Model Predictive Control of Power

Electronics Converter
Jiaying Wang
Master of Science in Electric Power Engineering
Supervisor: Lars Einar Norum, ELKRAFT
Department of Electric Power Engineering
Submission date: June 2012
Norwegian University of Science and Technology
Acknowledgment

First of all, I would like to thank my parents who raise me and support me to do further study
in Norway. I appreciate the help from my supervisor Lars Einar Norum and Phd student
Hamed Nademi. I benefit a lot from their meticulous attitude to study, abundant knowledge
and minded guidance. I am very grateful that NTNU Department of Electric Power
Engineering provides very good hardware resources. I would also like to thank my fellow
classmates through the pleasant master-education life. Again I am cheerful that I can graduate
from NTNU, a very nice university.
Recalling the past two years of life in Norway, I feel very happy that I found good teachers
and good friends, such as Hans, Shahbaz, Martin, Chen and so on, in the form of helping
people that I will always remember.



Trondheim, 12
th
J une 2012
J iaying Wang

2


3

Abstract

Voltage-source PWM (Pulse Width Modulation) rectifier can provide constant DC bus voltage
and suppress harmonic distortion of grid-side currents. It also has power feedback capability
and has a broad prospect in the field of DC power supply
[1]
, reactive power compensation,
active filtering and motor control system.
This dissertation studies the theory and implementation of PWM rectifier and completes the
following tasks:
1. Analyze three-phase voltage-source PWM rectifier (VSR), including its topology,
mathematical model and principle. Derive Clarke transformation and Park
transformation and analyze the mathematical model in the two-phase stationary
coordinate and dq rotating coordinate.
2. Make a detailed analysis on the principle and characteristics of Direct Power Control
(DPC) strategy and Model Predictive Control (MPC) strategy and study the
instantaneous active power and reactive power flow in the rectifier.
3. Based on the principle of traditional switching table of DPC, an improved table is
proposed. Then this project presents a further improved switching table to achieve
better control performance and the simulation model in Matlab/Simulink environment
is established to verify the algorithm of voltage-oriented direct power control strategy.
4. Based on different strategy studies and the simulation results from DPC system,
propose our model predictive control (MPC) algorithm.
5. Analyze the modulation principle of the space vector pulse width modulation
(SVPWM).
6. Build the MPC-SVPWM model in Matlab/Simulink to verify our MPC algorithm.
7. The simulation result shows that MPC-SVPWM performs better in harmonic
suppression, unity power factor, DC output voltage ripple coefficient and dynamic
response than DPC.

Key words: PWM rectifier, unity power factor, direct power control, model predictive control,
harmonic suppression, SVPWM
4


5

Contents
Abstract ...................................................................................................................................... 3
Contents ...................................................................................................................................... 5
Chapter 1 Introduction ............................................................................................................... 1
1.1 Background and significance of the study ....................................................................... 1
1.2 Current state of PWM converter research ........................................................................ 2
1.3 The application fields of PWM converters ....................................................................... 3
1.3.1 Active power filter and static var generator .............................................................. 3
1.3.2 Unified power flow controller ................................................................................... 4
1.3.3 Superconducting magnetic energy storage ................................................................ 4
1.3.4 Four-quadrant electrical drive ................................................................................... 4
1.3.5 Grid-connected renewable energy ............................................................................. 5
1.4 The main work of this thesis ............................................................................................ 5
Chapter 2 Three phase VSR mathematical model ...................................................................... 7
2.1 Derivation of coordinate transformation .......................................................................... 7
2.2 Principle of PWM VSR .................................................................................................... 9
2.2.1 Mathematical model in stationary coordinate system ........................................ 12
2.2.2 Mathematical model in dq rotating coordinate system ........................................... 13
Chapter 3 Principle of Direct power control ............................................................................ 15
3.1 Power theory and calculation of instantaneous power ................................................... 15
3.2 Voltage-oriented direct power control of PWM VSR .................................................... 16
3.2.1 System composition of VSR ................................................................................... 16
3.2.2 Principle of DPC ..................................................................................................... 17
3.3 Power flow in the converter ........................................................................................... 25
Chapter 4 Principle of Model predictive control ...................................................................... 28
4.1 Review of MPC in power converters ............................................................................. 29
4.2 Process, model and controller of VSR ........................................................................... 30
4.3 Two-level SVPWM modulation technique .................................................................... 32
4.3.1 Voltage space vector distribution of three-phase VSR ............................................ 32
4.3.2 Synthesis of voltage space vector ............................................................................ 34
Chapter 5 Simulation ................................................................................................................ 37
5.1 Simulation of direct power control system .................................................................... 37
5.1.1 Comparison of different switching tables ............................................................... 38
5.1.2 Dynamic response of DPC ...................................................................................... 42
5.1.3 Summary ................................................................................................................. 45
5.2 Simulation of model predictive control .......................................................................... 46
5.2.1 Startup and steady state ........................................................................................... 48
6

5.2.2 Dynamic performance of MPC-SVPWM ............................................................... 50
5.2.3 Summary ................................................................................................................. 63
Chapter 6 Conclusion and future work .................................................................................... 64
6.1 Conclusions from the Simulink results .......................................................................... 64
6.2 Suggested future work .................................................................................................... 64
Appendix .................................................................................................................................. 65
Deviation of power supply voltages in dq frame e
d
and e
q
.................................................. 65
Switching function waveforms in six sectors ....................................................................... 65
Codes in Embedded MATLAB Function ............................................................................. 67
References ................................................................................................................................ 72
1

Chapter 1 Introduction

1.1 Background and significance of the study

In modern industry, power conversion is needed in many occasions, converting AC power into
DC power and vice versa. The rectifier converts the AC into DC and the inverter converts DC
into AC. Since many electrical appliances and converters in the daily life require DC power
supply, the rectifier is indispensable. The traditional rectifiers are using the power diode or
thyristor to convert AC into DC, which are called uncontrolled rectifier or phase-controlled
rectifier. The diode rectifier draws distorted currents from the grid, resulting in harmonic
pollution, and the DC side energy can not be fed back to the grid. AC-side power factor is low
for phase-controlled rectifier, and DC output voltage has severe shorks. Structures of these
traditional rectifiers are shown as follows:
a
e
b
e
c
e
R
L
R
, , a b c
i
L
O
N
-
dc
v
dc
i
L
i
a
b
c
1
V
3
V
5
V
2
V
4
V
6
V

Figure 1-1 Diode voltage source rectifier
a
e
b
e
c
e
R
L
R
, , a b c
i
L
O
N
-
dc
v
dc
i
L
i
a
b
c
1
V
3
V
5
V
2
V
4
V
6
V

Figure 1-2 Thyristor voltage source rectifier
The traditional rectifiers cause serious distortion of currents in the grid and harmonic
pollution. Harmonic current generates heat loss in overhead lines and cables in power system;
2

harmonic currents flowing in the overhead line will cause serious electromagnetic interference
to adjacent communication equipment and interfere with the protection devices and cause
malfunction; harmonic currents cause dielectric loss in power capacitors and speed up aging;
much reactive power exchanges with the grid, producing large amount of additional energy
losses. These factors restrict its application in industry.
The most fundamental way to solve this harmonic pollution is to make the converter produce
no harmonics and realize the sinusoidal current in the grid side and unity power factor. With
the development of power electronics, the advanced full-controlled power semiconductor
devices and microprocessor technology and control theories promote the development of the
converter. A variety of converters emerge based on pulse width modulation control. Voltage-
source PWM rectifier (VSR) has the following advantages: low harmonics of the currents in
the grid side, unity power factor, constant DC voltage and bi-directional energy flow.
With certain topology of main circuit, to achieve the above advantages of VSR, various
control strategies have been proposed. Turn off and on the fully-controlled power devices in
accordance with a certain control strategy, we can control the magnitude and phase angle of
AC currents, supplying appropriate power to the load and AC currents close to sinusoidal
waveform, in phase with supply voltages. Then the power factor will be close to unity,
achieving the purpose to improve power factor and to suppress harmonics. The significance of
improvement of power factor on the actual production is enormous. For example, in a 10,000
ton-class chemical plant, DC current is needed to electrolyze the saline solution to produce the
most basic raw materials for the chemical industry: chlorine gas and sodium hydroxide. In this
energy-intensive industry, if the power factor can be increased by a few percentage points, a
very considerable part of electric energy can be saved. PWM converter can work in four
quadrants due to its fully-controlled switching devices (referring to chapter 2.2), so that
energy can be fed back into the supply grid when the motor works in regenerate mode.
1.2 Current state of PWM converter research

PWM rectifier research began in the 1980s, when the self-turn-off devices became mature,
promoting the application of the PWM technique. In 1982 Busse Alfred and Holtz J oachim
proposed the three-phase full-bridge PWM rectifier topology based on self-turn-off devices
and the grid-side current amplitude and phase control strategy
[2]
, and implemented unity
power factor and sinusoidal current control of the current-source PWM rectifier. In 1984,
Akagi Hirofumi with others proposed reactive power compensation control strategy based on
the PWM rectifier topology
[3]
, which was actually the early idea of voltage-source PWM
rectifier. At the end of the 1980s, Green A.W proposed continuous and discrete dynamic
mathematical models and control strategy of PWM rectifier based on coordinate
transformation, raising PWM rectifier research and development to a new level
[4]
.

PWM rectifier according to the output can be divided into the voltage-source and current-
source rectifier. For a long time, the voltage-source PWM rectifier (VSR) for its low losses,
simple structure and control strategy has become the focus of the study of PWM rectifier,
while the current-source PWM rectifier (CSR) is relatively complex due to the presence of
DC energy-storage inductor and AC LC filter inductor. Voltage-source inverter has a similar
topology with VSR but operating in opposite direction. The main power converter that is
3

investigated in this work is VSR.

There are several control strategies: current control (including indirect current control and
direct current control) and nonlinear control strategy (including instantaneous power control,
feedback linearization control, Lyapunov control and so on). Indirect current control strategy
controls grid currents indirectly by controlling the amplitude and phase angle of the
fundamental component of input voltage of rectifier. However, this strategy has some
disadvantages: bad stability, slow dynamic response and current overshoot in dynamic
process, restricting its application
[5]
. Direct current control strategy controls AC current
directly by following the given reference current. The typical example is the dual-loop PI
control. This strategy with fast dynamic response uses space vector modulation, increases the
utilization of DC voltage and has been applied in practical projects. PWM converter has the
following characteristics: nonlinear, multivariable and strong coupling. Its traditional control
algorithms adopt linearization based on small signal disturbance on steady operating point,
which may not maintain the stability of large range disturbances. So some proposed control
strategy based on Lyapunov stability theory. This novel control scheme builds the Lyapunov
function based on the quantitative relationship of inductor and capacitor energy storage. The
Lyapunov function combines the PWM mathematical model in dq rotating frame and
corresponding SVPWM constraints to deduce the control algorithm. This control strategy
solves the stability issue of large range disturbances
[6]
. To increase the performance of PWM
VSR, research on nonlinear control method and new control algorithm is a new challenge for
the reseachers now.

1.3 The application fields of PWM converters

The AC side of PWM rectifier has a characteristic of controlled current source, which makes
the development of the control strategies and topologies of PWM rectifier. Power converter
are used in many fields, such as static var generator (SVG), active power filter (APF), unified
power flow controller (UPFC), superconducting magnetic energy storage (SMES), high
voltage direct current transmission (HVDC), electrical drive and grid-connected renewable
energy
[7]
.
1.3.1 Active power filter and static var generator

The following diagram shows the topology of APF.
AC
Load
R
L
APF

Figure 1-3 Shunt active power filter topology
In this circuit, the LC filter and APF operate for grid harmonic suppression and reactive
power compensation together. The APF is regarded as a controlled current source and it
injects or draws current in such a way that the sum of harmonic part of load current and
current drawn by APF becomes zero. As a result the grid side current will be purely sinusoidal
and in phase with grid side voltage. The elimination of load harmonics will result into the
4

improvement of reactive power control as well
[8]
.
1.3.2 Unified power flow controller

The UPFC is the most promising power compensation device in flexible alternating current
transmission system. UPFC is used in the transmission grid, controlling active power flow
and absorbing or supplying reactive power. UPFC consists of combination of series active
power filter and shunt active power filter. The series APF is equivalent to a controlled voltage
source, compensating high frequency components of grid voltage, zero sequence component
and negative sequence component of fundamental component of grid voltage; while shunt
APF is equivalent to an APF, a controlled current source, absorbing or supplying reactive
power. Its topology is shown as follows:
AC
Grid

Figure 1-4 UPFC topology
1.3.3 Superconducting magnetic energy storage

SMES is mainly used for peak load regulation control, and other occasions where short-time
compensation of electrical energy is needed. When current consumption of electricity is
normal, the electricity in the grid is converted into energy in superconducting coils through
converter to store enough energy. During large power consumption, the energy in the
superconducting coil is fed to the grid through the converter, in order to achieve the purpose
of peak load regulation. Its topology is shown as follows:

Grid
Superconducting
coil

Figure 1-5 SMES topology
The main circuit of SMES is usually composed by current-source PWM rectifier. The lossless
superconducting coil is connected in series with the DC side of PWM rectifier. The coil itself
is both a DC buffer inductor and load of the DC side. This design simplifies the structure of
main circuit of current-source rectifier, and overcomes the shortcomingthe large loss of
conventional current-source rectifier.
1.3.4 Four-quadrant electrical drive

PWM rectifier replaces the diode rectifier, eliminates the energy dissipation device at the DC
side of inverter, achieves steady DC voltage at the DC side of rectifier and enhances the
actuating performance of motor. On the other hand, appropriate control strategy can reduce
5

the capacitance of the DC side capacitor. The four-quadrant electrical drive topology is shown
as follows:

AC
rectifier
M
inverter

Figure 1-6 Four-quadrant electrical drive topology
1.3.5 Grid-connected renewable energy

Grid-connected photovoltaic power system is composed of solar arrays and PWM converters.
While the PWM converter is used to boost the PV array voltage and ensure the maximum
power point tracking (MPPT), and inverts dc power into high quality ac power to the grid
[9]
.
The topology is shown in figure 1-7.

