Current Control and Reactive Power Minimization of

a Direct Matrix Converter Induction Motor Drive

with Modulated Model Predictive Control
Manjusha Vijayagopal*, Lee Empringham, Liliana de Lillo,
Luca Tarisciotti, Pericle Zanchetta, Patrick Wheeler
Department of Electrical and Electronic Engineering
University of Nottingham
Nottingham, UK
*[email protected]

Abstract— This paper investigates a novel model- predictive above all in high performance drives that require more and
control strategy to control an induction motor fed by a matrix more performing regulation.
converter with multi-objective control capability. The proposed
control method combines the features of the classical Model
Rd Matrix converter
Predictive Control and the Space Vector Modulation technique VA Isa
into a Modulated Model Predictive Control. This new solution
maintains all the characteristics of Model Predictive Control Cf
(such as fast transient response ,multi-objective control using VB Isb SAa SAb SAc
only one feedback loop, easy inclusion of nonlinearities and
constraints of the system, the flexibility to include other system Lf
requirements in the controller) adding the advantages of working VC Isc SBa SBb SBc
at fixed switching frequency and improving the quality of the
controlled waveforms. Simulation results employing the control
method to control the stator current and input reactive power of Bidirectional
switch SCa SCb SCc
a direct matrix converter induction motor drive are presented.

Keywords—matrix converter; reactive power control; Ioa Iob Ioc

current control; SVM; model predictive control Va Vb Vc
I. INTRODUCTION IM
Matrix converters (MC) are ac-ac converters with nine Fig. 1. Open loop schematic of matrix converter drive
bidirectional switches. The interesting feature of a MC is the
absence of an intermediate dc-link. The nine bidirectional Model Predictive Control (MPC), which considers the model
switches of the converter are connected in the form of a matrix of the system to predict its future behavior over a time horizon
in such a way that any input phase can be connected to any was introduced in the late seventies [6]. A cost function is
output phase. The output voltage of the converter is generated constructed in terms of the desired behavior of the system.
by selective switching of the power electronic switches to MPC is hence considered as an optimization problem where
generate signals with desired magnitude and frequency [1]. the future states of the system are obtained by minimizing the
The converter is mostly fed by a three phase supply on the cost function [7]. Due to the fast sampling times used in the
input side and can be connected to any AC load, like an control of power converters, solving the optimization problem
Induction Motor (IM), on the output side as shown in Fig.1. of MPC online is not practical. One approach is to use an
Several modulation methods have been developed for a MC explicit solution of MPC, solving the optimization problem
starting with the Alesina and Venturini modulation [2]. Later offline. The resulting controller is a search tree or a lookup
the Space Vector Modulation (SVM) method which provides a table and can be implemented without big computational
maximum voltage transfer ratio of 0.86 was proposed by effort. Considering that power converters are systems with a
Casadei for direct matrix converter [1]. Since then SVM has finite number of states, given by the possible combinations of
been used together with feedback control strategies to regulate the state of the switching devices, the MPC optimization
specific control variables in the case of electric drives or other problem can be simplified and reduced to the prediction of the
applications [3-5]. The absence of any energy storage device behaviour of the system for each possible state. Then, each
makes the converter smaller and more reliable compared to prediction is evaluated using the cost function and the state
any other topology. that minimizes it, is selected. This control strategy has been
Therefore, the use of these converters in electrical drives proposed in scientific literature to be used in many power
applications is an area of immense potential and interest, converter topologies for current and power control, among
which for three phase voltage source inverters [8] and for
current and torque control in induction motor and permanent period, four active vectors and three zero vectors are applied
magnet motor drives [9-11]. within one sampling interval in M2PC. The vectors are
The several advantages of MPC includes fast dynamic selected from a cost function minimization algorithm. This
response, multivariable control capability, easy inclusion of results in fixed switching frequency as opposed to variable
nonlinearities and constraints of the system and easy tuning of switching frequency in MPC. A classical SVM switching
the controller if the system model is known. The main sequence include four active and three zero vectors arranged
drawback of this control strategy is that the control chooses in a symmetrical pattern in order to minimize the switching
and applies one converter switching state for the entire losses. Each sequence is resulted from the choice of an input
sampling time; due to the absence of a modulator. This results sector and an output reference voltage sector. For all the
in variable switching frequency with higher switching ripple. possible combinations of input and output sectors there are 36
Various methods have been proposed to improve the vector sequences to select the best one from. The symmetrical
sequence [12] or to introduce a modulation scheme in the arrangement of the vectors with four active and three zero
MPC algorithm [13]. These methods however involve vectors is employed [3].
complicated expressions for the switching pattern and are not
flexible enough to include other system requirements in the
A. Model of the System
cost function. This is overcome by Modulated Model
Predictive Control (M2PC), which has been proposed for In order to design a model predictive control system, a precise
cascaded H-bridge converters [14] and active front rectifiers model is required. The model description can be divided into
[15]. matrix converter model and load model as presented in the
In this paper, the M2PC approach is utilized to control the following subsections.
current in an induction machine together with input reactive
power control. The objective is to obtain controlled power 1) Model of MC
factor, low-distortion input currents and good transient and
steady state drive performance using only an M2PC control The voltages at the output of a matrix converter are calculated
loop. from the input voltages and can be derived directly from Fig.1.

