mathematics

Article
Performance Comparison of Numerical Methods in a Predictive
Controller for an AC–DC Power Converter
Jazmin Ramirez-Hernandez , Oswaldo Ulises Juarez-Sandoval * , Leobardo Hernandez-Gonzalez * ,
Domingo Cortes , Juan C. Sanchez-Garcia and Pedro Guevara-Lopez

Escuela Superior de Ingeniería Mecánica y Eléctrica, Unidad Culhuacan, Instituto Politécnico Nacional,
Av. Santa Ana No. 1000, Col. San Francisco Culhuacan, Mexico City 04430, Mexico
* Correspondence: [email protected] (O.U.J.-S.); [email protected] (L.H.-G.)

Abstract: The use of model-based predictive control in power converters has substantially increased
in recent years. This control technique always needs a discrete system model to be implemented.
There are several methods to obtain a discrete model; in this paper, all common methods are examined
from a practical point of view. Their precision, simplicity, and implementation requirements are
analyzed to establish their advantages and disadvantages. From this analysis, it is shown that
different discretization methods result in different closed-loop converter performance. A model-
based predictive control AC–DC converter is used to show that different discretization procedures
result in different total harmonic distortion. For this evaluation, a simulation of a 1 kW three-phase
active rectifier was performed in Matlab-Simulink.

Citation: Ramirez-Hernandez, J.; Keywords: predictive control; discrete model; active filter; AC–DC converter
Juarez-Sandoval, O.U.;
Hernandez-Gonzalez, L.; Cortes, D.; MSC: 65K05
Sanchez-Garcia, J.C.; Guevara-Lopez,
P. Performance Comparison of
Numerical Methods in a Predictive
Controller for an AC–DC Power 1. Introduction
Converter. Mathematics 2022, 10, 2818.
Power converters are necessary in any application that requires regulating, storing,
https://doi.org/10.3390/
distribution, and changing the characteristics of electric energy. General performance of
math10152818
these converters is greatly determined by their controller. That is why the improvement of
Academic Editors: Juan Francisco control algorithms and their implementation represent an incessant research activity. As a
Sánchez-Pérez, Gonzalo García Ros consequence, many different control schemes have been developed for power converters.
and Manuel Conesa Hysteresis control compares the measured variable with a reference and determines
Received: 20 June 2022
the switching state of the semiconductor devices; however, it introduces variable switching
Accepted: 3 August 2022
frequencies, causing resonance problems [1]. Classical techniques, such as proportional–
Published: 8 August 2022
integral (PI) are widely used, but its inclusion in nonlinear systems increases the complexity
of the mathematical models and the controller design [2].
Publisher’s Note: MDPI stays neutral
The model-based predictive control (MPC) technique has become popular due to its
with regard to jurisdictional claims in
robustness, implementation simplicity, and capability to be applied to different kind of
published maps and institutional affil-
systems [3]. For example, in addition to power converters, predictive control works for
iations.
robotic systems [4] and industrial applications [5]. When applied to power converters, this
control technique uses a discrete time model to control a system variable that is commonly
voltage [6], current [7], or torque [8].
Copyright: © 2022 by the authors.
In general, the predictive control algorithm uses a system model to predict the behavior
Licensee MDPI, Basel, Switzerland. of the variable to be controlled and then select the best system operation such that the
This article is an open access article variable becomes as close as possible to the required behavior [9]. More precisely, it
distributed under the terms and requires a discrete system model to calculate the one-step-ahead value of a variable for all
conditions of the Creative Commons possible switching conditions of the system. Then, the best switch configuration is selected
Attribution (CC BY) license (https:// using a cost function that determines the lowest error between the calculated value and
creativecommons.org/licenses/by/ its reference. Finally, the appropriate control signals are sent to the system for the next
4.0/). switching period [10].

Mathematics 2022, 10, 2818. https://doi.org/10.3390/math10152818 https://www.mdpi.com/journal/mathematics

 selected  using  a  cost  function  that  determines  the  lowest  error  between  the  ca
value and its reference. Finally, the appropriate control signals are sent to the sy
the next switching period [10]. 
Until now, to obtain the required discretized model to implement predictive
Mathematics 2022, 10, 2818 2 of 19
most of the previously published works used the Euler method [11–14] or consid
method to be used irrelevant. This paper shows that the employed method to obt
a discretized model is not irrelevant, and that different methods give rise to diffe
Until now, to obtain the required discretized model to implement predictive control,
formances. To this end, an AC–DC converter is used. 
most of the previously published works used the Euler method [11–14] or considered
the method to be used irrelevant. This paper shows that the employed method to obtain
An AC–DC converter was selected due to its critical use in medium‐ and hig
such a discretized model is not irrelevant, and that different methods give rise to different
applications, e.g., in renewable energies, for battery recharge in electric and hyb
performances. To this end, an AC–DC converter is used.
cles [15–17]. Conventional AC–DC topologies first perform the AC–DC conversio
An AC–DC converter was selected due to its critical use in medium- and high-power
rect the power factor [18], followed by a DC–DC conversion to adjust the desired
applications, e.g., in renewable energies, for battery recharge in electric and hybrid ve-
hicles [15–17]. Conventional AC–DC topologies first perform the AC–DC conversion to
voltage, using boost or buck topologies [19]. 
correct the power factor [18], followed by a DC–DC conversion to adjust the desired output
In particular, a conventional three‐phase rectifier topology is presented in F
voltage, using boost or buck topologies [19].
which includes power transistors with antiparallel diodes as the main power s
In particular, a conventional three-phase rectifier topology is presented in Figure 1,
This rectifier operates with high switching frequency and is also known as an acti
which includes power transistors with antiparallel diodes as the main power switches.
end (AFE) rectifier. It overcomes all the problems and limitations of diode and t
This rectifier operates with high switching frequency and is also known as an active front-
end (AFE) rectifier. It overcomes all the problems and limitations of diode and thyristor
rectifiers [20]. Its main features are controlled DC voltage, controlled input curre
rectifiers [20]. Its main features are controlled DC voltage, controlled input currents with
sinusoidal waveform (reduced harmonics), operation with a very high power fac
sinusoidal waveform (reduced harmonics), operation with a very high power factor, and
bidirectional operation. 
bidirectional operation.

 
Figure 1. AC–DC converter to charge a battery. 
Figure 1. AC–DC converter to charge a battery.

