T

his article reviews the present state and trends in the devel-
MARIAN P. KAZMIERKOWSKI, opment of key parts of controlled induction motor drive sys-
LEOPOLDO G. FRANQUELO, tems: converter topologies, modulation methods, as well as
JOSE RODRIGUEZ, MARCELO control and estimation techniques. Two- and multilevel volt-
A. PEREZ, and JOSE I. LEON age-source converters, current-source converters, and direct
converters are described. The main part of all the produced
electric energy is used to feed electric motors, and the con-
version of electrical power into mechanical power involves motors ranges from
less than 1 W up to several dozen megawatts. The contemporary drive systems
are expected to meet a variety of requirements as follows:
n maximum conversion efficiency
n wide range steeples adjustment of angular speed, torque, acceleration,
angular, and linear position
Development n fast error elimination when the control and/or disturbance signals are
being changed
of New Converter n maximum utilization of motor power under reduced voltage and current
Topologies n reliable, user-friendly operation.
The present-day drive control engineering heavily draws from the theory of
electric motors, control theory, and industrial electronics; however, it is the last
of these disciplines that plays a decisive role in the development of automated
electrical drive systems. The two basic divisions of industrial electronics, digi-
tal signal processing and power electronics, have paved the way to replace dc
brush motors by ac motors in high-performance drives.
The roots of power electronics go back to 1901 when P.C. Hewitt invented the
glass bulb mercury-arc rectifier [1], [2]. However, the present era of semiconductor
power electronics started with the commercial introduction of a silicon-controlled
rectifier (SCR), popularly called a thyristor, by General Electric (GE) in 1958. Next,
Digital Object Identifier 10.1109/MIE.2011.942173
Date of publication: 23 September 2011

6 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011 1932-4529/11/$26.00&2011IEEE

the development continued in new semi- Therefore, a classification of the typi- Using a modulation strategy to gener-
conductor structures, materials, and cal applications of power electronic ate the gating pulses, it is possible to
fabrication, bringing to the market converters from a low- to high-power synthesize output voltages with the
many new devices with higher power group is given in Figure 1. desired fundamental component. Be-
ratings and improved characteristics. This article presents an overview cause of the high harmonic content in
Today, the most common power of power converter-fed high-perform- the output voltage that need to be fil-
electronic devices are power metal- ance ac drives in three basic parts: tered to produce nearly sinusoidal
oxide-semiconductor field-effect tran- power converter topologies, modula- currents, these converters are usually
sistor and insulated-gate bipolar tion techniques for power convert- applied to inductive loads such as
transistors (IGBTs), and in a very high ers, and control and estimation. motors, but sometimes, depending on
power range, the integrated gate- the application, an additional output
commutated thyristors (IGCTs) [3]. Power Converter Topologies filter could be required.
Also, integrated intelligent power The maximum output voltage
modules are available. The promis- Voltage-Source Converters when a maximum modulation index
ing and new era of high-voltage, high- Voltage-source converters are con- is used is defined by the dc-link volt-
frequency, and high-temperature tech- sidered as a mature technology and age value. To efficiently drive high-
nologies are semiconductor devices have become one of the most com- power loads, a large dc link voltage is
based on wide-band gap silicon car- mon power converter topologies in required, but in practice, this voltage
bide (SiC) material [4]. The new power industry. The basic two-level voltage- is limited by the blocking voltage of
semiconductor devices have always source inverter (2L-VSI) is shown in semiconductors and therefore its tec-
triggered the development of new con- Figure 2(a). It is composed of a dc ca- hnology. For instance, some industrial
verter topologies. At the beginning, the pacitor or voltage source, which gives drives use low-voltage IGBT (LV-IGBT)
topology started with diode- and thyr- its name, and an arrangement of two to provide up to 690-V output voltage.
istor-line-commutated converters and power semiconductors per phase. To avoid this voltage limitation, volt-
continued with modern forced com- The gating signals for both power age-source multilevel converter topol-
mutated converters controlled via pul- switches are complementary, and the ogies have been developed during the
sewidth modulation (PWM) methods. load is connected to the positive or last few decades [5]–[10]. These con-
The list of application of power elec- the negative bar and generates only verters are more complex than the
tronics is too long to list them. two possible output voltage levels. 2L-VSI in terms of topology, modulation,

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 7

 Low Power Medium Power High Power

Power Range Up to 2 kW 2–500 kW More Than 500 kW

Usual Converter
ac/dc, dc/dc ac/dc, dc/dc, dc/ac ac/dc, dc/ac
Topologies

Typical Power
MOSFET MOSFET, IGBT IGBT, IGCT, Thyristor
Semiconductors

Small Volume and Weight

High Power Density, High Nominal Power of the Converter
Technology Trend Low Cost and High
High Efficiency High Power Quality and Stability
Efficiency

Typical
Applications

Low-Power Appliances Electric Roof PV Renewable Transportation Power Industry

Devices Vehicles Energy Distribution

FIGURE 1 – Classification of the power converter applications.

and control, but the additional com- unequal distribution of losses, which in series. In addition, the unequal use
plexity can be considered as an op- causes the outer switches’ switching of the power devices of the topology
portunity to improve its power quality, losses to differ from the central ones (also present in the 3L-NPC for the
reliability, power density, perform- depending on the operation condition. inner power semiconductors) becomes
ance, and efficiency. This problem cannot be easily ad- critical. Finally, the reliability is also
The three-level neutral point dressed using the conventional scheme, reduced due to the increment of the
clamped (3L-NPC) converter is shown and modified topologies such as active components count [20]. The men-
in Figure 2(b) [11]. In this converter, NPC have been proposed [16]. In this tioned drawbacks limit the use of NPC
the dc link voltage is equally divided converter, the clamping diodes are with more than three levels in indus-
using two capacitors, and therefore, replaced by controlled switches; then, trial applications [21].
the phase output can be connected to by selecting the appropriate combina- Converters based on modular
the positive bar switching on the two tion of switches, it is possible to reduce power cells such as the cascaded H-
upper switches, to the midpoint using and equally distribute the losses. The bridge (CHB) and the flying capacitor
the two central switches, or to the neg- NPC converter can be scaled up to (FC) converter have been proposed
ative bar using the two lower switches achieve more than three levels by sim- so far to provide higher number of
[12]. Each semiconductor needs to ply dividing the dc link into more than output voltage levels than the 3L-NPC.
block only half of the total dc-link volt- two values using several capacitors The CHB converter topology, shown
age, allowing the increase of the power [18]. Each one of these partial dc-link in Figure 2(c), is a highly modular con-
rating using the same semiconductor voltages can be connected to the load verter based on several single-phase
technology than the conventional using an expanded arrangement of inverters, usually called power cells,
2L-VSI. The semiconductors normally switches and clamping diodes. Along connected in series to form a phase
used in these converters are the high- with the increased power rating, the [22], [23]. Each power cell is imple-
voltage IGBT (HV-IGBT) [13] and the advantages of several output voltage mented based on standard low-volt-
IGCT [14]. levels are its better power quality, age components, which provides an
A well-documented problem pre- smaller dv=dt, and the associated elec- easy and cheap replacement in case
sented in this converter is the unbal- tromagnetic interference [19]. How- of failure [24]. The main advantage of
ance of the capacitors produced by ever, when the NPC converter has this converter is that, using only low-
asymmetries in the converter hard- more than three levels, additional voltage components, it is possible to
ware and mainly the operational con- problems arise. From the point of view drive medium-voltage high-power loads.
ditions [15]. Several strategies to deal of the power topology, the clamping Although the switching frequency in
with this problem have been proposed, diodes require higher blocking voltage each cell is low, the equivalent switch-
mainly focusing on the modification than the main switches; therefore, it is ing frequency applied to the load is
of the modulation strategy [16], [17]. necessary to use different technology high, which reduces the switching
Another issue of this converter is the or several clamping diodes connected losses, produces low dv=dt, and helps

8 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

 P
P
sa sb sc sa1 sb1 sc1
+ Vdc ia + Vdc
2 va 2
ib s sb2 sc2
0 vb M da1 a2 db1 dc1
Vdc
ic 0
+ Vdc vc Vdc va vb vc
da2 db2 dc2
2 sa1 sb1 sc1
sa sb sc + Vdc
N 2
(a) sa2 sb2 sc2

0 N ia ib ic

sp sn sp sn sp sn M
Vdc Vdc Vdc
+ + +
Voa1 Vob1 Voc1 (b)
sp sn sp sn sp sn P

sa1 sb1 sc1

sp sn sp sn sp sn + Vdc
2 sa2 sb2 sc2
Vdc Vdc Vdc
+ + +
Voa2 Vob2 Voc2
sp sn sp sn sp sn Vdc Vdc sa3 Vdc Vdc sb3 Vdc Vdc sc3
Vdc
2 4 2 4 2 4
0 va vb vc

sp sn sp sn sp sn sa3 sb3 sc3

Vdc Vdc Vdc
+ + +
Voa3 Vob3 Voc3 sa2 sb2 sc2
+ Vdc
sp sn sp sn sp sn 2
sa1 sb1 sc1
ia ib
va vb ic
vc N ia ib ic
M
M
(c) (d)
va0 va0

Time Time

(e) (f)

va0 va0

Time Time

(g) (h)

FIGURE 2 – Topologies and phase voltages of the conventional two-level and multilevel voltage-source inverters. (a) Two-level VSI.
(b) Three-level NPC converter. (c) Seven-level CHB converter. (d) Five-level FC converter. (e) Two-level VSI. (f) Three-level NPC inverter.
(g) Five-level FC inverter. (h) Seven-level CHB inverter.

to avoid resonances [25]. Additionally, In the CHB converter, each power using a multipulse transformer with
several fault-tolerant strategies can be cell requires an isolated dc power several secondary diode rectifiers and
implemented by increasing their reli- source that can be obtained, for exam- a dc-link capacitor. The first option
ability and overall availability [26]. ple, from photovoltaic (PV) panels or provides high scalability and allows

