See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/280579571

Small Signal Modeling and Control Design of a

Three-phase ACDC Vienna Converter
RESEARCH JULY 2015
DOI: 10.13140/RG.2.1.1929.4561

READS

47

3 AUTHORS:
Nesrine Belhadj Youssef

Kamal Al-Haddad

19 PUBLICATIONS 94 CITATIONS

cole de Technologie Suprieure, Montral,

SEE PROFILE

510 PUBLICATIONS 5,918 CITATIONS

SEE PROFILE

Farhat Fnaiech
Universit de Tunis, Ecole Nationale Suprie
211 PUBLICATIONS 1,311 CITATIONS
SEE PROFILE

Available from: Nesrine Belhadj Youssef

Retrieved on: 25 November 2015

Small Signal Modeling and Control Design of a Three-phase ACDC

Vienna Converter
N. BEL HAJ YOUSSEF, F. FNAIECH, Senior member IEEE and K. AL-HADDAD*, Senior member IEEE

* Ecole de Technologie SupCrieure. Dept.Elect. Engineering, I LOO rue Notre Dame Ouest, H3C IK3 Montrkal Canada
E-mail: kamal, ffarhat, nyoussef@ele.etsmtl.ca
Abstract: In this paper, a small-signal linearized model
based on the state'space averaging method is proposed
for a particular three-phase unity power factor AC/DC
converter known as the Vienna rectifier. The resulting
model is a MIMO linear model. In order to control the
system, two control strategies have been developed. The
first one is based on a decoupling procedure and the
second one is a straightforward looping method. The
expected results of the proposed model and the related
control strategies are in good accordance with their
counterpart using a nonlinear MIMO model and a
feedback linearization control as reported in the
literature, with the main advantage of being easier to
implement.
Simulation results are also provided in order to validate
the theoretical approach carried ont through the paper.
Key Words: Vienna rectifier, unity power factor, dynamic
modeling, state-space averaging technique, multi-loop
control strategy.
I. Introduction

two complementary unidirectional switches were replaced

by a bi-directional one.

A great number of high power electronic interfaces have

been proposed in scientific literature. The traditional
approach used passive filters inserted between the electrical
source and the converter. However, this method suffers
from several disadvantages: susceptibility to resonance,
high fabrication costs, low robustness, etc...

Table -1- gives the states that can he taken by the switches
( Q I , Q ~ , Q ~during
)
the different intervals and their
corresponding values of the voltage V ~ m ( t ) .Due to the
spontaneous character of the diodes conduction, only the
two most possible occurring configurations in each interval
are considered.

Recently, a wide variety of active three-phase filters have

appeared like, for instance, the classical six-switch bridge,
the Minnesota, the current-injection based [I], [7] and the
Vienna converters [I], [2] and [3].

I l l . Small- signal modeling of the converter

Little attention has been given to ACDC Vienna rectifier in

spite of the numerous advantages offered by its topology.
First, it has a reduced number of components, which may
considerably decrease the complexity of structure and
control. Second, the middle capacitive point limits the
electrical constraints on the electronical devices, and
provides two independent stabilized supplies. Therefore,
modeling and control of such a converter is of a great
importance for power electronics designers. A nonlinear
model was proposed in [I], where the authors used a
feedback linearization technique to control the converter. In
this paper, we shall show that a small-signal model exhibits
good performances, too, in term of harmonic elimination
and voltage stabilization.

**

fjfl?\
VEZ."

a2

L2

is2

is3

&

I'0,h

-b
VS3."

a3

L3

-b

V0,i
C

01

Rod
i0.l

Fig.1 Vienna convener ulpology

The nonlinear equations d,:scribing the operation of the

Vienna rectifier have been already developed in literature
[I]. Therefore, for seek of space, they will not be detailed in
the present paper. We will directly use the (d,q.o) frame for
developing the small-signal. model and the control for the
system.

In the reference Park frame, the system state equations are

expressed as follows:

11. converter topology and operation

As shown in fig.1, the Vienna converter is built'by an

association of three identical elementary boosts, connected
each one to a phase of the network. Functionally, it is
equivalent to the classical six-switch converter, in which,

0-7803-7906-3/03/$17.00 02003 IEEE.