The mathematical models that describe the system behavior are commonly differ-
The mathematical models that describe the system behavior are commonly d
ential equations that must be discretized to work with sampling periods. It is not clear
tial  equations 
whether that  must  be 
different discretization discretized 
procedures result to  work  with 
in different valuessampling 
for the costperiods. 
function It  is  n
whether different discretization procedures result in different values for the cost 
and, hence, affect the converter performance. Discretization procedures are based on nu-
merical methods for solving differential equations, the most common of which is the Euler
and, hence, affect the converter performance. Discretization procedures are base
approximation due to its simplicity [21].
merical methods for solving differential equations, the most common of which is t
This paper presents a comparison of some numerical methods such as Euler, Runge–
approximation due to its simplicity [21]. 
Kutta, and trapezoidal approximation of first, second, and third order to obtain the discrete
modelsThis paper presents a comparison of some numerical methods such as Euler,
applied in the predictive controller for the active rectifier of Figure 1. Some aspects
to be evaluated are the exactitude, simplicity of the mathematical procedure, processing
Kutta, and trapezoidal approximation of first, second, and third order to obtain
time, and effect in the THD on the input currents. Comparison results show that the
crete models applied in the predictive controller for the active rectifier of Figure
discretization method employed to obtain the required discrete model to implement the
aspects to be evaluated are the exactitude, simplicity of the mathematical procedu
predictive control does affect the overall performance of power converters. The obtained
cessing time, and effect in the THD on the input currents. Comparison results sh
results also show that the first-order trapezoidal approximation is a good tradeoff between
the discretization method employed to obtain the required discrete model to im
the obtained performance and implementation simplicity. The comparison was only
performed through simulation because the effects that arise in practical implementation
the predictive control does affect the overall performance of power converters.
would affect all methods equally.
tained results also show that the first‐order trapezoidal approximation is a good
The remainder of the paper is organized as follows: in Section 2, predictive control
between the obtained performance and implementation simplicity. The compari
is precisely described; in Section 3, common methods for system model discretization
only performed through simulation because the effects that arise in practical im
are revisited; Section 4 describes the cost function employed to select the best switches
tation would affect all methods equally. 
The remainder of the paper is organized as follows: in Section 2, predictive c
precisely described; in Section 3, common methods for system model discretiza
 revisited; Section 4 describes the cost function employed to select the best switc
Mathematics 2022, 10, 2818
figuration; the comparison of methods to obtain the required discrete models for 3 of 19
tive control implementation is carried out in Section 5; simulation results are pres
Section 6; in Section 7, the simulations results are discussed; lastly, in Section 8, so
clusions are given. 
configuration; the comparison of methods to obtain the required discrete models for pre-
dictive control implementation is carried out in Section 5; simulation results are presented
in Section 6; in Section 7, the simulations results are discussed; lastly, in Section 8, some
2. The Predictive Control Technique 
conclusions are given.
A scheme of the general predictive control is shown in Figure 2, the aim is to
2.the variable X. The first step is to obtain a mathematical expression to model its b
The Predictive Control Technique
A scheme of the general predictive control is shown in Figure 2, the aim is to control
commonly a differential equation of n order determined by the specific system ch
the variable X. The first step is to obtain a mathematical expression to model its behavior,
istics. According to the state variables representation this model may be expresse
commonly a differential equation of n order determined by the specific system charac-
lows [22]: 
teristics. According to the state variables representation this model may be expressed as
follows [22]: . X  AX  Bu , 
X = AX + Bu, (1)
where A is the state matrix, B is the input matrix, and u is the input data. 
where A is the state matrix, B is the input matrix, and u is the input data.

 
Figure 2. General control scheme for predictive control. 
Figure 2. General control scheme for predictive control.

A discrete expression may be obtained from Equation (1) to calculate the one-step-
ahead A discrete expression may be obtained from Equation (1) to calculate the o
variable X k+1 and evaluate all the system operating modes. For example, in power
ahead variable X
converters, k+1 and evaluate all the system operating modes. For example, i
the possible switching combinations of the semiconductor devices produce
converters, 
diverse the modes.
operating possible  switching  combinations  of  the  semiconductor  devices 
The specific parameter to be evaluated is contrasted using the cost function, which
diverse operating modes. 
selects the operating mode that produces the closest value to the reference X kref and sends
The specific parameter to be evaluated is contrasted using the cost function
the control signals, SW k , to the system for the next sampling period.
selects the operating mode that produces the closest value to the reference X
An important step is the discrete model definition, such that a more accurate model kref an
the control signals, SW
results in less error produced k, to the system for the next sampling period. 
between the calculated value and the reference and, thus,
closer tracking to the desired reference behavior.
An important step is the discrete model definition, such that a more accurat
Figure 3 shows an example of X k and X kref signals. During the present sampling time
results in less error produced between the calculated value and the reference an
(k), the predicted values X k +1 obtained for each switching vector state of the control signals
closer tracking to the desired reference behavior. 
are evaluated; the one that generates the closer value to the reference is selected and applied
Figure 3 shows an example of X
in the next k and Xkref
sampling period. The process is performed  signals. During the present sampl
each sampling period.
(k), the predicted values Xk + 1 obtained for each switching vector state of the contro
are evaluated; the one that generates the closer value to the reference is selected
plied in the next sampling period. The process is performed each sampling perio
Mathematics 2022, 10, x FOR PEER REVIEW 
  Mathematics 2022, 10, 2818 4 of 19

 
Mathematics 2022, 10, x FOR PEER REVIEW 
Figure 3. Predictive control principle of operation. 5  of  22
  Figure 3. Predictive control principle of operation. 
3. Numerical Solutions to Obtain Discrete Model
3. Numerical Solutions to Obtain Discrete Model 
To obtain the discrete model of Equation (1), different numerical solution methods can
  such as forward and backward Euler, trapezoidal, Runge–Kutta, and compound
be used,
To obtain the discrete model of Equation (1), different numerical solution m
trapezoidal methods; these methods are analyzed below.
can be used, such as forward and backward Euler, trapezoidal, Runge–Kutta, a
3.1. Forward Euler Method 
pound trapezoidal methods; these methods are analyzed below. 
3.1. Forward Euler Method
This is the simplest method to solve a first‐order differential equation. The interva
   This is the simplest method to𝑇 solve
𝑡 is divided into subintervals  a first-order differential equation. nThe
obtaining a set of discrete points k interval t is
as shown in Figure 4
divided into subintervals T obtaining a set of discrete points kn as shown in Figure 4, where
where interval  𝑇  is the sampling period. At k exists a value Xk; considering that T is very
interval T is the sampling period. At k exists a value X k ; considering that T is very small,
small, an approximation of X
an  can be calculated considering a local approximation equiv
approximation of X k+1 can bek+1calculated considering a local approximation equivalent
alent to the local slope of X k, according to 
to the local slope of X k , according to
dX X  Xk
dX
= lim
klim
X k 1
+1 − Xk
. .  (2) (2
Tdt T
T  0
dt →0 T

 
Figure 4. Approximation using forward and backward Euler method. 
Figure 4. Approximation using forward and backward Euler method.

