HAL Id: hal-00525393
https://hal.archives-ouvertes.fr/hal-00525393
Submitted on 11 Oct 2010
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Hybrid Cascaded H-Bridge Multilevel-Inverter
Induction-Motor-Drive Direct Torque Control for
Automotive Applications
Farid Khoucha, Soumia Mouna Lagoun, Khoudir Marouani, Abdelaziz
Kheloui, Mohamed Benbouzid
To cite this version:
Farid Khoucha, Soumia Mouna Lagoun, Khoudir Marouani, Abdelaziz Kheloui, Mohamed Benbouzid.
Hybrid Cascaded H-Bridge Multilevel-Inverter Induction-Motor-Drive Direct Torque Control for Au-
tomotive Applications. IEEE Transactions on Industrial Electronics, Institute of Electrical and Elec-
tronics Engineers, 2010, 57 (3), pp.892-899. �10.1109/TIE.2009.2037105�. �hal-00525393�
892 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 3, MARCH 2010
Hybrid Cascaded H-Bridge Multilevel-Inverter
Induction-Motor-Drive Direct Torque Control
for Automotive Applications
Farid Khoucha, Soumia Mouna Lagoun, Khoudir Marouani, Abdelaziz Kheloui, and
Mohamed El Hachemi Benbouzid, Senior Member, IEEE
Abstract—This paper presents a hybrid cascaded H-bridge
multilevel motor drive direct torque control (DTC) scheme for
electric vehicles (EVs) or hybrid EVs. The control method is based
on DTC operating principles. The stator voltage vector reference
is computed from the stator flux and torque errors imposed by
the flux and torque controllers. This voltage reference is then
generated using a hybrid cascaded H-bridge multilevel inverter,
where each phase of the inverter can be implemented using a
dc source, which would be available from fuel cells, batteries, or
ultracapacitors. This inverter provides nearly sinusoidal voltages
with very low distortion, even without filtering, using fewer switch-
ing devices. In addition, the multilevel inverter can generate a high
and fixed switching frequency output voltage with fewer switching
losses, since only the small power cells of the inverter operate at
a high switching rate. Therefore, a high performance and also
efficient torque and flux controllers are obtained, enabling a DTC
solution for multilevel-inverter-powered motor drives.
Index Terms—Automotive application, direct torque control
(DTC), induction motor, multilevel inverters.
I. INTRODUCTION
M
ULTILEVEL voltage-source inverter topologies, in-
cluding diode-clamped, flying capacitor, and cascaded
H-bridge structures, are intensively studied for high-power
applications [1]–[5], and standard drives for medium-voltage
industrial applications have become available [6], [7]. Solutions
with a higher number of output voltage levels have the ability
to synthesize waveforms with a better harmonic spectrum and
to limit the motor-winding insulation stress. However, their
increasing number of devices tends to reduce the overall reli-
ability and efficiency of the power converter.
Manuscript received December 17, 2008; revised November 12, 2009. First
published December 4, 2009; current version published February 10, 2010.
F. Khoucha is with the Laboratoire Brestois de Mécanique et des Systèmes
(EA 4325), University of Brest, 29238 Brest, France, and also with the Elec-
trical Engineering Department, Polytechnic Military Academy, Algiers 16111,
Algeria (e-mail: fkhoucha04@yahoo.fr).
S. M. Lagoun is with the Laboratoire Brestois de Mécanique et des Systèmes
(EA 4325), University of Brest, 29238 Brest, France, and also with the Univer-
sity of Laghouat, Laghouat 03000, Algeria (e-mail: lagoun_mona@yahoo.fr).
K. Marouani and A. Kheloui are with the Electrical Engineering Department,
Polytechnic Military Academy, Algiers 16111, Algeria (e-mail: marouani_
khoudir@yahoo.fr; akheloui@caramail.com).
M. E. H. Benbouzid is with the Laboratoire Brestois de Mécanique et
des Systèmes (EA 4325), University of Brest, 29238 Brest, France (e-mail:
m.benbouzid@ieee.org).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIE.2009.2037105
On the other hand, solutions with a low number of levels
either need a rather large and expensive LC output filter to limit
the motor-winding insulation stress or can only be used with
motors that do withstand such stress.