Grid
Photovoltaic
(PV) array
Boosting
+MPPT+
DC/AC
converter

Figure 1-7 Grid connected PV system topology
Nowadays the wind turbine generator systems mainly include three kinds of generator:
squirrel cage induction generator (SCIG), doubly-fed induction generator (DFIG) and
permanent magnet synchronous generator (PMSG)
[10]
. For different generators, there are
different grid-connected methods. The direct-drive PMSG topology is shown as follows:

rectifier
M
inverter
Grid

Figure 1-8 Grid-connected PMSG topology
The blade is connected to the PMSG directly, and the system converts the wind energy into
time-variant frequency and amplitude AC, which can be converted into DC by rectifier, and
then the inverter converts the DC into three-phase AC with constant frequency. The active and
reactive power can be controlled by the appropriate strategy to the PWM converters.
1.4 The main work of this thesis

In recent years, novel control methods emerge one after another, such as fuzzy control,
artificial neural element control and predictive control. This dissertation, in point of interest,
chooses direct power control (DPC) and model predictive control (MPC), two algorithms to
control rectifier. A theoretical analysis is made on the DPC and MPC of PWM rectifier
system, their working principles and control methods are discussed. The theoretical analysis is
6

checked through simulation and at the end some conclusions are drawn. The main content is
as follows:
1. The first chapter describes that developing PWM rectifier becomes an important way
to solve the harmonic pollution and analyzes the status of PWM rectifier technology,
introduces some new PWM rectifier control strategies, and finally introduces the
applications of PWM rectifier in different fields.
2. In the first half of second chapter, the derivation of coordinate transformation is done
and in the second of the working principle of the PWM rectifier and its mathematical
model in different coordinate systems are discussed.
3. The third chapter analyzes the power control theory, the principle of direct power
control strategy and fundamentals of the traditional switching table and the
components of direct power control system, and improves the traditional switching
table.
4. The fourth chapter focuses on the principle of model predictive control, and gives the
mathematical prediction model of VSR.
5. The fifth chapter simulates the direct power control and model predictive control
strategy in the Matlab/Simulink environment, respectively, and simulation results are
compared and analyzed in detail. It draws the conclusion that model predictive control
of PWM rectifier performs better and MPC of VSR has the following advantages:
unity power factor, small AC current harmonics, and small DC voltage ripple
coefficient.
6. The content of the research is summarized and forecasted in the last chapter, mainly
summarizing the fulfillment of this project, conclusion and next-stage working plan.
7

Chapter 2 Three phase VSR mathematical model

This chapter mainly analyzes mathematical model of voltage-source rectifier, and then
describes its operating principle in favor of establishing simulation model in Simulink. In the
beginning of this chapter, the coordinate transformations namely Clarke transformation and
Park transformation which are used in building the mathematical model are formulated
briefly
[11]
.

2.1 Derivation of coordinate transformation

Derivation of the coordinate theory includes: abc stationary coordinate to stationary
coordinate system, abc stationary coordinate system to dq rotating coordinate system. There
are two standards for transformation: power invariance and amplitude invariance. In this
paper, power invariance is used in the modeling of direct power control. The following figure
shows the relationship of different coordinate systems.
a

c
b
axis o
axis |
d axis
q axis
o
x

Figure 2-1 Relationship of coordinates
In the beginning the amplitude invariant transformation is introduced. The amplitude invariant
coordinate transformation is that one common vector of one coordinate system is equal to
another common vector of other coordinate system. Power invariant transformation refers to
coordinate transformation before and after, the power does not change. The general vector x
will be taken for example to discuss two standards of coordinate transformation.
1 abc- transformation
We use a set of orthogonal axes affixed where -axis is aligned with the a-axis. The angle
between x and is . The projection of vector x along the abc-axis is obtained:
8


cos
cos( 2 / 3)
cos( 2 / 3)
a m
b m
c m
x x
x x
x x
o
o
o
=

= +

(2.1)
Where x
m
is the magnitude of vector x.
The projection of vector x along the -axis is as follows:

2 2
cos
sin
m
m
m
x x
x x
x x x
o
|
o |
o
o
=

= +

(2.2)
We know the following trigonometric relations:


2 1 1
cos cos cos( 2 / 3) cos( 2 / 3)
3 2 2
3 3
2
sin cos( 2 / 3) cos( 2 / 3)
3 2 2
o o o o
o o o
(
= +

(
= +
(

(2.3)
Combining the equations (2.1) (2.2) (2.3), we got

2 1 1
( )
3 2 2
3 3
2
( )
3 2 2
a b c
b c
x x x x
x x x
o
|

(2.4)
Set zero-axis component
1
( )
3
o a b c
x x x x = + + we got,

1 1
1
2 2
2 3 3
0
3 2 2
1 1 1
2 2 2
a
b
c
o
x x
x x
x
x
o
|
(

(
( ( (
( ( (
=
( ( (
( ( (


(
(

(2.5)

2 abc-dq transformation
The outstanding advantage of dq transformation is that the fundamental sinusoidal variables
are transformed into DC variables. In three-phase stationary abc coordinate system, e and i
represent grid emf vector and current vector respectively, and rotate with speed of
(fundamental angular frequency) in anticlockwise direction. When describing the three-phase
electrical parameters, to simplify the analysis, the d-axis of the two-phase rotating reference
frame aligns with the grid emf vector e.
As shown in figure 2-1, the angle between vector x and -axis is and the angle between d-
axis and -axis is , the projection of vector x along the dq-axis will be:
9


2 2
cos( )
sin( )
d m
q m
m d q
x x
x x
x x x
o
o
=

= +

(2.6)
We know the following trigonometric relations:

| |
| |
2
cos( ) cos cos cos( 2 / 3)cos( 2 / 3) cos( 2 / 3)cos( 2 / 3)
3
2
sin( ) sin cos sin( 2 / 3)cos( 2 / 3) sin( 2 / 3)cos( 2 / 3)
3
o o o o
o o o o

= + + + +

= + + + +

(2.7)
Also set zero-axis component
1
( )
3
o a b c
x x x x = + + we got,

cos cos( 2 / 3) cos( 2 / 3)
2
sin sin( 2 / 3) sin( 2 / 3)
3
1 1 1
2 2 2
d a
q b
c
o
x x
x x
x
x


(
( ( + (
( (
(
= +
( (
(
( (
(


(

(2.8)
Transformation matrix in the equations (2.5) and (2.8) is not orthogonal matrix, which makes
matrix operations difficult. So power invariant transformation is raised as follows:

1 1
1
2 2
2 3 3
0
3 2 2
1 1 1
2 2 2
a
b
c
o
x x
x x
x
x
o
|
(

(
( ( (
( ( (
=
( ( (
( ( (


(
(

(2.9)

cos cos( 2 / 3) cos( 2 / 3)
2
sin sin( 2 / 3) sin( 2 / 3)
3
1 1 1
2 2 2
d a
q b
c
o
x x
x x
x
x


(
(
( + (
(
(
(
= +
(
(
(
(
(
(


(
(

(2.10)

2.2 Principle of PWM VSR

The main circuit topology of three-phase voltage-source PWM rectifier is shown in figure 2-
2
[12]
, which is the most common topology. The project studies various control strategies
based on this three-phase half-bridge topology.
10

a
e
b
e
c
e
R
L
R
, , a b c
i
L
O
N
-
dc
v
dc
i
L
i
a
b
c
a
s
1
V
3
V
5
V
2
V
4
V
6
V
c
s
b
s

Figure 2-2 Topology of three-phase PWM rectifier
, ,
a b c
e e e e = are the three-phase power supply phase voltages, , ,
a b c
i i i i = are the three-phase
currents. V
1
-V
6
are the power electronic devices, for example, IGBT and diode. L is
inductance of the filter and R is the sum of resistance of filter and equivalent resistance of the
switch power losses. R
L
is the load resistance. The inductor of the filter suppresses harmonics
in AC currents and boosts the output voltage of VSR. , ,
r ao bo co
v v v v = are the input voltages of
VSR. Current i is controlled by the difference of power supply e and input voltage v
r
if we
neglect the influence of voltage across R. Since e is given, i is determined by v
r.
If the
amplitude and phase of v
r
can be controlled, then the amplitude and phase of i can be
controlled accordingly. It can operate in four quadrants, and its vector diagram is shown as
follows:

e
i
r
v
R
v
L
v
e
e
e
i
i
i
rectifier
L
v
L
v
L
v
R
v
R
v
R
v
r
v
r
v
r
v
inverter
capacitancecharacteristics inductancecharacteristics
) a
) b
) c ) d

Figure 2-3 Four-quadrant vector diagram
In figure 2-3 a), i is in phase with e, and its in unity power factor rectifier mode. In b), the
circuit operates as an inverter, achieving the feedback of energy. In c), the converter does not
absorb active power from the grid, and only absorbs the capacitive reactive power from the
grid. In d), the converter does not absorb active power from the grid, and only absorbs
inductive reactive power from the grid. By controlling the vector v
r
, the PWM VSR can
operate in any position of the four quadrants.

11

, ,
a b c
s s s are switching function of the power converter. Unipolar binary logic switch function
is defined as follows:

1
(k ,b,c)
0
k
s a

= =

(2.11)
1
k
s = means that upper switch is on and 0
k
s = means lower switch is on.
Our mathematical model is derived based on this switching function, and since it does not
neglect the role of high-frequency component, this model reflects the operation mechanism. It
can be used to get the high-accuracy dynamic simulation. It is also helpful for the
implementation of the physical hardware circuit.

In the equivalent mathematical model, conduction voltage drop and switching losses of the
switching devices are excluded, and AC side inductor saturation is excluded as well. Three-
phase balanced power supply voltages are given:


2 cos
2 cos( 2 / 3)
2 cos( 2 / 3)
a m
b m
c m
e U t
e U t
e U t

= +

(2.12)
Where U
m
is the Root-Mean-Square value of grid phase to phase voltage.
Apply Kirchhoff's voltage law on the AC side of rectifier, and we got the circuit equation of
phase A,

0
( )
a
a a aN N
di
L Ri e v v
dt
+ = + (2.13)
When V
1
is on and V
2
is off, S
a
=1,
aN dc
v v = ; when V
1
is off and V
2
is on, S
a
=0, 0
aN
v = . It
means
aN dc a
v v s = .
We can also get equations of phase B and C in the same way,

0
0
( )
( )
b
b b dc b N
c
c c dc c N
di
L Ri e v s v
dt
di
L Ri e v s v
dt

+ = +

+ = +

(2.14)
In three-phase neutral system, the sum of three-phase currents is zero.
0
a b c
i i i + + = (2.15)
For three-phase balanced grid voltage, we got
0
a b c
e e e + + = (2.16)
Adding the left side and right side of equations (2.13) (2.14) respectively, we got

0
( )
3
dc
N a b c
v
v s s s = + + (2.17)
In addition, apply Kirchhoff's current law at DC capacitor positive node, we got
12


dc
dc L
dv
C i i
dt
= (2.18)
Where
dc a a b b c c
i i s i s i s = + + and for resistive load,
L dc L
i v R = .
By solving equations (2.13), (2.14) and (2.18), we got the general mathematical model of
three-phase VSR under the three-phase stationary abc coordinate system:

( )
3
( )
3
( )
3
a a b c
a a a dc
b a b c
b b b dc
c a b c
c c c dc
dc dc
a a b b c c
L
di s s s
L e Ri s v
dt
di s s s
L e Ri s v
dt
di s s s
L e Ri s v
dt
dv v
C i s i s i s
dt R
+ +
=

+ +

+ +

= + +

(2.19)
The physical meaning of this mathematical model is clear, yet the variables in AC side are
time-varying quantities, which is inconvenient to design the control system. Therefore,
coordinate transformation is used.
2.2.1 Mathematical model in stationary coordinate system

Since we have two standards of coordinate transformation, we will list both.
Amplitude invariant transformation:

( )
3
2
dc
dc
dc
L
di
L e Ri v s
dt
di
L e Ri v s
dt
dv
C i s i s i
dt
o
o o o
|
| | |
o o | |

= +

(2.20)
Power invariant transformation:

( )
dc
dc
dc
L
di
L e Ri v s
dt
di
L e Ri v s
dt
dv
C i s i s i
dt
o
o o o
|
| | |
o o | |

= +

(2.21)
The mathematical model structure of VSR under stationary coordinate is shown as
follows:
13

1
R sL +
e
o

s
o

i
o

1
R sL +
e
|

+

s
|

i
|

+
+

+

L
i
1
sC
dc
v

Figure 2-4 Structure of VSR in stationary frame
It can be seen from above figure that if v
dc
is constant, there is no coupling betweeni and i
o |

in stationary frame. However, the voltages , e e
o |
and the currents , i i
o |
are still sinusoidal
variations, which are complex to control. To solve this problem, the model in dq rotating
coordinate frame is built.
2.2.2 Mathematical model in dq rotating coordinate system

Amplitude invariant transformation:

3
( )
2
d
d q d dc d
q
q d q dc q
dc
d d q q L
di
L e Li Ri v s
dt
di
L e Li Ri v s
dt
dv
C i s i s i
dt

= +

= +

(2.22)
Power invariant transformation:

( )
d
d q d dc d
q
q d q dc q
dc
d d q q L
di
L e Li Ri v s
dt
di
L e Li Ri v s
dt
dv
C i s i s i
dt

= +

= +

(2.23)
The mathematical model structure of VSR under dq rotating coordinate is shown as follows:

14

1
R sL +
d
e

d
s

d
i

1
R sL +
q
e

q
s

q
i

+
+

+

L
i
1
sC
dc
v
L
L
+

Figure 2-5 Structure of VSR in rotating frame
There is coupling between
d q
i and i in dq rotating frame. Coupling terms make complexity in
the design of control system, so the decoupling control is needed.
15

Chapter 3 Principle of Direct power control

From the energy point of view, when AC voltage is given, if the instantaneous power of PWM
rectifier can be controlled within the allowable range, the instantaneous current within the
allowable range can be controlled indirectly, and such control strategy is the direct power
control (DPC).