The value of the next sampling period is determined using a local approximation as
The value of the next sampling period is determined using a local approximation as
follows [23]:
follows [23]:   
∼ Xk + T dX .
Xk +1 = (3)
dt  dX 
X k 1  X k  T   (3
From Equation (1), it can be written that  dt  . 
From Equation (1), it can be written that 
 
dX
= AX + Buk . (4)
dt  dX  k
   AX k  Buk .  (4
 dt 
Substituting Equation (4) into Equation (3), the approximation to obtain the one‐step
ahead value is defined as 
X k 1  X k  T  AX k  Buk  (5
. 
Mathematics 2022, 10, 2818 5 of 19

Substituting Equation (4) into Equation (3), the approximation to obtain the one-step-
ahead value is defined as
Xk +1 ∼
= Xk + T (AXk + Buk ). (5)

3.2. Backward Euler Method

This method considers the derivation of the model from the k + 1 sampling period, as
shown in Figure 4, and is defined by
 
dXk+1
= AXk+1 + Buk+1 . (6)
dt

Substituting Equation (6) into Equation (3), the following expression is obtained:

Xk +1 ∼
= Xk + T (AXk+1 + Buk+1 ). (7)

The function of the next sampling period is calculated solving for Xk+1 .
−1
Xk +1 ∼
= (Xk − TBuk+1 )(Id − TA) , (8)

where Id is the identity matrix. Considering that T becomes very small and that the local
approximation of Equation (2) uses uk+1 ≈ uk , the known value uk may be used instead uk+1 .
−1
Xk +1 ∼
= (Xk − TBuk )(Id − TA) . (9)

3.3. Runge–Kutta Method

The Runge–Kutta method derives an approximate solution for a differential equation
using a finite number of terms of a Taylor series. In this paper, a fourth-order Runge–Kutta
method is analyzed to ensure a smaller error in the approximation. As shown in Figure 5,
slopes C1 , C2 , C3 , and C4 with weights 1, 2, 2, and 1, respectively, are used to obtain slope
CT that crosses the coordinate (k, X k ), which is determined as follows [24]:

1
CT = (C + 2C2 + 2C3 + C4 ), (10)
6 1
where slopes C1 to C4 are calculated according to the following equations:

dXk
C1 = , (11)
dt
dXk C
C2 = + 1, (12)
dt 2
dXk C
C3 = + 2, (13)
dt 2
dXk
C4 = + C3 . (14)
dt
In this way, the Runge–Kutta discrete model to get X k+1 is obtained as follows:

Xk +1 ∼
= Xk + TCT . (15)
 dX k C2
C3   ,  (13
dt 2
dX k
Mathematics 2022, 10, 2818 C4   C3 .  6 of 19
(14
dt

 
Figure 5. Approximation using Runge–Kutta method.
Figure 5. Approximation using Runge–Kutta method. 
3.4. Trapezoidal Method
In this way, the Runge–Kutta discrete model to get Xk+1 is obtained as follows: 
Mathematics 2022, 10, x FOR PEER REVIEW The trapezoidal method is used to solve an integral equation using an approximation
1  Xk 
X k defined inTC
  that corresponds to the area under the function the interval
T .  k and k + 1. According
  (15
to Figure 6, a trapezium is conformed with the vertices k + 1, k, and uk , uk+1 .

3.4. Trapezoidal Method 
The trapezoidal method is used to solve an integral equation using an approximation
that corresponds to the area under the function defined in the interval k and k + 1. Accord
ing to Figure 6, a trapezium is conformed with the vertices k + 1, k, and uk, uk+1. 

 
Figure 6. Approximation using trapezoidal method. 
Figure 6. Approximation using trapezoidal method.

The approximation of the integral is defined by the trapezium area as follows [24]:
     
Z+1
K
T
Xk +1 − Xk = u(t)dt ∼
= (uK +1 − uK ). (16)
2
K

The general discrete model for X k+1 is

Z+1
K
T
Xk +1 = u(t)dt ∼
= (uK +1 − uK ) + Xk . (17)
2
K

3.5. Compound Trapezoidal Method

In the compound trapezoidal method, an interval [a, b] is divided into n intervals that
are approximated by a line forming n number of trapeziums, as shown in Figure 7. The
total area is obtained by summing all the individual areas as follows [24]:

Zb
!
n −2
T
Xk +1 = u(t)dt ∼
= uk−(n−1) + 2 ∑ uk−i + uk+1 + Xk . (18)
2 i =0
a
 are approximated by a line forming n number of trapeziums, as shown in Figur
total area is obtained by summing all the individual areas as follows [24]: 
b
T n2

Mathematics 2022, 10, 2818
X k 1   u(t )dt  u
 k ( n 1)
2
 2i 0
uk i  uk 1   X k . 
 7 of 19
a

This model of order n considers n − 1 previous data. 

 
Figure 7. Compound trapezoidal principle. 
Figure 7. Compound trapezoidal principle.

This model of order n considers n − 1 previous data.

The  accuracy  of  this  numerical  method  depends  on  the  number  of  interv
The accuracy of this numerical method depends on the number of intervals into which
which [a, b] is divided; more intervals considered results in a closer approximatio
[a, b] is divided; more intervals considered results in a closer approximation.

4.4. Cost Function 
Cost Function
The cost function determines the error between the reference X (k+1)ref and the pre-
The cost function determines the error between the reference X(k+1)ref and the p
dicted value for all the possible operating points, X k+1 . The cost function is expressed in
value for all the possible operating points, X
orthogonal coordinates (α, β) [9] and measures the error k+1. The
between costthefunctionreferences is expressed
and the in or
coordinates
predicted (α, β) [9] and measures the error between the references and the predicted va
value.

gf − X   X
 + X((kk+11))βre X (k(k1)+1)refβ , X ( k 1) 
   
g = X(k+1)αre X((kk+1)1)αref f −X (19)

, 
where X
where X(k+1)α α and X
(k+1)and X(k+1)β  are the real and imaginary parts of the predicted load curr
are
(k+1)β the real and imaginary parts of the predicted load current
vector; for a given voltage vector, this prediction is obtained using load model.
tor; for a given voltage vector, this prediction is obtained using load model. 
The operating mode that generates the minimal error is selected. The future reference
The operating mode that generates the minimal error is selected. The future r
values are unknown; however, according to the linear approximation of Equation (19). it
values are unknown; however, according to the linear approximation of Equatio
can be assumed that X (k+1)ref ≈ Xkref , taking into account that the sample period T is very
can be assumed that X
close to zero when a system is   ≈ Xkref
working
(k+1)ref with , taking into account that the sample period T
high frequencies.
close to zero when a system is working with high frequencies. 
dXkre f
X(k+1)re f ∼
= T+X
dXkre f . (20)
dt kref
X ( k 1) ref  T  X kref
dt predicted from
When this assumption is not valid, the future value can be .  the present and
previous values with a second-order extrapolation.