Va SAa SBa SCa VA

II. MODULATED MODEL PREDICTIVE CONTROL FOR A
MATRIX CONVERTER DRIVE
[ Vb ] = [SAb SBb SCb ] [ VB ] (1)
Vc SAc SBc SCc VC
The objective of this control strategy is to develop an
algorithm with the positive features of both SVM and MPC.
The input and output voltages can be expressed as vectors as
This will result in a control with a built-in modulation scheme
follows
similar to the SVM. As in MPC the control depends on the
v𝒐 = S . v𝒊 (2)
model of the system and hence an accurate model is vital for
the best performace of the controller. A basic block diagram
where v𝒐 =[𝑉𝑎 𝑉𝑏 𝑉𝑐 ]𝑇 is the output voltage vector,
for the M2PC control strategy is shown in Fig.2.
v𝒊 =[𝑉𝐴 𝑉𝐵 𝑉𝐶 ]𝑇 is the input voltage vector, ‘T’ denotes
transpose of the vector and S is the instantaneous transfer
Vector
sequence
matrix of the direct matrix converter.
selector
* vectors Vsabc
ωr
*
+_
PI Isq
* Modulated
Matrix 2) Input Filter Model
Isd predictive Sijk Vabc
Vsabc Converter IM
control
ωr The model of the input filter is required for the prediction of
input voltages and currents, which are essential to calculate the
Qref

Isd predicted input reactive power. The input filter considered in

ωr
this paper is an LC filter with a damping resistor as shown in
Fig. 3.
Fig. 2 Block diagram for M2PC

Vs, Is Rd
As shown in Fig.2, the control strategy consists of two Vi, Ii
cascaded loops for speed and current-reactive power control. Lf
The speed of the machine is conrolled by a classical
proportional-integral (PI) controller whereras the stator Il Cf
currents in the dq coordinate frame and input reactive power
are controller by the M2PC algorithm.
In order to achieve a modulation-like behavior the M2PC uses
Fig. 3 Schematic of the input filter
the vector sequence employed in the SVM. Unlike MPC
where one voltage vector is applied for the entire sampling
The state space model in continuous time domain is given by co-ordinates and 𝑇𝑠 is the sampling time for the discretization.
the set of equations in (3) The stator currents can then be predicted in order to form the
cost function. The current prediction is also done in the dq
reference frame since a field oriented model is used.
−1 1 1 1
−
𝑅𝑑 𝐶𝑓 𝐶𝑓 𝑅𝑑 𝐶𝑓 𝐶𝑓
𝑥̇ (𝑡) = [ −1 ] 𝑥(𝑡) + [ −1 ] 𝑢(𝑡) (3) 𝑖𝑠𝑑 (𝑘 + 1) = 𝜖𝑉𝑠𝑑 (𝑘) + 𝛾 𝑖𝑠𝑑 (𝑘)
0 0
𝐿𝑓 𝐿𝑓
𝜆
where +𝑇𝑠 [𝜇 𝐼𝑠𝑞 2 + 𝜔𝑒 𝑖𝑠𝑞 (𝑘) + ] (10)
𝜏𝑟
𝑉𝑖 (𝑡) 𝑉 (𝑡)
𝑥(𝑡) = [ ] 𝑎𝑛𝑑 𝑢(𝑡) = [ 𝑠 ] 𝑖𝑠𝑞 (𝑘 + 1) = 𝜖𝑉𝑠𝑞 (𝑘) + 𝛾𝑖𝑠𝑞 (𝑘)
𝐼𝑙 (𝑡) 𝐼𝑖 (𝑡)

