Control of Circulating Current in Parallel Three-phase

Boost Rectifiers *
Zhihong Ye, Dushan Boroyevich, Jae-Young Choi, and Fred C. Lee
Center for Power Electronics Systems (CPES)
The Bradley Department of Electrical and Computer Engineering
Virginia Polytechnic Institute and State University
Blacksburg, VA 24061-01 11

-
Abstract In this paper, an averaged model predicting the
dynamics of a zero-sequence current in parallel three-phase
converters is developed. Based on the model, a new control
variable associated with space-vector modulation is introduced.
A strong zero-sequence current loop is designed to control and
suppress the circulating current. Simulation and experimental
results validate the developed model and the proposed control
scheme.
I. INTRODUCTION
The use of parallel power converters, particularly parallel
DC/DC converters, has become more common in the past
decade. However, the use of parallel three-phase converters
In section 11, an averaged model of the zero-sequence
current in two parallel three-phase boost rectifiers, as shown
in Fig. 1, is developed. Based on the model, section I11
introduces a new control variable, which is associated with
space-vector modulation. Then a zero-sequence current
control scheme is proposed. The control scheme is designed
within an individual converter in order to have modular
design. Section IV shows some simulation and experimental
results to validate the developed model and the proposed
control scheme. Section V summarizes the major
contributions of thls work and discusses ideas for future work.
idc.?
_ - Vdc
I I I - ? I I

has not yet been explored extensively. One distinguishing

r
Eq
feature when paralleling three-phase converters is a potential
zero-sequence circulating current [ 1- 121. To avoid the
circulating current, the following three approaches are used
commonly with present technology: Source
1. Isolation. Separate AC or DC power supplies [2], [ 111, or a
transformer isolated AC side [3], [ 5 ] is configured for the
-
overall parallel system. In this approach, the overall
parallel system is bulky and costly because of additional
io -i 2 l

Y idcl II -
power supplies or the AC line-frequency transformer.
2. High unpedance. Inter-phase reactors are used to provide
high zero-sequence impedance [l], [6]. However, the
reactors provide high impedance only at medium and high
frequencies. They cannot prevent a low-frequency
circulating current.
3. Synchronized control. This approach basically treats the
parallel converters as one converter [4-51, [7-lo]. For Fig. 1 A directly parallel three-phase boost rectifier system.
example, two parallel three-phase three-leg converters are
controlled as a three-phase six-leg converter. This approach
is not suitable for modular converter design. When more 11. MODELING
OF THE ZERO-SEQUENCE CURRENT
converters are in parallel, the system becomes very
complicated to design and control. A traditional modeling approach for a three-phase boost
This paper proposes a parallel system where individual rectifier is to transform stationary variables into rotating
converters connect both AC and DC sides directly without coordinates. Zero-sequence. components, such as zero-
additional passive components. The direct connection would sequence voltage, are not reflected in the model because they
reduce size and cost. Meanwhile, the parallel converters are do not affect control objectives, such as input line currents
controlled independently to facilitate modular design. and output DC voltage. In order to model the zero-sequence

This work was supported by the Office of Naval Research (ONR), and made use of the ERC shared facilities supported by the National Science Foundation
(NSF) under award number EEC-9731677.

0-7803-5864-3/001$10.00 0 2000 IEEE 506

current for the parallel rectifiers, a phase-leg averaging 5 Vdc
technique is used [ 121, as illustrated in Fig. 2, where the low
case symbols represent instantaneous variables, and the upper
case symbols represent averaged variables.
Fig. 2(a) shows one totem pole phase-leg of the rectifier. It
has a current source in one side and a voltage source in the
other. Both current and voltage are assumed continuous. Fig.
2(b) shows the pulse-width modulation (PWM) of the
ia
1
switches, where do is defined as the duty cycle of the top
switch sap, while San is complementary with sap. The
San T O
corresponding voltage and current relationships are also
shown in Fig. 2(b). Based on these relationshps, the averaged
model of the phase-leg is depicted in Fig. 2(c).
Applying phase-leg averaging to all three legs of the
I
1