X(k+1)re f = 3X(k)re f − 3X(k−1)re f + X(k−2)re f . (21)

Additionally, an extra term can be added to the cost function to consider two or more
  parameters to be controlled, such that the cost function may be defined as

g = λ a g a + λ b gb , (22)

where λa and λb are the weighting factors that determine the weight of the cost functions
ga and gb , respectively; the values of the weight factors determine the priority of the control
variable [25]. In this paper, only one variable is controlled; hence, weight factors are
not considered.
 g  a g a  b g b ,  (22) 

where  λa and  λb are the weighting factors that determine the weight of the cost functions 

ga and gb, respectively; the values of the weight factors determine the priority of the control 
Mathematics 2022, 10, 2818 8 of 19
variable [25]. In this paper, only one variable is controlled; hence, weight factors are not 
considered 

5.5. Comparison of Discrete Models to Obtain One‐Step‐Ahead Input Current Vector 
Comparison of Discrete Models to Obtain One-Step-Ahead Input Current Vector
The predictive current control technique used to drive the three-phase active rectifier
The predictive current control technique used to drive the three‐phase active rectifier 
of Figure 1 is shown in Figure 8. The aim of this system is to obtain a DC voltage V
of Figure 1 is shown in Figure 8. The aim of this system is to obtain a DC voltage V DC  and 
DC
and get sinusoidal input currents with a low THD; therefore, the three-phase line current
get sinusoidal input currents with a low THD; therefore, the three‐phase line current vec‐
vector is is considered the controlled variable. The algorithm is based on discrete samples
tor is is considered the controlled variable. The algorithm is based on discrete samples of 
of the single-phase output voltage of the matrix converter and the three-phase line current,
the single‐phase output voltage of the matrix converter and the three‐phase line current, 
which are processed in the alpha–beta plane. The inverse PQ theory [26] is used to obtain
which are processed in the alpha–beta plane. The inverse PQ theory [26] is used to obtain 
sinusoidal current reference; the amplitude depends on the magnitude of P.
sinusoidal current reference; the amplitude depends on the magnitude of P. 

 
Figure 8. Predictive control scheme for the AC–DC converter. 
Figure 8. Predictive control scheme for the AC–DC converter.

According to Table 1, eight combinations are possible, where switching states 1 and
According to Table 1, eight combinations are possible, where switching states 1 and 
generate io o= = 0 A. 
88 generate i 0 A.

Table 1. The switching states in active rectifier.

Table 1. The switching states in active rectifier. 

Switching State 
Switching State S1 
S1 S2 
S2 S3 
S3 S4 
S4 S5S5  S6S6 
SW1 1 
SW 0 
0 0 0 0 
0 11  1 1  1 1 
SW2 2 
SW 1 
1 0 0 0 
0 00  1 1  1 1 
SW3 
SW 1 
1 1 1 0 
0 00  0 0  1 1 
3
SW4  0  1  0  1  0  1 
SW4 0 1 0 1 0 1
SW5  0  1  1  1  0  0 
SW5 0 1 1 1 0 0
SW6  0  0  1  1  1  0 
SW6 0 0 1 1 1 0
SW7  1  0  1  0  1  0 
SW
SW7 8  1
1  1 0 1
1  00  1 0  0 0 
SW8 1 1 1 0 0 0
A dynamic model for the derivative of supply currents can be expressed as follows 
[9]:  A dynamic model for the derivative of supply currents can be expressed as follows [9]:

dis 1
= (vs − vc − R s is ), (23)
dt Ls
 
where Ls is the filter inductor, vs is the supply voltage vector, Rs is the filter resistance, and
vc is the converter voltage vector. To simplify calculations vs , vc , and is are considered in
the bidimensional α–β plane.

2 
vs = vsa + e j2π/3 vsb + e j4π/3 vsc . (24)
3
2 
is = isa + e j2π/3 isb + e j4π/3 isc . (25)
3
Mathematics 2022, 10, 2818 9 of 19

A discrete model of Equation (23) is required to define the line current for the next
sampling period, isk+1 , for all the possible switching modes. The converter voltage vector,
vc , changes as a function of the switching state vector SW = [S1 , S2 , S3 ] according to

2 
vc = S1 + e j2π/3 S2 + e j4π/3 S3 VDC , (26)
3
where S1 , S2 , and S3 are “0” when the switch is opened and “1” when it is closed; S4 , S5 ,
and S6 are complementary, and VDC is the output voltage.
According to the description given in the previous section, Equation (23) can be
approximated with the described discrete models to predict the current vector isk+1 .
The following considerations are used to obtain the approximations for Euler, Runge–Kutta,
and trapezoidal methods:
 
xk = isk = i ak + e j2π/3 ibk + e j4π/3 ick , (27)
 
xk+1 = isk+1 = i ak+1 + e j2π/3 ibk+1 + e j4π/3 ick+1 , (28)

Rs T
A=− , (29)
Ls
1
B= , (30)
Ls
uk = vs k − vc k . (31)

5.1. Prediction of the One-Step-Ahead Current Using Different Numerical Methods

The obtained expressions for isk+1 using the numerical methods detailed in Section 3
are presented below.
1. Forward Euler

Ls − Rs T
 
T
i s k +1 ∼


= vs k − vc k + is k , (32)
Ls Ls
where, vck is the variable to be tested for all the operating modes.
2. Backward Euler
  
∼ T
Ls
i s k +1 = is k + v s k +1 − v c k +1 . (33)
Ls Ls + Rs T
3. Runge–Kutta

disk
C1 = , (34)
dt

disk C
C2 = + 1, (35)
dt 2

disk C
C3 = + 2, (36)
dt 2

disk
C4 = + C3 , (37)
dt

1
CT = (C + 2C2 + 2C3 + C4 ), (38)
6 1

i s k +1 ∼
= isk + TCT . (39)
Mathematics 2022, 10, 2818 10 of 19

4. Trapezoidal first order

For the trapezoidal approximation, the integral of Equation (40) is considered to obtain
the input current vector:
kZ+1
1
i s k +1 = v L dt + isk . (40)
L
k

The voltage in the inductor vL is defined according to the difference of voltage supply
and voltage in the converter.
v L = vs − vc . (41)
Substituting Equation (41) into Equation (40) and solving to approximate the solution
using the first-order trapezoidal method, the predicted input current vector is obtained
as follows:
T
i s k +1 ∼


= v s k +1 − v c k +1 + v s k − v c k + i s k . (42)
2Ls
5. Trapezoidal second order
Using two intervals, the obtained approximation is

T 
i s k +1 ∼




= v s k −1 − v c k −1 + 2 v s k − v c k + v s k +1 − v c k +1 + i s k . (43)
2Ls

6. Trapezoidal third order

Using three intervals, the obtained approximation is

T 
i s k +1 ∼





= vsk−2 − vck−2 + 2 vsk−1 − vck−1 + 2 vsk − vck + vsk+1 − vck+1 + isk . (44)
2Ls

In Equations (42)–(44), vck+1 is the parameter to be evaluated using all the possible
switching modes. The voltage supply vector, vsk , may be used instead of vsk+1 because they
are almost equal. The second-order trapezoidal method uses data at k − 1, whereas the
third-order method additionally uses voltages at k − 2.

5.2. Cost Function Evaluation

The minimal error among the eight predicted currents isk+1 and the reference ikref
determines the best operating mode of the converter. This minimal error is obtained using
   
g = i(k)αre f − i(k+1)α  + i(k) βre f − i(k+1) β . (45)
   

Once the switching configuration that causes the minimal error is determined, the
appropriate control signals are sent to the semiconductor devices.