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 9

the PV energy to be directly connected using only a reduced number of power using, for example, single-phase
to the medium-voltage grid. The sec- cells [33]. It can produce up to 27 out- 3L-NPC as a power cell of the CHB [35]
ond option has become the standard put voltage levels using only three or combining an active NPC with an FC
for drive applications, providing power cells (only a nine-level conven- [36]. Recently, the modular multilevel
galvanic isolation and high-quality tional symmetrical CHB converter), converter has been proposed as a
input current, canceling out the low- but it requires the use of different devi- complete topology [37], [38]. This
order harmonics produced by the ces in each power cell and a control converter can operate with dc or sin-
diode rectifiers. The cancelation of system to avoid regeneration in the gle-phase ac sources. The cells are
current harmonics is produced by a low-voltage cells [29], [30]. based on dc-dc boost converter for
phase shift among the transformer The FC converter topology is shown the dc cell and a half-bridge inverter
secondary winding, which leads to a in Figure 2(d). The output voltage is for the ac cell, both cases with a
more complex transformer, espe- generated by directly connecting the floating dc capacitor. This converter
cially, when a high number of cells is phase output to the positive or nega- does not require isolated dc sources
used. This input transformer, how- tive bars or through the floating ca- because each cell has a floating dc
ever, reduces the scalability of the pacitors. The number of output voltage capacitor. The number of cell in series
converter. The dc link is designed to levels depends on the number of float- in this converter is very large com-
have a very large capacitance which, ing capacitors and the relation between pared with CHB (>200) generating
on the one hand, allows having a their different dc voltages. For example, very low dv=dt and a high-equivalent
good ridethrough capability, but on the topology shown in Figure 2(d) gen- switching frequency. This converter
the other hand, increases the size of erates five output voltage levels. This always requires a dc voltage control
each cell, where at least half of its size converter, as in the CHB case, also scheme because the floating dc vol-
are only capacitors and requires presents a modular topology, where tages must be kept at the reference
additional circuits to safely charge each cell is composed of a dc capacitor value. There are independent control-
and discharge the dc link. and two complementary switches. lers implemented to handle the input
As mentioned above, the standard However, on the contrary of the CHB and output currents and an additional
solution to provide the isolated dc case, the addition of extra power cells controller to minimize the circulating
sources is using a diode rectifier that in the FC converter does not increase current that flows only inside the con-
does not provide regenerative opera- the nominal power of the converter verter. This circulating current is a
tion and cannot be used in applica- but only reduces the dv=dt by improv- particular characteristic of this con-
tions such as downhill conveyors, ing the harmonic content of the out- verter [39], [40].
liquefied natural gas turbines where put waveforms. As a CHB converter,
the load is always regenerating or, at the modularity reduces the cost of Current-Source Converters
least, the great part of its operating component replacement and mainte- The standard topology of a current-
time. In these cases, a regenerative nance and allows implementing fault- source inverter (CSI) is shown in Fig-
converter, where the diode-based re- tolerant strategies. Finally, the FC ure 3(a). These converters always re-
ctifiers are replaced by controlled rec- converter also has its asymmetrical quire a controlled rectifier to provide
tifiers, has been proposed [27]. This version; for instance, if the floating a constant current flowing through
alternative can control bidirectional dc voltage relations of the topology the dc link. In the basic topology, a
active power, which additionally gives shown in Figure 2(d) are changed to 3 thyristor-based rectifier is usually
precise control of the dc-link voltage, Vdc/7 and Vdc/7, instead of Vdc/2 and chosen. A split inductor is used in the
the input reactive power, and provides Vdc/4, respectively, it is possible to gen- dc link to reduce the common mode
sinusoidal input currents that help to erate eight different symmetrical out- voltage in the load [40]. The inverter
simplify the input transformer configu- put voltage levels [33], [34]. As in the is an arrangement of power semicon-
ration. The drawbacks of the regenera- asymmetrical CHB case, the increase of ductors similar to the VSI but com-
tive topology are the requirement of levels is achieved at the expense of los- posed of integrated gate-commutated
inductive input filters and its associ- ing the converter modularity. thyristors (IGCTs). The output cur-
ated cooling system, the increased An FC converter requires only one rent has a PWM waveform and cannot
losses by the forced commutation of dc source to feed all the cells and be directly applied to an inductive
the controlled rectifier, and the reduc- phases. Therefore, the input trans- load such as the motor. An output
tion in reliability [28], [29]. former can be avoided, and the num- capacitive filter that eliminates the
Several topologies based on the ber of cells can be arbitrary increased di=dt and applies a very smooth volt-
cascaded approach have been pro- depending on the required output age to the load is mandatory [41]. This
posed in the literature. In particular, power. Similar to NPC, this converter converter can achieve medium volt-
the asymmetrical CHB converter, built requires a control strategy to regulate age operation and, moreover, is in-
using different values of the dc voltage the voltages in the capacitors. trinsically regenerative [42], [43]. A
for each power cell, allows increasing Along with the classical structures, back-to-back configuration as shown
the number of output voltage levels there are several mixed structures, in Figure 3(b) can be used when a

10 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

high-quality input current is required.
In this case, the input side also requires The torque control methods are presented in two
a second-order inductive-capacitive
(LC) filter to filter harmonics due to main groups: linear and nonlinear controllers.
the PWM currents presented at the
input terminals of the rectifier [44].
The matrix converter, in its direct a three-phase inverter [54], as shown
Direct Converters and indirect versions, also belongs to in Figure 4(e). The number of power
Direct converters transmit the energy the direct converter category. The direct semiconductors is the same as the
directly from the input side to the out- matrix converter (DMC) is shown in DMC if the bidirectional switch is con-
put side without using energy-storage Figure 4(c). The basic operation prin- sidered as two unidirectional switches,
elements. The main advantage of these ciple behind this converter is the con- but the number of possible switching
converters is their size reduction; how- nection of the phase output to any of states is different, and its analysis is
ever, the complexity of the control the input voltages [49], [50]. The con- simpler. Using the same configuration
scheme is increased. verter is composed of nine bidirec- of the IMC, it is possible to simplify its
The cycloconverter, shown in tional switches that can connect any topology and reduce the components
Figure 4(a), has been widely used in input phase to any output phase al- count restricting its operation to posi-
high-power applications such as grind- lowing the current flow in both direc- tive voltages in the virtual dc link [55].
ing mills and belongs to the direct tions. A second-order inductive- This reduced topology is called sparse
converter category [45]. This con- capacitive input filter is required to matrix converter (SMC), and it is
verter is composed by a thyristor- improve the input current [51]. The shown in Figure 4(f).
based dual converter per phase, output is directly connected to the
which can produce a variable dc inductive load. Not all the possible Modulation Techniques for
voltage controlled to follow a sinu- combinations of switches are al- Power Converters
soidal waveform reference [46]. The lowed, and they are restricted to The vast development of power con-
input of each converter is fed by a only 27 valid switching states. As verters for a large number of applica-
phase-shifted transformer, where low- mentioned earlier, the main advant- tions in the last 50 years has led to
order harmonics of the input current age of a matrix converter is the re- the creation of very important research
are canceled out. The output voltage duction in size, being specially suited lines being implemented all over the
results in a combination of segments for automotive and aircraft applica- world. Among these lines, modula-
of the input voltage in which the fun- tions [52], [53]. tion has been one of the most impor-
damental component follows a sinu- Several variations in the DMC tant topics in the last few decades
soidal reference as can be seen in have been proposed in the litera- because it is directly related to the
Figure 4(b). Due to its operation prin- ture. The most important one is the efficiency of the overall power
ciple, this converter is well suited to indirect matrix converter (IMC) that system affecting the economical
drive high-power low-frequency loads is composed of a bidirectional three- profit and performance of the final
[47], [48]. phase rectifier, a virtual dc link, and product.

idc idc

vsa isa sap sbp scp v ioa vsa isa vca voa ioa
oa
vsb isb ira vob iob vsb isb vcb ira vob iob
M M
vsc isc voc ioc vsc isc vcc voc ioc
san sbn scn

(a) (b)

ira voa ioa

Time Time Time

(c) (d) (e)

FIGURE 3 – Topologies and output waveforms of the conventional CSIs. (a) CSI. (b) Current-source back-to-back converter. (c) Output cur-
rent. (d) Output voltage. (e) Load current.

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 11

 vor
–20° ior
vor
vs

vos
vsa isa M
0° ios
vsb isb
vsc isc
Time
0
vot
+20° iot
ior

(a) (b)
vsa isa vca ica vsb
vor vsa
vsc
vsb isb vcb icb

vsc isc vcc icc

ior ios iot

vor vos vot
=
Time
M
ior
Bidirectional Switch
(c) (d)

P P

vsa isa vca vr vsa isa vca vr

vsb isb vcb vsb isb vcb
vdc vs M vdc vs M
vsc isc vcc vsc isc vcc
vt vt

N N
(e) (f)

FIGURE 4 – Topologies and output waveforms of the direct converters. (a) Cycloconverter. (b) Cycloconverter waveforms. (c) DMC.
(d) Matrix converter waveforms. (e) IMC. (f) SMC.

The fundamental objective of a A classification of the modula- modulation techniques will be ex-
modulation technique is to obtain the tion techniques for power convert- plained in detail.
best waveforms (voltages and cur- ers is shown in Figure 5. In this
rents) with minimum losses. Other classification, the modulation tech- Pulsewidth Modulation
secondary control objectives can be niques are divided into four main
dealt with the proper modulation tec- groups: PWM, space vector modula- Conventional Bipolar and
hnique such as common-mode volt- tion (SVM), harmonic control modu- Unipolar PWM
age reduction, dc voltage balancing, lation, and other variable switching Among the most commonly used mod-
input current harmonics minimiza- frequency methods. These modula- ulation techniques, PWM is the most
tion, and low dv=dt. Simultaneously tion techniques were applied to the successful modulation method because
achieving all the control targets is power converters since several dec- of its high performance, simplicity, fixed
impossible, so a tradeoff is needed. ades and are the most mature mod- switching frequency, and easy digital
Each power converter topology and ulation techniques implemented in and analog implementation [56]. The
each application has to be studied in the power converters as commer- PWM technique is based on the
depth to determine which modulation cial products. In the following sec- comparison of a reference signal with
technique suits most. tions, the proposed classification of frequency f1 , usually a sinusoidal