In the early 1990s, Tokuo Ohnishi proposed a new control strategy using instantaneous active
and reactive power in a closed loop control system of PWM converter, and then Toshihiko
Noguchi and other scholars studied and made progress.

Structure of DPC rectifier system contains the DC voltage outer loop and power control inner
loop, and it selects switches in the switching table according to the AC-side instantaneous
power, to achieve low total harmonic distortion (THD), high power factor, simple algorithm
and fast dynamic response.

3.1 Power theory and calculation of instantaneous power

To study the direct power control strategy of VSR, instantaneous power theory is used to
calculate the instantaneous values of active and reactive power. Instantaneous values of three-
phase voltages and three-phase currents are
a b c a b c
u u u and i i i respectively. After Clarke
transformation, we can get voltages u u
o |
and currents i i
o |
under two-phase coordinate
system.

(1) Three-phase abc stationary coordinate

In the three-phase circuit, instantaneous phase to phase voltages and the instantaneous phase
currents can compose the instantaneous voltage vector u and current vectori in the Cartesian
coordinate abc system.
[ ]
T
a b c
u u u u = (3.1)
[ ]
T
a b c
i i i i = (3.2)
Instantaneous active power is the scalar product while instantaneous reactive power is the
vector product. So we got,

a a b b c c
p u i u i u i u i = = + + (3.3)

1
[( ) ( ) ( ) ]
3
b c a c a b a b c
q u i u u i u u i u u i = = + + (3.4)
(2) Two-phase stationary coordinate

For power invariant transformation, we got

3 2 3 2
, ( ), , ( )
2 2 2 2
a b c a b c
i i i i i u u u u u
o | o |
= = = = (3.5)
From equations (3.3) (3.4) and (3.5), the formula of power can be expressed as follows:
16



p u i u i u i
q u i u i u i
o o | |
| o o |
= = +

= =

(3.6)
So similarly, we could find formula of power in dq coordinate system.
(3) Two-phase dq rotating coordinate


d d q q
q d d q
p u i u i
q u i u i
= +

(3.7)
For a three-phase balanced power system, we have


3 , 0
d m q
d d
d q
u U u
p u i
q u i
= =
=

(3.8)
Where U
m
is the Root-Mean-Square value of grid phase to phase voltage.
Power factor is cos = , and is the phase difference between the voltage and current when
the voltage and current are sinusoidal quantities. However, the phase difference between the
instantaneous voltage vector and instantaneous current vector is not constant in the process of
instantaneous power adjustment. We can use instantaneous power method to calculate the
power factor ,

2 2
p
p q
=
+
(3.9)
3.2 Voltage-oriented direct power control of PWM VSR

Take the angle of the rotating vector of the grid voltage as the reference angle of the controller
and then determine all the vectors position in the reference coordinate system, eventually
control the phase angle of the AC current. It is called voltage orientation control and this
control scheme needs to obtain the accurate phase angle of the grid voltage, usually obtained
by the direct detection of the grid voltage. Voltage-Oriented direct power control strategy uses
two options: with AC voltage sensors and without AC voltage sensors
[13]
, calculates the
instantaneous active and reactive power of the rectifier in real-time, compares them with a
given active and reactive power, and finally gives commands to keep the instantaneous power
as well as the instantaneous current in allowed limits. This report only covers AC voltage
sensor strategy.

3.2.1 System composition of VSR

Voltage-source PWM rectifier DPC system is mainly composed by the main circuit and the
control circuit. The main circuit is composed by the AC power supply, filter reactors,
rectifiers, capacitor and the load, as shown in figure 3-1. The control circuit is composed by
the AC voltage and current detection circuit, the DC voltage detection circuit (Hall sensor),
Power Calculator, sector division, power hysteresis comparator, switching table and the PI
regulator. Its diagram is shown as follows:
17


Figure 3-1 DPC block diagram
3.2.2 Principle of DPC

1. Calculation of power
We can use mathematical model in stationary coordinate system and equation (3.6) to
calculate instantaneous active power and reactive power.
2. Sector division
In order to find the position of grid voltage space vector u, u u
o |
is used,
arctan
u
u
|
o
o = (3.10)
Where is the angle between vector u and -axis. The voltage space is divided into 12 sectors
to optimize the performance of rectifier as shown in figure 3-2.

1

axis o
axis |
2

10

12

11

Figure 3-2 DPC sector division
n

is determined by the following equation.
( 1) , 1,2, ,12
6 6
n
n n n

= (3.11)
18

3. Power hysteresis comparator
The input of two hysteresis comparators are the difference =

of given value of
active power and actual value of active power and the difference =

of given
value of reactive power and actual value of reactive power.

is set by the product of the PI

regulator output and DC output voltage;

is set to be zero to achieve unity power factor.

The output of hysteresis comparators reflects the deviation of actual power from given power.
The power hysteresis comparator can be implemented by Schmitt circuit or software.

We define the following state values which reflect the deviation of actual power from given
power.

1,
0,
1,
0,
ref p
p
ref p
ref q
q
ref q
p p H
S
p p H
q q H
S
q q H
<

=

> +

<

=

> +

(3.12)
When the input of hysteresis comparator exceeds positive hysteresis band width H
p
or H
q
, the
output is 1, which means that the driving signals of PWM through modulation should increase
the power of rectifier. When the input is lower than negative hysteresis band width H
p
or
H
q
, the output is zero and the driving signals of PWM which will decrease the power of
rectifier should be chosen. When the input of comparator is between -H and +H, the output
will be the output of previous cycle.
The values of H
p
and H
q
have an important impact on the harmonic current and switching
frequency and power tracking capability. Based on equation (3.12), if the active power or
reactive power amplitude is not in their desired range, the selection of switches is made. The
logic of selection is mentioned in chapter 3 table 3-1 and the comparator model is drawn as
follows:

Figure 3-3 Power hysteresis comparators
Power hysteresis band affects the control precision of instantaneous power, DC voltage and
AC currents. From equation (3.6), there is cross coupling between the control of active and
reactive power. When the control system operates at border region of two sectors, wrong
switches can be chosen easily and with big hysteresis band, duration time of wrong switches
is long. It reveals that with a larger band, the power can vary over a larger range yet
increasing the instantaneous power ripple, DC voltage ripple and AC current distortion, which
is bad for converter and load. Some negative impact on performance of DPC is inevitable
with big ,
p q
H H . With small hysteresis band, the switch frequency increases and losses of
switches increase as well.

2
sq
1
sp
Relay1
Relay
4
P
3
P'
2
Q
1
Q'
19

Another important issue in DPC is the PI controller. The proportional gain and integral gain
also have significant impact on the performance of DPC. Usually these gains are obtained by
trial and error.

4. Switching table
Rewrite the first two equations in (2.21), and we got

dc
r
dc
di
L e Ri v s
di dt
L e Ri v
di dt
L e Ri v s
dt
o
o o o
|
| | |

(3.13)
Where , ,
a r dc dc
e e je i i ji v v s jv s
| o | o |
= + = + = + .
If the impact of R is neglected, we got

0
1
(0) ( )
T
r r
di
L e v i i e v dt
dt L
= = +

(3.14)
The switching table determines the values of S
a
, S
b
, S
c
based on equation (3.13) and the output
of comparators. V
r
is discrete values
0 1 2 3 4 5 6 7
, , , , , , , v v v v v v v v which are determined by , ,
a b c
S S S
and
dc
v . They are shown in figure 3-4.
1
(100) v
2
(110) v
3
(010) v
4
(011) v
5
(001) v
6
(101) v
7
(111) v
0
(000) v
r
e v
e
i
r
i

Figure 3-4 Grid voltage space vectors and Vr
We assume that e is in sector
12

,
r
i
corresponds to
ref
p
, when
i
lags behind and less than
r
i
, it
means that ,
ref ref
p p q q < > namely 1, 0
p q
S S = = . So appropriate V
r
is selected to make
i
close to
r
i
, p close to p
ref
and q close to q
ref
based on equation (3.14). So in the above example,

6
(101) v
is selected and 1, 0, 1
a b c
S S S = = = . When e is in other sectors, the same analysis can
be made, and then we got the switching table shown in table 3-1.






20

Table 3-1 DPC improved switching table
p
S
q
S , ,
a b c
S S S
1

2

3

4

5

6

7

8

9

10

11

12

1 0 101 100 100 110 110 010 010 011 011 001 001 101
1 1 110 010 010 011 011 001 001 101 101 100 100 110
0 0 100 100 110 110 010 010 011 011 001 001 101 101
0 1 110 110 010 010 011 011 001 001 101 101 100 100

The classical switching table is presented in the following:

Table 3-2 DPC classical switching table
p
S
q
S , ,
a b c
S S S
1

2

3

4

5

6

7

8

9

10

11

12

1 0 111 100 000 110 111 010 000 011 111 001 000 101
1 1 111 000 000 111 111 000 000 111 111 000 000 111
0 0 100 100 110 110 010 010 011 011 001 001 101 101
0 1 110 110 010 010 011 011 001 001 101 101 100 100

In the classical switching table, extensive use of vectors V
0
and V
7
weakens the control of
reactive power. Zero vectors can increase active power but the capability is weak. Its main
purpose is to reduce the switching frequency. The average switching frequency of classical
switching table is low, which is an advantage. Compared with classical table, this improved
switching table improves its ability to regulate the reactive power; however the area exists
where active power is out of control. See the figure below for specific analysis.

e
axis o
axis |
2

10

12

11

2
v

Figure 3-5 Analysis of uncontrollable area
When e is in sector
1

and 1, 1
p q
S S = = . According to the switching table, V
2
is selected. When
e is the green line, V
2
is correct. The vector angle between vector e and vector (e-v
r
) is acute
angle, which increases i and decreases the angle between i
r
and i. However, when e is the
purple line, the vector angle between vector e and vector (e-v
r
) is obtuse angle, which
decreases i and decreases the angle between i
r
and i. V
2
makes the active power even less. The
size of the uncontrollable area is determined by the ratio of radius of two circles:
21


3
3
2 2
3
m
m
r dc
dc
U
e U
v v
v
= = (3.15)
In the following, we will derive a new switching table, where the best basic voltage vector V
r

in each sector is selected. This table is synthesized by analyzing the change in the active and
reactive power
[14]
.
Adopt the discrete first order approximation on equation (3.13) and neglect the impact of R,
and we got the change of current vectors,

( 1) ( ) ( ( ) ( ))
( 1) ( ) ( ( ) ( ))
dc
dc
T
i i k i k e k v s k
L
T
i i k i k e k v s k
L
o o o o o
| | | | |

= + =

= + =

(3.16)
Where 1/T is the sampling frequency.
Rewrite the equation (3.6), and the instantaneous power can be expressed as follows:

p u i e i e i
q u i e i e i
o o | |
| o o |
= = +

= =

(3.17)
We assume that , e e
o |
are constant in one sampling period due to the high switching
frequency. So the changes of power in next period can be estimated by:

| |
| |
( 1) ( ) ( 1) ( )
( 1) ( ) ( 1) ( )
p e i k i k e i k i k
q e i k i k e i k i k
o o o | | |
| o o o | |
( = + + +

( = + +

(3.18)
Substituting , i i
o |
in equation (3.16) for equation (3.18), we got changes of active power
and reactive power in the next period ( 1) ( ), ( 1) ( ) p p k p k q q k q k = + = + ,

2 2
( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
dc dc
dc dc
T T
p e e e k v s k e k v s k
L L
T
q e k v s k e k v s k
L
o | o o | |
o | | o

( = +

( =

(3.19)
For power invariant transformation, we got,

3 cos
3 sin
m
m
e U
e U
o
|

(3.20)
For the basic vectors
2 2
0,1,2,...,6,7
( ) ( )
dc dc
v v s v s
o |
= + , the components are shown in the
following table:






22

Table 3-3 Basic space vectors in frame
( )
r a b c
v s s s
dc
v s
o

dc
v s
|

0
v (000)
0 0
1
v (100)
2
3
dc
v
0
2
v (110)
1
6
dc
v
1
2
dc
v
3
v (010)
1
6
dc
v
1
2
dc
v
4
v (011)
2
3
dc
v
0
5
v (001)
1
6
dc
v
1
2
dc
v
6
v (101)
1
6
dc
v
1
2
dc
v
7
v (111)
0 0

Combining equation (3.19) (3.20) and table 3-3, we can find the best basic vector among the
eight possible vectors for 12 sectors. For example,
1
2
, 0
3
r dc dc dc
v v v s v v s
o |
= = = the
changes of power can be expressed as follows:

2
3
2
( 3 cos 0)
3
2
(0 3 sin )
3
m
m dc
m dc
TU T
p U v
L L
T
q U v
L

(3.21)
For given values of U
m
, V
dc
and sector , the sign of , p q can be decided. Similarly, we can
get waveforms of changes in active and reactive power under other basic voltage vectors:



Figure 3-6 Changes of power under different voltage vectors
We assume that e is in sector
12

, the sign of changes in active power and reactive power is

shown in the following table:



0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
0
1
2
3
4
time 0 to 0.02s, corresponding to sector 1 to 12
S
i
g
n

o
f

c
h
a
n
g
e
,

p
o
s
i
t
i
v
e

z
e
r
o
,

n
e
g
a
t
i
v
e
Change in active power, yellow for v1, blue for v2, ...
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
-4
-2
0
2
4
time 0 to 0.02s, corresponding to sector 1 to 12
S
i
g
n

o
f

c
h
a
n
g
e
,

p
o
s
i
t
i
v
e
z
e
r
o
,

n
e
g
a
t
i
v
e
Change in reactive power, yellow for v1, purple for v2,...
23

Table 3-4 Sign of changes in power
p q
0 p > 0 p < 0 q > 0 q = 0 q <
0,3,4,5,7
v
1
v
1,2,3
v
0,7
v
4,5,6
v

When 1, 0
p q
S S = = , it means ,
ref ref
p p q q < > , the voltage vector which can increase active
power and decrease reactive power at the same time ( 0 p > 0 q < ) will be selected, and that
is
4 5
v and v . When e is in other sectors, and for other combinations of ,
p q
S S , the same
analysis can be made:

Table 3-5 Sign of changes in power in other sectors

1

2

3

4

5

6

7

8

9

10

11

12

0
p
>

0,7,3
,4,5
0,7,4
,5,6
0,7,4
,5,6
0,7,1
,5,6
0,7,1
,5,6
0,7,1
,2,6
0,7,1
,2,6
0,7,1
,2,3
0,7,1
,2,3
0,7,2
,3,4
0,7,2
,3,4
0,7,3
,4,5
0
p
<

1 2 2 3 3 4 4 5 5 6 6 1
0
q
>

2,3,4 2,3,4 3,4,5 3,4,5 4,5,6 4,5,6 1,5,6 1,5,6 1,2,6 1,2,6 1,2,3 1,2,3
0
q
<

1,5,6 1,5,6 1,2,6 1,2,6 1,2,3 1,2,3 2,3,4 2,3,4 3,4,5 3,4,5 4,5,6 4,5,6

Then we got the proposed switching table shown in table 3-6.