656

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.

where : vo=vo,h+vu.f
V0.h and

and

Avli=va.h-vd

With:

d-=

v d are respectively the high and low level voltages.

a is a theoretical constant parameter estimated to be

a =

x-=[ir,d-

v-=

0.64

[dd-

i3,q- v o -

AVa-p

dq- d0-p

[yYr.(- p

The state matrix A , the control matrix B and the disturbance

matrix E are defined by :

(16-18)

dk=l-&sgn(is.k)

,k={1,.?,3)

Wherefis the nonlinear function describing the system and

XO is the chosen static point.
We will consider the case of a balanced load, i.e:

vsl23.n, is121 , 6 1 2 3 are respectively the

voltage, the current and the control variables.

K is the Park transformation matrix defined,by:

Ro,h andRo./are measured off-line.

1 -0.5 -0.5

Applying the identities (l6)-(18), we obtain:

The desired steady state is characterized by:

v o.h
+ = v +o,/ = - E
2

With

Y.'=4&

AVO'= 0

=680V

(10)

IS* is the desired RMS value of the line current. It can be

computed using the power conservation law in steady slate:

3~L'=(v~,hj~,h+v.,,j~,,)

(II)

B. Transfer funclions

The converter model in the static regime is obtained by

replacing all the variables by their constant terms in the
system state equations, which leads to:

In the frequency domain, the dynamic model is expressed

by:

x ~ s ~ s ~ - ~ ~ ' B d ~ s ) + ( s l - A ~ ' (23)

E~s)
Below are givcn the twenty transfer functions relating the
system variables. The numerical values of the different
parameters are given in the appendix.
(W(14)

Note that the control variables allow the control of the

output voltage Vo, the current line I$ and the load unbalance
Avo
A.

Time domain small- signal modeling

The establishment of the small-signal dynamic model ofthe

system is based on linearizing the equations ( I ) around the
static point defined by (6)-(14). Thus, the non-linear state
equations of the system are put in the form:

X-(l)=AX-(/)+Bd-(t)+Ev-(/)

(15)

657

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.

1301

1V. Control strategy design

The

system

inputs

[dd- dq- do-pand

outputs

[is.,- is,q- vo- Av0-P are related by the transfer matrix

defined bv:

[ 4 w s*I
2

In order to obtain a square matrix, we choose to control

i s , d - , i s , q - and AVO- as the three fastest system variables.
The new transfer matrix is:

Gdi

G(di2

0
(42)

Note the existence of an interaction between the control

variables (d'd- , d'p) and the outputs (i s.d- , i ~,q-). To
develop a suitable design, a decoupling is necessaly.

A . Linear decoupling strafegy

(34)

G ~ V Z I (=S-)

Ki,m (L
63") + I )

(p 1)
wpi

(f

WP

+ 25s +

This approach consists of two steps.

In the first one, we have to find a pre-regulator bloc HI($
which aims to diagonalize h e transfer matrix G(s).
Hl(s) is chosen as:

0,

1431

658

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.

we will consider only the range of tkquencies in [ o , f . / 5 ],f.

being the sampling frequency, in order to impose a closed
loop bandwidth near tofslh. In this frequency band, the
following approximations can be made:

(45)- (471

In the useful frequency hand,

following expression:

H t(s)t.

has approximately the

WO

Kr; and Krdvare feedback scaling gains respectively equal to

0.05 and 5/680. These controllers are chosen so that they
guarantee an optimal damping ratio of 0.707, an overshoot of
4.32% and a rise time of 0.05ms.
Besides, the inner voltage loop handwidth must be chosen
such that: wbv < 2xd2fo
Where f. is the supplying network frequency. This
corresponds to an open loop bandwidth less than 0.64fu.
Note that the high frequencies filtering function is
incorporated in the controllers.
Two factors G i v / / c and Giv22c are also added in order to
comDensate the effects of the disturbances K,d- and b y - .
G,,,II(s)- 2
G;~I
i c =---G , ~ I I ( S ) V:

