Krzysztof KULIKOWSKI1, Andrzej SIKORSKI2

Bialystok University of Technology (1), (2)

Comparison of new DPC methods for two- and three-level

AC/DC converters
Abstract. The paper briefly describes and compares some new methods of Direct Power Control of a two- and three-level AC/DC converter. The
basis for comparing the methods was the equality of switching energy losses in each of the methods.
Streszczenie. Artyku zwile opisuje, oraz porwnuje nowe metody bezporedniej regulacji mocy dwu- i trjpoziomowego przeksztatnika AC/DC
wsppracujcego z sieci. Podstaw porwnania metod bya rwno strat na przeczanie dla kadej z metod. (Porwnanie nowych metod
bezporedniej regulacji mocy dwu- i trjpoziomowych przeksztatnikw AC/DC)

Keywords: DPC, three-level AC/DC inverter.

Sowa kluczowe: DPC, trjpoziomowy przeksztatnik AC/DC.

Introduction
In order to demonstrate the advantages of the application
of more expensive solutions i.e. a 3-level converter in
comparison with a 2-level one, a mere comparison of both
structures using basic methods seems to be inadequate.
New modified methods of direct power control [1],[2],[3],[4] of
an AC/DC converter eliminate some disadvantages of DPC
method [5],[6] and also reduce the frequency of switchings
for the same values of current deformation coefficients. This
paper presents a brief description and comparison of
modified methods of vector control of 2- and 3-level AC/DC
converters.
Mathematical model of converter
The relationships (1) and (2) [7] describe the input
voltage vectors of the converter in stationary reference
frame with respect to three- and two-level configuration [8].

(2)

2
j(n 1 )
3 , dla n {1,2...,6}
U e
U d [n] 3 dc

"0" , dla n {0,7}

where: Udc DC link voltage, n = 0,1,2... , 0 zero vector,

n vector number.
Vectors of current derivative of AC/DC converter
The AC/DC converter showed in Fig. 2 can be described
by relationship (2) in stationary reference frame and
relationship (3) in xy rotating reference frame.
(3)
(4)

E e j 1 t L
EL

d
i U d [ n ]
dt

d
i xy j1i xy U d [ n ] e j1t
dt

where: i vector of current in stationary reference

frame, ixy - vector of current in xy rotating reference frame,
L - inductance of network reactor, E - vector of net voltage.

Fig.2. Schematic diagram of AC/DC converter

Fig.1. Configuration of three-level (a) and two-level (b)converter

(1)

56

U d [n]

j(n 21 )
2
3 , for n {21,22...,26}
U dc e
3

j(n 15 )
3
3 , for n {15,16...,20}
U dc e
3

j(n 3 )
1
3 , for n {3,4,...,14}
U dc e
3
"0" , for n {0,1,2}

Fig.3. Vectors proportional to the vectors of current derivative D[n]

(a), and vectors of current derivative d[n] (b)

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

Dependences (5) and (6) of the above vectors are

described by transformation equation (4). The obtained
vectors of current derivatives determine the direction and
speed of current changes depending on the voltage vectors
of the converter [9].
(5)
(6)

D[ n ] L

By introducing the notion of the converter input voltage

vector, dependence (5) takes the shape of (8).
(8)

D[ n ] L

d
i xy U p U d [ n ] e j 1t
dt

d
i xy E j1 Li xy U d [ n ] e j 1t
dt
d
1
d [ n ] i xy D [ n ]
dt
L

where: D[n] - vectors proportional to the vectors of current

derivative, d[n] - vectors of current derivative.
DPC method
The DPC [5],[6] similarly to the DTC [10] method makes
it possible to control both current vector components by
selecting voltage vector Ud[n] from the voltage vector
selection table. The choice depends on sector N as well as
the current comparator states. The selection of sector N is
determined by the vector angle of virtual flux. The DPC
method assumes that the virtual flux is delayed with respect
to the vector of net voltage E by approximately /2.

