Вы находитесь на странице: 1из 6

A Three-Phase PWM Rectifier Using a Linear Active Disturbance Rejection

Control

Ali Boukhriss*. Tamou Nasser**. Ahmed Essadki ***.
Abderrahim Bennaoui****

* Laboratoire de Génie électrique, ENSET, Université Mohamed 5


Souissi Rabat, Morocco, (Ali_boukhriss@hotmail.fr)
** Ecole Nationale Suprieure d'Informatique et d'Analyse des Systemes,
Université Mohamed 5 Souissi Rabat, Morocco (tnasser@ensias.ma)
*** Laboratoire de Génie électrique, ENSET, Université Mohamed 5
Souissi Rabat, Morocco, (Aliahmed.essadki1@gmail.com)
**** Laboratoire de Génie électrique, ENSET, Université Mohamed 5
Souissi Rabat, Morocco, (Bennaouiabderrahim@yahoo.fr)

Abstract: AC/DC converters, based on the IGBT transistor are becoming more widespread in industrial
systems. The aim of this work is to present the active disturbance rejection control for the AC DC PWM
rectifier in order to minimize the effect of internal disturbances due to the filter parameters variations and
external disturbances due to the load change. ADRC based on the extended state observer ESO estimate
and compensate in real time all the internal and external disturbances of the physical plant. A voltage
oriented control VOC is used. The results are compared with PI based on the internal model controller
IMC. Simulations results are carried out with MATLAB / Simulink.
Keywords: Active disturbance rejection control, VOC, PWM, Extended state observer.

1. INTRODUCTION 2. PWM RECTIFIER MODEL.


Currently PWM rectifiers are widely used in industrial Fig.1 represents the rectifier connected by the dc link voltage
applications, to cite as example, the back-to-back converter and load resistor RLoad. if , irec and iload are respectively the
used in wind energy conversion systems. They have the current in the filter, the output current of the rectifier and the
advantage of bidirectional power transfer. This paper presents load current. Si indicates switching functions corresponding
the active disturbance rejection control ADRC based on the to the IGBT switches where (i denotes d and q in dq-axis
extended state observer ESO to minimize fluctuations in the reference).
voltage of the DC bus. Conventional control uses the PI
controller which are easy to implement, but presents a major
drawback when the filter parameters are changed. Indeed,
any change on the filter parameters affects mainly the time
constants of the PI controller. Consequently the performances
of regulators are seriously affected. ADRC method proposes
a control law which is not based on the accurate
mathematical model of the system; therefore all internal
disturbances such filter parameters variations and external
disturbances such load change are estimated and rejected in
real time, hence the name of the active disturbance rejection
control ADRC, which was proposed by Han (1998). This Fig. 1. PWM rectifier model.
paper presents the control of the DC bus voltage using
voltage oriented control, Malinowski et al. (2003). A PLL is
In dq-axis reference frame, relations between voltage and
implemented so as to align the voltage component Vq at the q-
current are given by:
axis, Kaura et al. (1997). The active and reactive powers are
controlled separately through the currents Idf and Iqf. In order dudc
C  irec  iLoad (1)
to have a unity power factor, reactive power is set to be zero. dt
 