−𝑇𝑠 [𝜇 𝐼𝑠𝑑 2 + 𝜔𝑒 𝑖𝑠𝑑 (𝑘) + 𝜆 𝜔𝑒 ] (11)

A discrete state space model of the system is required as the
predictions are not in continuous time. Considering a sampling 𝑅𝑠𝑖𝑔 𝑇𝑠 𝑇𝑠 𝐿𝑚
period Ts, the discrete-time system derived from (3). 𝛾 =1− ;𝜖= ; 𝜇= ;
𝜎𝐿𝑠 𝜎𝐿𝑠 𝜏𝑟 𝜑𝑟𝑑 (𝑘)

3) Load Model 𝑘𝑟 𝜑𝑟𝑑 (𝑘) 𝐿2𝑚

𝜆= ; 𝜎 =1−
𝜎𝐿𝑠 𝐿𝑠 𝐿𝑟
For an induction motor (IM) the output current control is
based on the model of the machine [16] in a field oriented These equations are then used to calculate the predicted stator
reference frame. The stator equations of an IM are given in (4) currents in order to formulate the cost function.
and (5).
B. Prediction of Control Variables
𝑑𝜑𝑠𝑑
𝑉𝑠𝑑 = 𝑖𝑠𝑑 𝑅𝑠 + − 𝜔𝑒 𝜑𝑠𝑞 (4) Once the model of the system is available, control variables at
𝑑𝑡
the future instants can be calculated from the present values.
𝑑𝜑𝑠𝑞 In this case, the control variables are the stator currents and
𝑉𝑠𝑞 = 𝑖𝑠𝑞 𝑅𝑠 + + 𝜔𝑒 𝜑𝑠𝑑 (5) the input reactive power. To implement the M2PC control
𝑑𝑡
strategy the predicted values of the stator current are
where 𝑉𝑠𝑑 and 𝑉𝑠𝑞 are the stator voltages, 𝜑𝑠𝑑 and 𝜑𝑠𝑞 the calculated for all the vectors (active and zero) in a given SVM
sequence as if they were applied for the whole sampling
stator fluxes, 𝑖𝑠𝑑 and 𝑖𝑠𝑞 the stator currents in the dq co-
interval just as in MPC. The same applies for the input
ordinate frame and 𝜔𝑒 is the speed of the stator flux. reactive power. The predicted values of stator currents in dq
Discretising the stator equations (4) and (5) by applying coordinate for the (k+1) instant are shown in (12) to (15).
Euler’s approximation method, the stator fluxes can be Equations (12) and (13) represents the current predictions for
estimated as shown in (6). the active vectors where j= 1,2,3,4 while (14) and (15) are the
current predictions for the three zero vectors.
𝜑𝑠𝑑𝑞 (𝑘 + 1) = 𝑉𝑠𝑑 (𝑘)𝑇𝑠 + 𝜑𝑠𝑑 (𝑘) + 𝜔𝑒 𝜑𝑠𝑞 (𝑘)𝑇𝑠
𝑖𝑠𝑑 (𝑗) (𝑘 + 1) = 𝜖𝑉𝑠𝑑 (𝑗) (𝑘) + 𝛾 𝑖𝑠𝑑 (𝑘)
−𝑖𝑠𝑑 (𝑘)𝑅𝑠 𝑇𝑠 (6)
𝜆
where the rotor flux is derived from the following equations. +𝑇𝑠 [𝜇𝐼𝑠𝑞 2 + 𝜔𝑒 𝑖𝑠𝑞 (𝑘) + ] (12)
𝜏𝑟
𝜑𝑟𝑑𝑞 = 𝐿𝑚 𝑖𝑠𝑑𝑞 + 𝐿𝑟 𝑖𝑟𝑑𝑞 (7)
𝑖𝑠𝑞 (𝑗) (𝑘 + 1) = 𝜖𝑉𝑠𝑞 (𝑗) (𝑘) + 𝛾 𝑖𝑠𝑞 (𝑘)
1
𝑖𝑟𝑑𝑞 = (𝜑𝑠𝑑𝑞 − 𝐿𝑠 𝑖𝑠𝑑𝑞 ) (8)
𝐿𝑚 −𝑇𝑠 [𝜇 𝐼𝑠𝑑 2 + 𝜔𝑒 𝑖𝑠𝑑 (𝑘) + 𝜆 𝜔𝑒 ] (13)