12 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

 Conventional dc/dc, ac/dc, dc/ac Modulation Techniques

PWM Two-Level, αβ Frame SHE SHM Predictive Control Hysteresis Control

i ref –Δ < i a < i ref +Δ

α h5 = h7 =…= 0 α hi = Ei < Li x(tk+2) = min{g}, i = 1,…,8

Phase-Shifted PWM N–Level in αβ Frame Staircase SHE or SHM Predictive Control

h5 = h7 =…= 0 or hi = Ei < Li x(tk+2) = min{g}, i = 1,…,m

Level-Shifted PWM Multiphase αβ–xy

Frame
Nearest Level Control

Single Phase

Hybrid Modulation Space Vector Control

High Switching Frequency

Mixed Switching Frequency
Low Switching Frequency
Variable Switching Frequency

Extension to Multilevel Converters

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 13

FIGURE 5 – Classification of the most common modulation techniques for power converters.
 medium-high power. It has to be
Implementation of any high-performance noticed that for high-power applica-
tions, the switching losses have to be
drive system requires a high accuracy taken into consideration and Rf has
to be as low as possible. This fact
estimation of the actual stator or/and rotor leads to create a low harmonic distor-
flux vector (magnitude and position) and tion because the carrier frequency is
inside the bandwidth, which has to
electromagnetic torque. be kept at a maximum. Therefore,
passive filters usually formed by in-
reference, with a triangular carrier can be obtained if a unipolar PWM is ductors and capacitors have to be
with frequency fc . If the reference is applied. part of the power system when PWM
above or below the triangular carrier, A way to achieve a better utiliza- is used for high-power applications.
some power semiconductors are tion of the dc voltage of the con- The design of these passive filters is
switched on and off or vice versa. verter can be obtained by a third extremely important because they are
As the power converter is inherently harmonic injection in the reference very bulky, heavy, and expensive. So,
a nonlinear system, the obtained out- voltage [57]. In this way, the modu- there is a tradeoff between switching
put voltage waveform is a switched lation index (related to the maxi- losses and cost of the filter.
signal with the possible output vol- mum voltage that can be generated
tages of the power converter. The har- by the converter) can be extended Extension of PWM to
monic spectrum of this switched up to 15.47%. The third harmonic Multilevel Converters
waveform has the fundamental har- injection does not represent a prob- PWM technique is suitable to be used
monic equal to the fundamental har- lem because it is canceled in the in simple dc/dc, ac/dc, and dc/ac con-
monic of the reference signal f1 but line-to-line voltages. An easier way verters with low number of power
some harmonic distortion appear to carry out this third harmonic semiconductors per phase. Additional
around the carrier frequency fc (and injection is to use a min–max se- considerations have to be taken into
their multiple values) due to the quence injection using the following account if a PWM technique is applied
switching of the converter. The fre- expression: to more complex power converter top-
quency ratio is defined as Rf ¼ fc =f1 ologies. Among these complex new
and has to be high enough to avoid vacomp ¼ converter topologies, the multilevel
the presence of low-order harmonic minfva , vb , vc gþ maxfva , vb , vc g converters are especially attractive
distortion. So, Rf is usually limited to va  , for high-power applications due to
2
values higher than ten. Ideally, to (1) the increase of nominal power at the
obtain an output waveform with mini- expense of a high number of medium-
vbcomp ¼
mum distortion and to avoid the low- low voltage power semiconductors.
minfva , vb , vc gþmaxfva , vb , vc g
frequency audible noise, Rf has to be vb  , This makes a conventional bipolar or
very large. However, if Rf is high, the 2 unipolar PWM not to be directly ap-
(2)
number of commutations of the power plied to the topology to determine the
semiconductors is large as well leading vccomp ¼ switching states of the power devices.
to inacceptable switching losses from minfva , vb , vc g þmaxfva , vb , vc g The PWM techniques for multile-
the point of view of efficiency and vc  , vel converters are an extension of the
2
power dissipation. (3) conventional PWM techniques there-
The basic PWM can be bipolar or by increasing the number of triangular
unipolar generating an output wave- where va , vb , and vc are the reference carriers to carry out the comparisons
form with two or three voltage levels, voltages for phases a, b, and c, with the reference voltage. In this way,
respectively, using one reference and respectively, and vacomp , vbcomp , and as is shown in Figure 5, the most com-
one carrier (bipolar PWM) or one vccomp are the modified reference volt- mon PWM methods for multilevel
reference and two opposite carriers age for phases a, b, and c, respectively. converters are the level-shifted and
(unipolar PWM). For instance, if a The min–max sequence injection has phase-shifted PWM techniques [58].
conventional two-level three-phase the advantage of avoiding the presence Each one of these multicarrier me-
VSI is considered, a bipolar PWM of a PLL because it is not necessary to thods is specifically well suited to be
technique can be applied obtaining a know the angle of the reference voltage applied to a multilevel converter topol-
phase voltage, which is a switched (in the conventional third-harmonic ogy. In fact, the level-shifted PWM tec-
voltage with values Vdc/2 and Vdc/2. injection, a synchronization is needed). hnique can be easily applied to the
On the other hand, if an H-bridge The PWM technique has been ex- NPC converter because the switching
is considered, an output voltage tensively applied to power convert- of the converter is directly related
with values V dc /2, 0, and V dc /2 ers for any application from low- to to the comparison of the reference

14 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

voltage with the triangular carriers. signals of a power converter. Unlike that, in the multilevel case, there are
On the other hand, the phase-shifted the PWM method, the SVM technique several redundant vectors different
PWM technique is especially well de- does not use triangular carriers and from the zero vector. This means
signed to be applied to the FC topol- comparators to generate the switch- that different switching states are
ogy. In this case, a bipolar PWM is ing of a power converter. The SVM represented using the same vector
applied to each power cell, but intro- uses a vector representation of the in the abc frame. Any of these redun-
ducing a phase shift in the triangular possible phase to neutral voltages dant vectors can be used in the
carrier of consecutive power cells than can be generated by the con- switching sequence without distort-
equal to p=(N  1), where N is the verter [6]. As the converter is a non- ing the line-to-line voltage, and there-
number of levels of the converter. In linear discrete system, it only has a fore, this concept can be used to
addition, the use of the phase-shifted discrete number of switching states improve some converter feature
modulation in a symmetrical flying- that generate a limited number of such as the dc voltage balancing and
capacitor converter leads to a natural phase voltages that can be plotted in the switching losses equalization or
balance of the FC voltages without the abc frame. This representation minimization.
using sensors. Also, the phase-shifted defines the possible vectors that can In the last few years, new SVM
PWM technique is well suited to be be used by the converter (i.e., seven techniques based on the natural Car-
applied to the CHB converter. In this different vectors for the eight possi- tesian abc coordinates have been
case, an unipolar PWM is imple- ble switching states of the conven- proposed [61]. The resulting modula-
mented in each H-bridge with a phase tional three-phase two-level converter). tion technique is a three-dimensional
shift in the triangular carriers of con- The c component of these vectors is SVM technique and is similar in terms
secutive H-bridges equal to p=(2m), zero in power systems where the neu- of computational cost compared with
where m is the number of H-bridges tral current is zero and this leads to a conventional SVM methods based on
per phase of the converter. two-dimensional representation using the abc representation. However, the
Both modulation techniques can the ab frame. The reference voltage is abc frame-based SVM techniques can
be compared in terms of harmonic represented as well using the same also be applied to converter topolo-
distortion and switching losses. It frame, and it can be generated as a lin- gies where the common-mode current
can be said that, assuming the same ear combination of the three nearest is not zero. In addition, it naturally
switching frequency of the output switching vectors. It has to be noticed eliminates the common-mode voltage
voltage, the level-shifted PWM meth- that the conventional SVM technique average over a sampling period at the
od achieves a better harmonic perform- has been claimed superior to PWM due expense of not dealing with the third
ance because it avoids the appearance to its better dc-link voltage utilization, harmonic injection. This is achieved
of side bands around the carrier fre- but, in fact, the SVM is a vector repre- determining the specific duty cycle
quency. However, it is achieved at the sentation of the PWM method with a of each redundant switching vector
expense of nonequally distributed third harmonic injection. Despite this of the switching sequence, which is
switching losses between the power fact, the SVM technique is very interest- the difference with the conventional
semiconductors and the consequent ing because the switching sequence abc frame-based SVM technique when
issues about asymmetrical power dis- can be manually changed by taking into is applied to power converters with
sipation and reliability reduction. This account control targets such as com- common-mode current equal to zero.
problem is not present if the phase- mon-mode voltage elimination, current On the other hand, in [62] it is demon-
shifted modulation is applied because ripple reduction, and switching losses strated that the abc frame-based SVM
it intrinsically imposes a very similar minimization. techniques are equivalent to a time-
switching pattern to each power cell based level-shifted PWM method ap-
of the converter. In addition, the effec- Extension of SVM for plied to each phase of the converter.
tive switching frequency of output Multilevel Converters
voltage is m times the carrier fre- The extension of the conventional Harmonic Control Techniques
quency used in the bipolar or unipolar SVM technique for multilevel con-
PWM method (where m is the number verters is straightforward because it Selective Harmonic
of power cells of the converter). This consists of the inclusion of all the Elimination Technique
makes the phase-shifted modulation additional switching states in the abc Selective harmonic elimination (SHE)
superior to the level-shifted method frame. As the number of switching is especially well suited for high-
in terms of harmonic distortion as well. states of a multilevel converter is power drives where the switching
higher than the two-level case, a high frequency is restricted by the switch-
Space Vector Modulation number of switching vectors appear. ing losses. The SHE technique is used
To generate the reference vector, to achieve very low switching fre-
Conventional SVM Technique again the linear combination of the quency by avoiding the presence of
The SVM is an alternative technique three closest vectors has to be used low-order harmonics. The conven-
to determine the switching gate [60]. However, it has to be noticed tional SHE algorithm is based on the

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 15

calculation of a reduced number of (SHM) [66]. The SHM method can be on a comparison between the mea-
switching angles of the output volt- used in applications where the SHE sured control variable xm (usually a
age pattern to eliminate a given num- technique operates obtaining better current) with a reference waveform
ber of harmonics using the Fourier results in terms of the resulting neces- xref with a hysteresis band Dx. The
decomposition [63]. The resulting ex- sary filter to be used in the power sys- aim of the controller is to keep the
pressions are nonlinear and very dif- tem. The SHM technique is based on control variable inside the hysteresis
ficult to solve online. For this reason, the offline calculation of the switching band, leading to a nonperiodic switch-
the switching angles are determined angles defined, and the cost function ing of the power devices of the con-
offline using mathematical search is to be minimized by a search algo- verter topology. The control variable
algorithms such as genetic, particle rithm. This cost function includes ripple is directly defined by the hyster-
swarm, Tabu, and simulated anneal- terms for the harmonic distortion com- esis band until the hysteresis band is
ing. The resulting angles are stored in ponents, the THD, and the minimum smaller than the dynamic response of
a lookup table for a range of modula- time between consecutive switching the load. To design a hysteresis con-
tion indexes and interpolated to gen- angles. If the harmonic distortion com- troller, the effect of the switching of
erate the switching pattern. ponents or the THD value are below the converter on the control variable
As the SHE technique is based on the limit imposed by the applied grid has to be known. In this way, the con-
the Fourier transform calculation of a code, its contribution to the cost func- troller design is very simple at the
period of the voltage waveform, it is tion is zero. expense of generating a spread har-
difficult to apply this modulation in monic spectrum [67].
fast dynamic close-loop applications. Variable Frequency
Several approaches to calculate the Modulation Techniques Nearest Level Control
switching angles online have been The fixed switching frequency mod- The nearest level control determines
proposed in the literature [64], [65]. ulation methods are superior to the the switching of the power converter
variable frequency ones in terms of comparing the phase voltage reference
Selective Harmonic better harmonic distortion distribu- with the possible voltages that can be
Mitigation Technique tion. The fixed frequency methods generated by the converter topology.
The grid requirements, imposed by the present a harmonic spectrum where Finally, the switching associated to the
governments and the electrical pro- the switching distortion is located nearest voltage level is applied, and, in
viders, usually define the maximum around the switching frequency and this way, as only one switching state is
limits of the harmonic distortion for its multiples. This does not happen used during each sampling period, the
each harmonic order up to 50 and a when a variable frequency modula- switching losses are reduced com-
maximum value of the total harmonic tion method is used when a spread pared with conventional PWM or SVM
distortion (THD). In this way, the elimi- harmonic spectrum is generated. techniques [68]. This is achieved at the
nation of low-order harmonic compo- However, in some specific conditions, expense of a more distorted output
nents by the SHE technique does not variable frequency modulation meth- voltage because the reference voltage
ensure to comply with harmonic re- ods that are mathematically and con- is not generated in an average over a
strictions imposed by the current grid ceptually simple can also be applied sampling period. The nearest level con-
codes. The noneliminated harmonics to achieve good results. It has to be trol can only be applied achieving
present when the SHE technique is noticed that the variable frequency high performance in power convert-
applied can take any value and are modulation techniques are not indeed ers, where the difference between
mostly above the limits imposed by the modulation methods because they do consecutive voltage levels is not
grid codes. This fact leads to design not synthesize a reference voltage large, which means that it is spe-
high cost and bulky and heavy filters to as an average of the discrete switch- cially well designed to be applied to
smoothen the output waveforms, keep- ing states of the power converter. In multilevel converters with a high
ing the harmonic distortion below the this way, the concept of switching number of levels (for instance, for a
limits and therefore fulfilling the grid sequence is not used in these tech- CHB converter with more than four
code requirements. niques, and the switching of the power cells per phase).
To overcome this problem, a modi- power devices only occurs when
fied modulation method, based on the some control conditions are ful- Space Vector Control
SHE modulation concept, has been filled. This is why the switching fre- As has been explained above, the near-
proposed, where the switching angles quency of the output waveforms is est level control is a PWM method
are calculated not to eliminate a given not constant. where only the closest switching state
number of low-order harmonic com- of the switching sequence is finally
ponents but to reduce the distortion Hysteresis Control applied to the converter. The space
below the limits imposed by the grid Also called bang–bang or on–off con- vector control applies the same con-
codes. This modulation technique is trol, it consists of the determination of cept to the SVM technique. In this way,
called selective harmonic mitigation the power devices switching depending instead of using a switching sequence

