int. Conf. Industrial Technology (ICIT Dec. 2014.
vol. 10, no. 4, pp. 1992–2002, Nov. 2014. 2009), pp. 1–6. 

december 2015 ■ IEEE industrial electronics magazine  21

View publication stats