Substituting (8) in (7), the expression of the rotor flux is 𝑖𝑠𝑑 (0) (𝑘 + 1) = 𝛾 𝑖𝑠𝑑 (𝑘)
obtained in (9).
𝜆
𝐿𝑠 𝐿𝑟 𝐿𝑟 +𝑇𝑠 [𝜇𝐼𝑠𝑞 2 + 𝜔𝑒 𝑖𝑠𝑞 (𝑘) + ] (14)
𝜏𝑟
𝜑𝑟𝑑𝑞 = (𝐿𝑚 − ) 𝑖𝑠𝑑𝑞 + 𝜑 (9)
𝐿𝑚 𝐿𝑚 𝑠𝑑𝑞
𝑖𝑠𝑞 (0) (𝑘 + 1) = 𝛾 𝑖𝑠𝑞 (𝑘)
where 𝐿𝑚 , 𝐿𝑠 , 𝐿𝑟 are the mutual, stator and rotor inductance
respectively, 𝑅𝑠 the stator resistance, 𝜑𝑟𝑑𝑞 the rotor flux in dq −𝑇𝑠 [𝜇 𝐼𝑠𝑑 2 + 𝜔𝑒 𝑖𝑠𝑑 (𝑘) + 𝜆 𝜔𝑒 ] (15)
Similarly the input reactive power is obtained from the The application times for all the vectors are then obtained
predictions of the source current according to the discrete from (18) as though each of the vector is applied for one
model of the input filter. The instantaneous reactive power can sampling resulting in a set of linear equations. Since the
be predicted by using calculations are done for one sampling period, the sum of the
application times for all vectors should be equal to the
𝑄(𝑘 + 1) = 𝑉𝑠𝛼 (𝑘 + 1)𝐼𝑠𝛽 (𝑘 + 1) sampling interval.
−𝑉𝑠𝛽 (𝑘 + 1)𝐼𝑠𝛼 (𝑘 + 1) (16)
𝑇𝑠 = 𝜏1 + 𝜏2 + 𝜏3 + 𝜏4 + 𝜏0 (19)
where the subscripts α and β represent real and imaginary
where τ1, τ 2, τ 3, τ 4 are the times for the four active vectors and
components of the associated vector.
t0 is the time for three zero vectors. Combining (18) and (19) it
is possible to obtain the normalizing coefficient ‘K’. The
C. Cost Function Calculation application times for zero vectors can be calculated from the
Similar to the MPC, the switching states are calculated based times for active vectors as shown in (20).
on a cost function minimization. The cost function can include
different performance factors according to the controlled 𝜏0 = 𝑇𝑠 − ( 𝜏1 + 𝜏2 + 𝜏3 + 𝜏4 ) (20)
variables and to the constraints required. More than one
variable can be controlled with the same control loop, The duty cycle ‘δ’ for all the vectors is then calculated using
providing a multi-objective control approach and avoiding (21) for j=0, 1, 2, 3, 4
nested loops.
The cost function for stator current control and input reactive 𝜏(𝑗)
𝛿(𝑗) = (21)
power control are computed for all the five vectors (active and 𝑇𝑠
zero) resulting in 5 cost functions 𝐺0 , 𝐺1 , 𝐺2 , 𝐺3 and 𝐺4 with
j=0, 1, 2, 3, 4. The values of the duty cycles thus obtained can be verified by
using the relation shown in (22).
𝐺(𝑗) = |𝐼𝑠𝑑𝑟𝑒𝑓 (𝑘 + 1) − 𝐼𝑠𝑑 𝑗 (𝑘 + 1)| +
𝛿 1 + 𝛿2 + 𝛿3 + 𝛿4 + 𝛿0 = 1 (22)
|𝐼𝑠𝑞𝑟𝑒𝑓 (𝑘 + 1) − 𝐼𝑠𝑞 𝑗 (𝑘 + 1)|
+𝐴 ∗ |𝑄𝑟𝑒𝑓 (𝑘 + 1) − 𝑄(𝑘 + 1)| (17) The active and zero vectors are then applied for their
respective duty cycles within one sampling interval.
In (17) 𝐼𝑠𝑑𝑟𝑒𝑓 (𝑘 + 1) and 𝐼𝑠𝑞𝑟𝑒𝑓 (𝑘 + 1) are the predicted The obtained duty cycles are directly related to the error
values of the stator current references in dq frame. 𝐼𝑠𝑞𝑟𝑒𝑓 (𝑘 + between the reference and the control variable; thereby
1) is generated by a Proportional-Integral (PI) speed achieving modulation and control functions simultaneously.
controller. For minimization of reactive power, 𝑄𝑟𝑒𝑓 (𝑘 + The intrinsic modulation strategy in the M2PC ensures fixed