Most investigations concerned topologies with the same volt-
age rating for all devices. The advantages of such symmetric
multilevel converters are modularity and control s implicity. Hy-
brid multilevel inverters use different types of voltage sources
in various parts of the inverter. These sources, which can be
batteries, ultracapacitors, or fuel cell, allow this structure to be
narrowly well adapted to electric vehicles (EVs) and hybrid
EVs (HEVs) [8]–[11]. By addition and subtraction of these
voltages, more different output voltage levels can be generated
with the same number of components, compared to a symmetric
multilevel inverter [12]–[17]. Higher output quality can be
obtained with fewer cascaded cells and control complexity, and
output filters can be remarkably shrunk or even eliminated.
One of the methods that have been used by one major man-
ufacturer in multilevel inverters is direct torque control (DTC),
which is recognized today as a high-performance control
strategy for ac drives [18]–[24]. Several authors have addressed
the problem of improving the behavior of DTC ac motors,
particularly by reducing the torque ripple. Different approaches
have been proposed [12]: improving the lookup table; varying
the hysteresis bandwidth of the torque controller; and using
flux, torque, and speed observers. Although these approaches
are well suitable for the classical two-level inverter, their exten-
sion to a greater number of levels is not easy. Throughout this
paper, a theoretical background is used to design a simple and
practical strategy that is compatible with hybrid cascaded H-
bridge multilevel inverter [25]. It allows not only controlling the
electromagnetic state of the motor with improved performance
(minimization of the torque ripple) but also reducing flux and
current distortion caused by flux position sector change. An
induction motor has been adopted for the vehicle propulsion
because it seems to be the candidate that better fulfills the
major requirements for EV or HEV electric propulsion [26].
II. C
ASCADED H-BRIDGE STRUCTURE AND OPERATION
The hybrid cascaded H-bridge inverter power circuit is
shown in Fig. 1. The inverter is composed of three legs, in
each one is a series connection of two H-bridge inverters fed
by independent dc sources that are not equal (V
1
<V
2
).
0278-0046/$26.00 © 2010 IEEE
Authorized licensed use limited to: Universite de Bretagne Occidentale. Downloaded on February 15,2010 at 06:18:21 EST from IEEE Xplore. Restrictions apply.
KHOUCHA et al.: MULTILEVEL-INVERTER INDUCTION-MOTOR-DRIVE DTC FOR AUTOMOTIVE APPLICATIONS 893
Fig. 1. Asymmetric cascaded H-bridge multilevel inverter.
Indeed, it may be obtained from batteries, fuel cells, or
ultracapacitors in EVs or HEVs [11], [27], [28].
The use of asymmetric input voltages (inverter fed by a set of
dc-voltage sources where at least one of them is different from
the other one) can reduce, or when properly chosen, eliminate
redundant output levels, maximizing the number of different
levels generated by the inverter. Therefore, this topology can
achieve the same output voltage quality with fewer number of
semiconductors [13].
The maximum number of redundancies is equal to (3
K
−
2K − 1) and can be obtained when the partial dc voltages
are equal to E/(N − 1). If there is K connected cells per
multilevel-inverter phase leg, 3
K
switching configurations are
possible. The multilevel-inverter output voltage depends on the
partial voltage feeding each partial cell. The possible number
of redundant switching states can be reduced if the cells are
fed by unequal dc-voltage sources. This also reduces volume
and costs and offers inherent low switching losses and high
conversion efficiency [29]. When cascading two-level inverters
like H-bridges, the optimal asymmetry is obtained by using
voltage sources proportionally scaled to the two- or three-
H-bridge power.
Particular cell i can generate three levels (+V
i
, 0, −V
i
).
The total inverter output voltage for a particular phase j is
defined by
ν
jN
=
m
i=1
ν
ji
=
m
i=1
V
i
(S
i1
− S
i2
),j∈{a, b, c} (1)
where ν
ij
is the i cell output voltage, m is the number of cells
per phase, and (S
i1
,S
i2
) is the switching state associated to the
i cell. Equation (1) explicitly shows how the output voltage of
one cell is defined by one of the four binary combinations of
switching state, with “1” and “0” representing the “ON” and
“OFF” states of the corresponding switch, respectively.
The optimal asymmetry is obtained with dc links scaled
in powers of two or three, generating seven (Fig. 2) or nine
(Fig. 3) different output levels. Nine different output levels can
be generated using only two cells (eight switches), while four
cells (16 switches) are necessary to achieve the same amount of
level with a symmetric-fed inverter.
Fig. 2. Asymmetric multilevel inverter with seven-level output voltage
synthesis.
Fig. 3. Asymmetric multilevel inverter with nine-level output voltage
synthesis.