Table 3-6 Further improved switching table
p
S
q
S , ,
a b c
S S S
1

2

3

4

5

6

7

8

9

10

11

12

1 0 001 001 101 101 100 100 110 110 010 010 011 011
1 1 011 011 001 001 101 101 100 100 110 110 010 010
0 0 100 100 110 110 010 010 011 011 001 001 101 101
0 1 110 110 010 010 011 011 001 001 101 101 100 100

Comparing the above two tables, we can also find the flaw of this table that the active power
is not regulated timely when 0
p
S = . In each sector, there is only one voltage vector which can
be chosen to decrease the active power. This flaw exists in the three switching tables: classical
table, improved table and further improved table. It is important to calculate the quotient
2
3
3
m dc
U v to determine the change in active power p , as we can see form equation
(3.21), there is a shift in the sinusoidal waveform of p .

5. PI regulator.
When losses of R and switches are neglected, the system operates in steady state with unity
power factor, we got
24


2
dc dc
dc
L
dv v
p cv
dt R
= + (3.22)
Set
0 dc dc dc
v v v = + , the above equation can be written as

0
( )
dc dc
dc
L
dv v
p cv
dt R c
+ (3.23)
With Laplace transforms, we got

0
( )
( ) 1
L
dc dc
L
R
v s v
p s R cs
=
+
(3.24)
Due to high switching frequency of power inner loop, the power inner loop can be regarded as
a small inertia link,

1
( )
1
p
p
G s
T s
=
+
(3.25)
Where T
p
is the equivalent time constant of power inner loop.
To eliminate the ripple in the difference of dc voltage, a low-pass filter with cut-off frequency
1/T
c
can be added to the control system.
From the above, we got control system structure as followed, and ( )
PI
G s in figure 3-7 is the
transfer function of voltage PI regulator.

1
1
c
T s +
1
1
c
T s +
1
1
p
T s +
0
1
L
dc
L
R
v
R cs +
0 dc
v ( )
PI
G s
dc
v
0 dc
v

Figure 3-7 DPC control system structure

( )
i
PI p
K
G s K
s
= + so open-loop transfer function of the control system is,

1 1
( ) ( )
1 1 1
i L
op p
c p L
K R
G s K
T s s T s R cs
= +
+ + +
(3.26)
If ,
p c p L
T T T R c , equivalent small inertia link of power inner loop can be neglected. If the
parameters satisfy
c c
T
cut-off angular frequency, equation (3.26) can be simplified to the
standard transfer function of classic II system
[15]
,

2
( )
( )
( 1)
p i L
op
c L
K s K R
G s
T s R cs
+
=
+
(3.27)
From the above equation, the control system is three-order system and three-order optimum
tuning can be used to determine K
p
and K
i
,
25


2
3 2
0.6
0.12
c
p
L
c
i
L
T
K
R c
T
K
R c

(3.28)
3.3 Power flow in the converter

Started with mathematical model of PWM rectifier in the two-phase stationary coordinate
frame, the power flow in the converter is studied.

dc
dc
di
L e Ri v s
dt
di
L e Ri v s
dt
o
o o o
|
| | |

(3.29)
Both ends of the equations are multiplied by i
|
and i
o
respectively, and we got,

dc
dc
di
L i e i Ri i v s i
dt
di
L i e i Ri i v s i
dt
o
| o | o | o |
|
o | o | o | o

(3.30)
Subtracting the two equations gives
( ) ( ) ( )
dc dc
di
di
L i i v s i v s i e i e i
dt dt
|
o
| o | o o | | o o |
= (3.31)
According to the definition of instantaneous power, introduce the concept of instantaneous
active and reactive power of AC side of the rectifier.

r dc dc
r dc dc
p v s i v s i
q v s i v s i
o o | |
| o o |
= +

(3.32)
Disregarding the filter resistance, the energy in the AC-side inductor flows in a form of
reactive power. Define the active and reactive power of AC-side inductor:

0
i
i
di
di
p L i L i
dt dt
di
di
q L i L i
dt dt
|
o
o |
|
o
o |

= + =

(3.33)
,
di
di
L L
dt dt
|
o
are projection values of the resultant vector of three-phase voltages across the
AC-side inductors along the axes respectively. The equation (3.31) can be expressed as
follows:

i r
q q q = (3.34)
The above equation shows that the AC-side reactive power of the PWM rectifier is
determined by the instantaneous power of inductor and the reactive power that the grid
supplies. The instantaneous reactive power of AC inductor is not only in connection with the
26

reactive current, but also active current, which reflects the coupling between the active power
and reactive power, and this coupled relationship exists in the form of the inductor power. In
particular, when the PWM rectifier operates at unity power factor, sum of AC-side reactive
power of rectifier and the reactive power of inductor is zero. Then where does the AC-side
reactive power of rectifier go?

Both ends of the equation (3.29) are multiplied by i
o
and i
|
respectively, and we got,

dc
dc
di
L i e i Ri i v s i
dt
di
L i e i Ri i v s i
dt
o
o o o o o o o
|
| | | | | | |

(3.35)
Adding the two equations gives

2 2
( ) ( ) ( )
dc dc
di
di
L i L i e i e i R i i v s i v s i
dt dt
|
o
o | o o | | o | o o | |
+ = + + + (3.36)
The above equation can be written in the form of the instantaneous power:

i R r
0 p p p p = = (3.37)
Where
R
p means active power consumed by the resistor R.
The equation (3.37) clearly shows that the AC-side active power flows from the grid to the
AC-side of rectifier.
Now analyze the DC-side power flow.
Both ends of the equation
( )
dc
L
dv
C i s i s i
dt
o o | |
= + are multiplied by v
dc ,
and we got,

( )
dc
dc dc L dc
dv
C v i s i s v i v
dt
o o | |
= + (3.38)
The left side of the equation is the instantaneous energy of the capacitor C, and the right part
can be written as follows:

dc dc L dc r L
v s i v s i i v p p
o o | |
+ = (3.39)
The equation (3.38) can be written in the form of the instantaneous power:

C r L
p p p = (3.40)
Where
L
p and
C
p are instantaneous load energy and DC capacitor energy respectively.

The equation (3.40) shows that the AC-side active power of the rectifier is equal to the load
energy plus the capacitor energy and active power goes to the DC side from the AC side of the
rectifier. Then we can answer the last question. The AC-side reactive power of rectifier
charges the DC-side capacitor to boost the DC voltage.

Now we can get the power flow diagram from equations (3.34), (3.37) and (3.40).

27

R
e
c
t
i
f
i
e
r
p
r
p
R
p
C
p
r
p
L
p
AC side
q C
q
DC side
r
q
i
q
r
q

Figure 3-8 Power flow in rectifier

28

Chapter 4 Principle of Model predictive control

Model predictive control strategy has become an advanced process control technology in
chemical process industry, and its usage is spreading to other application areas.

Model predictive control uses the model to compute a trajectory of a future manipulated
variable u to optimize the future behavior of the plant output x (controlled variable). MPC
uses the model and the current measurements of the process to calculate the future actions of
manipulated variables and ensures the controlled variables and manipulated variables to
satisfy the constraints, and then MPC controller puts the first element of the calculated
manipulated variable sequences to the process plant.



Figure 4-1 MPC scheme
The detailed control strategy is shown in figure 4-2.
x is the controlled variable and u is manipulated variable. At the time instant k, with initial
state x
k
, the manipulated variables, at the next n sampling instants, u(k), u(k+1), , u(k+n-1)
are computed. n means prediction horizon. These manipulated variables are calculated to
minimize a cost function J, for example, the predicted deviations from the reference trajectory
over the next n sampling instants. The first element u(k) will be applied to the system for
interval from k to k+1. At the next sampling instant, k+1, the above steps are repeated.
Typically, quadratic programming is used to calculate the manipulated variables at each
sampling instant.
(4.1)
Where Q and R are weighting matrices used to weight prediction error and control actions. If
Q and R are positive definite, the quadratic problem would be convex.
29

future
k
Sampling instant
( ) u k m +
past
1 k + 2 k + 3 k + k n +

( ) x k m +
future output
future action
reference value
( )
r
x k m +

Figure 4-2 Control strategy of MPC
One advantage of the model predictive control is to prevent violations of input and output
constraints, for example, limitation on rate of change in input, product quality and quantity.
Usually, the cost function J is subject to the following inequality constraints:

,0 1
,1
k
k
u u u k n
x x x k n

(4.2)
Where u

and x

are the lower bound on u and x; u

and x

are the upper bound on u and x. It

is important to justify the need for nonlinear MPC before it is applied, because nonlinear
optimization is more time-consuming than linear MPC
[16]
. In our project, linear MPC will be
the focus. The implementation of MPC for power converters may be difficult due to the high
amount of computations to solve optimization problem. With the development of high-speed,
high-precision microprocessor, now we have high speed DSP and FPGA, and that problem
can be solved.
4.1 Review of MPC in power converters

There are four kinds of model used in MPC, and they are Impulse response models, Step
response models, Transfer function models and State-space models. There are two main
methods using MPC to control the power converters, and one of them is predictive direct
power control
[17]
and the other is finite control set MPC
[18]
. The model predictive control
takes advantage of discrete time model of power converter.

This finite control set (FCS) means finite number of switching states in power converter.
There are eight switching states in three-phase half-bridge rectifier, and the switching state
which minimizes the cost function is selected. In reference [19], two-level voltage-source
inverter is tested to compare the performances of classical linear controller and FCS-MPC.
This method does not need modulation stage and the controller exports the states of switches.
Predictive direct power control computes the best voltage vectors and their durations of action
to minimize power errors in a prediction time interval based on the model of power
prediction. P-DPC procedure is implemented by space vector pulse width modulation.

Several model predictive algorithms have been applied in inverter
[19],[20]
; but the application is
30

limited to the control of three-phase rectifier. In this project, our main goal is to investigate
the MPC of rectifier using discrete-time state-space model for prediction.
4.2 Process, model and controller of VSR

The discrete-time state-space model is used to predict the future behaviors. The required
manipulated variables (input voltage of PWM rectifier) are calculated to minimize the cost
function and then the desired voltage is generated by the converter via space vector pulse
width modulation, which will solve the drawback of finite control set MPC, variable
switching frequency. There is no PI controller, so this project saves the controller tuning
procedures. However, there are a number of design parameters in MPC, such as prediction
horizon n, weighting matrices Q, R and sampling period.

The main circuit topology of three-phase voltage-source PWM rectifier is shown in figure 4-
3.
a
e
b
e
c
e
R
L
R
, , a b c
i
L
O
N
-
dc
v
dc
i
L
i
a
b
c
a
s
1
V
3
V
5
V
2
V
4
V
6
V
c
s
b
s

Figure 4-3 Topology of three-phase PWM rectifier
, ,
a b c
e e e e = are the three-phase power supply phase voltages, , ,
a b c
i i i i = are the three-phase
currents. , ,
r ao bo co
v v v v = are the input voltages of VSR.

According to the above figure, the dynamic model for three-phase rectifier is given by:

( )
3
( )
3
( )
3
a a b c
a a a dc
b a b c
b b b dc
c a b c
c c c dc
dc dc
a a b b c c
L
di s s s
L e Ri s v
dt
di s s s
L e Ri s v
dt
di s s s
L e Ri s v
dt
dv v
C i s i s i s
dt R
+ +
=

+ +

+ +

= + +

(4.3)
In the above model, , ,
a b c
s s s are the manipulated variables and , , ,
a b c dc
i i i v are the controlled
variables according to our control goal, which is to suppress the harmonics in AC currents and
regulate the DC output voltage.

