Você está na página 1de 6

115

MODELLING AND CONTROL OF SINUSOIDAL PWM RECTIFIERS

S. Fukuda, Y. Iwaji, and T. Aoyama

Department of Electrical Engineering, Hokkddo University, Japan

PWM rectifiers are promising because h


te?! can supply DC power while keeping
a sinusoidal AC current with unity in the fundamental power factor. There are two
types, a vdtage-fed rectifier and a current-fed rectifier. The former is required to
simulta-sly control both the fundamental power factor and DC voltage, while the
latter does 80 for both the fundamental power factor and DC current.
State feedback control is essentially suited for multi-input and multi-output systems
such as those. Here a di8crete-tjme optimal regulatoy is applied to their control
because it provides a microprocessor-based robust feedback system without steady-state
errors in response to a step reference and/or disturbance change.
To obtain a mathenatical model, only the fundamental component of the rectifier
switchblg function is consldered. The obtained nonlinear equations are linearized and
then modified in the form that the regulator requires.
Flndy, the regulator is implemented using a digital-signal-processor. Experimental
results demonstrate the validity.

Kevwords. Rectifier, PWH. Power factor, State feedback contrd, Optimal regulator

1. INTRODUCTION change, and the effects of a time-lag in the output


due to the computation by DSP are taken into accoqnt.

An increase in current harmonics and a decrease in the Finally digital-signal-processor (DSP)-controlled


displacement power factor in AC power lines generated setups were buat and the validity of the proposed
by diode or thyristor rectifiers are becoming serious method w a s demonstrated by experimental results.
problems. To solve these prob1ems sinusoidal PWM
rectifiers are promising because they oan supply DC
power while keeping sinusoidal AC current with unity 2. MATHEUTICAL MODEL OF VOLTAGE-FED RECTIPIER
in the fundamend power fact". Furthermore, as
their power flow is bilateral, they can convert DC
power to AC pawer as well. Fig. 1 shows a voltage-fed rectifier. BY transforming
the voltages and currents on the AC side into the d-q
T h e r e are two t y p e s , a v o l t a g e - f e d r e c t i f i e r wordinate which is rotating with the source angular
(VFR)(U-O) and a current-fed rectifier (CFR)(4H6). frequenby,w, the voltage equations are given by(q)
VFR Is required to simultaneously control both the DC
voItage and the fundamental power factor, while CFR is LP i d = - R i d m L i , - v , + e d
required to simultaneously control both t h e DC current (1)
LP i .=--wL id-R i *-v,+e,
and t h e fundamental power Pactor. So far P-I regula-
tors have been used for rectifier control (lHZ).
But as these two controlled variables are coupled with Decide the d and q axes 8s @=E (rms value of the
each other, designing the gains of P-I regulators maips voltage) and eq=O, respectively. Then if id and
requiras empirical knowlbdge. 4 are controlled so that they are constant and zero,
respectively, sinusoidal AC current with unity in the
S t a t e feedback control is, however, essentially suited fundamental power factor will be accomplished. By
for systems with multi-input and multi-output. In assuming that the DC load consists of RJ and LI, the
this paper, a discretetime optimal linear quadratic voktage equations on t h e DC side will be
integral regulator (LQI regulator) i s employed. This
is because an LQI regulator provides the following
desirable features:
*guarantee of a stable system,
*insens,itivity to s a d parameter variations,
*no steady-state errors in response to 8 atep
cbange in the referawe and/or disturbance, The rectifier Input terminal voltages are related to
*a flexible control system based on the DC voltage through t h e switching functions, Sd and
a microprocessor, Sq (see the definition of the switching function in
*consideration of a time-lag in the output due to Section 4) as:
microprocessor-based control.
In particular, as the DC load varies, robustness i s a 2 / z v d = s d v d ,, 2 / z v , = s . v * , (3)
very important characteristic for rectifier mptrol-
lers. hjol'eove?, if the s t a t e equations of a system The rectifier input voltage Waveforms vdr and vpr are
are known, determining feedback g+s is easy since a pulse-shaped bat as harmonic components influence
systematic approach to the r e w l a t o r design exists. little the average power they are neglected. This is
equivalent to considering only the DC component of the
In this paper, a control s y s t e m design for both VFR
and CFR is presented. First, modeling of the recrtifi-
ers is described. By consMering only the fundamental
components of voltages or currents at t h e rectifier
input terminals, 4-th and 5 t h order nonlinear' equa-
tions with 2-input Z'output are obtained for CFR ahd
VFR, respectively. Then t h e equations are linearized
around an operating point so that they are suitable
for an optimal regulator.