III. INDUCTION MOT OR DTC
DTC is an alternative method to flux-oriented control. The
basic principle is the selection of the electromagnetic torque
and stator flux references by choosing the appropriate inverter
state [30]. Several advantages may be considered, namely,
nearly sinusoidal stator flux and current waveforms, higher ro-
bustness regarding motor parameter variations except the stator
winding resistance, higher torque dynamics, and easier flux and
speed estimator implementation, since coordinate transforma-
tion is not required. However, in the standard version, important
torque ripple and high switching losses are obtained due to the
use of hysteresis bands and the small number of applicable
voltage vectors. Moreover, the converter switching frequency
is inherently variable and very dependent on the torque and
shaft speed. This produces torque harmonics with variable fre-
quencies and acoustic noise with disturbance intensities that are
very dependent on these mechanical variables and particularly
grating at low speed [18], [31].
The additional degrees of freedom (space vectors, phase
configurations, etc.) provided by the multilevel inverter should
therefore be exploited by the control strategy in order to reduce
the previously cited drawbacks.
A. Nomenclature
ν
s
(i
s
) Stator voltage (current) vector.
φ
s
(φ
r
) Stator (rotor) flux vector.
Authorized licensed use limited to: Universite de Bretagne Occidentale. Downloaded on February 15,2010 at 06:18:21 EST from IEEE Xplore. Restrictions apply.
894 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 3, MARCH 2010
Fig. 4. Influence of ν
s
over φ
s
during a simple interval Δt.
T
e
Electromagnetic torque.
R
s
Stator resistance.
L
s
(L
r
) Stator (rotor) inductance.
L
m
Magnetizing inductance.
σ Total leakage coefficient, =1− L
2
m
/L
s
L
r
.
θ
sr
Angle between stator and rotor flux vectors.
p Pole pair number.
B. Torque and Flux Estimation
The stator flux vector of an induction motor is related to the
stator voltage and current vectors by
dφ
s
(t)
dt
= ν
s
(t) − R
s
i
s
(t). (2)
Maintaining ν
s
constant over a sample time interval and
neglecting the stator resistance, the integration of (2) yields
Δφ
s
(t)=φ
s
(t) − φ
s
(t − Δt)=
t
t−Δt
ν
s
Δt. (3)
Equation (3) reveals that the stator flux vector is directly
affected by variations on the stator voltage vector. On the
contrary, the influence of ν
s
over the rotor flux is filtered by the
rotor and stator leakage inductances [30] and is therefore not
relevant over a short-time horizon. Since the stator flux can be
changed quickly while the rotor flux rotates slower, the angle
between both vectors θ
sr
can be controlled directly by ν
s
.A
graphical representation of the stator and rotor flux dynamic
behaviors is shown in Fig. 4. The exact relationship between
the stator and rotor fluxes shows that keeping the amplitude of
φ
s
constant will produce a constant flux φ
r
[6].
Since the electromagnetic torque developed by an induction
motor can be expressed by [6]
T
e
=
3
2
p
L
m
σL
s
L
r
φ
s
φ
r
sin θ
sr
. (4)
It follows that a change in θ
sr
due to the action of ν
s
allows for
direct and fast change in the developed torque.
DTC uses this principle to achieve the induction motor de-
sired torque response by applying the appropriate stator voltage
vector to correct the flux trajectory.
C. Voltage Vector Selection
Fig. 5 shows one of the 343 voltage vectors generated by
the inverter at instant t = k, denoted by ν
k
s
(central dot). The
Fig. 5. Possible voltage changes Δν
k
s
that can be applied from certain ν
k
s
.
Fig. 6. Voltage selection Δν
k
s
in sector 2.
next voltage vector to be applied to the load ν
k+1
s
can be
expressed by
ν
k+1
s
= ν
k
s
+Δν
k
s
(5)
where Δν
k
s
= {ν
i
|i =1,...,6}. Each vector ν
i
corresponds to
one corner of the elemental hexagon illustrated in gray and by
the dashed line in Fig. 5. The task is to determine which ν
k+1
s
will correct the torque and flux responses, knowing the actual
voltage vector ν
k
s
, the torque and flux errors e
k
φ
and e
k
T
, and
the stator flux vector position (sector determined by θ
s
), except
if the voltage vector is in the end of the external hexagon.
In this case, a trajectory correction is necessary because the
voltage vector that should be chosen by the algorithm cannot be
achieved by the inverter (this will be clearly shown in Fig. 8).