6. Simulation Results
A simulation of the three-phase active rectifier of Figure 8 was performed in Matlab-
Simulink using the parameters listed in Table 2 [27] under ideal conditions to evaluate
the impact of the use of the different numerical approximations in the predictive con-
troller model. The obtained line current is shown in Figure 9 for Euler, Runge–Kutta,
and trapezoidal methods contrasting the input current is with the current reference iref in
phase a.
 6. Simulation Results 
A simulation of the three‐phase active rectifier of Figure 8 was performed in Matlab‐
Simulink using the parameters listed in Table 2 [27] under ideal conditions to evaluate the 
Mathematics 2022, 10, 2818 11 of 19
impact of the use of the different numerical approximations in the predictive controller 
model. The obtained line current is shown in Figure 9 for Euler, Runge–Kutta, and trape‐
zoidal methods contrasting the input current is with the current reference iref in phase a. 
Table 2. T Simulation parameters.
Table 2. T Simulation parameters. 
Parameter Value
Parameter 
Sampling period Ts Value 
10 µs
Sampling period T
Voltage source
s  10 μs 
127 Vrms
Voltage source  127 Vrms 
Filer resistor Rs 0.1 Ω
Filer resistor Rs  0.1 Ω 
Input filter inductance Ls 10 mH
Input filter inductance Ls  10 mH 
Source frequency 60 Hz
Source frequency  60 Hz 
DC output voltage VDC
DC output voltage V DC 
300 V
300 V 

 
Figure 9. Comparison of is and iref in phase a for Euler, Runge-Kutta and firs-order trapezoidal
Figure 9. Comparison of is and iref in phase a for Euler, Runge‐Kutta and firs‐order trapezoidal meth‐
methods (a) Reference source current waveform iref and measured current waveform, is , in phase
ods (a) Reference source current waveform iref and measured current waveform, i s, in phase a using 
a using Euler, Runge–Kutta, and first-order trapezoidal approximations, for Ts = 10 µs; (b) error
Euler, Runge–Kutta, and first‐order trapezoidal approximations, for Ts = 10 μs; (b) error magnitude. 
magnitude.
For this example, the results from the forward and backward Euler methods are al‐
For this example, the results from the forward and backward Euler methods are
most the same since the resistance 
almost the same since the resistance 𝑅   is very small; therefore, the values obtained by solv‐
Rs is very small; therefore, the values obtained by
ing Equations (32) and (33) are very close, and the difference in the i
solving Equations (32) and (33) are very close, and the difference in the sa waveforms is im‐
isa waveforms is
perceptible. When R s has a greater weight, the Euler backward method tends to produce 
imperceptible. When Rs has a greater weight, the Euler backward method tends to produce
a closer tracking to the reference. 
a closer tracking to the reference.
The error between i
The error between is and i ref in phase a is shown in Figure 9b; to measure it, the mean 
s and iref in phase a is shown in Figure 9b; to measure it, the mean
square error (MSE) is calculated (Table 3) using Equation (46), where N corresponds to 
square error (MSE) is calculated (Table 3) using Equation (46), where N corresponds to
the number of samples [28]. The Runge–Kutta method produces a higher error in contrast 
the number of samples [28]. The Runge–Kutta method produces a higher error in contrast
with the Euler and trapezoidal approximations, with the trapezoidal method being the 
with the Euler and trapezoidal approximations, with the trapezoidal method being the
strategy 
strategy that  produces a a 
that produces closer 
closer tracking 
tracking to reference.
to the the  reference.  Therefore, 
Therefore, the  trapezoidal 
the trapezoidal method
method provides a higher exactitude, resulting in a smaller current ripple. This is achieved 
provides a higher exactitude, resulting in a smaller current ripple. This is achieved due to
the amount of information that is considered to predict the one-step-ahead current, i.e., the
data at k and k − 1, in contrast with Euler and Runge–Kutta methods that only employ the
data at k.
  1 Nh i2
MSE = ∑ ire f ( j) − is ( j) . (46)
N j =1
Mathematics 2022, 10, 2818 12 of 19

Mathematics 2022, 10, x FOR PEER REVIEW  13  of  22 

 

Table 3. Mean square error between is and iref in phase a for the evaluation of numerical methods in
the predictive controller.
due to the amount of information that is considered to predict the one‐step‐ahead current, 
i.e., the data at k and k − 1, in contrast with Euler and Runge–Kutta methods that only 
Trapezoidal Trapezoidal
Euler Runge–Kutta Trapezoidal 1st-Order
employ the data at k.  2nd-Order 3rd-Order
N
MSE for Ts = 10 µs 0.129520 0.232941 0.038633 1 0.051673 0.051729
 i
2
MSE  ref ( j )  is ( j )  .  (46) 
MSE for Ts = 100 µs 3.152851 1.894599 1.366505 N j 1 3.246525 2.574761

Figure 10 shows a comparison among the first‐, second‐, and third‐order trapezoidal 
approximations. According to Table 3, the first‐order approximation provides the smallest 
Figure 10 shows a comparison among the first-, second-, and third-order trapezoidal
MSE between the reference current and the measured current in comparison with the sec‐
approximations. According to Table 3, the first-order approximation provides the smallest
ond‐  and  third‐order 
MSE between techniques, 
the reference currentsuch 
and that  the  current 
the measured ripple 
current inbecomes  smaller. 
comparison with These 
the
second- and third-order techniques, such that the current ripple becomes smaller. These
methods use data at k, k − 1, and k − 2; as the order of the trapezoidal approximation in‐
methods use data at k, k − 1, and k − 2; as the order of the trapezoidal approximation
creases, more previous information is required, increasing the memory use. 
increases, more previous information is required, increasing the memory use.

 
Figure  10. Comparison
Figure 10. Comparison ofof isis and
and iref
iref in
in phase
phase a a for
for  compound 
compound trapezoidal 
trapezoidal methods 
methods (a)(a)  Reference 
Reference
source current waveform, i , and measured current waveform, i , in phase a using first‐, second‐, 
source current waveform, iref , and measured current waveform, is , in phase a using first-, second-,
ref s

and third‐order trapezoidal approximations, for Ts = 10 μs; (b) error magnitude. 
and third-order trapezoidal approximations, for Ts = 10 µs; (b) error magnitude.

Table 3. Mean square error between i s and i
To validate the impact of the use  in phase a for the evaluation of numerical methods in 
ofrefthe numerical approximations in the predictive
the predictive controller. 
controller model under different conditions, the sampling period was modified to 100 µs. In
Figure 11a, the comparison between the reference current and the line current in phase a ob-
Trapezoidal 1st‐ Trapezoidal 2nd‐ Trapezoidal 3rd‐
 
tained from Euler 
Euler, Runge–Kutta, Runge–Kutta 
and first-order trapezoidal methods is shown; Figure 11b
Order  Order  Order 
shows
MSE for Tthe error. It can
s = 10 μs  be seen that0.232941 
0.129520  the error increases with the sampling
0.038633  0.051673  period,0.051729 
but the
use of thes = 100 
MSE for T trapezoidal approximation allows a better performance.
3.152851  1.894599  1.366505  3.246525  2.574761 
Figure 12a, a comparison between the current reference and current in phase a is
Inμs 
presented when first-, second-, and third-order trapezoidal approximations are used in the
predictive model for a Ts = 100 µs. Table 3 reveals that the MSE increases with Ts , but the
To validate the impact of the use of the numerical approximations in the predictive 
first-order trapezoidal approximation allows the minimum MSE.
controller model under different conditions, the sampling period was modified to 100 μs. 
To verify the predictive control capability to operate under dynamic conditions, a
In Figure 11a, the comparison between the reference current and the line current in phase 
power step from 1.5 kW to 2 kW at 50 ms was applied, as shown in Figure 13; the discrete
a obtained from Euler, Runge–Kutta, and first‐order trapezoidal methods is shown; Figure 
model obtained with the first-order trapezoidal method given by Equation (42) was used
11b shows the error. It can be seen that the error increases with the sampling period, but 
in the predictive controller. The current demand increases from 5.8 A to 7.5 A, and the
the use of the trapezoidal approximation allows a better performance. 
controller continues tracking the reference current.
Mathematics 2022, 10, x FOR PEER REVIEW  14  of  22 
  Mathematics 2022, 10, 2818 13 of 19