1) = 0. ‘A’ is the weighting factor for the reactive power switching frequency.
control. The selection of this variable is done by trial and III. SIMULATION RESULTS
error. The weighting factor decides the controllability of one
control variable over another when more than one control The M2PC control strategy proposed in this paper is applied to
variable exists. Once the cost functions for each of the active a 30kW induction machine fed by a direct matrix converter.
and zero vectors are calculated; the predictive control will The simulations are carried out in MATLAB-Simulink
choose the best sequence of vectors which produces the lowest environment using the schematic diagram in Fig.1 The
error. These vectors are then applied within the sampling parameters for the input filter and load are given in Table.1.
interval with different duty cycles. The duty cycles are TABLE I. SYSTEM PARAMETERS
calculated based on the cost functions values of the active and Parameter Symbol Value
zero vectors. Filter inductance Lf 0.7mH
Filter capacitance(star) Cf 24.9 µF
D. Calculation of Duty Cycles Damping resister Rd 10 Ω
The expressions for the duty cycles are derived based on the Stator resistance Rs 0.081 Ω
characteristic of MPC where the vector with the least error Rotor resistance Rr 0.0788 Ω
will be applied for the longest time. Hence the application Stator leakage inductance Lls 1.36mH
time of a vector is inversely proportional to the cost function Mutual inductance Lm 36.8mH
associated with that vector. This can be expressed by (18), Rotor leakage inductance Llr 0.916mH
Rated Speed ω 1470rpm
𝐾 Rated Voltage V 380V
𝜏(𝑗) = (18)
𝐺(𝑗) Rated Torque Te 195Nm
where ‘G(j)’ is the cost function and ‘𝜏 (j)’ is the time for which Sampling time Ts 50µs
the respective vector is applied.
 Torque and stator current control are achieved by the predictive of speed at full load when controlled by M2PC are shown in
control algorithm while the speed control is implemented with Fig. 5. The simulation is carried out for different values of
a PI regulator. weighting factor, A and the performance is analysed. In Fig.5
Fig.4.a) and 4.b) shows the steady state three phase stator a), there is no input reactive power control and the stator
currents at no load for both MPC and M2PC respectively. As currents, flux, and torque are controlled using the M2PC
seen from the figure, the stator currents of the motor controlled algorithm. However the input currents related to this conditons
using MPC present higher and more spread harmonic content have a considerable amount of distortion as seen from Fig.6a).
than the one controlled by M2PC with the same simulation When the input reactive power control is added to the M2PC
parameteres and conditions. algorithm with a weighting factor of 0.0002 (VA)-1, the
50 40
performance of the machine remained the same and the input
30
current quality is considerably improved as noticeable from
20 Fig.6.b).This proves that there is good control of both control
Stator Currents (A)