Second, a discretetime LQI regulator is applied to


the rectifier control, where an integral action 1s
introduced SD a s not to leave steady state errors in
response to a step reference and/or disturbance Fig. 1. PWM vdtage-fed rectifier.

01993 The European Power ElectronicsAssociation


116

switching function. Sd and Sq are the input variables and twooutput.


controlling the DC voltage and power factor. Actual- An LQI regulator minimizes the following cost func-
ly, however, the modulation index, M, and phase angle, tion:
Q , are used for VFR control. The relations among
them are('): ~ = ~ [ x ~ ( j ) ~ Q X ~ ( l ) + A U ( f ) ~ H h U(9)( j ) l

M = d m , ?b=wt+6 where the diagonal matrices Q and H denote weights for


lohere d=ton-l& (4) the output and input, respectively. Denote & and
Qv& a s the weight values for iq and Vdc, respective-
S d
ly; and Hsd and Hsq for s d and Sq, respectively.
Linearizing (1) and (2) after the substitution of (3) Responses are adjustable by changing the weight
into (2). the state equations will be summarized in values, i.e., larger weights correspond to smaller pe-
the form: sponses. For example, if a larger value of is a
chosen, the transient variation in in becomes smaller.
PX ( t ) =AvX( t ) +BvU ( t ) The weight values are decided by simulation-based
examinations of transient responses.
l o h m e X ( t ) = [ i d i. vsc z , p , (5)
U(t)=[S,S.1' The input of an LQI regulator that minimizes the cost
function will be in the feedback form:
where X ( t ) and U ( t ) denote small deviations from the
operating point. The matrices AV and Bv are shown in AU(&)
=-[H+BorPBoI-'BoTPAoXo ( k )
the Appendix. Sd and Sq are the input variables, and =FXo(k) (10)
i, and Vdc are the output variables. By controlling
vdr and vqr through Sd and Sq, the phasor relationship =~.e ( k ) + F x A X ( k ) + F u A W ( k )
shown in Fig. 2 is maintained a t a steady state. The
output vector Y ( t ) is given by where F is the feedback gain vector and P is the
semi-definite matrix which satisfies the following
Y (t ) = C X ( t ) = [ i V d 2 Riccati equation:

met=[ 0 1 0 0
0 0 1 0 ] P ~ Q + A o r P A o - A o T P B o [ H + BoTPBo]-'BoTPAo
(11)
From (10) the input, U(k), is obtained in the form:
U (k)=F.&e (j ) + F ~ x (+~F)~ (Uk -1) (12)
3. LQI REGULATOR DESIGN
The first t e r m of (12) represents the integral-action
which ensures that the steady-state errors in response
Considering that the system consists of multi-input to a step change in the references and/or disturbances
and multi-output, the application of state feedback are reduced to zero. Configuration of the regulator
control is suitable for regulator design. In this is shown in Fig. 3. Solving the Riccati equation and
paper, an optimal regulator is selected because it is calculating feedback gains are complicated but if
insensitive to small parameter variations and guaran- software tools such a s MATLAB are used, they are quite
tees a stable feedback system. With an optimal regu- easily done.
lator, however, a s its input is given by a proportion-
al action, steady-state e r r o r s in response to a step
reference and/or disturbance change remain in the 4. PWM CONTROL OF VFR

;p
output. To remove the errors an integral-type optimal
regulator (LQI)@) is employed here.
A space-vector-based quasi-circular locus method(1Q)
In an LQI regulator, the object is to find an optimal is employed because it is suitable for micro
output, U(t). that quickly reduces the error between
output, Y(t), and reference, R(t); the derivative of
state vector, pX(t); and the derivative of input,
p U ( t ) . to zero. Thus, the system is expanded in a
discrete-time new state space consisting of the error R'id
vector e(k). and the first-order difference of each .................
state and input vectors:
e (&)=I2 ( k )-Y ( k )
~