To demonstrate efficiency of ADRC, results are compared to 3
the PI based on the internal model controller IMC. irec  S q I qf  S d I df (2)
2
dI df This can be controlled by a simple proportional controller.
 R f I df  L f s I qf  Vd  Vcd (3)
u0  k p r  z1 
Lf (12)
dt
dI qf Where, r is the input signal reference to track.
Lf  R f I qf  L f s I df  Vq  Vcq (4)
dt The controller tuning is chosen as kp= c, where c is the
Vcd  Sd udc (5) desired closed loop frequency.
The combination of linear ESO and the controller is the linear
Vcq  Squdc (6)
ADRC. Generally we choose 0=3~7 c, and consequently,
3. ACTIVE DISTURBANCE REJECTION CONTROL c is the only tuning parameter. Fig. 2 represents the
implementation of the linear ADRC.
The active disturbance rejection control is designed to deal
with systems having a large amount of uncertainty in both the
internal dynamics and external disturbances, Han (2009). The
particularity of the ADRC design is that the total disturbance
is defined as an extended state of the system; and estimated
using a state observer, known as the extended state observer
(ESO). It was also simplified to linear ADRC using the linear
ESO, Zhiqiang Gao (2006), which makes it easy and
convenient to implement. We consider the case of first order
system for the illustration of the method.
Fig. 2. Architecture of the developed ADRC controller.
y  f  y, d , t   b0u
.
(7)
Where u and y are input and output variables. d is the external 4. PROPOSED CONTROL STRATEGY
disturbance, and f(y,d,t) represents the combined effect of
internal dynamics and external disturbance and b0 parameter The main target for the control strategy is to keep the
to estimate. DC bus voltage constant regardless of the disturbances, to
The basic idea is the estimation and compensation of f. have a nearly sinusoidal input current, to have low harmonic
Equation (7) can be written in an augmented state space form distortion of line current and maintain a unity power factor.
as: Indirect active and reactive power control based on current
. vector orientation with respect to the line voltage vector
 x1  x2  b0 u (voltage oriented control VOC) is used.
. .
x 2  h where h f
(8)
y  x 1.1 Regulation of the Voltage Loop
 1

 If we neglected losses in three phase PWM rectifier, the input


A state observer of (8) will estimate the derivatives of y and f active power Pf is equal to the DC link power Pdc, that is:
since (8) is now a state in the extended state model. Pf  Pdc
This observer, denoted as a Linear Extended State Observer A phase locked loop PLL is used to orient the voltage on the
LESO, is constructed as: q-axis and so the voltage on the d-axis is equal to zero. The
.
 z1  z 2  1 x1  z1   b0u input active Pf and reactive power Qf is writing in dq- axis as:
. 3
 z 2   2 x1  z1  (9) Pf  Vq I qf (13)
2
Where, 1 and 2 are the observer gains parameters. To 3
simplify the tuning process, the observer gains are Q f  Vq I df (14)
2
parameterized as 3
 1  20 
Pdc  udcirec  Vq I qf
2
    2 
 2  0  Thus let to
Where, 0 is the bandwidth of the observer determined by the du 3
cu dc dc  Vq I qf  PLoad (15)
pole placement technique, Zheng et al. (2009). The estimate dt 2
is more precisely by increasing the bandwidth of the Letting w=udc2, then (15) can be expressed as:
observer; however, a wide bandwidth increases the sensitivity dw 3 2
 Vq I qf  PLoad (16)
to noise. In practice, a compromise is made between the dt c c
speed at which the observer tracks the states and its Equation (16) can be written in the canonic form:
sensitivity to sensor noise, Gao (2003). With a properly
 f d , w, t   bv0u
dw (17)
designed ESO, z1 and z2 are tracking respectively y and f. dt
The control law is given by:
Where
u z (10)
u 0 2  2 3 
b0  f   PLoad   Vq  bv 0  I qf
The original plant in (7) is reducing to a unit gain integrator.  c  c 
u  I
y   f  z 2   u0  u0
.
(11)  qf
Where, f represents the generalized disturbance, w and Iqf are
Ci s  
w 2
i0 
 2wi 0 wic s  wi20 wic (23)
respectively the output and the control input of the plant. bv0 bi 0 ss  2wi 0  wic 
the parameter to approximate (the value of 3Vq/c can be
chosen). From (17) and (19), transfer function of a physical plant is
given as
So the linear ADRC can be used in the voltage loop.
Rf
Gi s   
1 (24)
where 
1.2 Regulation of the Current Loop L f s    Lf