Fig.6. Scheme of the DPC 3x2- control method

To adjust the control in relation to Up, voltage vector of

virtual flux should be turned by angle and the current
errors in the xy rotating reference frame - by angle -
respectively. Finally, the control structure of the DPC-3x2-
will take the shape introduced in Fig. 6.
Assuming that iy0, the angle between vectors of
voltages E and Up (Fig. 5) can be described by dependence
(9) [9].
(9)

Fig.4. Schematic diagram of DPC control method

Fig.5. Graphic interpretation of the vector of virtual flux in DPC (a)

and DPC- (b) control method

Li x
E

arctg

Table 1. Voltage vector selection table for DPC 3x2- method

N=1
N =2
N =3
N =4
N =5
N =6
vector number (n)
d x= 1 6
1
2
3
4
5
d y= 1
d x= 0 1
2
3
4
5
6
dx= -1 2
3
4
5
6
1
d x= 1 5
6
1
2
3
4
dy= -1 dx= 0 4
5
6
1
2
3
dx= -1 3
4
5
6
1
2

DPC 5x2- method

The DPC 5x2- [11] algorithm
is , in fact, the
DPC 3x2- method supplemented by the use of the threelevel converter capabilities. The schematic diagram of the
method is shown in Fig. 7.

DPC 3x2- method

The DPC-3x2 method does not to take into account the
voltage drop caused by the passage of current through the
network reactor, which results in voltage shift on the
converter feeder clamps with respect to the vector of net
voltage E by angle . It also has a larger amplitude (see Fig.
5). The shift between these voltage vectors may also bring
about an incorrect sector selection and, in consequence,
deteriorate the quality of control. To minimize this negative
effect we introduce a converter input voltage vector
described by formula (7), to take over the control process
[11].
(7)

U p E j1 Li xy

where: Up converter input voltage vector

Fig.7. Scheme of the DPC 5x2- control method

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

57

Due to an increased number of available voltage

vectors, the number of current comparator levels in axis x
rose from 3 to 5. Additional current comparator levels are
used to control current by means of short current and long
current derivatives in the static and dynamic state
respectively.
Table 2. voltage vector selection table for DPC 5x2- method
N= 1
N= 2
N= 3
N= 4
N= 5
N= 6
vector number (n)
d x= 2
26
21
22
23
24
25
d x= 1
8,14
3,9
4,10
5,11
6,12
7,13
d y= 1
d x= 0
3,9
4,10
5,11
6,12
7,13
8,14
dx= -1
4,10
5,11
6,12
7,13
8,14
3,9
dx= -2
22
23
24
25
26
21
d x= 2
25
26
21
22
23
24
d x= 1
7,13
8,14
3,9
4,10
5,11
6,12
dy= -1
d x= 0
6,12
7,13
8,14
3,9
4,10
5,11
dx= -1
5,11
6,12
7,13
8,14
3,9
4,10
dx= -2
23
24
25
26
21
22

DPC-3A

Fig.10. Graphic representation of optimum error distribution area

As each operation of turning and shifting of the whole

error area would be troublesome, it can be replaced by an
inverse operation performed on the error vector, i.e. shifting
it by vector -Wi and turning by angle - (Fig. 11).

Fig.8. Schematic diagram of DPC-3A method

The DPC-3A [3] method is a modification of DPC 3x2-

analogous to that of DTC-3A. The purpose of this
modification is to decrease the switching frequency with
similar deformation coefficients of phase currents.
As shown in [10], the division of the error area in the
DPC method is not optimal. The determination of the
optimal boundary requires a delimitation of such a dividing
line, where the impact of the two adjacent voltage vectors
are equivalent. The DPC-3A method has been devised to
optimize the division of the error area.
It was shown [10] that the boundaries of a optimum
error distribution area are made up of three half-lines of
common origin, and turned against each other by 120. The
boundaries of error distribution area are shifted by angle
with respect to the x-axis and the common origin of
boundaries are shifted from the origin of the coordinate
system by Wi vector (Fig. 10).

Fig.11. Graphic representation of the equivalence operation of the

turn by angle and shift by vector Wi made on the error area (a)
and the shift by vector -Wi and turn by angle - of the error vector
(b)

Wi vector is proportional to voltage vector W that

constitutes the difference between voltage vector Up and
the centre of delta composed of three voltage vectors (Wb1,
Wb2, 0).
In DPC-3A method the real value of
current is
compared to set values of current in both axes of rotating
reference frame. Error vector is thus created from x,y and
subsequently transformed as shown in Fig. 11b, i.e. vector
Wi is subtracted from error vector and then turned by
angle [10].
(10)

" ( Wi ) e j " e j

Vector Wi is related to vector by formula (11) [10].