We can apply the Park transformation, and get dq coordinate model from the abc three-phase
coordinate system model,
31


( )
d
d q d dc d
q
q d q dc q
dc
q q d d L
di
L e Li Ri v s
dt
di
L e Li Ri v s
dt
dv
C i s i s i
dt

= +

= +

(4.4)
Now ,
d q
s s are the manipulated variables and , ,
d q dc
i i v are the controlled variables. As we can
see from the equation, it is a nonlinear system. We use some approximation to simplify the
system to get a linear system to apply the linear model predictive control.

To achieve the rapid regulation of the DC voltage, ignore the losses on AC side resistor and
switches, and we use energy conservation law: input power from the grid should be equal to
the load power and capacitor charging power. Assuming that the system is in steady state, and
we have
dc dco
v v = , which is the reference value of output DC voltage.
( )
d d q q dco q q d d
p e i e i v i s i s = + = + (4.5)
From equation (4.4), we have ,
dc dc
q q d d L L
L
dv v
i s i s C i i
dt R
+ = + = ,
Substituting for
q q d d
i s i s + from above equation,
( )
dc dc
d d q q dco
L
dv v
p e i e i v C
dt R
= + = + (4.6)
So we have,



dc d d dc
dco L
dv e i v
C
dt v R
= (4.7)
Then the dynamic model of the power rectifier is expressed by:

d
d q d dc d
q
q d q dc q
dc d d dc
dco L
di
L e Li Ri v s
dt
di
L e Li Ri v s
dt
dv e i v
C
dt v R

= +

(4.8)
Equation (4.8) can be written in a state-space form as follows:

1
0 0
1
0 0
0 0 1
0
d
d
d dc d
q q
q dc q
dc
d dc
dco L
R
i
L L
i
e v s
R
i i
e v s
L L
v
e v
Cv CR

(
(

(
(
(
(
( (
(
( (
( (
(
= +
( (
( (
(


(
( (
(

(
(
(

(
(

(4.9)
A discrete-time form with sampling time T can be used to predict the future value of
32

controlled variables. Using Euler approximation
( 1) ( ) di i k i k
dt T
+
= , we obtain a discrete-time
model,

1 0 0
( 1)
( )
( 1) 1 0 ( ) 0
( )
0 0 ( 1)
0 1
d
d
d dc d
q q
q dc q
dc
d dc
dco L
RT T
T
i k
L L
i k
e v s
RT T
i k T i k
e v s
L L
v k
e T T v k
Cv CR

(
(

(
(
(
+
(
( (
(
( (
( (
(
+ = +
( (
( (
(


(
( (
(

+
(
(
(

(
(


(4.10)
We restate the aim of the model predictive control here, suppressing the harmonics of the
input current with unity power factor operation and regulating the DC output voltage quickly,
which means 0,
q dc dco
i v v = = and i
d
supplying exact required power to the load. So we want
our MPC controller to track these references. Then our cost function is,

| | | | | | | |
1
0
1
min ( 1) ( 1) ( 1) ( 1) ( ) ( ) ( ) ( )
2
n
T T
r r r r
k
J x k x k Q x k x k u k u k R u k u k

=
= + + + + +

(4.11)
Where
( )
( ) ( )
( )
d
q
dc
i k
x k i k
v k
(
(
=
(
(

, ( )
d dc d
q dc q
e v s
u k
e v s
(
=
(

,
2 0 0
2 0
0 2 0 ,
0 2
0 0 2
Q R
(
(
(
= =
(
(

(

.
A quadratic objective function with some linear constraints can be solved by a quadratic
program
[21]
. MPC controller can compute the control action u(k) which minimizes the cost
function, and then ( ) ( )
dc d dc q
v s k and v s k are derived as the desired input voltage to the PWM
rectifier.

We get the input voltage to the PWM rectifier in dq frame. We have to transform
input voltage from dq frame to frame, and then space vector pulse width modulation
(SVPWM) is used to modulate this desired space vector:
2 2
( ) ( ) ( tan )
dc d
dc d dc q
dc q
v s
V v s v s t arc
v s

= + +
4.3 Two-level SVPWM modulation technique

4.3.1 Voltage space vector distribution of three-phase VSR

Space vector PWM (SVPWM) control strategy is a novel idea to control the converter. Space
vector control strategy was introduced by the J apanese in the early 1980s for AC motor drive
system. SVPWM compared to conventional Sinusoidal PWM method has the following
advantages: increasing the voltage utilization rate by 15.47%, having a lower switching
frequency, easier to implement for microprocessor due to simple vector mode switching.

In figure 4-3 and for abc coordinate system, we have
33


( )
3
( )
3
( )
3
a b c
ao a dc
a b c
bo b dc
a b c
co c dc
s s s
v s v
s s s
v s v
s s s
v s v
+ +
=
+ +
=
+ +
=
(4.12)
There are six switches and the two switches on the same bridge can not close or open at the
same time, so only 2
3
=8 switch combinations exist. The combinations are shown as follows:

Table 4-1 Voltage values of the different switch combinations
Voltage vector
switch voltage
a
s
b
s
c
s
ao
v
bo
v
co
v
0
V 0 0 0 0 0 0
1
V 1 0 0
2
3
dc
v
1
3
dc
v
1
3
dc
v
2
V 1 1 0
1
3
dc
v
1
3
dc
v
2
3
dc
v
3
V 0 1 0
1
3
dc
v
2
3
dc
v
1
3
dc
v
4
V 0 1 1
2
3
dc
v
1
3
dc
v
1
3
dc
v
5
V 0 0 1
1
3
dc
v
1
3
dc
v
2
3
dc
v
6
V 1 0 1
1
3
dc
v
2
3
dc
v
1
3
dc
v
7
V 1 1 1 0 0 0


It is not difficult to find that the AC side input voltage of different switch combinations can be
expressed by a space vector with length
2
3
dc
v . Take
1
V for example, using power invariant
transformation from abc to coordinate system, we got

1 1
2
1
2 2
3
2
2 3 3 1
2 3
0
3
3 2 2 3
0
1 1 1 1
3 2 2 2
dc
dc jo
dc dc
o
dc
v
V
V v
V v V v e
V
V
v
o
o
| o|
|
(
(

(
(
( (
(

=
( (
(
= =

( (
(
=
( (
(

(
(

(
(


(4.13)
Number of different switch combinations is limited, so there are only eight basic space
vectors
0 1 2 3 4 5 6 7
, , , , , , , V V V V V V V V .

( 1) /3
0,7
2
( 1,...,6)
3
0
j k
k dc
V v e k
V

= =

(4.14)
34

4.3.2 Synthesis of voltage space vector

The objective of the SVPWM control is to synthesize the desired AC side input voltage space
vector. Six non-zero vectors form six sectors. For any given voltage space
vector
*
V , it can be synthesized by the eight basic space vectors, as shown in figure 4-4.

I
II
III
IV
V
VI
1
(100) v
2
(110) v
3
(010) v
4
(011) v
5
(001) v
6
(101) v
7
(111) v
0
(000) v
*
V
2 2
S
V T
T
1 1
S
VT
T

Figure 4-4 basic vectors and sectors
We will focus on the voltage space vector in sectorand later on generalize the discussion to
other sectors.

1 2
, V V are applied for intervals
1 2
, T T respectively, and zero vectors are applied
for interval
0,7
T to synthesize V

over a time period

s
T . The voltage vector can be expressed
as,

*
1 1 2 2 s
TV T V V T + = (4.15)
The angle between V

and
1
V is , according to the sinusoidal law, we got

2 1
* 2 1
2
sin
sin sin( )
3 3
s s
T T
V V
V
T T

o
o
= =

(4.16)
And we have
dc
v V V
3
2
2 1
= = , the time interval can be calculated as follows:

1
2
0,7 1 2
2
sin( )
3
2
sin
s
dc
s
dc
s
V
T T
v
V
T T
v
T T T T

o
o

(4.17)
Constraint of linear SVPWM modulation is,

1 2 s
T T T + (4.18)
Combining equation (4.17) and (4.18), we got
35


2 2
sin( ) sin
3
s s s
dc dc
V V
T T T
v v

o o

+ (4.19)
The above equation should be tenable for any possible value of , then we got,

2
dc
v
V

(4.20)
If V

rotates at a constant speed in the complex plane, three-phase balanced sinusoidal

voltages will be modulated. In fact, V

can only rotates at a stepper speed due to the

switching frequency and limited switch combinations. However, the amplitude of the voltage
space vector V

should be limited to the circle within the hexagon in Fig.4-4 to prevent

distortion in the AC side currents. It is possible for V

bigger than
2
dc
v
, and this condition
is called over saturation of SVPWM. To solve this problem, let
2
dc
v
V

= .
In practice, there are several ways to synthesize the desired voltage space vectorV

. In this
project, seven-segment SVPWM control strategy
[22]
is used, which reduces the amplitude of
harmonic at the switching frequency and improves the waveform quality, however it increases
the switching frequency. The seven-segment SVPWM strategy divides the interval
0,7
T into
three segments, two
0
V s lie in the desired vectors start and end, and
7
V lies in the middle of
V

. Two repeated non-zero basic vectors compose two triangles. The order of on and off of
six switches should satisfy the following principle: only one of six switches can change the
position, close or open, for change of one basic voltage vector to another vector. For example,
supposing the voltage vectorV

in sector 1 as shown in figure 4-4. V

is made up of vectors
0 1 2 7 2 1 0
, , , , , V V V V V V and V . The rectifier with this strategy will switch six times in one PWM
cycle. For the choice of zero vectors, making fewer changes of switches and reducing the
switching losses are the main consideration.
1
(100) V
2
(110) V
2 2
2
s
V T
T
1 1
2
s
VT
T
*
V
1 1
2
s
VT
T
2 2
2
s
V T
T

Figure 4-5 Synthesis of seven-segment SVPWM
Corresponding switch , ,
a b c
s s s values and switching function waveforms for sectorare
shown as follows:



36

Table 4-2 Switch state values
switch


0 1 1 1 1 1 0

0 0 1 1 1 0 0

0 0 0 1 0 0 0


Figure 4-6 Switching function waveforms in sector 1
The flowchart of MPC-SVPWM is drawn below:


Figure 4-7 Flowchart of MPC-SVPWM
For the implementation of SVPWM and MPC controller in Matlab/Simulink, please see the
appendix.

37

Chapter 5 Simulation

This chapter builds models of DPC and MPC-SVPWM power rectifier in Matlab/simulink
environment. Based on these models, their static and dynamic responses, parameter errors on
performance and other factors affecting performances are studied and compared.

The inductor and capacitor are important components of the rectifier. The AC-side inductor
stores energy and filters harmonic currents. It isolates grid voltages from input voltages of
VSR and uses the stored energy to boost the DC output voltage. There are two principles for
determining value of inductor: meeting the requirement of output power of VSR and good
current tracking performance. DC-side capacitor provides a buffer for energy exchange
between DC side of VSR and load, and reduces the DC voltage ripple coefficient. There are
also two principles for determining value of capacitor: fast voltage tracking capability and
good anti-disturbance ability. Combined with engineering practice, trial-and-error method is
used to determine the value of the inductor and capacitor.
5.1 Simulation of direct power control system

Based on figure 3-1 control block diagram and analysis of system composition in 3.2.1 model
of voltage-oriented PWM direct power control system is built in Matlab/simulink
environment. Figure 5-1 shows the model, and part of the simulation parameters are as
follows:

Table 5-1 Simulation parameters
AC side Voltage source peak amplitude 110v
Frequency 50Hz
Inductance 0.022H
Resistance 1
DC side Capacitance 0.0022F
Load resistance 50
Given DC voltage 200v
Sampling frequency 20k Hz

38



Figure 5-1 Model of voltage-oriented PWM direct power control
5.1.1 Comparison of different switching tables

There are three different switching tables, and the performances of different tables are studied.
Capacitor initial voltage is 190v. Simulation results are as follows:
Classical switching table:


Figure 5-2 Instantaneous reactive power and active power
UA
UB
UC
Ualf
Ubet
u3/u2
theta sector
sector
Sp
Sq
sector
g
pulses
Continuous
powergui
Ualf
Ubet
Ialf
Ibet
P
Q
power
Q'
Q
P'
P
sp
sq
hysteresis
Ubet
Ualf
theta
angle
g
A
B
C
+
-
Universal Bridge
z
1
Unit Delay
v
+
-
v
+
-
Uc
v
+
-
Ub
v
+
-
Ua
A
B
C
A
B
C
Three-Phase
Series RL Branch
Product
PI
i
+
-
Ic
i
+
-
Ib
i
+
-
Ia
IA
IB
IC
Ialf
Ibet
I3/I2
[Gate]
[Gate]
200
Constant1
0
Constant
1st-Order
Filter
-100
0
100
200
300
Instantaneous reactive power (upper) and active power, classical table
1.15 1.155 1.16 1.165 1.17 1.175 1.18 1.185 1.19 1.195 1.2
600
700
800
900
Time[s]
39


Figure 5-3 Instantaneous reactive power and active power, startup
Ability of zero vectors for adjusting reactive power is poor. Due to the coupled property
between active power and reactive power, the ability for regulating active power is reduced.
During system startup, reactive power has big overshoot.


Figure 5-4 Phase C current and voltage waveform
The figure shows that the DPC with classical table can achieve unity power factor, yet with
obvious harmonics. FFT tool is used to find the total harmonic distortion which is 9.27%.


Figure 5-5 DC output voltage
There are high ripples in DC output voltage.
-100
0
100
200
300
400
Instantaneous power waveforms, startup, upper for reactive power(classical table)
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
-500
0
500
1000
1500
2000
Time[s]
1.15 1.2 1.25
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase C current and voltage waveforms, classical table
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
186
187
188
189
190
191
192
193
194
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, classical table
40

Improve switching table


Figure 5-6 Instantaneous reactive power and active power


Figure 5-7 Instantaneous reactive power and active power, startup

The improved switching table decreases the reactive uncontrollable area. During system
startup, active power has big overshoot.