Equations (3) and (4) that represent the currents in the filter wic and wi0 denote respectively controller and state observer
can be written as: bandwidth of the current loop. Ci(s) and Gi(s) are respectively
controller and physic plant transfer function. Hi(s) is
dI df

1
Vd  R f I df  L f s I qf   1 Vcd (18) considered as a pre-filter function. The block diagram of the
dt Lf Lf closed-loop control can be represented by Fig. 4. The open
dI qf

1
Vq  R f I qf  L f s I df   1 Vcq (19) and closed loop transfer functions are
dt Lf Lf Fi _ ol s   Ci s Gi s  (25)
This led to put the current Idf, and the same studies can be
H i s Ci s Gi s 
used for Iqf, into the canonic form: Fi _ cl s   (26)
1  Ci s Gi s 
 f I df , d , t   bi 0u t 
dI df
(20)
dt
Where
  
f 
1
Vd  R f I df  L f  s I qf    1  bi 0 Vcd
 Lf  Lf 

u  V cd

Where, f represents the generalized disturbance, Idf and vcd are


respectively the output and the control input of the plant. bi0
the parameter to approximate (the value of -1/Lf can be Fig. 4. Block diagram of the current closed-loop control
chosen).
As above a linear ADRC can be applied. From (26), the characteristic polynomial is
i s   L f bio s 3  L f bio s 2 2wi 0  wic     swi20  2wi 0 wic   2wi 0  wic   wi20 wic
(27)
Using the Routh-Hurwitz criterion, it is easy to show that the
system is stable. According to Fig. 4, the transfer function of
disturbance is given as
d s 
Fid s  
I df s 
L f bi 0 ss  2wi 0  wic  (28)


L f bi 0 ss   s  2wi 0  wic   s w  2wi 0 wic  w wic
2
i0  2
i0

It is clear from (26) that as s goes to zero or infinite, Fid(s)


will converge to zero. Bode diagram in Fig. 6, plotted with
Fig. 3. schematic block diagram for the proposed control value given in appendix; reveal that de disturbances are
attenuated since the gain in db is under 0db.
5. STABILITY AND ROBUSTNESS ANALYSIS In order to reveal robustness of the controller against
uncertainties, bode diagram of the open loop Fi_ol(s) is plotted
2.1 Current Loop in Fig. 7 with variation into the filter parameters (Rf, Lf). We
chose the nominal value of  and the extreme values
Using Laplace transformation for ESO and controller – ie (9) corresponding to  As can be seen, the phase margin is
and (12) – applied for Ifd current (and the same studies for not affected, and the system is perfectly stable.
Ifq). We can have:
2.2 Voltage Loop
Vcd s   Ci (s)( H i (s) I fd  I fd )
ref
(21)
The same procedure used for the study of the current loop
Where
will be used for treating the voltage loop. It follows
wic s  wi 0 
2
H i s   (22)
 
w  2wi 0 wic s  wi20 wic
2
i0
I qf s   Cv (s)( H v (s)udc
ref
 udc ) (29)
Where
wvc s  wv 0  wvc and wv0 denote respectively controller and state observer
2
H v s   (30)
 w  2wv 0 wvc s  wv20 wvc
2
v0  bandwidth of the voltage loop. As the current loop, Fig. 5
represents the block diagram of the voltage closed-loop.

Cv s  
w 2
v0 2wv 0 wvc s  w2 v 0 wvc  (31)
bv 0 ss  2wv 0  wvc 
Bode Diagram
100

50

From (17), a physical plant transfer function is given by

Magnitude (dB)
0 c-0.2c
c
3Vq c+0.2c

Gv s   (32) -50
cs -90

-120

Phase (deg)
-150

-180
0 1 2 3 4
10 10 10 10 10
Frequency (rad/s)

Fig. 9. Bode plots of open loop voltage Fv_ol(s)

The open and closed loop function transfer are given


Fig. 5. Block diagram of the voltage closed-loop control Fv _ ol s   Cv s Gv s  (33)