(11)

Wi

W
Tp
L

where: Tp sampling time.

W vector is calculated using formula (12) as shown in
Fig. 9.
(12)

Fig.9. Graphic interpretation of the vector W in DPC-3A method

58

W S U

where S is the vector of the centre of the equilateral delta

built of three voltage vectors characterized by the shortest
current derivatives in actual N sector.

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

Vector S can be calculated using formula (13) [10].

(13)

1
S ( Wb1e j Wb2 e j )
3

where: Wb1, Wb2 currently used voltage vectors Ud[n]

(Fig. 9)
Table 3. voltage vector selection table for DPC-3A method

(0,2/3)
(2/3, 4/3)
(4/3, 2)

N= 1

N= 2

N= 3 N= 4 N= 5
vector number (n)
5
6
1
0 (7)
4
5
6

N= 6
2
1

DPC-3L-3A
The DPC-3L-3Am [12] (Direct Power Control 3 Level
3 Area with modifications) method is a modification of
DPC-3A method making use of an increased number of
active vectors of three level converter.
Fig.14. Error distribution area in triangles 0 (a), I (b), II (c), III (d) in
sector N = 1

Fig.12. Scheme of the DPC-3L-3A method

In three-level converter each N sector can be divided

into four deltas that, in turn, are divided into 3 voltage
vectors Ud[n] (Fig. 13). In each of the equilateral triangles
formed by the voltage converter an optimal error area (Fig.
14) analogous to of DPC-3A method can be determined.

Fig.15. Error distribution area in triangle II before (a) and after (b)
the change of the sign of component x of the error vector in sector
N=1

In DPC-3L-3A control method the real value of current

is compared to set values of current in both axes of
rotating reference frames. Error vector is thus created out
from errors x,y and subsequently transformed in
agreement with formula 14, i.e. vector Wi is subtracted from
error vector and then turned by angle . In the case when
the end of converter input voltage vector Up is in delta II the
sign of component x is changed into the opposite [12].
"
(14) e

Fig.13. The vectors of the voltage creating equilateral triangle in the

sector N = 1 for two-level (a) and three-level (b) converters

As can be seen in Fig. 14, error distribution areas in

triangles 0, I, and III are the same. Also the error area in
triangle II can by expressed in a similar form, however it
requires a change of the sign of component x of the error
vector into an opposite one (Fig. 15). Ultimately, DPC-3am
control system for the three-level converter takes the form
shown in Fig. 12.

( j y Wi ) e j , for deltas 0, I, III

x
j
( x j y Wi ) e , for delta II

Vectors Wi and W can be calculated from formulas

(11),(12), similarly to DPC-3A method. Vector S is
dependent on delta where the converter input voltage
vector Up is currently used. The vector is described by
formula (15) [12].

(15)

1 3 ( Wb0 Wb1 Wb2 ), for triangle 0

1 3 ( W W W ), for triangle I

b1
b3
b4
S
W
W
W
1
3
(

b1
b2
b4 ), for triangle II

1 3 ( Wb2 Wb4 Wb5 ), for triangle III

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

59

Table 5. Results of simulations

Set U
V
dc
value

Table 4. voltage vector selection table for DPC-3L-3A method

N= 1 N= 2 N= 3 N= 4 N= 5 N= 6

vector number (n)

triangle
0
5,11 6,12 7,13 8,14
3,9
4,10
I
16
17
18
19
20
15
0; 23
II
5,11 6,12 7,13 8,14
3,9
4,10
III
23
24
25
26
21
22
0
0,1,2 0,1,2 0,1,2 0,1,2 0,1,2 0,1,2
2 4
, 3
I
4,10 5,11 6,12 7,13 8,14
3,9
3
II
16
17
18
19
20
15
III
5,11 6,12 7,13 8,14
3,9
4,10
0
4,10 5,11 6,12 7,13 8,14
3,9
4 , 2
I
22
23
24
25
26
21
3
II
4,10 5,11 6,12 7,13 8,14
3,9
III
16
17
18
19
20
15