AX(R)=X(&)-X(k-I)
(7)
AU ( k ) =U ( k )-U ( k - 1 ) d
AW(k)=AU(k-l)
where R(k)=[i,M kl is a reference vector, and "9 q
AW(k) is introduced to take account of the time-lag of Fig. 2. Phasor diagram of voltages
one sampling interval in the DSP output needed for
computation. The new state equation in the expanded and currents for VFR.
system is:
x o ( k+ I ) = A ~ x+ ~B (~~A) U ( ~ )
&re x o ( & ) = [ e ( k ) A X ( & ) A W ( k ) l r v
I -CA -CB
1. (a)

-til
Bo=[O 0 I]'
where matrices A and B are the transformed version of id iq "dc 1'
A. and Bv in (5) for a discrete-time system. Accord-
ingly, PWM voltage-fed rectifiers are modeled by the Fig. 3. Configuration of the control system
8-th order linear difference equations with two-input based on an optimal regulator.
117

controlled PWM. There are eight spacevoltage vectors By assuming that the load on the DC side consists of
depending on the switching function of each phase as RI and LI, the voltage equations on the DC side will
shown in Fig. 4. The switching function for phase "a" be given by
is defined as C p ud= i .,+oCua- i d.
S.=1 when Sa+ is ON and Sa- is OFF, (14)
Se-1 when Lt is OFF and Sa- is ON. CPU,= i .-oCuI- id.
With the quasi-circular locus method, the locus of
time-integrated rectifier input voltage vectors is By making that the instantaneous power on the AC and
adjusted so as to follow the reference circle as DC sides is equal:
closely as possible by selecting a proper space vector
and applying it for a proper time interval. The locus
of the time-integrated vectors is shown in Fig. 5.
The rectifier input currents, & and iQr, are pulse-
5. CURRENT-FED RECTIFIERS shaped but as harmonic currents influence little the
average power, they are neglected; and only fundamen-
tal current is considered. The relationship between
Fig. 6 shows a current-fed rectifier. The voltage the AC and DC currents is:
equations on the AC side transformed into the d-q
coordinate will be given by(*) idr=SdIl , iq.=SqI, (16)

L p i d = - R i .,+a L i .-U ,+e4 where Sd and S, are the DC components of the d- andq-
(13)
~p i , = - - 0 L i .-R i q - u ,+e,

A s ed=E and eq=O, ifi d and i, are controlled so that


they are constant and zero respectively, a sinusoidal
AC current with unity in the fundamental power factor
will be obtained.
Im

I I

Fig. I. Configuration of the experimental


VFR controller.

Fig. 4. Space voltage vectors for VFR.


-
I
<
25.0 ........................................
I I
I I

0.
Im -rl

> 0.0
.:.....? .........:................
I
-25.0 ........ I ............... I . . . . . . . . . . . . . . . .
I

,c ".p- >
.........................................
........................ ............