16 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

formed by the state vectors closest to Control and Estimation permanent magnet synchronous motors
the reference vector, only the closest with sinusoidal electromotor force and
one is applied to the power converter Introduction currents.
[69]. As in the nearest level case, the The general block diagram of high-
space vector control reduces the performance speed controlled ac Linear TC
switching losses at the expense of a motor drives is shown in Figure 6. The linear torque controllers operate
higher distortion and a spread har- The core of the scheme are internal in association with voltage PWMs.
monic spectrum of the output voltages flux and torque control (TC) loops The controllers calculate the required
and currents. In the same way, this with the estimator block that can be stator voltage vector averaged over a
technique has to be used in power implemented in different ways, al- sampling period. The voltage vector is
converter topologies where the geo- though the outer speed control loop finally synthesized by a PWM tec-
metrical distance between the refer- is rather unified and generates com- hnique, which in most cases is the
ence vector and the closest vectors is mand values for torque Mc and flux SVM. So, different from the nonlinear
not large. Therefore, as in the nearest Wc (via flux program block) control- TC schemes where signals are proc-
level control, it is suitable to be used in lers. The speed feedback signal can essed on instantaneous values, in a
multilevel converters with a high num- be measured by a mechanical motion linear TC scheme, the linear [propor-
ber of levels. (speed/position) sensor Xm or deliv- tional-integral (PI)] controllers oper-
ered from the estimator Xm creating ate on values averaged over the
Model Predictive Control possibility of the speed sensorless sampling period. Therefore, the sam-
Model predictive control (MPC) is operation. pling frequency can be reduced from
performed in discrete time, where in about 40 kHz in nonlinear TC, to
each sample time, the output cur- Control Methods 2–5 kHz in linear TC schemes. In the
rent is predicted for each one of the Several TC methods have been devel- linear group, the following TCs are de-
valid switching states using a model oped in the last 40 years. Not all of scribed: field-oriented control (FOC),
of the load and the converter. All the them have found wide industrial appli- direct TC with voltage SVM (DTC-SVM),
predicted values are evaluated by a cations. Therefore, we present only the and DTC with flux SVM (DTC-FVM).
cost function, and the switching most popular strategies commercially
state that minimizes this function is used and some of the future trends. Field-Oriented Control
selected to be applied at the next The existing TC methods can be classi- The FOC, proposed in the 1970s by
sample time (see also the section fied in different ways [72], [73]. In this Hasse [74] and Blaschke [75], is
‘‘Predictive DTC Scheme’’) [70]. Sim- article, the TC methods are presented based on an analogy to the mechani-
ilarly to hysteresis, this control has in two main groups: linear and nonlin- cally commutated dc brush motor. In
a time-based operation, producing a ear controllers. The discussed generic this motor, owing to separate excit-
variable switching frequency and TC methods are presented for the ing and armature winding, flux is con-
consequently a spread harmonic induction motor drives; however, they trolled by the exciting current and
spectrum. One of the main advan- can be easily expanded for control of torque is independently controlled
tages of this algorithm is its flexibil-
ity because the cost function can be
easily modified, including several
Udc
primary and secondary control objec-
tives simultaneously, such as a Ψc Flux
frequency-related term to improve Control
–
its frequency behavior, the dc volt-
age balance in an NPC converter, Ωmc Mc
Speed Torque
switching frequency, and common- Control Control
–
mode voltage reduction. At the pre-
"

"

sent state of the art, the coefficients Me Ψ

(usually called weighting factors) of Usαβ
"

the different terms that form the Ωm Estimation

Isαβ
cost function are determined by an
empirical procedure. There is no
analytical or numerical solution pro- IM
posed yet to obtain an optimal solu- Ωm d γm 3~
dt
tion to the problem for finite control
set-MPC (FCS-MPC). In [71], some
guidelines are presented to help the FIGURE 6 – Block diagram of the speed-controlled ac motor drive with internal flux and tor-
weighting factor design process. que loops.

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 17

by adjusting the armature current. So ab=dq transformation and reference scheme, torque and flux are directly con-
the flux and torque currents are elec- voltage vector components Vsac and trolled in closed loops, and therefore, an
trically and magnetically separated. Vsbc by using inverse dq=ab transfor- accurate estimation of motor flux and
Contrarily, the cage-rotor induction mation. So, in the FOC, torque and torque is necessary. Contrary to classi-
motor (IM) has only a three-phase flux are controlled indirectly by field- cal hysteresis-based DTC, the DTC-SVM
winding in the stator, and the stator oriented current vector components. operates at a constant switching fre-
current vector, Is, is used for both quency. This considerably improves the
flux and TC. So, the exciting and DTC with Voltage SVM drive performance in terms of reduced
armature current is coupled (not sep- A block scheme of DTC-SVM with torque and flux pulsations, reliable start-
arated) in the stator current vector closed-loop torque and flux control up, and low speed operation.
and cannot be controlled separately. operating in Cartesian stator flux coor-
The decoupling can be achieved by dinates [76], [77] is presented in Fig- DTC with Flux SVM
the decomposition of the instantane- ure 8. The output of the PI flux and Further simplification can be achieved
ous stator current vector, Is, into two torque controllers is interpreted as the in a block scheme of DTC with flux
components: flux current, Isd , and reference stator voltage components, vector modulation (FVM), as shown in
torque-producing current, Isq , in the VW c and VMc , in S-FOC (dq). These dc Figure 10. For torque regulation, a PI
rotor-flux-oriented coordinates (R- voltage commands are then transformed controller is applied, and its output
FOC) dq (see the vector diagram in into stationary coordinates (ab), and produces an increment in the torque
Figure 7). In this way, the control of the commanded values, Vsac and Vsbc , angle, DdW (see the vector diagram in
the IM becomes identical with a sepa- are delivered to the SVM block. Note Figure 10) [78], [81]. Assuming that the
rately excited dc brush motor and that this scheme can be seen as a sim- rotor and flux magnitudes are approxi-
can be implemented using a current- plified stator flux-oriented control (S- mately equal, the torque is controlled
controlled PWM inverter with linear FOC) without current control loops by changing the torque angle, dW ,
PI controllers and voltage SVM (see [77] or as a classical ST-DTC scheme which corresponds to the increment
the block scheme in Figure 7). The (see Figure 9) in which the switching of the stator flux vector DWs . The com-
core of the FOC scheme are coordi- table (ST) is replaced by a modulator manded stator flux vector is calculated
nate transformation blocks that allow (SVM) and a hysteresis torque and by addition of the estimated flux posi-
calculation of field-oriented current flux controllers are replaced by linear tion cs and change of the torque angle
components Isd and Isq by using PI [76], [77], [79]. So, in the DTC-SVM DdW . Its value is compared with the

β
q

Is

Isq LM
M= q r ψ ⋅ Iq
Flux Lr rc
Isd
γr d
Stator Vdc
α

Vsdc Vsαc
ψrc ∼ Isd PI d-q SA
– SB
SVM
Vsqc Vsβ c SC
Mc ∼ Isq PI αβ
– Vdc

Voltage
"

"

Vsα Vsβ Calculation

Isα
"

Isq Isd γr Flux

αβ
Speed
Estimator ABC
Isβ

d-q
IM
αβ 3~

FIGURE 7 – Vector diagram and block diagram of rotor FOC. Torque is controlled indirectly via torque current Isq control loop.

18 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

 β
q

VS Flux
d ψS
VM M= ⋅V
Rs M
Vψ
γr
Stator Vdc
α

ψsc Vψ c Vsα c
PI d–q SA
–
SB
SVM
VMC Vsβ c SC
Mc
PI αβ
– Vdc
"
"

Ωm Ψs Voltage
Calculation
Vsα Vsβ

γs
"
"

Ψs
Flux Isα
Torque αβ
"

Me Speed Isβ
ABC
Estimator
IM
"

Ωm
3~

FIGURE 8 – Vector diagram and block scheme of DTC-SVM. Torque is controlled directly via voltage vector component VM .

estimated flux, and the stator flux n coordinate transformation is not appropriate voltage vector from the
error DWs is directly used for the cal- required switching selection table. Thus, pulses
culation of VSI switching states in the n there is no separate voltage PWM SA , SB , and SC for control the inverter
FVM block [81], [82]. Thanks to the n speed sensor is not required power switches are generated from
internal stator flux loop used for the n accurate stator flux vector and the vector selection table. The charac-
calculation of DWs in the flux PWM, torque estimation is required. teristic features of the ST-DTC scheme
the flux PI controller of Figure 8 is This section includes the ST-based of Figure 9 include:
eliminated. DTC (ST-DTC), direct self control (DSC), n sinusoidal stator flux and current
and online-optimized model-predictive waveforms with harmonic con-
Nonlinear TC DTC. Also, neural networks (NNs) and tent determined by the flux and
The presented nonlinear TC group fuzzy logic controllers (FLCs) belong to torque controller hysteresis tol-
departs from the idea of coordinate the class of nonlinear control [72], [84]. erance bands
transformation and the analogy with n excellent torque dynamics (depend-
dc motor control, which is the basis Classical ST-DTC Scheme ing on voltage reserve)
for the FOC. It proposes to replace The block diagram of the classical n flux and torque hysteresis bands
the decoupling control with the ST-DTC scheme is shown in Figure 9 determine the inverter switching
bang–bang control, which meets very [83]. The stator flux magnitude Wsc frequency, which varies with the syn-
well with on–off operation of the and the motor torque Mc are the chronous speed and load changes.
inverter semiconductor power devi- command signals that are compared Many modifications of the classical
ces. Compared with the conventional with the estimated W ^ e values,
^ s and M ST-DTC scheme aimed at improving
FOC (Figure 7), the DTC schemes respectively. The digitized flux and starting, overload conditions, very low
have the following features: torque errors generated by the hyster- speed operation, torque ripple reduc-
n simple structure esis controllers dW , dM , and the posi- tion, variable switching frequency
n there is no current control loops, tion sector N(cs ) of the stator flux functioning, and noise level attenua-
and current the is not regulated vector obtained from the angular posi- tion have been proposed during the
directly tion cs ¼ arctg(Wsb =Wsa ) selects the last few decades [77], [79].