Ix(puls)RMS

Ia(puls)RMS

I*x

Iy(puls)RMS

Fig.16. Graphic interpretation of the vector W in DPC-3L-3A control

method

Method
DPC 3x2-
DPC-5x2-
DPC-3A
DPC-3L3A
DPC-3x2-
DPC-5x2-
DPC-3A
DPC-3L3A
DPC-3x2-
DPC-5x2-
DPC-3A
DPC-3L3A
DPC-3x2-
DPC-5x2-
DPC-3A
DPC-3L3A

800

1400

10

20

30

10

20

30

A
A
A
A
A
A
A
A
A
A
A
A
kHz
kHz
kHz
kHz

0,375
0,371
0,378
0,361
0,741
0,366
0,343
0,340
0,456
0,421
0,426
0,383

0,393
0,391
0,391
0,371
0,560
0,377
0,346
0,357
0,502
0,475
0,422
0,437

0,369
0,371
0,359
0,349
0,552
0,370
0,328
0,356
0,551
0,501
0,379
0,371

0,972
0,961
0,966
0,872
1,016
0,856
0,849
0,940
1,200
1,059
1,036
0,794

0,952
0,901
0,941
0,915
0,975
0,821
0,835
0,951
1,232
1,021
1,031
0,873

0,983
0,974
0,974
0,915
1,054
1,007
0,845
0,845
1,311
1,116
1,079
0,979

60
82
44
22

60
74
41
23

60
69
41
24

60
74
30
11

60
72
31
11

60
64
35
10

Comparison of DPC control methods

The simulations were performed in Matlab Simulink
environment for two values of DC link voltage i.e. 800 V and
1400 V, and for three set values of current i.e.10, 20 and
30A. The inductance value of network reactors used in the
simulation amounted to 20 mH. The basis for comparing the
methods in the static state was the similarity of the
coefficient of network phase current deformations described
by dependence (16). The values determined in simulations
performed for DPC 3x2- method at 60 kHz switchings
were adopted as reference values. To compare the
methods coefficients of current deformations in xy rotating
reference frame were determined and defined by formulas
(17), (18).
T

(16)

I a( puls )RMS

1 o
( ia ia1 )2 dt
To 0

where: Ia(puls)RMS rms value of all current harmonics

obtained by subtracting the first harmonic from the actual
value of phase current, To net voltage period, ia actual
value of phase current, ia1 actual value of the first
harmonic of phase current.
T

(17)

I x( puls ) RMS

1 o
( I x I *x )2 dt
To 0

(18)

I y ( puls ) RMS

1 o
( I y I *y )2 dt

To 0

where: Ix(puls)RMS, Iy(puls)RMS rms values of current

components in xy rotating reference frame after subtracting
the set values.

60

Fig.17. Time courses of current vector components in xy rotating

reference frame in the dynamic state for methods: DPC 3x2- (a),
DPC 5x2- (b) DPC-3A (c), DPC-3L-3A (d)

It be should emphasized here that in order to correctly

interpret the simulation results, a tree-level converter was
used in both DPC 5x2- and DPC-3L-3A methods. The

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

switching of the switching devices in three-level converters

occur at twice lower voltage than in two-level ones. As a
result, switching energy losses in tree-level converters
operating at the same switching frequency is twice smaller.
It follows from Table V that the use of the DPC-3A
method reduces the number of two-level converter
switchings by approximately 27-50% compared with the
DPC 3x2- method.
The use of the DPC 5x2- method will cause an
increase of the number of switchings by about 7-37% when
compared to the DPC 3x2- method. Keeping in mind that
all switchings take place at twice lower voltage, switching
energy losses when using this method, in most cases,
should be a little lower than with the DPC-3A method.
Basing on the simulation results ( see: Table V), we can
state that the DPC-3L-3A method is characterized by the
lowest switching frequency.
In the dynamic state the methods were compared on
the basis of the time courses of current vector components
in xy rotating reference frame during step changes of the
set value of current component in axis x from 30 A to -30 A
(Fig. 17).
All the methods in the dynamic state have shown similar
properties. This is due to the fact that each of the presented
control methods requires the identification dynamic states
where voltage vectors giving the longest current derivatives
are selected.
Conclusion
The results of simulations have shown that using new
modified methods of Direct Power Control in two- as well
three-level converters is more advantageous. In static
states, with similar values of current deformations, it is
possible to operate converters with lower frequency
switchings than in DPC methods for the same converter
configuration, thus reducing switching energy losses.
In some applications using two-level converters
controlled by the DPC-3A method would be more costeffective than using more expensive three-level solutions
controlled by the DPC 5x2- method. It was the
DPC-3L-3A method that showed the lowest switching
frequency among all the other methods compared.
The dynamics of all the presented control methods is
comparable owing to the introduction of the identification of
the actual state and the use of voltage vectors giving
longer current derivatives.