Figure 5-8 Phase C current and voltage waveform
The figure shows that the DPC with improved table can achieve unity power factor with
obvious harmonics. FFT tool is used to find the total harmonic distortion which is 7.06%.
Harmonics in currents are reduced and the current waveform is improved using this improved
-100
0
100
200
300
Instantaneous reactive power (upper) and active power, improved table
1.15 1.155 1.16 1.165 1.17 1.175 1.18 1.185 1.19 1.195 1.2
600
700
800
900
Time[s]
-100
0
100
200
300
400
Instantaneous power waveforms, startup, upper for reactive power (improved table)
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
-500
0
500
1000
1500
2000
Time[s]
1.15 1.2 1.25
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase C current and voltage waveforms, improved table
41

table.


Figure 5-9 DC output voltage
The simulation results indicate that feasibility of DPC with improved table is verified.

Further improved switching table


Figure 5-10 Instantaneous reactive power and active power


Figure 5-11 Phase C current and voltage waveform
The figure shows that the DPC with further improved table can achieve unity power factor
with obvious harmonics. FFT tool is used to find the total harmonic distortion which is
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
186
187
188
189
190
191
192
193
194
195
196
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, improved table
-100
0
100
200
300
Instantaneous reactive power (upper) and active power, further improved table
1.15 1.155 1.16 1.165 1.17 1.175 1.18 1.185 1.19 1.195 1.2
600
650
700
750
800
850
900
950
Time[s]
1.15 1.2 1.25
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase C current and voltage waveforms, further improved table
42

10.27%.


Figure 5-12 DC output voltage
The waveform of instantaneous power shows that active power control is not good enough,
since the error exceeds the lower limit of the hysteresis. This is due to the vector selected by
the switching table whose ability to regulate the active power is too strong. Another reason is
the coupled property between active and reactive power. In addition, in the calculation of the
impact of eight vectors on the instantaneous power in different sectors, some approximate
treatments are made, for example, neglecting the impact of resistor.

The result of the simulation indicates that the further improved switching table is not better
than this improved table.

5.1.2 Dynamic response of DPC

Changing load resistance from 50 ohm to 45 ohm at time instant 2.0s enables the dynamic
simulation. We are using improved table and simulation results (DC output voltage and three-
phase currents) are shown as follows:


Figure 5-13 DC output voltage, load changes at 2.0s
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
186
187
188
189
190
191
192
193
194
195
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, improved table
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, load changes at 2.0s(DPC)
43


Figure 5-14 Three-phase currents, load changes at 2.0s


Figure 5-15 Phase C voltage and current waveforms, load changes at 2.0s
The following table summarizes the harmonics of current C of steady state before and after
load changes:
Table 5-2 THD and value of current C
Ic Peak value(A) Rms value THD
Before change 4.714 3.333 7.58%
After change 5.198 3.675 6.99%

Magnitude of three-phase currents increased and the DC output voltage barely changed as we
decreased the load resistor. The AC-side power is equal to DC-side power. We know that the
DC-side power is increased when the load resistance decreases. To increase the AC side
power, the AC side currents should increase.
Above simulation results are obtained with a small load change, and test with a big change in
load resistance is investigated, load resistance from 50 ohm to 25 ohm at time instant 2.0s.
Simulation results are shown as follows:
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
-10
-8
-6
-4
-2
0
2
4
6
8
10
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase current, load changes at 2.0s(DPC)
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase C voltage and current waveforms, load changes at 2.0s(DPC)
44


Figure 5-16 DC output voltage, load changes to 25 ohm at 2.0s


Figure 5-17 Three-phase currents, load changes to 25 ohm at 2.0s


Figure 5-18 Phase C voltage and current waveforms, load changes to 25 ohm at 2.0s
The following table summarizes the harmonics of current C of steady state before and after
load changes:

1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, load changes to 25 Ohm at 2.0s(DPC)
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
-15
-10
-5
0
5
10
15
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, load changes to 25 Ohm at 2.0s(DPC)
1.95 2 2.05 2.1 2.15 2.2 2.25 2.3 2.35 2.4 2.45
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase C voltage and current waveform, load changes to 25 Ohm at 2.0s(DPC)
45

Table 5-3 THD and value of current C
Ic Peak value(A) Rms value THD
Before change 4.672 3.303 6.54%
After change 9.132 6.457 7.54%

Table 5-2 and table 5-3 indicate that when load changes, AC-side currents react to the
changes. DC output voltage goes down for both cases, but with smaller change in load side,
DC output voltage changes less.
5.1.3 Summary

During system startup, active and reactive power of direct power control have big overshoot,
which is bad for the converter and grid. DPC has good dynamic performance. The switching
frequency of the direct power control system using hysteresis comparator is not fixed. To
improve the control precision, the hysteresis band should be small, resulting in increase of
switching frequency and higher switching loss. Total harmonic distortion is dissatisfactory.

There is reactive uncontrollable area in two-level comparator scheme. To improve the
performance of DPC, three-level comparator can be used. The hysteresis characteristic is
shown below:
P
s
1
P
0
P
H
P
H
1

Figure 5-19 Three-level comparator Hysteresis characteristic
State values reflecting the deviation of actual power from given power is shown as follows:

1,
0,
1,
ref p
p ref p ref p
ref p
p p H
S p H p p H
p p H
<

= < < +

> +

(5.1)
When
ref p ref p
p H p p H < < + , voltage vector which basically adjusts reactive power is
selected; When
ref p
p p H > + , select the voltage vector which basically adjusts active power.
46

5.2 Simulation of model predictive control

Based on the flowchart of MPC-SVPWM, the model of MPC-SVPWM control system is built
in Matlab/simulink environment. Figure 5-20 shows the model, and part of the simulation
parameters are the same with table 5-1.

The PWM VSR is a hybrid system, which is modeled by a discrete-time model in our project.
The accuracy of the discrete-time model is dependent on the sampling period T as we can see
from equation (4.10). The sampled signal should be a sufficient representation of this hybrid
system and at the same time, we dont want the sampling frequency to be very large. The
control system includes sampling operation, calculation operation and modulation operation,
and we want the control system to finish these jobs in one sampling period.

Figure 5-20 Model of MPC-SVPWM system
So first, we will find the best sampling frequency for our MPC algorithm. To get DC voltage
closer to reference value, unity power factor and fewer harmonics in currents are the selection
criteria for sampling frequency.

Begin with 10k Hz, the simulation results are shown as follows:

wt
theta
Continuous
powergui
i
+
-
idc
iaibic
ia
[B]
gate1
[B]
gate
[A]
[A]
Vdc
g
A
B
C
+
-
Universal Bridge
v
+
-
v
+
-
A
B
C
A
B
C
Three-Phase
Series RL Branch
Scope2
Scope1
Scope
abc,theta
Id
Iq
Park
i
+
-
Ic
i
+
-
Ib
i
+
-
Ia
10
Gain
u y
fcn
Embedded
MATLAB Function
47


Figure 5-21 Three-phase currents


Figure 5-22 DC output voltage


Figure 5-23 Phase A current and voltage waveforms

From above figures, we can see that MPC-SVPWM system has fewer harmonic in AC
currents, smaller ripple coefficient of DC output voltage and fast start-up speed. The same
simulations with different sampling frequencies are tested. FFT Spectrum Analysis is
performed and a table is made to compare the performances of MPC under different sampling
frequencies.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-10
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents
Sampling frequency 10KHz
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage
Sampling frequency 10KHz
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-100
-80
-60
-40
-20
0
20
40
60
80
100
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
e
r
)
Phase A voltage and current waveforms
Sampling frequency 10KHz
48

Table 5-4 DC output voltage and THD under different sampling frequencies
Frequency (Hz) Actual DC output (v) THD in phase current a,b,c
12k 197.0 0.52% 0.51% 0.51%
10k 198.3 0.62% 0.61% 0.61%
8k 199.7 0.76% 0.76% 0.76%
5k 203.3 1.16% 1.16% 1.16%

To solve the following quadratic programming problem is time-consuming.
| | | | | | | |
1
0
min ( 1) ( 1) ( 1) ( 1) ( ) ( ) ( ) ( )
n
T T
r r r r
k
J x k x k Q x k x k u k u k R u k u k

=
= + + + + +

For carefully consideration of power factor, DC output voltage, THD in AC currents and
calculation in digital control system, the sampling frequency 8 KHz is chosen for our case.

5.2.1 Startup and steady state

We can see from equation (4.8) that the MPC model is built based on steady state, which
means the capacitor voltage is the reference value 200 v. It is needed to consider whether to
pre-charge the DC capacitor in the startup process; it is related to system response speed and
transient AC current overshoot problem. It can be predicted that the response speed is slow
without pre-charging. In the case of the same simulation model parameters, two situations are
simulated: pre-charging DC capacitor and not pre-charging, respectively, and the simulation
results are as follows:

The capacitor initial voltage is zero:

Figure 5-24 DC output voltage without pre-charging

When the initial voltage across the capacitor is zero, the DC output voltage takes approximate
0.35s to reach the reference value 200v.
0 0.1 0.2 0.3 0.4 0.5 0.6
0
50
100
150
200
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage without pre-charging
Sampling frequency 8KHz
49


Figure 5-25 Grid-side current waveforms without pre-charging


Figure 5-26 Phase A voltage and current waveforms without pre-charging

The capacitor initial voltage is 200v:


Figure 5-27 DC output voltage with pre-charging

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-25
-20
-15
-10
-5
0
5
10
15
20
25
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents without pre-charging
Sampling frequency 8KHz
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms without pre-charging
Sampling frequency 8KHz
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
110
120
130
140
150
160
170
180
190
200
210
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage with capacitor initial voltage 200v
Sampling frequency 8KHz
50


Figure 5-28 Grid-side current waveforms with pre-charging


Figure 5-29 Phase A voltage and current waveforms with pre-charging

When the capacitor initial voltage is 0, the response speed of DC output voltage and AC-side
currents is slow; the grid-side current overshoot is much more serious; the speed of the current
tracking voltage phase is slow, taking about a few line cycles. Pre-charging is superior to non-
pre-charging. In the following, all the simulations are tested with pre-charging.

5.2.2 Dynamic performance of MPC-SVPWM

Performance of MPC depends on the accuracy of the open-loop prediction, which depends on
the accuracy of the plant model. Prediction model can not be completely the same with actual
plant. Simplification of plant model brings uncertainty. In addition, MPC controller
parameters are not same with the main circuit parameters such as resistance, inductance and
capacitance. Measurement noise and parameter uncertainty (or variation under operation)
degrade the control system performance and even affect system stability. In order to assure the
system performances, these factors must be considered in the design of MPC controller, for
example, building a system parameter estimator.
Some important issues regarding these factors which affect the performance of model
predictive control will be discussed in this project.

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-10
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase current with capacitor initial voltage 200v
Sampling frequency 8KHz
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms with pre-charging
Sampling frequency 8KHz
51

Load changes

This part mainly tests anti-disturbance ability of the algorithm. First, we will investigate the
performances of the load changes without parameter estimator. In the prediction matrix, there
is one element related to the load resistance. The performance will deteriorate due to the load
changes if we do not adapt the prediction model accordingly. Changing load resistance from
50 ohm to 45 ohm at time instant 0.2s enables the dynamic simulation. Simulation result is
obtained using Matlab/Simulink, and the sampling period is Ts=125us; weighting matrices are
as follows:
2 0 0
2 0
0 2 0 ,
0 2
0 0 2
Q R
(
(
(
= =
(
(

(




Figure 5-30 DC output voltage, load changes at 0.2s


Figure 5-31Grid-side current waveforms, load changes at 0.2s

0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, load changes at 0.2s (MPC)
0.15 0.2 0.25 0.3 0.35
-10
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, load changes at 0.2s(MPC)
52


Figure 5-32 Phase A voltage and current waveforms, load changes at 0.2s
There is no overshoot in currents in the transient process of load changes. Magnitude of three-
phase currents doesnt change and the DC output voltage decreases as we decrease the load
resistor. This makes sense. The AC-side power is equal to DC-side power. We can see from
figure that the AC-side power is constant when the load resistance decreases. To maintain the
DC side power, the DC output voltage should decrease.

The MPC controller has important tuning parameters, matrices Q and R. Increasing the
elements in matrix Q brings a higher penalty on plant output changes which are I
d
, I
q
, V
dc
,
meanwhile increasing the elements in matrix R brings a higher penalty on plant manipulated
variables which are input voltage of rectifier in our case. Second simulation results (including
startup and load changing process) are obtained under the following weighting matrices,
higher penalty on I
q
and V
dc
.

0 0 0
2 0
0 2000 0 ,
0 2
0 0 2000
Q R
(
(
(
= =
(
(

(


Q (1,1)=0 means no penalty on I
d
and higher penalty on Iq and V
dc
. MPC-SVPWM algorithm
uses law of conservation of energy, and makes I
d
equal to the value which forces Iq=0,
V
dc
=200v and supply the exact power to the load side.


Figure 5-33 DC output voltage, startup, high penalty on Iq and Vdc
0.15 0.2 0.25 0.3 0.35
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, load changes at 0.2s(MPC)
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, startup, high penalty on Iq and Vdc(MPC)
53


Figure 5-34 Grid-side current waveforms, startup, high penalty on Iq and Vdc

Figure 5-35 Phase A voltage and current waveforms, startup, high penalty on Iq and Vdc
The above results are about the startup process and the following figures are about load
changing process.