z
V?,,,.," v2
U
.cI
>

Fig. 5. Quasi circular locus method. 150.0

t t t L1

'1

STEP
R1
I I
0 40 80 120
TIUE Imsl

Fig. 8. Measured responses of VFR to step changes


Fig. 6. PWM Current-fed rectifier. in DC voltage reference.
118

axis switching functions, respectively. The defini- 6. EXPERIMENTAL RESULTS


tion of the switching function for phase "a" is:
Sa=l when Sa+ is ON and Sa- is OFF,
S.=-l when Sa+ is OFF and Sa- is ON, Fig. 7 shows the configuration of the VFR controller
Se+ when both Sa+ and Sa- are ON, or OFF. used for experiment. DSP executes algorithms such as
the coordinate transformation and input calculation,
Substituting (16) into (14), then linearizing (13) and and a single-chip microprocessor generates PWM signals
(14). the state equations will be obtained in the to the power devices of the rectifier. Experiments
form: were carried out a t the same sampling and carrier
frequencies of 1.8[kHzl. The average switching fre-
p X ( t ) = A c X ( t )+E&( t ) quency of the power devices is l.Z[kHz]. The load w a s
&reX(t)=[i, i. ug U, 1 1 1 ~ ~ (11) composed of Ri=3Z[Ql and Li=l,6[1nH]. Other parameters
were: E=lOO[Vl, VdelSO[Vl, C=33O[pFl, R=l[Q] and
U ( t ) =[M. M J L=5[1nHl. The weight values in the cost function are
selected so as to provide satisfactory responses by
It is clear that i9 and Ii are the output variables: simulation-based examinations: &loo, Qvdc=l, and
iq is for fundamental power factor control and L for H~,FH.~~O~.
load current control, while s d and Sq are the input
variables. Actually CFR is controlled through the Fig. 8 shows the responses of VFR to step changes in
modulation index, M, and phase angle, 6 The rela- . the DC voltage reference, from 16O[Vl to l8O[Vl at 50
tions among them are: steps, then back to 16O[Vl again a t 150 steps (1 step
equals one sampling interval, 556[psl). It is seen
that the response is rapid; steady state errors are
zero; the AC current is kept sinusoidal and there is a
unity fundamental power factor.

Fig. 9 shows the responses of VFR to step changes in


CFR is modeled by 5th-order nonlinear differential the load resistance, from 64[Q] to 32[Ql, then back to
equations with two-input and two-output. The matrices 641Ql again. The DC voltage deviations are suppressed
A. and BO are shown in the Appendix. The output vector within 9%, while the power factor is kept at unity,
Y ( t ) is given by which demonstrates the robustness of the LQI regulator
because a system parameter, RI, is changed in this
situation.
(19) Fig. 10 shows the step responses when the weights of
the cost function were changed, e l 0 0 and & d e l in
Fig. 9 but &=1 and Qvdc=lOO in Fig. 10; the other
Equation (17) is transformed in a discrete-time sys- conditions were the same as Fig. 9. We can observe a
tem, then modified to meet the LQI regulator require- quicker response of Vdc as Qvdc is increased, while a
ment in the manner described for VFR. Finally, 9-th slower response of iq as gd, is decreased. Thus,
order linear difference equations are obtained for responses can be adjusted by the weight values.
CFR, which is an expression very s i m i l a r to (8) for Fig. 11 shows the waveforms of the AC current and the
VFR.
.................................... -
deviations of the DC voltage, AVdc, a t steady state.
< 25.0 .......................................
2 25'0"'' i I
,
........................
I
I ................
e
; 0.0
~,
. 7 --
I
! io
I ...............
-25.0 1.. . . . . . i . . . . . . . . . . . . . . .i . ...............

f-"i-l
....................................
.I... . . . . .I
e
U
> . . . . . . . .................................
. . . . . . . , . . . . . . . . . . . . . . . I . . . . . . . . . . . . . . . .
I0
100.0

10.0 ........................................ 10.0 r . ' . . . ' ...............................


-
<
l i :
I
....... I ...............I . . . . . . . . . . . . . . . . - i
....... ............... . . . . . . . . . . . . . . . .
........................................

-~ ......................................
CI

....... ................................

1
0.0 50 100 150 200 250 0.0 50 1W 150 200 250
STEP STEP

I I t I
0 40 80 120 0 40 80 120
TIME Ins1 TIHE IS.[
Fig. 9. Measured responses of VFR to step Fig. 10. Measured responses of VFR to step
load change. & = l o 0 and QvdFl. load change. &=1 and Qvde100.

____. ..... ...... .- .-___I_


119

Experiments for CFR were also carried out a t the same


sampling and carrier frequency, l.B[kHz]. E=lO0[Vl9
R=l[QI, L=lO[mHl, C=15[flFI, Li=5O[mHl and R1=10[Ql,
The weight values in the cost function are: &r, =@I
=1, and Hs,i=H~.~1000. Fig. 12 shows the experimental
configuration of the CFR controller. Fig. 13 shows
the responses of CFR to step changes in the DC current

Fig. 14 Waveforms of CFR at steady state. AC


source voltage, ea (upper, lOOV/div.); source
current L, (middle, 5A/div.); rectifier
input voltage, V. (lower, 100V/div). 5ms/div.

reference, from 2[Al to 8[Al a t 50 steps, then back to


2[A] again at 150 steps. I t is seen that the DC
current responds within 2 cycles, keeping sinusoidal
AC current a t the unity fundamental power factor.
i. ( u p p e r , 5A/div.); DC v o l t a g e d e v i a t i o n : A V d e Fig. 14 shows voltage and current waveforms a t steady
(lower, 5V/div.). 2ms/div. State.