Mathematics 2022, 10, x FOR PEER REVIEW  15  of  22 

 

 
In Figure 12a, a comparison between the current reference and current in phase a is 
11. Comparison
Figure 11. 
Figure  Comparison ofof  is iand irefiref
s  and  in  in 
phase a fora Euler,
phase  Runge-Kutta
for  Euler,  and firs-order
Runge‐Kutta  trapezoidal
and  firs‐order  meth-
trapezoidal 
presented when first‐, second‐, and third‐order trapezoidal approximations are used in 
ods with sampling period modified (a) Reference source current waveform iref and measured
methods with sampling period modified (a) Reference source current waveform i current
ref and measured 
the predictive model for a Ts = 100 μs. Table 3 reveals that the MSE increases with T s, but 
current waveform, i
waveform, is , in phases, in phase a using Euler, Runge–Kutta, and first‐order trapezoidal approxima‐
a using Euler, Runge–Kutta, and first-order trapezoidal approximations, for
the first‐order trapezoidal approximation allows the minimum MSE. 
Ts = 100 µs;s = 100 μs; (b) error magnitude. 
tions, for T (b) error magnitude.

   

 
Figure 12. Comparison of iss and i
Figure 12. Comparison of i and iref in phase a for compound trapezoidal methods with sampling
ref in phase a for compound trapezoidal methods with sampling 
period modified (a) Reference source current waveform, iref
period modified    (a) Reference source current waveform, i , and measured current waveform,s, 
ref, and measured current waveform, i
is , in phase a using first-, second-. and third-order trapezoidal approximations, for Ts = 100 µs;
in phase a using first‐, second‐. and third‐order trapezoidal approximations, for Ts = 100 μs; (b) error 
magnitude. 
(b) error magnitude.

    The bidirectional capability of the AC–DC converter can be observed in Figure 14,
where a power step from 1.5 kW to −1.5 kW was applied at 50 ms; at that instant, the
current flow changes the direction in the line currents, and the controller continues tracking
the new references.
To validate the performance of the controller under dynamical changes in nominal
values, some parameters were modified in one phase to unbalance the system. Figure 15
shows the line currents in phase a obtained with Euler and first-order trapezoidal approxi-
mations in the controller for a line resistor variation of ±50%. It can be seen that the control
action compensates for the unbalance to continue tracking the reference current, and the
THD is not significantly altered.
 To  verify  the  predictive  control  capability  to  operate  under  dynamic  conditions,  a 
power step from 1.5 kW to 2 kW at 50 ms was applied, as shown in Figure 13; the discrete 
model obtained with the first‐order trapezoidal method given by Equation (42) was used 
Mathematics 2022, 10, 2818 in the predictive controller. The current demand increases from 5.8 A to 7.5 A, and the 
14 of 19
controller continues tracking the reference current. 

Mathematics 2022, 10, x FOR PEER REVIEW  17  of  22 

 

 
Figure 13. Reference source current waveform, i , and measured current waveform, is (red, blue
ref
Figure 13. Reference source current waveform, i
  ref, and measured current waveform, is (red, blue 
and violet), when a power step from 1.5 kW to 2 kW is applied at 50 ms.
and violet), when a power step from 1.5 kW to 2 kW is applied at 50 ms. 

The bidirectional capability of the AC–DC converter can be observed in Figure 14, 
where  a  power  step  from  1.5  kW  to  −1.5  kW  was  applied  at  50  ms;  at  that  instant,  the 
current flow changes the direction in the line currents, and the controller continues track‐
ing the new references. 
   

 
Figure 14. Reference source current waveform, iref , and measured current waveform, is (red, blue
Figure 14. Reference source current waveform, iref, and measured current waveform, is (red, blue 
and violet), when a power step from 1.5 kW to −1.5 kW is applied at 50 ms.
and violet), when a power step from 1.5 kW to −1.5 kW is applied at 50 ms. 
The line currents in phase a obtained with Euler and first-order trapezoidal approxima-
tionsTo validate the performance of the controller under dynamical changes in nominal 
for an input voltage variation of ±10% are presented in Figure 16. The control action
values, some parameters were modified in one phase to unbalance the system. Figure 15 
compensates for the variation with the turning on and turning off times in the switching
shows the line currents in phase a obtained with Euler and first‐order trapezoidal approx‐
devices, allowing tracking of the reference current without a significant impact on the THD
imations in the controller for a line resistor variation of ±50%. It can be seen that the control 
of the input current.
action compensates for the unbalance to continue tracking the reference current, and the 
THD is not significantly altered. 
   
 
Mathematics 2022, 10, x FOR PEER REVIEW  18  of  22 
 

Mathematics 2022, 10, 2818 15 of 19

 

Mathematics 2022, 10, x FOR PEER REVIEW  19  of  22 

 

 
Figure 15. Currentsa ifor variations in the filter resistor using Euler and first‐order trapezoidal ap‐
Figure 15. Current i sa for variations in the filter resistor using Euler and first-order trapezoidal
 
approximations.
proximations. 

The line currents in phase a obtained with Euler and first‐order trapezoidal approxi‐
mations for an  input  voltage  variation of  ±10% are presented  in Figure 16.  The  control 
action  compensates  for  the  variation  with  the  turning  on  and  turning  off  times  in  the 
switching devices, allowing tracking of the reference current without a significant impact 
on the THD of the input current. 
   

 
Figure 16. Current
Figure 16. Current i isa for variations in the input voltage using Euler and first-order
sa for variations in the input voltage using Euler and first‐order trapezoidal ap‐
trapezoidal approximations.
proximations. 
7. Discussion
7. Discussion 
Table 4 shows the number of mathematical operations solved in each discrete model
and Table 4 shows the number of mathematical operations solved in each discrete model 
the cycles needed for them to be executed on a DSPIC; the processor taken as an
and the cycles needed for them to be executed on a DSPIC; the processor taken as an ex‐
example was the 32 bit DSPIC33FJ256MC710, where addition and multiplication take one
ample was the 32 bit DSPIC33FJ256MC710, where addition and multiplication take one 
cycle, whereas division takes 182 cycles. Additionally, the time required to complete the
cycle, whereas division takes 182 cycles. Additionally, the time required to complete the 
evaluation for the eight switching states is shown.
evaluation for the eight switching states is shown. 
The discrete model obtained with the Runge–Kutta method requires a higher number
of cycles to obtain isk+1 compared with the Euler strategy, which is the simplest. The number
Table 4. Number of mathematical operations and cycles required. 
of cycles in the trapezoidal approximation does not increase considerably upon increasing
the order; however, the amount of previous information becomes higher, and this should
Time for the Eight 
be Numerical 
considered in the design of
Addition  Multiplication  controller. ItCycles Re‐
the predictive Division  can be seen that the selection
Switching States Evalua‐
Method  quired 
of a first-order trapezoidal approximation is suitable to obtain accurate results with a low
tion (μs) 
computational
Euler  cost. 3  1  1  328  32.8 
Runge–Kutta  9  5  3  986  98.6 
Trapezoidal 1st‐or‐
4  2  1  330  33 
der 
Trapezoidal 2nd‐
6  3  1  333  33.3 
order 
  Trapezoidal 3rd‐
8  4  1  336  33.6 
order 
 Mathematics 2022, 10, 2818 16 of 19

Table 4. Number of mathematical operations and cycles required.