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 19

 DSC Scheme
β The block diagram of the DSC
V3 V1 method [85] is shown in Figure 11.
V4 V2 Based on the command stator flux
Sector 3 ψ V V6 Wsc and the actual phase components
S 5
δψ WsA , WsB , and WsC , the flux compara-
ψ r tors generate digital variables dA , dB ,
Sector 4 LM
Sector 1 α M= ψrψs sin δψ and dC , which corresponds to active
σLrLs voltage states (V1  V6 ). The hystere-
Sector 5 Sector 6 sis torque controller generates signal
Vdc dM , which determines zero states. So,
the stator flux controller imposes the
ψsc dψ time duration of the active voltage
SA
– Vector states, which move the stator flux
SB
Selection along the commanded trajectory,
Mc dM Table SC
and the torque controller determines
– the time duration of the zero voltage
states, which keep the motor torque
"

N (γs) Vdc
in the defined-by-hysteresis toler-
"

Vsα
ance band.
"

ψs Voltage
"

Flux Vsβ Calculation

The characteristic features of the
Torque DSC scheme of Figure 11 are:
"

Speed Isα n nonsinusoidal stator flux and cur-

Me
Estimator αβ
rent waveforms that, with the
Isβ
ABC exception of the harmonics, are
IM
3~ identical for both PWM and six-
step operation
n the stator flux vector moves
FIGURE 9 – Vector diagram and block scheme of ST-DTC. Torque is controlled directly via stator
flux vector movement by selection of appropriate forward/backward active voltage vector (V1 or
along a hexagon path also under
V6 ) and stops by selection of zero voltage vector V0 . Stator flux vector moves on a circular path. PWM operation

β
Δψs = eVsdt
ψsc
Δδψ
ψs LM 1
M≅ q r ψr Δψs
Lr σ Lr
δψ
γs ψr
Stator
α Vdc
ψsc

Δψsα c Δψsα c SA
Mc Δδψ
SB
"

PI ψsc⋅ej(γs + Δδψ) – FVM

– Δψsβ c Δψsβ c SC

– Vdc
"

"

ψsα Vsα
Voltage
"

ψsβ Flux
"

Vsβ Calculation
Torque
"

γs Speed Isα
Estimator αβ
"

Me Isβ
ABC
IM
3~

FIGURE 10 – Vector diagram and block scheme of DTC-FVM. Torque is controlled directly via stator flux vector increment DWs with FVM.

20 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

n no voltage supply reserve is nec-
essary and the inverter capability There is a strong trend to avoid mechanical
is fully utilized
n the inverter switching frequency motion (speed/position) sensors because it
is lower than in the ST-DTC of
Figure 9
reduces cost and improves reliability and
n excellent torque dynamics in con- functionality of the drive system.
stant and weakening field regions.
Note that the behavior of a DSC
scheme can be reproduced by an ST- n easy to implement n the accuracy of the predictive
DTC scheme from Figure 9 for flux n open to include modifications and model has a direct influence on the
hysteresis of 14% wide. extensions depending on specific quality of the predictive controller.
applications. As reported in the works by Papa-
Predictive DTC Scheme However, these advantages are fotiou et al. (see in [86] and [87]) by
The current and future trend in the offset by problems when implement- using a long-output horizon-predic-
control of ac motors is to incorporate ing predictive DTC: tive DTC, the average switching
more advanced techniques such as n larger amount of online calculations frequency, and as consequence also
the MPC [86], [87]. This trend is sup- compared with classical controllers inverter switching losses, has been
ported by the DSP producers that not
only increase the operation speed and
computation capacity but also expand
their offer by specialized circuits. Pre- β
dictive control is a very wide class of
controller that uses the model of the
V3 V1
system (called predictive model) for Sector 2
the prediction of the future behavior V V2
Sector 3 ψS 4
of the controlled variables. This infor- V5 V6
δψ
mation is used for the controller to α
ψr
calculate the optimal actuation that Sector 4
Sector 6 LM
M= ψrψs sin δψ
minimizes a predefined cost function. σLrLs
The simplified block diagram of Sector 5
the predictive DTC scheme is shown Torque
Controller
in Figure 12. This type of MPC is clas-
sified as MPC with FCS (seven switch- Mc dM
ing states are available in two-level
ψsc –
VSI) and, contrarily to a MPC with
dA Vdc
"

continuous control see [86], [87], ψSA

"

Me
operates without PWMs because the
inverter switching state is calculated SA
dB
"

online by minimization of a cost func- ψSB SB

tion. In most cases, the cost function SC
is defined as a weighted sum of torque
"

and flux errors; however, other compo- ψSC dC Vdc

nents can be added (e.g., reducing
unnecessary switching, shaping fre-
Flux Voltage
quency spectrum, maintaining the volt- Controller Calculation
"

age balance of the dc-link capacitors in Vsα

"

Vsβ
case of multilevel VSI, reducing com-
mon voltage) increasing the flexibility Flux Isα
and priority in predefinition of system Torque αβ
performances. The important advan- Speed Isβ
Estimator ABC
tages of the MPC are: IM
n intuitive and easy to understand 3~
concept
n easy inclusion of nonlinearities in
FIGURE 11 – Vector diagram and block scheme of DSC. Torque is controlled directly in simi-
the model and simple treatment lar way as in the ST-DTC scheme; however, stator flux vector moves on hexagonal path
of constraints because of different sector definition.

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 21

 rotor flux vector (magnitude and posi-
tion) and electromagnetic torque.
M = min J ⇒ Vn(k)
(k+1) Once the flux vector is accurately
Vdc
estimated, the torque estimation is
ψsc (k) performed easily as a cross product
SA(k)
Mc(k) Online of the flux and measured stator cur-
SB(k)
Minimalization rent vectors.
of Cost Function SC(k)
Vdc
Flux Vector Estimation
Voltage
To avoid the use of flux sensors or
"

"

ψ s(k +1) M (k +1) measuring coils in the IM, methods of

"

"
Vsα Vsβ Calculation
indirect flux vector generation have
been developed, known as flux mod-
"

ψ s(k)
Flux Isα els or flux estimators. These are mod-
"

Predictive Me(k) Torque αβ

els of motor equations that are
Model Speed
"

Ωm(k) Estimator Isβ ABC excited by appropriate easily measur-

IM able quantities, such as stator vol-
3~
tages and/or currents (Vs , Is ), angular
shaft speed (Xm ), or position angle
FIGURE 12 – Block diagram of the predictive DTC. Torque is directly controlled by online (cm). There are many types of flux vec-
minimization of the cost function in every sampling period.
tor models, which usually are classi-
fied in terms of the input signals used
reduced by 16.5% with respect to the proven to be especially powerful as it [73], [88]–[90]. Such models generate
classical ST-DTC used by ABB while offers advantages for both techni- the stator or/and rotor flux vector
maintaining the same control quality ques. Several vector control schemes which, in an ideal case, rotates syn-
of the motor torque, magnitude of with neurofuzzy controller has been chronously with the motor magnetic
stator flux, and three-level inverter’s proposed and described in the litera- field. Because the motor parameters
neutral point voltage. This illustrates ture [1], [72], [89]. However, at very often are known roughly and
a very high improvement potential of present, they only represent some change during motor operation, an
the MPC. alternative solutions to the other error therefore appears between the
existing techniques, and their spe- motor flux and that estimated in the
Neurofuzzy Control cific application areas cannot be model. The error depends on model
In the last few decades, an interest in clearly defined. variants, parameter deviation between
the development of artificial intelli- motor and model, accuracy of input
gence-based controllers in the area of Estimation signal measurement, and motor point
power electronics and drive control Implementation of any high-performance of operation. To minimize the sensitiv-
has been observed. The combination drive system requires a high accuracy ity of flux models to motor parameters
of fuzzy logic and artificial NNs has estimation of the actual stator or/and changes, the model adaptive reference
systems and the observer technique
[76], [89], [91] can be used. Also, slid-
ing mode approach for robust flux esti-
Is mation has been proposed [92].
An example of an early developed
Rs [93] and simple flux vector observer
based on voltage model with com-
ψs
"

– ψr
"

Vs
ψs (ψr ) pensation signal calculated from the
– commanded WRc and estimated WR^
ψs
"

Ka,b rotor flux vector is presented in

Figure 13. It operates without the
ψrc ψrc measured speed signal and can there-
e j γψ r
"

fore be used in speed sensorless

– drives. At very low speed, the flux
γψ r
"

vector is generated by the reference

arctg(ψrβ /ψrα ) value WRc because the stator voltage
signal is almost zero. For higher
FIGURE 13 – Simple flux vector observer based on voltage model with compensation signal speed level (determined by selection
calculated from the commanded WRc and estimated WR^ rotor flux vector. of the compensation gain K a, b ), the

22 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

 Vs
Rs
–
Is Voltage
Current Model Compensator
Model
I

"
I
ψs

"

"

"
Is Isdq ψrdq ψr ucomp ψs
e–jγψR e jγψR ψS(ψR)
–

"
ψs

"
γψ r
"

γψ r

"
ψr

"
ψr
arctg(ψRβ /ψRα ) ψr (ψs)

Is

FIGURE 14 – Speed sensorless flux vector observer based on voltage model compensated by current model.