Figure 5-36 DC output voltage, high penalty on Iq and Vdc, load changes at 0.2s

When the load resistance decreases, DC output voltage goes from 200.5v to 197.6v, a drop of
1.45% for MPC-SVPWM.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, startup, high penalty on Iq and Vdc(MPC)
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, startup, high penalty on Iq and Vdc
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, high penalty on Iq and Vdc(MPC)
Load changes at 0.2s
54


Figure 5-37 Grid-side current waveforms, high penalty on Iq and Vdc, load changes at 0.2s


Figure 5-38 Phase A voltage and current waveforms, high penalty on Iq and Vdc, load changes at 0.2s

The following table summarizes the harmonics of current A of steady state before and after
load changes:
Table 5-5 THD and value of current A
Ia Peak value(A) Rms value THD
Before change 5.196 3.674 0.75%
After change 5.637 3.986 0.69%

When Q =2*diag([0, 1000, 1000]), R =[2 0; 0 2], higher penalty on V
dc
and I
q
makes the
control more aggressive so that the output tracks the setpoints aggressively and faster. Since a
higher value of Q (a lower value of R relatively) gives a smaller punishment on the
manipulated variables.
Above simulation results are obtained with a small load change, and test with a big change in
load resistance is investigated under the following weighting matrices, load resistance from 50
ohm to 25 ohm at time instant 0.2s. Simulation results are shown as follows:
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, startup, high penalty on Iq and Vdc(MPC)
Load changes at 0.2s
0.15 0.2 0.25 0.3 0.35
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, high penalty on Iq and Vdc, load changes at 0.2s
55

0 0 0
2 0
0 2000 0 ,
0 2
0 0 2000
Q R
(
(
(
= =
(
(

(


Figure 5-39 DC output voltage, load changes to 25 ohm at 0.2s, high penalty on Iq and Vdc

Figure 5-40 Grid-side current waveforms, load changes to 25 ohm at 0.2s, high penalty on Iq and Vdc
The following table summarizes the harmonics of current A of steady state before and after
load changes:
Table 5-6 THD and value of current A


Table 5-5 and table 5-6 indicate that when load changes, AC-side currents react to the
changes. DC output voltage goes down for both cases, but with bigger change in load side,
DC output voltage changes more.
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65
170
180
190
200
210
220
230
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, High penalty on Iq and Vdc(MPC)
Load changes to 25 Ohm at 0.2s
0.15 0.2 0.25 0.3 0.35 0.4
-15
-10
-5
0
5
10
15
Time[s]
T
w
o

s
t
e
a
d
y

s
t
a
t
e
s

a
n
d

o
n
e

o
b
v
i
o
u
s

t
r
a
n
s
i
e
n
t

s
t
a
t
e
Three-phase currents, High penalty on Iq and Vdc(MPC)
Load changes to 25 Ohm at 0.2s
Ia Peak value(A) Rms value THD
Before change 5.196 3.674 0.75%
After change 9.481 6.704 0.40%
56


Figure 5-41 Phase A current and voltage waveforms, load changes to 25 ohm at 0.2s, high penalty on I
q
and V
dc

This big change in load causes non-unity power factor.

Parameter error

Model Predictive control is a parametric model-based approach and it is sensitive for
parameter changes. We have invested the effects of load changes and now we will investigate
the filter parameter errors (inductance and resistance) on performances of MPC-SVPWM.
The following table summarizes the parameter errors which affects the performances of MPC-
SVPWM.
Table 5-7 Parameter errors
Error Filter inductance L AC-side resistance R
0 0.022H 1
-50% 0.011H 0.5
+50% 0.033H 1.5

Simulation results are shown as follows:


Figure 5-42 DC output voltage, startup, 0.5L
0.15 0.2 0.25 0.3 0.35 0.4
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t
(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A current and voltage waveforms, High penalty on Iq and Vdc(MPC)
Load changes to 25 Ohm at 0.2s
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
120
130
140
150
160
170
180
190
200
210
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 0.5L
57


Figure 5-43 Phase A voltage and current waveforms, startup, 0.5L
There is a phase shift between voltage and current.


Figure 5-44 Grid-side current waveforms, startup, 0.5L


Figure 5-45 DC output voltage, startup, 1.5L
0 0.05 0.1 0.15 0.2 0.25 0.3
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 0.5L
0 0.05 0.1 0.15 0.2 0.25 0.3
-5
-4
-3
-2
-1
0
1
2
3
4
5
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, parameter error 0.5L
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
180
190
200
210
220
230
240
250
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 1.5L
58


Figure 5-46 Phase A voltage and current waveforms, startup, 1.5L


Figure 5-47 Grid-side current waveforms, startup, 1.5L


Figure 5-48 DC output voltage, startup, 0.5R
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 1.5L
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-15
-10
-5
0
5
10
15
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, parameter error 1.5L
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
188
190
192
194
196
198
200
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 0.5R
59


Figure 5-49 Phase A voltage and current waveforms, startup, 0.5R


Figure 5-50 Grid-side current waveforms, startup, 0.5R


Figure 5-51 DC output voltage, startup, 1.5R
0 0.05 0.1 0.15 0.2 0.25
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 0.5R
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-8
-6
-4
-2
0
2
4
6
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, parameter error 0.5R
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
190
192
194
196
198
200
202
204
206
208
210
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 1.5R
60


Figure 5-52 Phase A voltage and current waveforms, startup, 1.5R


Figure 5-53 Grid-side current waveforms, startup, 1.5R

Load changes with parameter error

Next we will investigate the dynamic performance when load changes from 50 ohm to 45
ohm at time instant 1.0s with parameter error.


Figure 5-54 DC output voltage, parameter error 0.5L, load changes at 1.0s
0 0.05 0.1 0.15 0.2 0.25
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 1.5R
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
-10
-8
-6
-4
-2
0
2
4
6
8
Time[s]
A
m
p
l
i
t
u
d
e

o
f

c
u
r
r
e
n
t
[
A
]
Three-phase currents, parameter error 1.5R
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4
120
125
130
135
140
145
150
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 0.5L, load changes at 1.0s
61


Figure 5-55 Phase A voltage and current waveforms, parameter error 0.5L, load changes at 1.0s
There is a phase shift between voltage and current.


Figure 5-56 DC output voltage, parameter error 1.5L, load changes at 1.0s


Figure 5-57 Phase A voltage and current waveforms, parameter error 1.5L, load changes at 1.0s
0.95 1 1.05 1.1 1.15 1.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 0.5L, load changes at 1.0s
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4
220
225
230
235
240
245
250
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 1.5L, load changes at 1.0s
0.95 1 1.05 1.1 1.15 1.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 1.5L, load changes at 1.0s
62


Figure 5-58 DC output voltage, parameter error 0.5R, load changes at 1.0s


Figure 5-59 Phase A voltage and current waveforms, parameter error 0.5R, load changes at 1.0s


Figure 5-60 DC output voltage, parameter error 1.5R, load changes at 1.0s
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4
180
185
190
195
200
205
210
Time[s]
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 0.5R, load changes at 1.0s
0.95 1 1.05 1.1 1.15 1.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 0.5R, load changes at 1.0s
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 1.35 1.4
180
185
190
195
200
205
210
A
m
p
l
i
t
u
d
e

o
f

D
C

o
u
t
p
u
t

v
o
l
t
a
g
e
[
v
]
DC output voltage, parameter error 1.5R, load changes at 1.0s
63


Figure 5-61 Phase A voltage and current waveforms, parameter error 1.5R, load changes at 1.0s
The following table summarizes the harmonics of current of steady state before and after load
changes:
Table 5-8 THD and value of current A
Ia Peak value(A) Rms value THD
Figure5-57 Before change 7.545 5.335 0.56%
After change 7.534 5.327 0.55%
Figure5-59 Before change 5.138 3.633 0.78%
After change 5.005 3.539 0.77%
Figure5-61 Before change 5.207 3.682 0.77%
After change 5.207 3.682 0.73%

As we can see from table 5-8, when the load has a small change, the AC-side currents almost
dont change and even become smaller. The system does not work with parameter errors.

5.2.3 Summary

The performance is sensitive for parameter changes and parameter errors, especially for the
filter inductance. Additional control block with online parameters estimation should be added
in order to decrease control sensitivity to parameter changes. Adaptive control can overcome
the impact of changes in system parameters and adaptive control applied to the PWM
converter is worth studying.


0.95 1 1.05 1.1 1.15 1.2
-150
-100
-50
0
50
100
150
Time[s]
P
i
n
k

f
o
r

c
u
r
r
e
n
t

(
m
a
g
n
i
f
i
e
d

t
e
n
-
f
o
l
d
)
Phase A voltage and current waveforms, parameter error 1.5R, load changes at 1.0s
64

Chapter 6 Conclusion and future work

This thesis carried out the theoretical study and Matlab simulation on the direct power control
and model predictive control of PWM rectifier, and completed the following tasks:
Study the principles of DPC and MPC-SVPWM
Build DPC and MPC-SVPWM models in Matlab/Simulink environment
Compare results from respective models
Through simulation results, conclusions are drawn:
6.1 Conclusions from the Simulink results

1. Sampling rate of MPC-SVPWM scheme is low, reducing the switch loss and
increasing its lifecycle. Voltage-orientation direct power control could not achieve the
same experimental performance even if the sampling rate is higher.
2. Both MPC-SVPWM and DPC have good dynamic performance when load changes.
Model predictive control is a parametric model-based approach and it is sensitive for
parameter changes but there is no parameter error issue in direct power control.
3. Synchronous phase angle and frequency are needed for Park transformation in MPC-
SVPWM. There is no complicated park transformation in DPC.

Voltage-source PWM rectifier with model predictive control strategy performs better in the
following aspects:
Fewer harmonics in AC currents
Smaller DC voltage ripple coefficient
Unity power factor
Good static and dynamic performance
Higher utilization of the DC bus voltage

The performance of MPC-SVPWM is sensitive for parameters. To increase the performance,
the accuracy of a model should be increased, and more exact model should be built. Make less
approximate treatment to get the prediction model. Parameter estimator and adaptive control
applied to the PWM converter are worth studying to overcome the impact of parameter
uncertainty.

Though this project achieves its expected purpose, there are a lot of work has to be done, for
example, increasing system stability.
6.2 Suggested future work

1. Solving the MPC problem takes a substantial amount of computation time. The
computation time should be investigated and compared with the sampling period.
2. Select an appropriate digital signal processor for this demanding power electronics
control application.
3. Establish the simulation model in DSP in the lab to verify the algorithm of model
predictive control.
4. Energy feedback and the inverter aspects should be included in the experiment.
5. Test other load rather than resistive load in the experiment.

65

Appendix
Deviation of power supply voltages in dq frame ed and eq

Three phase balanced power supply voltages are given:

2 cos
2 cos( 2 / 3)
2 cos( 2 / 3)
a m
b m
c m
e U t
e U t
e U t

= +

(1.2)
Power invariant transformation,

cos cos( 2 / 3) cos( 2 / 3)
2
sin sin( 2 / 3) sin( 2 / 3)
3
1 1 1
2 2 2
d a
q b
c
o
x x
x x
x
x


(
(
( + (
(
(
(
= +
(
(
(
(
(
(


(
(

(1.3)
Note that t = , therefore,

| |
cos 2 cos cos( 2 / 3) 2 cos( 2 / 3)
2
3
cos( 2 / 3) 2 cos( 2 / 3)
cos cos cos( 2 / 3) cos( 2 / 3)
2
cos( 2 / 3) cos( 2 / 3)
3
cos cos cos cos(2 / 3) sin sin(2 / 3)
cos
2
3
m m
d
m
m
m
U t U t
e
U t
t t
U
t
t
U





(
+ +
= (
+ + (

+ + (
=
(
+ +

+ +
=
| |
| |
| |
cos(2 / 3) sin sin(2 / 3)
cos cos(2 / 3) sin sin(2 / 3)
cos cos(2 / 3) sin sin(2 / 3)
2 3 3
cos cos sin sin
2 2 3
3 cos( )
3
m
m
m
t t
t t
U
t t
U t
U







+ +

`

)
(
= +
(

=
=
(1.4)
In the same way, we got e
q
=0.
Switching function waveforms in six sectors

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-1 Switching function waveforms in sector 1
66

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-2 Switching function waveforms in sector 2

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-3 Switching function waveforms in sector 3

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-4 Switching function waveforms in sector 4

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-5 Switching function waveforms in sector 5
67

0
4
T
0
2
T
1
2
T
0
4
T
2
2
T
1
2
T
2
2
T
a
s
b
s
c
s
1
1
1

Figure 0-6 Switching function waveforms in sector 6

Codes in Embedded MATLAB Function

%MPC cont r ol l er
f unct i on y = f cn( u)
%Decl ar e f ol l owi ng f unct i ons t o be ext r i nsi c
eml . ext r i nsi c( ' spar se' ) ;
eml . ext r i nsi c( ' bl kdi ag' ) ;
eml . ext r i nsi c( ' opt i mset ' ) ;
eml . ext r i nsi c( ' quadpr og' ) ;
%
per si st ent f cn_cal t i me;
i f i sempt y( f cn_cal t i me)
f cn_cal t i me = doubl e( zer os( 1, 1) ) ;
end
%nu: number of cont r ol s
per si st ent nu;
i f i sempt y( nu)
nu = doubl e( 2) ;
end
%Syst emmat r i ces Ad and Bd
per si st ent Ad;
i f i sempt y( Ad)

Ad = doubl e( [ 0. 9915 0. 0393 0;
- 0. 0393 0. 9915 0;
0. 0383 0 0. 9989] ) ;

end
per si st ent Bd;
i f i sempt y( Bd)

Bd = doubl e( [ 0. 0057 0;
0 0. 0057;
0 0] ) ;
end
%n: Lengt h of cont r ol hor i zon
per si st ent n;
i f i sempt y( n)
n = doubl e( 3) ;
end
%Wei ght i ng mat r i x Q and R i n cost f unt i on
per si st ent I _n;
i f i sempt y( I _n)
I _n = zer os( 3) ;
68