O i

The new system transfer matrix becomes:

The output voltage vo- must be expressed respectively in

terms of the current components i3.d- and k y - in the
frequency band [ 0 ,$/2500 I. As the currents loops are made
much faster than that of the voltage, is,q goes fast to its desired
value i s , p = 0.
Therefore, only the transfer function relating is,d- to vo- is
considerable.
S

K"(,I (%,I

The effects of the perturbations Vs.d- and Vs,q- respectively on

is.y- and ixd- are also negligible. Consequently, Gidz(s) and
Gzvz/(s) can be neglected.
The second step, a post-regulator bloc must be designed to
ensure the control of two independent current loops obtained,
the third loop being already independent.

- 1)

W"d2

(s)

." ..
(571-(58)

An optimal voltage controller is chosen as:

ff

wv=

If"

(59)

K H " = ~Kv=, 5

2500

680
I>*is obtained using the power conservation principle which
becomes in this case:

Where vcv is the control signal delivered by the voltage

controller and:

659

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.

B. Direct looping control strategy

Note that, for the considered frequency band:

IG, ,

( i w ) l > > lGidlz( i w ) l

IGid22(jW)I

lGidZl (jw)l

(62)

(63)

This means that, in steady state, we can neglect the

interactions between the control variables ( d d , d q - ) and the
currents ( k d - , k q - ) . As AVO- only depends on $0-, this is
equivalent to neglecting the transfer functions Gid12is~and

xo

Gid21is).

The system is reduced and we obtain three independent

subsystems. Each one is controlled separately.
The controllers are exactly chosen as in A.

f -tm

-m002

V. Simulation results

tm
003 004 O D s 005

002

003 004 0 . E 006

In order to validate the theoretical results, a numerical version

of the system is carried out using MATLAB Power System
Blockset.
The numerical values of the system parameters and the
nominal operating conditions are indicated below:

VS,,=1 2 n v
Nominal source voltage:
Nominal load voltage:
van = 680V
Nominal power:
P 0 . n = 25KW
Network frequency:
o = 60Hz
Sampling frequency:
s = SOW2
Inductances:
L = ImH, cach
Capacities:
CO= I mF. each
The power elements L and COare chosen such that they
satisfy the two following conditions:
i) The maximum current ripple Ahmm equals 1% of the line
current pick value.
ii) The maximum low frequency ripple of the output voltage
equals 1 0 % of its mean value vd2.
Fig. 2 gives the simulation results conceming the first method.
The obtained THD is about
0.846%and the power factor
is about 0.997. The output voltages oscillate around 340V
with a 9.1 18% oscillation rate.

Fig. 2 Simulation results in normal operating condition far linear decaupling

(At thc tap), and dircct looping (in thc bottom)

Fig. 3 gives the simulation results for the second method. The
THD obtained is about 0.656% and the power factor is unity.
The output voltages oscillate around 340 V with a 8.823%
ripple rate.

l(S1

11s)

Fig2 Responsc of thc systcm to a decrease of the network voltage (a), to M

incrcasc of the network voltage (b) and to a load unbalance (c).

The decrease of the network voltage bd ar t=O.O67s causes the

increase of the line current id and of the THD from 0.656% to
10.03%. The ripple of die output voltage increases kom
0.147%to 2.94%around C17OV which corresponds to a falling
of 1.47%. On the other hand, the increase of the network
voltage Vd at 0.067s leads to the decrease of id.The THD
increases to 6.71%. The output voltage ripple rate goes to
1.18% around 676V which corresponds to a falling of0.59%.

660

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.

A severe load unbalance (Ro,~l=Ro.n:

Rd=Ro,n/2) causes the
increase of the line
current to 150 A. The output voltage
increases also with a 1.47% rate.