Time for the Eight Switching

Numerical Method Addition Multiplication Division Cycles Required
States Evaluation (µs)
Euler 3 1 1 328 32.8
Runge–Kutta 9 5 3 986 98.6
Trapezoidal 1st-order 4 2 1 330 33
Trapezoidal 2nd-order 6 3 1 333 33.3
Trapezoidal 3rd-order 8 4 1 336 33.6

A better-approximated solution of the mathematical model is reflected in the THD

Mathematics 2022, 10, x FOR PEER REVIEW  20  of  22 
  obtained in the supply currents. Figure 17 contrasts the obtained THD when different
numerical approximations are used; it can be seen that the Runge–Kutta and first-order
Trapezoidal methods generate a low THD, but the Runge–Kutta method requires a higher
computational
trapezoidal  cost. It can
methods  also be the 
increases  observed
THD that
in the use of second-
contrast  andfirst‐order 
with  the  third-order trapezoidal 
trape-
zoidal methods increases the THD in contrast with the first-order
method, because the supply currents are deformed a little more.  trapezoidal method,
because the supply currents are deformed a little more.

 
Figure 17. THD obtained using the obtained approximations for different output power. 
Figure 17. THD obtained using the obtained approximations for different output power.

Table 5 summarizes the advantages (4)and

Table 5 summarizes the advantages ( disadvantages of all the numerical meth-
) and disadvantages of all the numerical meth‐
ods analyzed to obtain the discrete model used in the predictive control algorithm. Accord-
ods analyzed to obtain the discrete model used in the predictive control algorithm. Ac‐
ing to the results, the optimal numerical method is the first-order trapezoidal approximation;
cording to the results, the optimal numerical method is the first‐order trapezoidal approx‐
it offers exactitude, fast processing time, and a low THD.
imation; it offers exactitude, fast processing time, and a low THD. 

Table 5. Comparison of numerical methods. 
Trapezoidal 1st‐ Trapezoidal 2nd‐ Trapezoidal 3rd‐
  Euler  Runge–Kutta 
Order  Order  Order   
Exactitude           
Simplicity           
Fast processing time           
Low THD         

8. Conclusions 
Mathematics 2022, 10, 2818 17 of 19

Table 5. Comparison of numerical methods.

Trapezoidal Trapezoidal
Euler Runge–Kutta Trapezoidal 1st-Order
2nd-Order 3rd-Order
Exactitude 4 4 4 4
Simplicity 4 4
Fast processing time 4 4 4 4
Low THD 4 4

8. Conclusions
According to the obtained results, the performance of the predictive control tech-
nique can be improved with the use of different numerical methods in the discretization
expression used to predict the future behavior of the control variable.
In this paper, the evaluation of five different numerical methods used to obtain the
discrete function was performed for application in the predictive controller of a three-phase
active rectifier. The simulation results revealed that the use of different numerical methods
impacted the THD of the input currents. For example, the THD could be reduced by
using more complex numerical methods such as Runge–Kutta; however, the mathematical
procedure and computational cost also increased. Other numerical methods offered more
exactitude and a lower THD using previous data, such as the trapezoidal approximation;
however, the processing time increased to evaluate all possible combinations due to the
data to be saved and included in the model.
All the evaluated parameters must be taken into account in the design and model
conditions, depending on the application of the predictive control technique and the
specific system requirements. The processing time of the mathematical procedure in each
numerical method needs to be evaluated for prototype implementation, especially when a
high-frequency operation in the power converter is required.
Table 5 revealed that the first-order trapezoidal approximation is a good tradeoff
between obtained performance and implementation simplicity; it also allows a low THD,
fast processing time, and exactitude for this example, representing a good choice to improve
the predictive controller performance in power converter applications.

Author Contributions: Conceptualization, J.R.-H. and O.U.J.-S.; methodology, L.H.-G., D.C. and
P.G.-L.; software, J.R.-H. and O.U.J.-S.; validation, L.H.-G., D.C., J.C.S.-G. and P.G.-L.; formal analysis,
L.H.-G., J.C.S.-G. and J.R.-H.; investigation, J.R.-H., O.U.J.-S. and L.H.-G.; resources, J.R.-H., O.U.J.-S.,
L.H.-G., D.C., J.C.S.-G. and P.G.-L.; data curation, D.C., L.H.-G. and J.C.S.-G.; writing—original draft
preparation, J.R.-H., O.U.J.-S. and L.H.-G.; writing—review and editing, D.C., P.G.-L. and J.C.S.-G.;
visualization, J.R.-H., O.U.J.-S. and J.C.S.-G.; supervision, L.H.-G. and D.C.; project administration,
L.H.-G. and J.C.S.-G.; funding acquisition, L.H.-G. All authors have read and agreed to the published
version of the manuscript.
Funding: This research was funded by Instituto Politécnico Nacional.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Acknowledgments: The authors are grateful to the Instituto Politécnico Nacional (IPN) for their
encouragement and kind economic support of the research project.
Conflicts of Interest: The authors declare no conflict of interest.
Mathematics 2022, 10, 2818 18 of 19