I _n = ( eye( 3) ) ;
end
per si st ent Qk;
i f i sempt y( Qk)
Qk = doubl e( 2*di ag( [ 1, 1, 1] ) ) ;
end
per si st ent Rk;
i f i sempt y( Rk)
Rk = doubl e( [ 2 0;
0 2] ) ;
end
%Equal i t y const r ai nt
per si st ent Aeq_c1;
i f i sempt y( Aeq_c1)
Aeq_c1 = zer os( 9) ;
Aeq_c1 = ( eye( 3*3) ) ;
end
per si st ent Aeq_c2;
i f i sempt y( Aeq_c2)
Aeq_c2 = zer os( 9) ;
Aeq_c2 = ( kr on( di ag( ones( 2, 1) , - 1) , - Ad) ) ;
end
per si st ent Aeq_c3;
i f i sempt y( Aeq_c3)
Aeq_c3 = zer os( 9, 6) ;
Aeq_c3 = kr on( I _n, - Bd) ;
end
per si st ent Aeq;
i f i sempt y( Aeq)
Aeq = zer os( 9, 15) ;
Aeq = [ ( Aeq_c1 + Aeq_c2) , Aeq_c3] ;
end
%I nequal i t y const r ai nt s
%Lower bound on x
per si st ent x_l b;
i f i sempt y( x_l b)
x_l b = doubl e( - I nf ( 3*3, 1) ) ;
end
%Upper bound on x
per si st ent x_ub;
i f i sempt y( x_ub)
x_ub = doubl e( I nf ( 3*3, 1) ) ;
end
%Lower bound on u
per si st ent u_l b;
i f i sempt y( u_l b)
u_l b = doubl e( - I nf ( 3*2, 1) ) ;
end
%Upper bound on u
per si st ent u_ub;
i f i sempt y( u_ub)
u_ub = doubl e( I nf ( 3*2, 1) ) ;
end
%Lower bound on z
per si st ent l b;
i f i sempt y( l b)
l b = doubl e( [ x_l b; u_l b] ) ;
end
%Upper bound on z
per si st ent ub;
i f i sempt y( ub)
69

ub = doubl e( [ x_ub; u_ub] ) ;
end
per si st ent Q_mpc;
i f i sempt y( Q_mpc)
Q_mpc = doubl e( kr on( I _n, Qk) ) ;
end
per si st ent R_mpc;
i f i sempt y( R_mpc)
R_mpc = kr on( I _n, Rk) ;
end
per si st ent G_mpc;
i f i sempt y( G_mpc)
G_mpc = zer os( 15) ;
G_mpc = doubl e( bl kdi ag( Q_mpc, R_mpc) ) ;
end
per si st ent Aeq_mpc;
i f i sempt y( Aeq_mpc)
Aeq_mpc = doubl e( [ Aeq_c1 + Aeq_c2, Aeq_c3] ) ;
end
per si st ent opt ;
i f i sempt y( opt )
opt = opt i mset ( ' Di spl ay' , ' of f ' , ' Di agnost i cs' , ' of f ' ,
' Lar geScal e' , ' on' ) ;
end
per si st ent beq_mpc;
i f i sempt y( beq_mpc)
beq_mpc = doubl e( [ zer os( 3, 1) ; zer os( 2*3, 1) ] ) ;
end
%Sampl i ng f r equency
per si st ent f r eq_si gnal ;
i f i sempt y( f r eq_si gnal )
f r eq_si gnal = doubl e( 8000) ;
end
per si st ent f cn_out st at e;
i f i sempt y( f cn_out st at e)
f cn_out st at e = doubl e( [ 0, 0, 0, 0] ) ;
end
per si st ent f cn_wt ;
i f i sempt y( f cn_wt )
f cn_wt = doubl e( zer os( 1) ) ;
end
per si st ent f cn_gat e;
i f i sempt y( f cn_gat e)
f cn_gat e = doubl e( zer os( 2, 6) ) ;
end
per si st ent f cn_l i mi t T;
i f i sempt y( f cn_l i mi t T)
f cn_l i mi t T = doubl e( zer os( 1, 6) ) ;
end
per si st ent f cn_t now;
i f i sempt y( f cn_t now)
f cn_t now = doubl e( zer os( 1) ) ;
end

i f u( 3) <= 0. 01
y = [ 0; 1; 0; 1; 0; 1] ;
el se
%Sol ve opt i mi zat i on pr obl emat each sampl i ng i nst ant
i f u( 4) / ( 2*pi *50/ f r eq_si gnal ) >= f cn_cal t i me
x = u( 1: 3) - [ 5. 938157; 0; 200] ;
beq_mpc( 1: 3) = Ad*x;
70

z = zer os( 15, 1) ;
%f val = zer os( 1) ;
%exi t f l ag = zer os( 1) ;
%out put =
%l ambda =
st r uct ( ' l ower ' , zer os( 15, 1) , ' upper ' , zer os( 15, 1) , ' eql i n' , zer os( 9, 1) , ' i neql i n'
, zer os( 0, 1) ) ;
[ z, f val , exi t f l ag, out put , l ambda] =
quadpr og( G_mpc, [ ] , [ ] , [ ] , Aeq_mpc, beq_mpc, l b, ub, [ ] , opt ) ;
u_mpc = doubl e( z( 10: 11) ) ; %Onl y f i r st el ement i s used
u_mpc = [ 125. 8147, - 41. 0416] ' - u_mpc;
u_mag = sqr t ( u_mpc( 1) ^2 + u_mpc( 2) ^2) ;
u_angl e = acos( u_mpc( 1) / u_mag) ;
i f u_mpc( 2) < 0
u_angl e = - u_angl e;
end
u_angl e = u( 4) + u_angl e;
i f u_mag > u( 3) / sqr t ( 2)
u_mag = u( 3) / sqr t ( 2) ;
end
whi l e u_angl e < 0
u_angl e = u_angl e + 2*pi ;
end
par t _val ue = u_angl e/ ( pi / 3) ;
par t _i n = f i x( par t _val ue) ;
par t _angl e = u_angl e - par t _i n*pi / 3;
par t _i n = r em( par t _i n, 6) ;
par t _i n = par t _i n + 1;
T_t emp = sqr t ( 2) *u_mag/ u( 3) *1/ f r eq_si gnal ;
T1 = T_t emp*si n( pi / 3- par t _angl e) ;
T2 = T_t emp*si n( par t _angl e) ;
T0 = 1/ f r eq_si gnal - T1 - T2;
f cn_out st at e = [ T1, T2, T0, par t _i n] ;
swi t ch ( par t _i n)
case 1
f cn_gat e = [ 1, 0, 0, 1, 0, 1;
1, 0, 1, 0, 0, 1] ;
case 2
f cn_gat e = [ 1, 0, 1, 0, 0, 1;
0, 1, 1, 0, 0, 1] ;
case 3
f cn_gat e = [ 0, 1, 1, 0, 0, 1;
0, 1, 1, 0, 1, 0] ;
case 4
f cn_gat e = [ 0, 1, 1, 0, 1, 0;
0, 1, 0, 1, 1, 0] ;
case 5
f cn_gat e = [ 0, 1, 0, 1, 1, 0;
1, 0, 0, 1, 1, 0] ;
ot her wi se
f cn_gat e = [ 1, 0, 0, 1, 1, 0;
1, 0, 0, 1, 0, 1] ;
end
%SVPWM modul at i on
f cn_l i mi t T( 1) = T0/ 4;
f cn_l i mi t T( 2) = f cn_l i mi t T( 1) + T1/ 2;
f cn_l i mi t T( 3) = f cn_l i mi t T( 2) + T2/ 2;
f cn_l i mi t T( 4) = f cn_l i mi t T( 3) + T0/ 2;
f cn_l i mi t T( 5) = f cn_l i mi t T( 4) + T2/ 2;
f cn_l i mi t T( 6) = f cn_l i mi t T( 5) + T1/ 2;
f cn_cal t i me = f cn_cal t i me + 1;
71

f cn_t now = u( 4) / ( 100*pi ) ;
end
y0 = [ 0, 1, 0, 1, 0, 1] ;
t _now = 0;
t _now = u( 4) / ( 100*pi ) - f cn_t now;
i f t _now < f cn_l i mi t T( 1)
y0 = [ 0, 1, 0, 1, 0, 1] ;
el se
i f t _now < f cn_l i mi t T( 2)
y0 = f cn_gat e( 1, : ) ;
el se
i f t _now < f cn_l i mi t T( 3)
y0 = f cn_gat e( 2, : ) ;
el se
i f t _now < f cn_l i mi t T( 4)
y0 = [ 1, 0, 1, 0, 1, 0] ;
el se
i f t _now < f cn_l i mi t T( 5)
y0 = f cn_gat e( 2, : ) ;
el se
i f t _now < f cn_l i mi t T( 6)
y0 = f cn_gat e( 1, : ) ;
el se
y0 = [ 0, 1, 0, 1, 0, 1] ;
end
end
end
end
end
end
y = y0' ;
end
end


72

References


[1]
Sun Zhou; Yin Zhongdong; Chen Anyuan; Zhen Xiaoya; , "Application of voltage PWM rectifier in the
charger of electric vehicles based on power feed-forward decoupling control," Electric Utility Deregulation and
Restructuring and Power Technologies (DRPT), 2011 4th International Conference on , vol., no., pp.554-557, 6-
9 J uly 2011.

[2]
Busse Alfred, Holts J oachim, "Multiloop control of a unity power factor fast switching AC to DC converter,"
Proceeding of Powre Electronics Specialist Conference, 1982, pp.171-179.


[3]
Akagi, Hirofumi; Kanazawa, Yoshihira; Nabae, Akira, Instantaneous Reactive Power Compensators
Comprising Switching Devices without Energy Storage Components in Industry Applications, IEEE
Transactions on, pp. 625-630. 1984.

[4]
Green, A.W; Boys, J .T.; Gates, G.F. 3-phase voltage sourced reversible rectifier, in Electric Power
Applications, IEE Proceedings B, Nov 1988, pp. 362 370.

[5]
Dixon, J .W; Boon-Teck Ooi, Indirect current control of a unity power factor sinusoidal current boost type
three-phase rectifier, Industrial Electronics, IEEE Transactions on, Nov 1988, pp.508-515.

[6]
Komurcugil, H.; Kukrer, O.; , "Lyapunov-based control for three-phase PWM AC/DC voltage-source
converters," Power Electronics, IEEE Transactions on , vol.13, no.5, pp.801-813, Sep 1998.

[7]
Chenwei, "Research and realization on direct power control of three-phase voltage-source PWM rectifier",
Huazhong University of scisece & Technology, 2009.

[8]
Marei, M.I.; El-Saadany, E.F.; Salama, M.M.A.; , "A novel control scheme for STATCOM using space vector
modulation based hysteresis current controller", Harmonics and Quality of Power, 2004. 11th International
Conference on , vol., no., pp. 58- 65, 12-15 Sept. 2004.

[9]
J ain, S.; Agarwal, V.; , "A Single-Stage Grid Connected Inverter Topology for Solar PV Systems With
Maximum Power Point Tracking", Power Electronics, IEEE Transactions on , vol.22, no.5, pp.1928-1940, Sept.
2007.

[10]
Xin-yan Zhang; Xiao-bo Zhang; Wei-qing Wang; Suonan J iale; , "The Study of on Grid Wind Turbine
Generator Made in China", Power and Energy Engineering Conference (APPEEC), 2010 Asia-Pacific , vol., no.,
pp.1-4, 28-31 March 2010.

[11]
Ned Mohan, Advanced Electric Drives: Analysis, Control and Modeling Using Simulink, Mnpere, October
2001.

[12]
Ned Mohan, Tore M. Undeland, William P. Robbins, Power Electronics: Converters, Applications, and
Design, Wiley; 3 edition, October 10, 2002.

[13]
Noguchi, T. ; Tomiki, H. ; Kondo, S. ; Takahashi, I. Direct power control of PWM converter without
power-source voltage sensors in Industry Applications, IEEE Transactions on, Volume: 34, Issue: 3, pp. 473
479, May/J un 1998.

[14]
Bouafia, A.; Gaubert, J .-P.; Krim, F, Analysis and design of new switching table for direct power control of
three-phase PWM rectifier, in EPE-PEMC, pp.703 - 709, Sept. 2008.

[15]
Brogan,W. Modern Control Theory, Prentice-Hall, 1985.

[16]
Lars Imsland. Introduction to Model Predictive Control, May. 2007.

[17]
YANG Xingwu, J IANG J ianguo, Predictive Direct Power Control for Three-phase Voltage Source PWM
Rectifiers, Bulletin of Polish Academy of Science, vol. 54, no. 3, pp. 287292, 2006.

73


[18]
Aguilera, R.P. and Quevedo, D.E, On stability and performance of finite control set MPC for power
converters, Predictive Control of Electrical Drives and Power Electronics, pp. 55-62, Oct. 2011.

[19]
Larrinaga, S.A.; Vidal, M.A.R.; Oyarbide, E.; Apraiz, J .R.T.; , "Predictive Control Strategy for DC/AC
Converters Based on Direct Power Control," Industrial Electronics, IEEE Transactions on , vol.54, no.3,
pp.1261-1271, J une 2007.

[20]
Kouro, S.; Cortes, P.; Vargas, R.; Ammann, U.; Rodriguez, J . "Model Predictive ControlA Simple and
Powerful Method to Control Power Converters," Industrial Electronics, IEEE Transactions on , vol.56, no.6,
pp.1826-1838, J une 2009.

[21]
J orge Nocedal, Stephen J . Wright, Numerical Optimization, pp449, J ul 27, 2006.

[22]
Shutuan Zhang, Kai Zhang, Zhongshan J iang and Fang Lu, Modeling and Simulation of Three-phase
Rectifier Based on SVPWM, 2010 Third International Conference on Information and Computing.