10 parameters without compromising the performance of either of

0 0
them. However more tests are done to analyse the perfomance
-10

-20
of the system at different value of A and it is found that as the
-30 value of A increases, the importance on the reactive power
-50
0.6 0.61 0.62 0.63 0.64 0.65 -40
0.4 0.41 0.42 0.43 0.44 0.45
control becomes dominant, which in turn affects the
Time (s)
performance of the load current controller. To demonstrate this,
Time (s)

(a) (b)
1 0.5
when A=0.0006 (VA)-1 , as seen in Fig.6.c) the input currents
THD=11.8% THD=4.8%
0.8 0.4 starts gettting more distorted and the performance of the load is
also affected. This is visible from the distortions in torque and
Amplitude (A)

0.6 0.3
stator currents in Fig. 5.c). The selection of the weighting
0.4 0.2
factor A, as mentioned before is done on a trial and error basis.
0.2 0.1 Therefore the optimum value of A which gives a satisfactory
0 0
control overt the input and output of the drive has to be
0 2000 4000 6000 8000 10000 12000 14000 16000 0 0.5 1 1.5 2 2.5 3 3.5 4
Frequency (Hz) Frequency (Hz)
x 10
4
selected.
(c) (d)
Fig.4 (a) Steady state three phase no load stator currents for M2PC, (b) Steady
state three phase no load stator currents for MPC ,(c) FFT of stator current
IV. CONCLUSIONS
controlled by M2PC , (d) FFT of stator current controlled by MPC The use of M2PC to control an induction motor fed by a direct
To investigate this further, the fast fourier transform (FFT) of matrix converter is discussed in this paper. The prospect of
both the currents shown in Fig.4.c) and 4.d) is analysed. In the using more than one control variable in the cost function of the
case of the MPC controlled drive, a total harmonic distorion control is also investigated with promising simulation results.
(THD) of 11.8% is calculated with more harmonic content in By varying the weighting factor in the cost function, the
the low frequency range. However, in the case of the M2PC relevance of the control variable can be altered according to
controlled drive the THD is about 4.8%. Fig.4.d) shows the the design requirements. A comparison of quality of the stator
harmonics in sidebands multiples of switching frequency currents controlled by MPC and M2PC indicates that M2PC
(20kHz) which is the result of a fixed switching frequency can provide better current quality with fixed switching
operation. The three phase stator currents, startor flux , speed frequency.
and electromagnetic torque of the machine during the reversal
Flux (Wb) currents (A)

200 200 200

0 0
Stator

-200 -200 -200

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7
2 2 2
Stator

0 0 0

-2 -2 -2
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7
2000 2000 2000
(rpm)
Speed

0 0 0

-2000 -2000 -2000

1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7
200
Te (Nm)

200 200
0 0 0
-200 -200 -200
-400 -400 -400
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Time (s) Time (s) Time (s)
(a) (b) (c)
Fig.5: Simulation results for M2PC of IM drive: Three Phase stator currents, Stator flux in alpha-beta , speed of the motor and electromagnetic torque during
speed reversal for (a) A=0 (VA)-1, (b) A=0.0002 (VA)-1and (c) A=0.0006 (VA)-1with a load torque of 195Nm.
 100 100 100

Sourve voltage (V) *1/5 ,

Source current (A)
Source voltage (V) ,

Source Voltage (V),

50
source current (A)

Source Current (A)

50 50

0 0 0

-50 -50 -50

-100 -100 -100

0.9 0.95 1 1.05 1.1 0.9 0.95 1 1.05 1.1 0.9 0.95 1 1.05 1.1
Time (s) Time (s) Time (s)

(a) (b) (c)

Fig.6: Source current and voltage of the matrix converter IM drive controlled by M2PC with (a) A=0 (VA)-1, (b) A=0.0002 (VA)-1 and (c) A=0.0006 (VA)-1 at
steady state with no load.