References
1. Khazraei, M.; Sepahvand, H.; Ferdowsi, M.; Corzine, K.A. Hysteresis-Based Control of a Single-Phase Multilevel Flying Capacitor
Active Rectifier. IEEE J. Emerg. Sel. Top. Power Electron. 2013, 28, 154–164. [CrossRef]
2. Kobaku, T.; Jeyasenthil, R.; Sahoo, S.; Dragicevic, T. Experimental Verification of Robust PID Controller Under Feedforward
Framework for a Nonminimum Phase DC–DC Boost Converter. IEEE J. Emerg. Sel. Top. Power Electron. 2021, 9, 3373–3383.
[CrossRef]
3. Mehreganfar, M.; Hosein-Saeedinia, M.; Davari, S.A.; Garcia, C.; Rodriguez, J. Sensorless Predictive Control of AFE Rectifier With
Robust Adaptive Inductance Estimation. IEEE Trans. Ind. Inf. 2019, 15, 3420–3431. [CrossRef]
4. Satorres-Martínez, S.; De la Casa-Cárdenas, J.; Gámez-García, J.; Gómez-Ortega, J. Position predictive control of an anthropomor-
phic robotic arm using a time-of-flight camera. IEEE Sens. 2014, 2014, 1718–1721. [CrossRef]
5. Carron, A.; Seccamonte, F.; Ruch, C.; Frazzoli, E.; Zeilinger, M.N. Scalable Model Predictive Control for Autonomous Mobility-on-
Demand Systems. IEEE Trans. Control Syst. Technol. 2021, 29, 635–644. [CrossRef]
6. Kumar, C.H.; Mishra, M.K. Predictive Voltage Control of Transformerless Dynamic Voltage Restorer. IEEE Trans. Ind. Electron.
2015, 62, 2693–2697. [CrossRef]
7. Bahrami, A.; Narimani, M.; Norambuena, M.; Rodriguez, J. Current Control of a Seven-Level Voltage Source Inverter. IEEE Trans.
Power Electron. 2020, 35, 2308–2316. [CrossRef]
8. López, M.; Rodriguez, J.; Silva, C.; Rivera, M. Predictive Torque Control of a Multidrive System Fed by a Dual Indirect Matrix
Converter. IEEE Trans. Ind. Electron. 2015, 62, 2731–2741. [CrossRef]
9. Cortés, P.; Wilson, A.; Kouro, S.; Rodriguez, J.; Abu-Rub, H. Model Predictive Control of Multilevel Cascaded H-Bridge Inverters.
IEEE Tran. Ind. Electron. 2010, 57, 2691–2699. [CrossRef]
10. Ramirez-Hernandez, J.; Araujo-Vargas, I.; Rivera, M. An AC-DC converter with trapezoidal predictive controller for wireless
electric vehicle applications. In Proceedings of the IEEE International Conference on Industrial Technology (ICIT), Seville, Spain,
17–19 March 2015; pp. 2236–2240. [CrossRef]
11. Li, X.; Wang, Y.; Guo, X.; Cui, X.; Zhang, S.; Li, Y. An Improved Model-Free Current Predictive Control Method for SPMSM
Drives. IEEE Access 2021, 9, 134672–134681.
12. Cortes, P.; Rodriguez, J.; Quevedo, D.E.; Silva, C. Predictive Current Control Strategy With Imposed Load Current Spectrum.
IEEE Trans. Power Electron. 2008, 23, 612–618. [CrossRef]
13. Huu-Cong, V.; Hong-Hee, L. Model-Predictive Current Control Scheme for Seven-Phase Voltage-Source Inverter With Reduced
Common-Mode Voltage and Current Harmonics. IEEE Trans. Emerg. Sel. Top. Power Electron. 2021, 9, 3610–3621.
14. Shahrouz, E.; Qihong, C.; Liyan, Z. Model Predictive Current Control with Duty Cycle Optimization for Two-Level Three-Phase
Grid-Tied Inverter with Output LCL Filter Based on Forward Euler Approximation. In Proceedings of the 2017 International Con-
ference on Industrial Informatics—Computing Technology, Intelligent Technology, Industrial Information Integration (ICIICII),
Wuhan, China, 2–3 December 2017; pp. 155–158. [CrossRef]
15. Viatkin, A.; Hammami, M.; Ricco, M.; Grandi, G. Analysis of a three-phase four-leg front-end converter for EV chargers with
Balanced and unbalanced grid currents. In Proceedings of the 45th Annual Conference of the IEEE Industrial Electronics Society
(IECON 2019), Lisbon, Portugal, 14–17 October 2019; pp. 3442–3449. [CrossRef]
16. Dini, P.; Saponara, S. Electro-Thermal Model-Based Design of Bidirectional On-Board Chargers in Hybrid and Full Electric
Vehicles. Electronics 2022, 11, 112. [CrossRef]
17. Ramos, L.A.; Van Kan, R.F.; Mezaroba, M.; Batschauer, A.L. A Control Strategy to Smooth Power Ripple of a Single-Stage
Bidirectional and Isolated AC-DC Converter for Electric Vehicles Chargers. Electronics 2022, 11, 650. [CrossRef]
18. Hongfei, W.; Meng, H.; Kai, S. Dual-Voltage-Rectifier-Based Single-Phase AC–DC Converters With Dual DC Bus and Voltage-
Sigma Architecture for Variable DC Output Applications. IEEE Trans. Power Electron. 2019, 34, 4208–4222. [CrossRef]
19. Jiangfeng, W.; Hongfei, W.; Tianyu, Y.; Li, Z.; Yan, X. Bidirectional Three-Phase DC–AC Converter With Embedded DC–DC
Converter and Carrier-Based PWM Strategy for Wide Voltage Range Applications. IEEE Trans. Ind. Electron. 2019, 66, 4144–4155.
[CrossRef]
20. Eun-Su, J.; Sangshin, K.; Taehyung, K. Performance Comparison of Model Predictive Control Methods for Active Front End
Rectifiers. IEEE Access 2018, 6, 77272–77288. [CrossRef]
21. Quevedo, D.E.; Aguilera, R.P.; Perez, M.A.; Cortes, P.; Lizana, R. Model Predictive Control of an AFE Rectifier With Dynamic
References. IEEE Trans. Power Electron. 2012, 27, 3128–3136. [CrossRef]
22. Khalilzadeh, M.; Vaez-Zadeh, S.; Rodriguez, J.; Heydari, R. Model-Free Predictive Control of Motor Drives and Power Converters:
A Review. IEEE Access 2021, 9, 105733–105747. [CrossRef]
23. Huang, Z.; Li, H.; Li, W.; Liu, J.; Huang, C.; Yang, Z.; Fang, W. A New Trajectory Tracking Algorithm for Autonomous Vehicles
Based on Model Predictive Control. Sensors 2021, 21, 7165. [CrossRef]
24. Gutierrez Robles, J.A.; Olmos Gomez, M.A.; Casillas Gonzalez, J.M. Análisis Numérico, 1st ed.; Roig Vazque, P.E.,
Delgado, R.A.L., Eds.; McGrawHill: Mexico City, México, 2010; Volume 1, pp. 207–217, 247–260.
25. Thielemans, S.; Vyncke, T.J.; Melkebeek, J. Weight factor selection for model-based predictive control of a four-level flying-
capacitor inverter. IET Power Electron. 2012, 5, 323–333. [CrossRef]
26. Czarnecki, L.S. Instantaneous reactive power p-q theory and power properties of three-phase systems. IEEE Trans. Power Del.
2006, 21, 362–367. [CrossRef]
Mathematics 2022, 10, 2818 19 of 19

27. Rodriguez, J.; Cortes, P. Control of an Active Front-End Rectifier. In Predictive Control of Power Converters and Electrical Drives,
1st ed.; Wiley: Chichester, UK, 2012; Volume 1, pp. 81–98.
28. Sai, Y.; Jinxia, R.; Zhongxia, L. Learning of Neural Networks Based on Weighted Mean Squares Error Function. In Proceedings of
the Second IEEE International Symposium on Computational Intelligence and Design, Changsha, China, 12–14 December 2009;
pp. 241–244. [CrossRef]