REFERENCES
[1]

G. Escobar, M. F. Martinez-Montejano, R. E.
T o r r e s O l g u i n , A . A . V a l d e z , An adaptive direct power
control for three-phase pwm rectifier in the unbalanced case,
Power Electronics Specialists Conference 2008, June 2008,
3150-3155.
[2] A . B a k t a s h , A . V a h e d i , M . A . S . M a s o u m , Improved
Switching Table For Direct Power Control Of Three-Phase
PWM Rectifier, Australasian Universities Power Engineering
Conference 2007, Australasian Universities, December 2007,
1-5.
[3] K . K u l i k o w s k i , A . S i k o r s k i , Sterowanie przeksztatnika
AC/DC wsppracujcego z sieci metod DPC-3A, Przegld
Elektrotechniczny, 88 (2010), No.4, 130-134.
[4] J . E l o y - G a r c i a , S . A r n a l t e s , J . L . R o d r i g u e z A m e n e d o , Extended direct power control for multilevel
inverters including DC link middle point voltage control, Electric
Power Applications, No.4, July 2007, 571-580.
[5] S . V a z q u e z , J . A . S a n c h e z , J . M . C a r r a s c o , J . I .
L e o n , E . G a l v a n , A Model-Based Direct Power Control for
Three-Phase Power Converters, IEEE Transactions on
Industrial Electronics , 55(2008), No. 4, 1647-1657.
[6] M . M a l i n o w s k i , M . P . K a z m i e r k o w s k i , S . H a n s e n ,
F . B l a a b j e r g , G . D . M a r q u e s , Virtual-flux-based direct
power control of three-phase PWM rectifiers,
Industry
Applications, IEEE Transactions on , 37(2001), No. 4, 10191027.
[7] M . K o r z e n i e w s k i , Nowe algorytmy bezporedniej regulacji
momentu i strumienia silnika indukcyjnego zasilanego z
trjpoziomowego przeksztatnika DC/AC, Rozprawa Doktorska,
Politechnika Biaostocka, Biaystok 2009.
[8] B . - R . L i n , T . - Y . Y a n g , Three-phase AC/DC converter with
high power factor, Electric Power Applications, IEE
Proceedings, 152(2005), No. 3, 757- 764.
[9] A . S i k o r s k i , Problemy minimalizacji strat czeniowych w
przeksztatnikach AC/DC/A PWM zasilajcych maszyn
indukcyjn, Wydzia Wydawnictw i Poligrafii, Biaystok 1998.
[10] A . S i k o r s k i , Bezporednia regulacja momentu i strumienia
silnika indukcyjnego, Oficyna Wydawnicza Politechniki
Biaostockiej, Biaystok 2009.
[11] K .
Kulikowski,
A.
S i k o r s k i , Regulacja mocy
trjpoziomowego przeksztatnika AC/DC wsppracujcego z
sieci, Przegld Elektrotechniczny, 86 (2010), No. 2, 279-284.
[12] K .
Kulikowski,
A.
Sikorski,
Sterowanie
trjpoziomowego przeksztatnika AC/DC wsppracujcego z
sieci metod DPC-3L-3Am, XIII Sympozjum Podstawowe
Problemy Energoelektroniki, Elektromechaniki i Mechatroniki
2009, Polska, Wisa 2009.
Authors: mgr in. Krzysztof Kulikowski, Bialystok University of
Technology, Division for Power Electronics and Electric Drivers, ul.
Wiejska 45D, 15-351 Biaystok, E-mail: [email protected];
dr hab. in. Andrzej Sikorski, Bialystok University of Technology,
Division for Power Electronics and Electric Drivers, ul. Wiejska
45D, 15-351 Biaystok, E-mail: [email protected].

PRZEGLD ELEKTROTECHNICZNY (Electrical Review), ISSN 0033-2097, R. 87 NR 1/2011

61