Sap I
rectifier, an averaged model of the rectifier is developed, as
shown in Fig. 3.
For a carrier-based PWM rectifier, the duty cycles do, db
and dc are sinusoidal in steady state under balanced condition.
Therefore, the sum of da, db and dc is zero. A space-vector
modulated rectifier, however, usually has triple harmonics in
order to reduce switching losses, increase maximum
modulation index and improve waveforms total harmonic
distortion (THD). Therefore, the sum of the duty cycles is not
equal to zero, and it is defined as a zero-sequence duty-cycle
dz:
+ +
dz = dn db dc . (1)
From (I), the following equation can be easily derived:

(do - dA)+ (db- dA)+ (dc- dA)= 0. (2)

That is
do' + db' + dc' = 0 , (3) (b) Switching pulses and the relationships between input and output
variables.
where do', db' and dc' are

do'= do -dA,
db' db-dA, dc' dc- d A .
= = (4)

Therefore, da, db and dc can be expressed as

-
da=da'+dA, db=db'+dA, dc=dc'+dA. (5)
As a result, Fig. 4 shows the averaged model of the three- (c) An averaged phase-leg model.
phase rectifier with zero-sequence components. For a single Fig. 2 A totem pole phase-leg averaging technique.
rectifier, the sum of l a , 16 and IC has to be zero because there
is no zero-sequence current path. Although a zero-sequence

x
voltage dz. vdc exists in the converter, it does not affect the
input currents and output voltage control.
Idc vde

With two rectifiers in parallel, a zero-sequence current path

is formed and a circulating current may occur. Fig. 5 shows
the averaged model of the two parallel three-phase rectifiers.
In Fig. 5, the zero-sequence current Io is defined as
1 0 = 1 0 1 = 1014- Ibl + IC]
= - 1 0 2 = -(Io2 + Ib2 + 1c2) . (6) Fig. 3 The averaged phase-leg model of a three-phase boost rectifier.

507
 dz .(Ia i-
Ib -I-I4/3
Fig. 4 The averaged model of a three-phase PWh4 boost rectifier with zero-sequence components.

dzl * (In1 + I b l + Ic1)/3

Fig. 5 The averaged model of the parallel rectifiers based on phase-leg averaging.
In the AC side, there are three loops forming three equations:

508
where RI and R2, which are not identified as a separate circuit
element in Fig. 5 , are equivalent series resistors (ESRs) of the
inductors Lt and Lz, respectively.
Summing up (7), (8) and (9), and using (3) and (6), the
following equation can be derived:
dI o
vdc . (dzl - d z ~=) (LI+ L2) -+ (RI+ R2) .I O. ( 10)
dt
Equation (10) describes the dynamic of the zero-sequence
components. As a result, the averaged model of the zero-
sequence dynamic is developed, and is depicted in Fig. 6.
L l + LZ Ri+R2
Although different SVM schemes have the same dl and d2
in the synthesis of a reference vector, dz can differ. The
distribution of the zero vectors can vary without affecting the
control objectives, such as the input AC currents and the
output DC voltage. This indicates that dz can be controlled by
the distribution of do. Based on t h s idea, a new control
variable k is introduced as follows:
dPPP
k=-,
do
where dppp is the time period for applying the zero vector
ppp, as illustrated in Fig. 8.

A kdo?
Io

I J

Fig. 6 An averaged model of the zero-sequence dynamic.

: PPP
Fig. 8 The definition of the new control variable k.
111. CONTROL OF THE ZERO-SEQUENCE CURRENT Usually, for the scheme with alternative zero vectors,
k0.5,as shown in Fig. 7. With the definition in (13), (12)
Different space-vector modulation (SVM) schemes can be rewritten as
produce different triple harmonics, and therefore have
different duty cycles da, db and dc. For example, Fig. 7 shows dz=dn+db+dc=3kdo+d1+2d~. (14)
one PWM pattern of the SVM scheme with alternative zero Then, the difference of dz of the two rectifiers is expressed
vectors. The da, db and dc are the duty cycles of sap, sbp and as
scp, respectively.
dz1 - dz2 = 3do(kl- k2). (15)
d o = d , + d i + d 2 , d b = d d + d 2 , d c = d , , (11) As a result, the new averaged model of the zero-sequence
dynamic with the new control variable k is shown in Fig. 9.
where dl and d2 are the duty cycles of the active vectors pnn
and ppn, respectively, and do is the total duty cycle of the zero Ll+ L2 R I + R2
vectors ppp and nnn.
___, M

-
do -
di d2_ & d2 d i -
do Io
4 2 2 2 2 2 4

Sap

Fig. 9 The averaged zero-sequence model with the new control

-
variable k.