voltage signal becomes dominant, Speed Estimation and area of applications. In FOC and
and the flux vector is calculated by There is a strong trend to avoid DTC-SVM schemes, the control action
the voltage model. This observer al- mechanical motion (speed/position) is usually synchronized with PWM
lows starting up and breaking the sensors because it reduces cost and generation and executed with the
drive system; however, control tor- improves reliability and functionality sampling time equal to the switching
que operation at zero speed is not of the drive system. Various methods time ranging from 50 to 500 ls. The
possible. of the rotor speed estimation has typical torque rise time is about four
Another flux vector observer also been developed for induction motors; to six sampling and is limited by the
operating without measured speed however, the detailed analysis is VSI switching frequency. For compari-
signal is shown in Figure 14. The sta- beyond the scope of this article. A son, the parameters of scalar constant
tor flux vector is calculated by the good review of IM speed sensorless (V/Hz) control are given.
voltage model and compensated by control schemes is presented in [84]
the current model operating in the and [88]. Summary
R-FOCs dq. At higher speed, the volt- Power electronics controlled high-
age model guarantees better accu- Closing Remarks and Overview performance ac drives belong to
racy because the stator resistance is The overview of the basic parameters high-tech industry and is one of main
small while, at low speeds, the cur- and application areas of main dis- factors for energy saving and produc-
rent model is better and operates cussed control strategies are pre- tivity growth. This article reviews the
even at zero frequency [76]. The tran- sented in Table 1. It can be concluded present state and trends in the devel-
sition between the two models is that the group of high-performance opment of key parts of controlled
defined by the design of the PI con- control achieves similar parameters drive systems: converter topologies,
troller that generates the compensa-
tion voltage ucomp .
In Figure 15, a flux vector observer
"

based on measured stator voltage V s , Is – is – 1

current I s , and mechanical speed Xm Lσ
with compensation signal calculated
from measured I s and estimated I ^s 1 + 1 RR
Rs Lσ
stator current vectors is shown. In Ls Lσ
this observer, the motor model is fed ucomp
with current vector estimated from Vs –
RR –
′
"

ψR
the stator and rotor flux. The mea- Lσ
ψ μ′
"

sured stator current is delivered only

Ωm
to PI controller for calculation the X
compensation voltage ucomp . There-
fore, the model can be easily used for j
testing and investigation. With me-
chanical motion sensor it operates FIGURE 15 – Flux vector observer based on measured Vs , Is , and Xm with compensation
well at zero speed [85], [94]. signal calculated from measured Is and estimated I^s stator current vectors.

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 23

 TABLE 1–OVERVIEW OF THE MAIN CONTROL STRATEGIES IN LOW AND MEDIUM POWER DRIVES.
SPEED CONTROL STATIC SPEED TORQUE RISE STARTING
CONTROL STRATEGY RANGE ACCURACY (%) TIME (ms) TORQUE COST TYPICAL APPLICATIONS
Scalar control 1:10 (open loop) 5–10 Not available Low Very low Low performance: pumps, fans
(constant V/Hz) compressors, HVAC, etc.
FOC >1:200 (closed loop) 0 <1–2 High High High performance: crane, lifts,
transportation, etc.
DTC-SVM >1:200 (closed loop) 0 <1–2 High High High performance: crane, lifts,
transportation, etc.
ST-DTC >1:200 (closed loop) 0 <1 High High High performance: crane, lifts,
transportation, etc.
DSC >1:200 (closed loop) 0 <1–2 High High High performance: traction, fast
field weakening
Servo drives >1:10 (closed loop) 0 <1 High High High performance, speed rise time
10 ms: robots, manipulators,
industrial automation, etc.

PWM methods, as well as control and he has been the head of the Centre Electronics since 2007. He has been
estimation techniques. of Excellence on Power Electronics the president of IES since 2010. He
The following topologies were de- and Intelligent Control for Energy is a Fellow of the IEEE. His current
scribed: two- and multilevel voltage- Conservation at ICIE. He received an research interests include modula-
source converters, current-source honorary doctorate degree from tion techniques for multilevel inver-
converters, and direct converters. Aalborg University in 2004 and the ters and its application to power
Also, PWM techniques were briefly Institut National Polytechnique de electronic systems for renewable
presented: conventional bipolar and Toulouse, France, in 2010. In 2005, he energy systems.
unipolar PWM, SVM with extension received the Dr.-Ing. Eugene Mittel- Jose Rodriguez (jrp@usm.cl) re-
for multilevel converters, harmonic mann Achievement Award from the ceived his engineering degree in
control techniques, and variable IEEE Industrial Electronics Society electrical engineering from the Uni-
frequency modulation (hysteresis, (IES) and, in 2007, the SIEMENS Re- versidad Tecnica Federico Santa
nearest level, and model predictive search Award in Poland. In 2007, he Maria (UTFSM), Valparaiso, Chile, in
technique). was elected as the corresponding 1977 and the Dr.-Ing. degree in elec-
Finally, the generic TC methods member of the Polish Academy of Sci- trical engineering from the Univer-
were described in two groups: linear ence. He was the editor-in-chief of sity of Erlangen, Germany, in 1985.
and nonlinear controllers. The linear IEEE Transactions on Industrial Elec- He has been with the Department of
group was presented with FOC, DTC tronics (2004–2006). Currently, he is Electronics Engineering, UTFSM since
with voltage SVM, and DTC with FVM the dean of the Technical Science 1977, where he is currently a full
(DTC-FVM). The group of nonlinear Department, Polish Academy of Sci- professor and rector. He has coau-
control includes: classical ST-based ence. He is a Fellow of the IEEE. thored more than 250 journal and
hysteresis DTC, DSC, and future Leopoldo G. Franquelo (lgfranquelo@ conference papers. He has been an
trend solutions based on predictive ieee.org) received the M.Sc. and Ph.D. associate editor of IEEE Transactions
DTC and neurofuzzy schemes. Also, degrees in electrical engineering from on Power Electronics and IEEE Transac-
selected simple flux vector observers/ the University of Seville, Spain, in tions on Industrial Electronics since
estimators were discussed. 1977 and 1980, respectively. He is 2002. He received the 2007 Best Paper
currently a full professor and head Award from IEEE Transactions on
of the Power Electronics Group with Industrial Electronics as well as the
Biographies the University of Seville. He was the 2008 Best Paper Award from IEEE
Marian P. Kazmierkowski (mpk@ vice president of the IES Spanish Industrial Electronics Magazine. He is a
isep.pw.edu.pl) received his M.S., Chapter (2002–2003) and a member member of the Chilean Academy of
Ph.D., and Dr.Sci. degrees in electri- at large of the IES Administrative Engineering and a Fellow of the IEEE.
cal engineering from the Institute of Committee (2002–2003). He was His main research interests include
Control and Industrial Electronics the vice president for conferences multilevel inverters, new converter
(ICIE), Warsaw University of Technol- (2004–2007), in which he has also topologies, control of power convert-
ogy, Poland, in 1968, 1972, and 1981, been a Distinguished Lecturer since ers, and adjustable speed drives.
respectively. From 1987 to 2008, he 2006. He has been an associate edi- Marcelo A. Perez (marcelo.perez@
was the director of ICIE. Since 2003, tor for IEEE Transactions on Industrial usm.cl) received his engineering