H v s Cv s Gv s 
Fv _ cl s  
Bode Diagram
-70
(34)
-80
1  Cv s Gv s 

Magnitude (dB)

-90

-100 Here also, Routh criterion is used to the characteristic

-110 polynomial v(s). It is easy to show that the system is stable
90
by proceeding in the same manner as above.

v s   cbv0 s 3  s 2 cbv0 2wv0  wvc   s3Vq wv20  2wv0 wvc   3Vq wv20 wvc (35)
45
Phase (deg)

-45
The capability to reject disturbances will be also demonstrate
-90
10
1
10
2
10
3

Frequency (rad/s)
10
4
10
5 by calculating the transfer function of disturbance Fvd.
Fig. 6. Bode plots of Fid(s) with uncertainties in  d s 
Fvd s  
150
Bode Diagram
u dc s 
100
cbv 0 ss  2 wv 0  wvc  (36)

 
s 2 s  2 wv 0  wvc cbv 0  s wv20  2 wv 0 wvc 3Vq  wv20 wvc 3Vq
Magnitude (dB)

50


0 

-50
It is clear that Fvd(s) converge to zero when s goes to zero or
-100
-90 infinite. Bode diagram in Fig. 8, plotted with value given in
appendix; reveal that de disturbances are attenuated since the
Phase (deg)

-135 gain in db is under 0db.


Bode diagram of the open loop transfer function Fv_ol(s)is
-180
0 1 2 3 4 5 6
plotted in Fig. 9, considering the uncertainties on the value of
10 10 10 10 10 10 10
Frequency (rad/s)
capacitor (c and c±0.2c), in order to check the controller
Fig. 7. Bode plots of open loop current Fi_ol(s) robustness. As a current loop, voltage loop is widely stable.

-40
Bode Diagram
6. SIMULATION AND RESULTS
-50

Simulations are carried out using Matalb Simulink. PWM


Magnitude (dB)

-60 

-70
 rectifier settings used in the simulation are listed in the


-80
appendix below. Performances of the proposed model are
90
compared with those of PWM rectifier using PI based on the
45
internal model control PI IMC. The system is subjected to a
voltage step change of 100V and -100V, respectively, at
Phase (deg)

-45
times t = 0.8s and t = 1.2s, then an increase in the load
-90
resistance from Rload to 5/4 of Rload at t = 1.6s is setting.
10
0
10
1
10
2

Frequency (rad/s)
10
3
10
4
Voltage drop is smaller in ADRC model, as illustrated in
Fig. 8. Bode plots of Fvd(s) with uncertainties in the Fig.10, than PI IMC model, time response is faster, and
capacitor c external uncertainties are rapidly cancelled. Fig. 11 shows a
set of unity power factor, and a smaller THD in the current's
waveform (around of 2.7%). In order to highlight the
robustness of proposed control scheme, the internal 1250
L=1.75mH
parameters are subject to uncertainties. Fig. 12, Fig. 14 and L=2.25mH
L=1.25mH
1240
Fig. 16 show the system response due to changes respectively
in the filter inductance Lf (1.25mH, 1.75mH, 2.25mH), the 1230

filter resistance Rf (25m, 50m, 75m) and the capacitor

Udc (V)
1220
(3.5mF, 5mF, 6.5mF). (Fig.13, Fig. 15 and Fig. 17 are their
zoom respectively). It is clear that the system presents a good 1210

robustness against uncertainties of internal parameters. 1200


Eventually the system has undergone a series of load
variations Rload (5 at t=0s, 3.75 at 0.6s, 5 at 0.8s, 1190
1.5 1.55 1.6 1.65 1.7 1.75