6. CONCLUSION

Modelling of PWM rectifiers with sinusoidal AC current


is carried out under consideration of only the funda-
mental component of the input terminal voltage for
VFR or the input current for CFR. The obtained non-
linear equations are linearized around an operating
point, then transformed into a discrete-time system.
A s a result, the mathematical models of VFR and CFR
are represented by 4-th and 5-th order linear differ-
ence equations, respectively.
A s the models have two-input and two-output, state
feedback control is suitable. In this paper, a dis-
crete-time optimal linear quadratic integral regula-
Fig. 12. Configuration of the experimental tor is applied for the controller, since it guarantees
CFR controller. a robust feedback system without steady-state errors
in response to a step reference and/or disturbance
change.
Finally, small-capacity DSP-based VFR and CFR systems
were built. The validity of the proposed method was
confirmed by experimental results.

A s a systematic method for control system design is


provided by using an optimal regulator, this approach
could be expanded to larger systems, e.g., a whole AC
drive system, "AC mains - Rectifier - DC link - In-
verter - I.M.." in the future.

References

-
2-
I
150.0
1. Sugimoto, H. e t al, 1988, "A high performance con-
trol of a voltage-type PWM converter," IEEE PESC'88
Record, pp.360-368.

2. Sakai, K. e t al, 1990, "All digital control method


of voltage source converter," IEE of Japan Static
-150.0 I I I I Power Converter Conf. SPC-90-33, pp.5-14.
10.0, 1
3. Tozuka, H. e t al, 1989, "Voltage source converter-
inverter system with instantaneous power feedback,"
Conf. Rec. of IEE of Japan Ind. Appl. National Conv.,
pp.473-474.

I I 4. Hombu, M. e t al. 1985. Trans. Ind. -


IA-
0.0 50 100 150 200 250 &, 1192-1,198.
STEP
5. Fukuda, S. and Takada, N., 1987, "PWM current
I I source rectifier with sinusoidal line current," IEEE
0.0 40.0 80.0 120.0
TIME [msl IAS Annual Meet. pp.679-684.

Fig. 13. Measured responses of CFR to step change 6. Nonaka, S. e t al. 1989, IEE of Japan Trans. vol.D-
in DC current reference.. los, 90-97.
120

7. Iwaji, Y. and Fukuda, S., 1 9 p , "A control method


of PWM voltage source rectifier, Conf. Rec. of IEE of
Japan Ind. Appl. National CMV., pp.412-417. - -_R 1
L a --L o
8. Fukuda, S. and Iwaji, Y., 1992, "Moping and con-
trol of current 8ource PWM rectifiers, Conf. Rec. of
IEE of Japan Ind. Appl. National Conv., pp.384-389.
--w -_R
L
0 _ LI_ o
9. Anbe. T., 1991, "Power elecronics and system con- 1
- 0 0 --
SdO
A== 0
trol," Ohm-Sha Pubishing, Tokyo, Japan. C C
10. Fukuda, S. et al, 1090, m E Trans. Power 0 c1 --w 0 -c s.0
Electronics, pp.54-61.
o ~ s . o - l
APPENDIX L1 Ll LI
- 0 0
1. Voltage-Fed Rectifier 0 0
_- I10
C O
BC=
I10
0 --C

VdO
- v.0
-
. Ll LI

The values of state or input variables at the operat-


ing point are given by

B+dE2-4RRrIioZ
v do= 2
-wL
v.o=- (v.0-E)
R
sdo=-i1 1 do+acvPol
I IO
OC
s .(I=--
I10
v do
i do=- 1 (E - U 40)
where the subscript "0" indicates the values of state
or input variables a t the operating point. They are R
given by

E - - . J E ~ - - ~ R,ovaCO
I
i do=
2R
Vdd
I Io=-
RI
S d O =.&
-(-R i do+E)
VdCO

45
Sq0=--aL i do
Vd.0

Você também pode gostar