SCP Since it is a first-order system, the control bandwidth of the

zero-sequence current loop can be designed to be very high
Pnn PPn PPP PPnPnn nnn and a strong current loop suppressing the zero-sequence
Fig. 7 The duty cycle relationship between phase-legs and space current can be achieved.
vectors. One rectifier needs only two current sensors to implement
In this case, dz is power factor correction, because the sum of the three line
currents is always zero. With two rectifiers in parallel, three
dz = do + d b + d c = l . 5 d 0 + dl + 2d2. (12) current sensors are needed in order to obtain the zero-

509
sequence current. Fig. 10 shows the implementation of the
zero-sequence current control. In a two-parallel converter
system, it is sufficient to control one of the two converters
since there is only one zero-sequence current. The shaded
block is the zero-sequence current controller added onto the
other control parts of the rectifier.
This control scheme is advantageous over the one that
treats the parallel converters as one converter. Since this
scheme is implemented within the individual converter and
does not need any additional interconnected circuitry, it
allows modular design.
frequency circulating current exists in the system. The io1 and
io2 are zero-sequence currents for rectifiers 1 and 2,
respectively. The circulating current causes distorted input
line currents iai and ia2, as shown in Fig. ll(b). By applying
the zero-sequence current control, the waveforms in Fig. 12
show that the circulating current is almost gone. Only high-
frequency current ripples still exist, and they can be easily
attenuated.

id~2 Vdc

. I . .
Fig. 1 1 Simulated waveforms without zero-sequence current control
6 ~ ~ kH~,fsw~=16
3 2 ~Hz).

Fig. 10 The implementation of the common-sequence control.

Iv. SIMULATION AND EXPERIMENTAL

RESULTS

The simulation model was developed using SABER.

Without the zero-sequence current control, any discrepancies
between two rectifiers (different switching frequencies,
different power stage parameters or different switching
deadtime, for example, each of whch was demonstrated in
simulation) may cause a large circulating current.
Fig. 11 shows two rectifiers with switching frequencies of
32 kHz and 16 kHz. Fig. Il(a) shows that a significant low-

510
 Tek 50 O k S / S 4 Acqs
I r 1

5 A Idiv
i
1 I
t i I

. . .

o s o m om oms 0- ID.I o s om ex OJLI on a m D.O m am o m

m 01
4.1 (a) The zero-sequencecurrents.
(b) The input phase currents.
Fig. 12 Simulated waveforms with zero-sequence current control TeK so.oks/s
I
3 ACqS
t 1
cfsw/=32kHz,fsw2=16WZ).
5A/div I t

The experimental results are shown in Fig. 13 and Fig. 14. ii

A two-parallel three-phase boost rectifier breadboard system
was built and tested. Fig. 13 shows the zero-sequence currents
and the line currents without zero-sequence current control,
whereas Fig. 14 shows the waveforms with control. Due to
non-uniform practical conditions (such as different delays of
clock signals), the experimental waveforms are less uniform
than simulation waveforms. Also, three noticeable ripples in
Fig. 14(a) are due to distortion introduced by zero-crossing cn1 l o o m v o &B l o om v n bl5 . O O m r Chl 1 1.amv

detection.
Tek SO O k S A 2 ACqS (b) The input phase currents.
15A Idiv
I 7
I - 1
" " " " " " "

Fig. 14 Experimental waveforms with zero-sequence current control

(fsw/=32kHz,fs~r=16kHz, unsynchronized).