6.25 at 1s and 5 at 1.2s). Fig. 18 presents the response


Time (s)

of active and reactive powers. It is clear that the reactive Fig. 13. Zoom of udc under Lf uncertainties
power remains zero, in spite of changes in the active power,
and a unity power factor is respected. 1320
R=50m 
R=75m 
1300
R=25m 
1320
ADRC
PI IMC 1280
1300

1260

Udc (V)
1280

1240
1260
Udc (V)

1220
1240

1200
1220

1180
0.6 0.8 1 1.2 1.4 1.6 1.8 2
1200
Time (s)

1180
0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
Fig. 14. udc voltage under Rf uncertainties
Fig. 10. Response of ADRC and PI IMC under udc_ref step 1250
R=50m 
change and load changed. R=75m 
R=25m 
1240

600
Va 1230
Ifa
Udc (V)

400
1220

200
1210
Va (V) & Ifa (A)

0
1200

-200
1190
1.5 1.55 1.6 1.65 1.7 1.75
Time (s)
-400

-600
0.4 0.41 0.42 0.43 0.44 0.45 0.46 0.47 0.48
Fig. 15. Zoom of udc under Rf uncertainties
Time (s)

Fig. 11. Line voltage and current wave form


1320
C=5 mF
C=6.5mF
1320 C=4.5mF
L=1.75mH 1300
L=2.25mH
L=1.25mH
1300
1280

1280
1260

1260
1240
Udc (V)

1240
1220

1220
1200

1200
1180
0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
1180
0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)

Fig. 16. udc voltage under capacitor uncertainties


Fig. 12. udc voltage under Lf uncertainties
1260
C=5 mF
Appendix
C=6.5mF
C=4.5mF
1250

Grid voltage line to line rms U=690V f=50Hz


Filter parameters Rf=50m Lf=1.75mH
1240

1230
Capacitor c=5mF
Udc (V)

1220 Load resistance 5


1210
Voltage controller bandwidth wvc=60rd/s
Current controller bandwidth wic=900rs/s
1200
Voltage observer bandwidth wv0=300rd/s
1190
1.45 1.5 1.55 1.6
Time (s)
1.65 1.7 1.75 1.8 Current observer bandwidth wio=6300rd/s

Fig. 17. Zoom of udc under capacitor uncertainties


5
x 10
4.5
Ps (W)
Qs (VAR)
4

3.5

3
Active & Reactive Power

2.5

1.5

0.5

-0.5
0.4 0.6 0.8 1 1.2 1.4
Time (s)

Fig. 18. Active and reactive power under load changes

6. CONCLUSIONS
In this paper, we introduced a PWM rectifier using active
disturbance rejection control. The simulations show that the
proposed control scheme has a fast dynamic response than PI
IMC and a good robustness against parameters uncertainties.
The unity power factor is achieved and the line current
presents very low harmonics.

REFERENCES
Gao, Z. (2003). Scaling and Bandwidth-Parameterization
Based Controller, Proceedings of the 2003 American
Control Conference, Vol. 6, 4989-4996.
Gao, Z. (2006). Active disturbance rejection control: a
paradigm shift in feedback control system design.
In American Control Conference, 2399-2405.
Han, J. (1998). Auto disturbance rejection control and its
applications. Control and Decision, volume (1), 19-23.
Han, J. (2009). From PID to Active Disturbance Rejection
Control. IEEE Transactions on Industrial Electronics,
Volume (56), Issue (3), 900-906.
Kaura, V. and Blasko, V. (1997). Operation of a phase locked
loop system under distorted utility conditions. IEEE
transactions on industry applications, volume (33), issue
(1), 58-63.
Malinowski, M., Kazmierkowski, M.P. and Trzynadlowski,
A. (2003). Review and comparative study of control
techniques for three-phase PWM rectifiers. Mathematics
and Computers in Simulation, volume (63), 349–361.
Zheng, Q., Chen, Z. and Gao, Z. (2009). A practical approach
to disturbance decoupling control. Control Engineering
Practice, volume (17), 1016-1025.

Вам также может понравиться