Kb,d12=0.019, W.d3=1831
&22=0.12,
-2

In this paper, a dynamic model of a three-phase ACiDC

PWM-controlled Vienna converter has been proposed. The
linearized version of this model has been used to derive the
different transfer functions. Two control strategies, namely
the linear decoupling and the direct looping were used.
Simulation results using MATLAB Power System Blockset
have shown that the two strategies are equivalent in steady
state but the second one is more performant at power factor
and response time. These two methods have the advantage of
being simply implementable and robust to source
perturbations, however they exhibit less robustness to big load
variations. A study carried in [ I ] stated out a close
relationship between the load unbalance and the parameter a
variations: Therefore, works on adaptive control are being
carried out to deal with these inconveniences.

Appendix
-5

rads,

MidZn2.78

rads,

-1

Wpi=44.76 rads, Wp=569 rads, 5 = 5 . 0 3 e ,

-5

Kid12=

-5

l.3& , W;d3=198.66 rads, K;d2/=0.7Iet Wid4=103.81

-3

rads, UMvdl3=204 rad's, Kiv/~=0.52e,

Wvl=86.6 rads,
-5

$VI

I =0.59, Wvz =198.66

rads, K;VIZ=
1.378

-7

, K i d = 3e ,

-2

Wv3=5.37 rad/s, Kiv22=Kid22=

KVV/I=3 . 9 e ,

- 1 . 8 9 ~, 0~,,=1831.09 rads,

VI. Conclusion

K i d i ~ = 4 e, Widl=207.78

rad's,

1 .22 e ,

WidS=420.63

-4

radis, Kw/ll=- 6e , W.dl =1688,29rad/s, dhd=6.01 rads,

Wv4=420.63

rads,

-3

W.vl=77.62 rads, K v v l Z =

&&z =0.117, K,,,,

=313.6

acknowledgments
This work has been supported by M Chaire de recherche du
Canada
en conversion de I'tnergie electrique et de
I'electronique de puissance (CRC-C3EP) n. The first authors
acknowledge the Tunisian ministry of higher education and
technology and the (( Mission Universitaire Tunisienne en
Amerique du Nord )) for their support.
References
[ 1 ] H. Kanaan,"Contribution I la modelisadon et a" rbglags des
converti~se~rstriphases actifs non polluanu unidirectionnels de type
klkvateur et I frkquence de eommulation fixe", Ph-D thesis, ETS, Montrhl,
Canada, 2002.

[ 2 ] J.W. Kolar and F. Slogerer (2001): "A new concept for reeOnltNCtion
of the input phase currenu of a three-phaseiswitchllevel PWM (Vienna)
rectifier based on neutral point current measurements" PESCZOOO, 31"
Annual, I, pp 139-146.
[ 3 ] J. Minnibock, F. Slogerer and J.W. K o l a (2001), "A novel concept for
mains voltage proportional input current shaping of a Vienna rectifier

eliminating controller multipliers ",Applied Power Electronics Conference

and Exposition (APEC2001) 16''Annual. 1.pp.582-591.
[ 4 1 1. Sun, D.M. Mitchell, M.F. Greuel, P.T. Kreinand R.M. Bass, (2001),
"Averaged modeling of PWM conveners operating in discontinuous
conduction mode ", IEEETrans. Power Electronics 16(4), pp.482-492.
[ 5 ] J.W. Lee, E. Nowicki and C. Alfred (2000). "A compulational smallsignal modeling technique far switch mode conveners ", Canadian
Conference on Elsctrical and Computer Engineering, 1, pp.587-591.
I6 I Y. Ye. M. Kazerami and V.H. Ouintana (2001)
,
, .,"A novel modcline and
control method of a three-phase P W M converlers" IEEE PESCi001,
Vancouver, BC.
.

[ 7 ] H. Kanaan, K. AI-Haddad, R. Chaffai, L. Duguay and F. Fnaiech

(2001). "A small signal modcl of a three - phase current-in)cction-basssd
rectifier. IEEE PESC2001, Vancouver, BC.

Table. I

Authorized licensed use limited to: ECOLE DE TECHNOLOGIE SUPERIEURE. Downloaded on February 26, 2009 at 14:16 from IEEE Xplore. Restrictions apply.