V. REFERENCES
[9] T. Geyer, G. A. Beccuti, G. Papafotiou, and M. Morari,
[1] D. Casadei, G. Grandi, G. Serra, and A. Tani, "Space "Model Predictive Direct Torque Control of permanent
vector control of matrix converters with unity input power magnet synchronous motors," in IEEE Energy Conversion
factor and sinusoidal input/output waveforms," in Fifth Congress and Exposition (ECCE), 2010, pp. 199-206.
European Conference on Power Electronics and [10] P. Correa, M. Pacas, and J. Rodriguez, "Predictive Torque
Applications, 1993, pp. 170-175 vol.7. Control for Inverter-Fed Induction Machines," IEEE
[2] A. Alesina and M. Venturini, "Analysis and design of Transactions on Industrial Electronics, vol. 54, pp. 1073-
optimum-amplitude nine-switch direct AC-AC converters," 1079, 2007.
IEEE Transactions on Power Electronics, vol. 4, pp. 101- [11] R. Vargas, U. Ammann, B. Hudoffsky, J. Rodriguez, and P.
112, 1989. Wheeler, "Predictive Torque Control of an Induction
[3] D. Casadei, G. Serra, and A. Tani, "The use of matrix Machine Fed by a Matrix Converter With Reactive Input
converters in direct torque control of induction machines," Power Control," IEEE Transactions on Power Electronics,
IEEE Transactions on Industrial Electronics, vol. 48, pp. vol. 25, pp. 1426-1438, 2010.
1057-1064, 2001. [12] H. Jiabing and Z. Q. Zhu, "Improved Voltage-Vector
[4] P. Wheeler, J. C. Clare, M. Apap, D. Lampard, S. Sequences on Dead-Beat Predictive Direct Power Control
Pickering, K. J. Bradley, et al., "An Integrated 30kW of Reversible Three-Phase Grid-Connected Voltage-Source
Matrix Converter based Induction Motor Drive," in 36th Converters," IEEE Transactions on Power Electronics,
IEEE Power Electronics Specialists Conference, PESC vol. 28, pp. 254-267, 2013.
'05, pp. 2390-2395. [13] P. Antoniewicz and M. P. Kazmierkowski, "Virtual-Flux-
[5] T. F. Podlesak, D. C. Katsis, P. W. Wheeler, J. C. Clare, L. Based Predictive Direct Power Control of AC/DC
Empringham, and M. Bland, "A 150-kVA vector-controlled Converters With Online Inductance Estimation," IEEE
matrix converter induction motor drive," IEEE Transactions on Industrial Electronics, vol. 55, pp. 4381-
Transactions on Industry Applications, vol. 41, pp. 841- 4390, 2008.
847, 2005. [14] L. Tarisciotti, P. Zanchetta, A. Watson, J. Clare, and S.
[6] E. F. Camacho and C. B. Alba, Model predictive control: Bifaretti, "Modulated Model Predictive Control for a 7-
Springer, 2013. Level Cascaded H-Bridge back-to-back Converter," IEEE
[7] J. Rodriguez and P. Cortes, Predictive control of power Transactions on Industrial Electronics, vol. PP, pp. 1-1,
converters and electrical drives vol. 37: John Wiley & 2014.
Sons, 2012. [15] L. Tarisciotti, P. Zanchetta, A. Watson, J. Clare, M.
[8] J. Rodriguez, J. Pontt, C. A. Silva, P. Correa, P. Lezana, P. Degano, and S. Bifaretti, "Modulated model predictve
Cortes, et al., "Predictive Current Control of a Voltage control (M2PC) for a 3-phase active front-end," in 2013
Source Inverter," IEEE Transactions on Industrial IEEE Energy Conversion Congress and Exposition
Electronics, vol. 54, pp. 495-503, 2007. (ECCE),pp.1062-1069.