24 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011

degree in electronic engineering, multilevel converters,’’ IEEE Trans. Ind. ‘‘Survey on fault operation on multilevel
Electron., vol. 57, no. 8, pp. 2553–2580, inverters,’’ IEEE Trans. Ind. Electron.,
M.Sc. degree in electrical engineer- Aug. 2010. vol. 57, no. 7, pp. 2207–2218, July 2010.
ing, and the D.Sc. degree in electri- [10] S. Fazel, S. Bernet, D. Krug, and K. Jalili, [27] J. Rodriguez, J. Pontt, E. Silva, J. Espinoza,
‘‘Design and comparison of 4-kV neutral- and M. Perez, ‘‘Topologies for regenerative
cal engineering from the University point-clamped, flying-capacitor, and series- cascaded multilevel inverters,’’ in Proc.
of Concepción, Chile, in 2000, 2003, connected H-bridge multilevel converters,’’ IEEE 34th Annu. Power Electronics Specialist
IEEE Trans. Ind. Applicat., vol. 43, no. 4, Conf., PESC, June 2003, vol. 2. pp. 519–524.
and 2006, respectively. In 2006, he pp. 1032–1040, July–Aug. 2007. [28] M. Perez, J. Espinoza, J. Rodriguez, and
joined the Electronics Department [11] A. Nabae, I. Takahashi, and H. Akagi, ‘‘A P. Lezana, ‘‘Regenerative medium-voltage
new neutral-point-clamped PWM inverter,’’ ac drive based on a multicell arrangement
at UTFSM, Valparaiso, Chile, as a IEEE Trans. Ind. Applicat., vol. IA-17, no. 5, with reduced energy storage require-
postdoctoral researcher. Since 2009, pp. 518–523, Sept. 1981. ments,’’ IEEE Trans. Ind. Electron., vol. 52,
[12] J. Rodriguez, S. Bernet, P. Steimer, and I. no. 1, pp. 171–180, Feb. 2005.
he has been an associate researcher Lizama, ‘‘A survey on neutral-point- [29] J. Rodriguez, J. Pontt, M. Perez, P. Lezana,
at the same institution. He is a Mem- clamped inverters,’’ IEEE Trans. Ind. Appli- and P. Hammond, ‘‘High power synchro-
cat., vol. 57, no. 7, pp. 2219–2230, July nous machine fed by a cascaded regenera-
ber of the IEEE. 2010. tive inverter,’’ in Proc. IEEE Industry
Jose I. Leon (jileon@gte.esi.us.es) [13] M. Rahimo and S. Klaka, ‘‘High voltage Applications Society Annu. Meeting, IAS,
semiconductor technologies,’’ in Proc. 13th Oct. 2008, pp. 1–7.
received his B.S., M.S., and Ph.D. European Conference on Power Electronics [30] J. Dixon, J. Pereda, C. Castillo, and S.
degrees in telecommunications engi- and Applications, EPE, Sept. 2009, pp. 1–10. Bosch, ‘‘Asymmetrical multilevel inverter
[14] Y. Suh and P. Steimer, ‘‘Application of for traction drives using only one dc sup-
neering from the University of Seville, IGCT in high-power rectifiers,’’ IEEE Trans. ply,’’ IEEE Trans. Veh. Technol., vol. 59,
Spain, in 1999, 2001 and 2006, respec- Ind. Applicat, vol. 45, no. 5, pp. 1628–1636, no. 8, pp. 3736–3743, Oct. 2010.
Sept.–Oct. 2009. [31] M. Perez, P. Cortes, and J. Rodriguez, ‘‘Pre-
tively. Currently, he is an associate [15] K. K. Tan, F. Gao, P. C. Loh, and F. Blaab- dictive control algorithm technique for
professor with the Department of jerg, ‘‘Enhanced buck-boost neutral- multilevel asymmetric cascaded H-bridge
point-clamped inverters with simple inverters,’’ IEEE Trans. Ind. Electron.,
Electronic Engineering, University of capacitive-voltage balancing,’’ IEEE Trans. vol. 55, no. 12, pp. 4354–4361, Dec. 2008.
Seville. He received the 2008 IEEE Ind. Applicat., vol. 46, no. 3, pp. 1021–1033, [32] J. Dixon, A. Breton, F. Rios, J. Rodriguez, J.
May–June 2010. Pontt, and M. Perez, ‘‘High-power machine
Industrial Electronics Magazine Best [16] S. R. Pulikanti, M. S. A. Dahidah, and V. G. drive, using no redundant 27-level inver-
Paper Award. He is a Member of the Agelidis, ‘‘Voltage balancing control of ters and active front end rectifiers,’’ IEEE
three-level active NPC converter using Trans. Power Electron., vol. 22, no. 6,
IEEE. His research interests include SHE-PWM,’’ IEEE Trans. Power Delivery, pp. 2527–2533, Nov. 2007.
electronic power systems, modulation vol. 26, no. 1, pp. 258–267, Jan. 2011. [33] T. Meynard, M. Fadel, and N. Aouda,
[17] C. Wang and Y. Li, ‘‘Analysis and calcula- ‘‘Modeling of multilevel converters,’’ IEEE
and control of power converters, and tion of zero-sequence voltage considering Trans. Ind. Electron., vol. 44, no. 3, pp. 356–
industrial drives. neutral-point potential balancing in three- 364, June 1997.
level NPC converters,’’ IEEE Trans. Ind. [34] L. Xu and V. Agelidis, ‘‘Flying capacitor
Electron., vol. 57, no. 7, pp. 2262–2271, July multilevel PWM converter-based UPFC,’’
2010. IEE Proc. Electric Power Applicat., vol. 149,
References [18] J. Pou, R. Pindado, and D. Boroyevich, no. 4, pp. 304–310, July 2002.
[1] B. K. Bose, ‘‘Power electronics and motor ‘‘Voltage-balance limits in four-level diode- [35] J. Holtz and N. Oikonomou, ‘‘Optimal con-
drives recent progress and perspective,’’ clamped converters with passive front trol of a dual three-level inverter system
IEEE Trans. Ind. Electron., vol. 56, no. 2, ends,’’ IEEE Trans. Ind. Electron., vol. 52, for medium-voltage drives,’’ IEEE Trans.
pp. 581–588, Feb. 2009. no. 1, pp. 190–196, Feb. 2005. Ind. Applicat., vol. 46, no. 3, pp. 1034–1041,
[2] M. Liserre, ‘‘Dr Bimal K. Bose: A reference [19] T. Chaudhuri, A. Rufer, and P. Steimer, May–June 2010.
for generations,’’ IEEE Ind. Electron. Mag., ‘‘The common cross-connected stage for [36] F. Kieferndorf, M. Basler, L. Serpa, J.-H.
vol. 3, no. 2, pp. 2–5, June 2009. the 5l ANPC medium voltage multilevel Fabian, A. Coccia, and G. Scheuer, ‘‘A new
[3] A. K. Chattopadhyay, ‘‘Alternating current inverter,’’ IEEE Trans. Ind. Electron., medium voltage drive system based on
drives in steel industry,’’ IEEE Ind. Electron. vol. 57, no. 7, pp. 2279–2286, July 2010. ANPC-5l technology,’’ in Proc. IEEE Int.
Mag., vol. 4, no. 4, pp. 30–42, Dec. 2010. [20] J. Fang, C. Mao, J. Lu, D. Wang, and Q. Conf. Industrial Technology, ICIT, Mar.
[4] J. Wang, A. Q. Huang, W. Sung, Y. Li, and Zhang, ‘‘The research on protection of 2010, pp. 643–649.
B. J. Baliga, ‘‘Smart grid technologies. over-voltage about NPC three-level [37] M. Hagiwara, K. Nishimura, and H. Akagi,
Development of 15-kV SiC IGBTs and their medium voltage motor drives,’’ in Proc. ‘‘A modular multilevel PWM inverter for
impact on utility applications,’’ IEEE Ind. Int. Conf. Electrical Machines and Systems, medium-voltage motor drives,’’ in Proc.
Electron. Mag., vol. 3, no. 2, pp. 16–23, June ICEMS, Oct. 2008, pp. 1595–1599. IEEE Energy Conversion Congress and Expo-
2009. [21] N. Hatti, Y. Kondo, and H. Akagi, ‘‘Five- sition, ECCE, Sept. 2009, pp. 2557–2564.
[5] J. Rodriguez, S. Bernet, B. Wu, J. Pontt, level diode-clamped PWM converters con- [38] A. Lesnicar and R. Marquardt, ‘‘An innova-
and S. Kouro, ‘‘Multilevel voltage-source- nected back-to-back for motor drives,’’ tive modular multilevel converter topol-
converter topologies for industrial IEEE Trans. Ind. Applicat., vol. 44, no. 4, ogy suitable for a wide power range,’’ in
medium-voltage drives,’’ IEEE Trans. Ind. pp. 1268–1276, July–Aug. 2008. Proc. IEEE Power Tech Conf., Bologna, June
Electron., vol. 54, no. 6, pp. 2930–2945, [22] M. Malinowski, K. Gopakumar, J. Rodri- 2003, vol. 3. pp. 1–6.
Dec. 2007. guez, and M. Perez, ‘‘A survey on cas- [39] M. Hagiwara, K. Nishimura, and H. Akagi,
[6] L. G. Franquelo, J. Rodriguez, J. I. Leon, S. caded multilevel inverters,’’ IEEE Trans. ‘‘A medium-voltage motor drive with a
Kouro, R. Portillo, and M. M. Prats, ‘‘The Ind. Electron., vol. 57, no. 7, pp. 2197–2206, modular multilevel PWM inverter,’’ IEEE
age of multilevel converters arrives,’’ IEEE July 2010. Trans. Power Electron., vol. 25, no. 7,
Ind. Electron. Mag., vol. 2, no. 2, pp. 28–39, [23] M. Marchesoni, M. Mazzucchelli, and pp. 1786–1799, July 2010.
June 2008. S. Tenconi, ‘‘A nonconventional power [40] J. Dai, D. (David) Xu, B. Wu, and N. R. Zar-
[7] J. Rodriguez, L. Franquelo, S. Kouro, J. Leon, converter for plasma stabilization,’’ IEEE gari, ‘‘Unified dc-link current control for
R. Portillo, M. Prats, and M. Perez, ‘‘Multi- Trans. Power Electron., vol. 5, no. 2, low-voltage ride-through in current
level converters: An enabling technology pp. 212–219, Apr. 1990. source-converter-based wind energy con-
for high-power applications,’’ Proc. IEEE, [24] P. Hammond, ‘‘A new approach to version systems,’’ IEEE Trans. Power Elec-
vol. 97, no. 11, pp. 1786–1817, Nov. 2009. enhance power quality for medium volt- tron., vol. 26, no. 1, pp. 288–297, Jan. 2011.
[8] F. Peng, W. Qian, and D. Cao, ‘‘Recent age ac drives,’’ IEEE Trans. Ind. Applicat., [41] M. Naguib and L. Lopes, ‘‘Harmonics
advances in multilevel converter/inverter vol. 33, no. 1, pp. 202–208, Jan./Feb. 1997. reduction in current source converters
topologies and applications,’’ in Proc. 2010 [25] S. Lu and K. Corzine, ‘‘Advanced control using fuzzy logic,’’ IEEE Trans. Power Elec-
Int. Power Electronics Conf. (IPEC), June and analysis of cascaded multilevel con- tron., vol. 25, no. 1, pp. 158–167, Jan. 2010.
2010, pp. 492–501. verters based on p-q compensation,’’ IEEE [42] H. Bilgin and M. Ermis, ‘‘Design and imple-
[9] S. Kouro, M. Malinowski, K. Gopakumar, J. Trans. Power Electron., vol. 22, no. 4, mentation of a current-source converter
Pou, L. G. Franquelo, B. Wu, J. Rodriguez, pp. 1242–1252, July 2007. for use in industry applications of D-STAT-
M. A. Perez, and J. I. Leon, ‘‘Recent advan- [26] P. Lezana, J. Pou, T. Meynard, J. Rodri- COM,’’ IEEE Trans. Power Electron., vol. 25,
ces and industrial applications of guez, S. Ceballos, and F. Richardeau, no. 8, pp. 1943–1957, Aug. 2010.