Note that the next voltage vector ν
k+1
s
applied to the load will
always be one of the six closest vectors to the previous ν
k
s
;this
will soften the actuation effort and reduce the high dynamics in
torque response due to possible large changes in the reference.
Using (4) and (5) and analyzing, for example, sector 2
shown in Fig. 6, the application of ν
1
increases the stator
flux amplitude but reduces θ
sr
, leading to a torque reduction.
Conversely, ν
4
reduces the flux magnitude, while it increases
θ
sr
and, thus, the torque. If ν
3
is applied to the load, both torque
and flux increase, and it is clear that ν
6
produces the inverse
effect. Table I summarizes the vector selections according
to the aforementioned criterion for the different sectors and
comparator outputs (desired φ
s
and T
e
corrections).
To implement the DTC of the induction motor fed by a
hybrid H-bridge multilevel inverter, one should determine at
each sampling period the logic state of the inverter switches
Authorized licensed use limited to: Universite de Bretagne Occidentale. Downloaded on February 15,2010 at 06:18:21 EST from IEEE Xplore. Restrictions apply.
KHOUCHA et al.: MULTILEVEL-INVERTER INDUCTION-MOTOR-DRIVE DTC FOR AUTOMOTIVE APPLICATIONS 895
TAB LE I
V
OLTAGE VECTOR SELECTION LOOKUP TABLE
Fig. 7. Optimal space vector tracking and trajectory correction in the station-
ary αβ frame.
as a function of instantaneous values of torque and flux for the
selection of the space vector in the α−β frame.
The proposed control algorithm was divided into two major
tasks which are independent and executed in cascade.
The first task aims at the control of the electromagnetic state
of the machine. The instantaneous values of torque, flux, and
their variations will be taken into account for the selection of
the space vector in the α−β plane. Once the s pace is chosen,
the sequence of phase l evels can be selected. To ensure this task,
one should detect the position of the space vector in the α−β
frame (Q
k
at sampling time t
k
). The proposed algorithm must
then select the next position Q
k+1
to be achieved before the
next sampling instant t
k+1
(Fig. 7) in order to reduce voltage
step magnitude. Only one step displacement in the α−β frame
is authorized per sampling period T
s
. Hence, in the absence
of inverter saturation, Q
k+1
must coincide with one of the six
corners of the elementary hexagon centered at Q
k
.Thesame
procedure will be carried out at the next period in order to
determine the next trajectory direction, yielding Q
k+2
, which,
in turn, will coincide with one of the six corners of the new
elementary hexagon centered at Q
k+1
.
The second task exploits the degree of freedom related to
the multilevel topology to choose the sequence of phase levels
that synthesizes the voltage vector that was selected previously.
There are several sequences of phase levels that are able to gen-
erate the same vector shown in Fig. 8; this degree of freedom
can therefore be exploited to reduce the voltage step magnitude
Fig. 8. Space vector and sequences of a seven-level cascaded H-bridge
inverter.
according to one of the following criteria: 1) minimize the com-
mutation number per period; 2) distribute commutations for the
three phases per period; or 3) choose a vector which minimizes
the homopolar voltage. This task allows losses and torque ripple
minimization. Finally, the configuration of each phase will be
selected and must be able to generate the phase levels.
In the case of inverter saturation, which happens if Q
k
gives an unreachable point for Q
k+1
, a trajectory correction
is necessary. Fig. 7 shows the three cases which lead to an
unreachable point. In cases 2) and 3), the closest displacement
direction is selected. Case 1) illustrates a particular situation
in which no switching should be performed since the nearest
reachable trajectory goes roughly toward the opposite sense of
the favored one given by the lookup table (Table I).
IV. S
IMULATION RESULTS
For the validation of the previously discussed control ap-
proach, simulations on a seven-level inverter-fed induction
motor have been carried out using Matlab–Simulink. The sim-
ulated induction motor ratings are given in the Appendix.
Figs. 9–14 show the simulation results for torque, flux,
currents, voltages, and the corresponding fast Fourier transform
(FFT) spectrums, when the square torque reference waveform
of Fig. 9 is applied.
When analyzing the simulation results, the following conclu-
sions are drawn.
1) Torque pulsations are reduced compared to a two-
and five-level inverter DTC strategy, as shown in
Figs. 15 and 16.
2) DTC also controls the stator flux that is shown in Fig. 10.
With this strategy, it has been possible to improve the flux
Authorized licensed use limited to: Universite de Bretagne Occidentale. Downloaded on February 15,2010 at 06:18:21 EST from IEEE Xplore. Restrictions apply.