V. CONCLUSIONS

This work has developed an averaged model to predict

zero-sequence dynamics in two parallel three-phase boost
rectifiers. To control the zero-sequence current, a new control
variable associated with space-vector modulation was
(a) The zero-sequence currents. introduced. Since the zero-sequence dynamic is a fist-order
system, a h g h bandwidth control loop was designed to
TeK 50 Oks&
I
5 ACqS
-T __ 1 effectively suppress the circulating current. Both simulation
1 SAIdiv ''''1''" " " " " " " ' . "
and experimental results validated the proposed control
scheme. The implementation requires only one additional
current sensor. The control algorithm can be easily
programmed in a digital signal processor (DSP).
This modeling approach and control concept can be
generalized for any parallel multi-phase converters, such as
fill bridge rectifiers and inverters, three-phase three-leg
t 1 10.0mVR
ia 2

mm
I
1 0 Oinvn
i
M 5. 0 0 m s C h l 1
: I
1 .6m
rectifiers and inverters, and three-phase four-leg rectifiers and
inverters. These converters cover most medium and high
power applications, such as motor dnves, AC power supplies
(b) The input phase currents. and DC power supplies. The generalization of this concept
Fig. 13 Experimental waveforms without zero-sequence current will be reported in a separate paper.
control (fsw/=32kHz,fswr=l6kHz, unsynchronized).

51 1
 REFERENCES
[I] K. Matsui, "A pulse width modulated inverter with parallel-
connected transistors by using sharing reactors," IEEE Industry
Applications Society Annual Meeting, pp. 1015- 1019, 1985.
[2] Takao Kawabata and Shigenori Higashino, "Parallel operation
of voltage source inverters," IEEE Transactions on Industry
Applications, vol. 24, no. 2, pp. 281-287, MarcWApril 1988.
[3] Juan W. Dixon and Boon T. Ooi, "Series and parallel operation
of hysteresis current-controlled PWM rectifiers," IEEE
Transactions on Industry Applications, vol. 25, no. 4, pp. 644-
65 1 , JulyIAugust 1989.
[4] S. Ogasawara, J. Takagaki, H. Akagi, "A novel control scheme
of a parallel current-controlled PWM inverter," IEEE
Transactions on Industry Applications, vol. 28, no. 5, pp. 1023-
1030, September/October 1992.
[SI Yoshihiro Komatsuzaki, "Cross current control for parallel
operating three-phase inverter," 25th Annual IEEE Power
Electronics Specialists Conference, pp. 943-950, 1994.
[6] Y . Sato and T. Kataoka, "Simplified control strategy to
improve ac-input-current waveform of parallel-connected
current-type PWh4 rectifiers," IEE Proceedings on Electric
Power Applications, vol. 142, no. 4, pp. 246-254, July 1995.
512
[7] L. Matakas Jr. and E. Masada, "Analysis of the parallel
connection of 3-phase VSC converters," International Power
Electronics Congress (IPEC), pp. 854-859, Yokohama, 1995.
[8] L. Matakas Jr. and W. Kaiser, "Low harmonics, decoupled
histeresis type current control of a multi-converter consisting of
a parallel transformerless connection of VSC converters," IEEE
Industry Applications Society Annual Meeting, pp. 1633-1640,
New Orleans, Louisiana, October 5-9, 1997.
[9] S. Fukuda and K. Matsushita, "A control method for parallel-
connected multiple inverter systems," Power Electronics and
Variable Speed Drive, September 1998, Conference
Publication No. 456, pp. 175-180, IEE 1998.
[IO] R. Abe, Y. Nagai and K. Tsuyuki, "Development of multiple
space vector control for direct connected parallel current source
power converters," Proceedings of the Power Conversion
Conference, vol. I , pp. 283-288, Nagaoka, 1997.
[I 13 C. S. Lee, et al, "parallel UPS with an instantaneous current
sharing control," Proceedings of the 24th Annual Conference
of the IEEE Industrial Electronics Society, Vol. 1 , pp.568-573,
1998.
[ 121 Kun Xing, "Modeling, analysis and design of distributed power
electronics system based on building block concept," Ph.D.
dissertation, Virginia Polytechnic Institute and State
University, May 1999.