SEPTEMBER 2011 n IEEE INDUSTRIAL ELECTRONICS MAGAZINE 25

[43] Y. W. Li, M. Pande, N. Zargari, and B. Wu, IEEE Ind. Electron. Mag., vol. 3, no. 1, machines,’’ Siemens Rev., vol. 34, no. 1,
‘‘An input power factor control strategy pp. 20–31, Mar. 2009. pp. 217–220, 1972.
for high-power current-source induction [60] N. Celanovic and D. Boroyevich, ‘‘A fast [76] I. Boldea and S. A. Nasar, Electric Drives.
motor drive with active front-end,’’ IEEE space-vector modulation algorithm for 2nd ed. Boca Raton, FL: CRC, 2006.
Trans. Power Electron., vol. 25, no. 2, multilevel three-phase converters,’’ IEEE [77] G. S. Buja and M. P. Kazmierkowski, ‘‘Direct
pp. 352–359, Feb. 2010. Trans. Ind. Applicat., vol. 37, no. 2, torque control of PWM inverter-fed AC
[44] F. Gao, P. C. Loh, F. Blaabjerg, and D. M. pp. 637–641, Mar./Apr. 2001. motors—A survey,’’ IEEE Trans. Ind. Elec-
Vilathgamuwa, ‘‘Five-level current-source [61] M. M. Prats, L. G. Franquelo, J. I. Leon, R. tron., vol. 51, no. 4, pp. 744–757, Aug. 2004.
inverters with buck-boost and inductive- Portillo, E. Galvan, and J. M. Carrasco, ‘‘A [78] L. Xu and M. Fu, ‘‘A sensorless direct tor-
current balancing capabilities,’’ IEEE SVM-3D generalized algorithm for multile- que control technique for permanent mag-
Trans. Ind. Electron., vol. 57, no. 8, vel converters,’’ in Proc. 29th Annu. Conf. net synchronous motors,’’ in Proc. IEEE
pp. 2613–2622, Aug. 2010. IEEE Industrial Electronics Society, IECON, Ind. Appl. Conf., 1999, vol. 1. pp. 159–164.
[45] B. Wu, J. Pontt, J. Rodriguez, S. Bernet, Nov. 2–6, 2003, vol. 1, pp. 24–29. [79] D. Casadei, F. Profumo, G. Serra, and A.
and S. Kouro, ‘‘Current-source converter [62] J. I. Leon, S. Vazquez, J. A. Sanchez, R. Tani, ‘‘FOC and DTC: Two viable schemes
and cycloconverter topologies for indus- Portillo, L. G. Franquelo, J. M. Carrasco, for induction motors torque control,’’
trial medium-voltage drives,’’ IEEE Trans. and E. Dominguez, ‘‘Conventional space- IEEE Trans. Power Electron., vol. 17, no. 5,
Ind. Electron., vol. 55, no. 7, pp. 2786–2797, vector modulation techniques versus the pp. 779–787, 2002.
July 2008. single-phase modulator for multilevel con- [80] T. Ohtani, N. Takada, and K. Tanaka, ‘‘Vec-
[46] J. Vobecky, ‘‘Future trends in high power verters,’’ IEEE Trans. Ind. Electron., vol. 57, tor control of induction motor without
devices,’’ in Proc. 27th Int. Conf. Microelec- no. 7, pp. 2473–2482, July 2010. shaft encoder,’’ IEEE Trans. Ind. Applicat.,
tronics, MIEL, May 2010, pp. 67–72. [63] H. S. Patel and R. G. Hoft, ‘‘Generalized vol. 28, no. 1, pp. 157–165, 1992.
[47] M.-K. Nguyen, Y.-G. Jung, H.-Y. Yang, and techniques of harmonic elimination and volt- [81] A. Tripathi, A. M. Khambadkone, and S. K.
Y.-C. Lim, ‘‘Single-phase z-source cyclo- age control in thyristor inverters: Part 1— Panda, ‘‘Stator flux based space vector
converter with safe-commutation Harmonic elimination,’’ IEEE Trans. Ind. modulation and closed loop control of the
strategy,’’ IET Power Electron., vol. 3, no. 2, Applicat., vol. IA-9, no. 3, pp. 310–317, stator flux vector in overmodulation into
pp. 232–242, Mar. 2010. May/Jun. 1973. six-step mode,’’ IEEE Trans. Power Elec-
[48] J. Pontt, J. Rodriguez, J. Rebolledo, K. [64] R. Ray, D. Chatterjee, and S. Goswami, tron., vol. 19, no. 3, pp. 775–782, May 2004.
Tischler, and N. Becker, ‘‘Operation of ‘‘Harmonics elimination in a multilevel [82] A. Steimel, ‘‘Direct self-control and
high-power cycloconverter-fed gearless inverter using the particle swarm opti- synchronous pulse techniques for high-
drives under abnormal conditions,’’ misation technique,’’ IET Power Elec- power traction inverters in comparison,’’
IEEE Trans. Ind. Applicat., vol. 43, no. 3, tron., vol. 2, no. 6, pp. 646–652, Nov. IEEE Trans. Power Electron., vol. 51, no. 4,
pp. 814–820, May–June 2007. 2009. pp. 810–820, 2004.
[49] C. Ortega, A. Arias, C. Caruana, J. Balcells, [65] A. I. Maswood and S. Wei, ‘‘Genetic-algo- [83] I. Takahashi, I. Takahashi, and T. Noguchi,
and G. Asher, ‘‘Improved waveform qual- rithm-based solution in PWM converter ‘‘A new quick response and high-efficiency
ity in the direct torque control of matrix- switching,’’ IEE Proc. Electric Power Appli- control strategy of an induction motor,’’
converter-fed PMSM drives,’’ IEEE Trans. cat., vol. 152, no. 3, pp. 473–478, May 2005. IEEE Trans. Ind. Applicat., vol. IA-22, no. 5,
Ind. Electron., vol. 57, no. 6, pp. 2101–2110, [66] L. G. Franquelo, J. Napoles, R. Portillo, J. I. pp. 820–827, Sept./Oct. 1986.
June 2010. Leon, and M. A. Aguirre, ‘‘A flexible selec- [84] L. M. Grzesiak and M. P. Kazmierkowski,
[50] R. K. Gupta, K. K. Mohapatra, A. Somani, tive harmonic mitigation technique to ‘‘Improving flux and speed estimators for
and N. Mohan, ‘‘Direct-matrix-converter- meet grid codes in three-level PWM con- sensorless AC drives,’’ IEEE Ind. Electron.
based drive for a three-phase open end- verters,’’ IEEE Trans. Ind. Electron., vol. 54, Mag., vol. 1, no. 3, pp. 8–19, Fall 2007.
winding ac machine with advanced fea- no. 6, pp. 3022–3029, Dec. 2007. [85] M. Depenbrock, ‘‘Direct self control of
tures,’’ IEEE Trans. Ind. Electron., vol. 57, [67] M. Mohseni and S. Islam, ‘‘A new vector- inverter-fed induction machines,’’ IEEE
no. 12, pp. 4032–4042, Dec. 2010. based hysteresis current control scheme Trans. Power Electron., vol. PE-3, no. 4,
[51] J. Monteiro, J. F. Silva, S. F. Pinto, and J. for three-phase PWM voltage-source inver- pp. 420–429, Oct. 1988.
Palma, ‘‘Matrix converter-based unified ters,’’ IEEE Trans. Power Electron, vol. 25, [86] M. P. Kazmierkowski, R. M. Kennel, and
power-flow controllers: Advanced direct no. 9, pp. 2299–2309, Sept. 2010. J. Rodriguez, ‘‘Special section on predictive
power control method,’’ IEEE Trans. Power [68] S. Kouro, R. Bernal, H. Miranda, C. A. Silva, control in power electronics and drives—
Delivery, vol. 26, no. 1, pp. 420–430, Jan. and J. Rodriguez, ‘‘High-performance tor- Part I,’’ IEEE Trans. Ind. Electron., Part I,
2011. que and flux control for multilevel vol. 55, no. 12, pp. 4309–4429, Dec. 2008.
[52] S. Kwak, ‘‘Fault-tolerant structure and inverter fed induction motors,’’ IEEE [87] M. P. Kazmierkowski, R. M. Kennel, and
modulation strategies with fault detection Trans. Power Electron., vol. 22, no. 6, J. Rodriguez, ‘‘Special section on predic-
method for matrix converters,’’ IEEE pp. 2116–2123, Nov. 2007. tive control in power electronics and
Trans. Power Electron., vol. 25, no. 5, [69] J. Rodrı́guez, L. Morán, P. Correa, and drives–Part II,’’ IEEE Trans. Ind. Electron.,
pp. 1201–1210, May 2010. C. Silva, ‘‘A vector control technique for vol. 56, no. 6, pp. 1823–1963, June 2009.
[53] S. Bernet, S. Ponnaluri, and R. Teichmann, medium-voltage multilevel inverters,’’ IEEE [88] J. Holtz, ‘‘Sensorless control of induction
‘‘Design and loss comparison of matrix Trans. Ind. Electron., vol. 49, no. 4, machines—With or without signal injec-
converters, and voltage-source converters pp. 882–888, Aug. 2002. tion?’’ IEEE Trans. Ind. Electron., vol. 53,
for modern ac drives,’’ IEEE Trans. Ind. [70] S. Kouro, P. Cortes, R. Vargas, U. Ammann, no. 1, pp. 7–30, Feb. 2006.
Electron., vol. 49, no. 2, pp. 304–314, Apr. and J. Rodriguez, ‘‘Model predictive con- [89] B. K. Bose, Modern Power Electronics and
2002. trol—A simple and powerful method to AC Drives. Englewood Cliffs, NJ: Prentice-
[54] R. Pena, R. Cardenas, E. Reyes, J. Clare, control power converters,’’ IEEE Trans. Ind. Hall, 2001.
and P. Wheeler, ‘‘A topology for multiple Electron., vol. 56, no. 6, pp. 1826–1838, June [90] M. P. Kazmierkowski and H. Tunia, Auto-
generation system with doubly fed induc- 2009. matic Control of Converter Fed Drives. New
tion machines and indirect matrix [71] P. Cortes, S. Kouro, B. La Rocca, R. Var- York: Elsevier, 1994.
converter,’’ IEEE Trans. Ind. Electron., gas, J. Rodriguez, J. I. Leon, S. Vazquez, [91] R. Nilsen and M. P. Kazmierkowski,
vol. 56, no. 10, pp. 4181–4193, Oct. 2009. and L. G. Franquelo, ‘‘Guidelines for ‘‘Reduced order observer with parameter
[55] P. Correa, J. Rodriguez, M. Rivera, J. Espi- weighting factors design in model predic- adaption for fast rotor flux estimation in
noza, and J. Kolar, ‘‘Predictive control of tive control of power converters and induction machines,’’ Inst. Elect. Eng. Proc.
an indirect matrix converter,’’ IEEE Trans. drives,’’ in Proc. IEEE Int. Conf. Industrial D, vol. 136, no. 1, pp. 35–43, 1989.
Ind. Electron., vol. 56, no. 6, pp. 1847–1853, Technology, ICIT, Gippsland, Australia, [92] C. Lascu, I. Boldea, and F. Blaabjerg,
June 2009. Feb. 10–13, 2009, pp. 1–6. ‘‘Variable-structure direct torque con-
[56] J. Holtz, ‘‘Pulsewidth modulation—A sur- [72] M. P. Kazmierkowski, R. Krishnan, and F. trol—A class of fast and robust control-
vey,’’ IEEE Trans. Ind. Electron., vol. 39, Blaabjerg, Control in Power Electronics. lers for induction motor drives,’’ IEEE
no. 5, pp. 410–420, Oct. 1992. New York: Academic, 2002. Trans. Ind. Electron., vol. 51, no. 4,
[57] G. Buja and G. Indri, ‘‘Improvement of pul- [73] R. Krishnan, Electric Motor Drives. Engle- pp. 785–792, 2004.
sewidth modulation techniques,’’ Archiv wood Cliffs, NJ: Prentice-Hall, 2001. [93] T. Ohtani, N. Takada, and K. Tanaka, ‘‘Vec-
für Elektrotechnik, vol. 57, no. 5, pp. 281– [74] K. Hasse, ‘‘Drehzahlgelverfahren fur tor control of induction motor without
289, 1975. schnelle Umkehrantriebe mit strom-rich- shaft encoder,’’ IEEE Trans. Ind. Applicat.,
[58] B. P. McGrath and D. G. Holmes, ‘‘Multi- tergespeisten Asynchron-Kurzchlusslau- vol. 28, no. 1, pp. 157–165, 1992.
carrier PWM strategies for multilevel fermotoren,’’ Reglungstechnik, vol. 20, [94] A. Steimel, ‘‘Direct self-control and
inverters,’’ IEEE Trans. Ind. Electron., no. 2, pp. 60–66, 1972. synchronous pulse techniques for high-
vol. 49, no. 4, pp. 858–867, Aug. 2002. [75] F. Blaschke, ‘‘The principle of field-orienta- power traction inverters in comparison,’’
[59] L. R. Limongi, R. Bojoi, G. Griva, and A. tion as applied to the transvector closed- IEEE Trans. Power Electron., vol. 51, no. 4,
Tenconi, ‘‘Digital current-control schemes,’’ loop control system for rotating-field pp. 810–820, 2004.

26 IEEE INDUSTRIAL ELECTRONICS MAGAZINE n SEPTEMBER 2011