Sunteți pe pagina 1din 6

Multirate Perfect Tracking Control of Single-phase Inverter

with Inter Sampling for Arbitrary Waveform


Hironori Abe

and Hiroshi Fujimoto

Department of Electrical and Computer Engineering


Yokohama National University

79-5 Tokiwadai, Hodogaya-ku, Yokohama, 240-8501 Japan


Phone: +81-45-339-4107, Fax: +81-45-338-1157
E-mail: abe@h.dnj.ynu.ac.jp, hfuji@ynu.ac.jp
AbstractIn this paper, multirate perfect tracking control
is proposed for a single-phase inverter. Although conventional
single-rate deadbeat control could not guarantee zero tracking
error for arbitrary reference signals, the proposed multirate
control can achieve perfect tracking at every sampling point.
Feedback characteristic is enhanced by inter sampling tech-
nique. Simulations and experiments are carried out to compare
the proposed method with a single-rate deadbeat controller
for non-sinusoidal reference waveform. Finally, the proposed
method is applied to a nonlinear rectier load.
Index Termsinter sampling, single-phase inverter, multi-
rate control, deadbeat control, arbitrary waveform
I. INTRODUCTION
The deadbeat control is well known and widely used
technique in high speed and high precision control. Several
papers have tried to apply the deadbeat control to inverter
systems [1][2][3]. However, it is proven theoretically that the
conventional single-rate deadbeat control could not guarantee
zero tracking error for arbitrary reference signals [4][5].
We have applied multirate perfect tracking control [4] to a
single-phase inverter [5][6][7]. The terminology of perfect
tracking control(PTC) is originally dened in [8], which
means the plant output perfectly tracks the desired trajectory
with zero tracking error at every sampling point. In the
perfect tracking control, the tracking error of plant state be-
comes completely zero at every sampling period of reference
input for a nominal plant without disturbance. Moreover,
by combining the proposed feedforward controller with a
robust feedback controller such as disturbance observer or
H

controller, high tracking performance is preserved even


if the plant has modeling error and disturbance.
In [6], there was a possibility that the feedback characteris-
tic of PTC worsens more than the single-rate control because
the output period became longer than the input period, by
synchronizing the sampling period of reference signal for the
simplication. In [7], the feedback characteristic of PTC is
equal to single-rate control by synchronizing the sampling
period of the output with the career period. In both [6] and
[7], the output samples only at the bottom vertices of the
triangle career. However, there are development environment
that the output samples both at the top and bottom vertices of
the career recently. On the other hand, it was proposed multi-
sampling method with FPGA based on hardware controller
in [9].
In this paper, we proposed the error suppression technique
by increasing and decreasing the pulse width every half
cycle of the career. Then, the PWM pulse can become an
asymmetry type when the tracking error is caused by the
modeling error and disturbance. Thus, it is possible that
the bandwidth of feedback system becomes twice, because
sampling period of the feedback loop is half. The proposed
method is applied to arbitrary AC power supply in the
simulations and experiments. Also we apply the proposed
method to a nonlinear load.
II. SINGLE-RATE CONTROL
A. Plant Model of Single-phase Inverter
In this section, the plant model is introduced in order
to apply the PTC to the voltage control of a single-phase
inverter. Tracking performance is very important not only in
motor drive but also in active lter and UPS. Fig. 1 shows
the controlled plant. As shown in Fig. 2, the PWM inverter
bridge can generate output voltage of V
DC
or 0 as v
inv
(t).
The load of v
inv
(t) is considered as plant P(s), which is
modeled as
x(t) = A
p
x(t) +b
p
v
inv
(t), y(t) = c
p
x(t). (1)
It is assumed that this plant is n-th order linear load, i.e.,
the number of inductance L and capacitance C is n. In the
example of Fig. 1, the plant system has a LC lter and a
resistive load R. Hence n = 2. The plant coefcients are
represented as
A
p
=

0 1

1
LC

1
RC

, b
p
=

0
1
LC

c
p
=

1 0

, x(t) =

v
c
v
c

. (2)
In order to design the digital controller in discrete-time
domain, we need to discretize the continuous-time plant
model (1). The discrete-time state space model is formulated
with the period T
u
as
x[k + 1] = A
s
x[k] +b
s
u[k], y[k] = c
s
x[k], (3)
L=1mH
C=10F
R=
23.8
vc vinv
VDC=
60V
DSP
Gate Signal
v
c
Fig. 1. Inverter system.
u[k]
kTu (k+1)Tu
Tu
r[i]
r[i+1]
VDC
T[k]
Fig. 2. Single-rate PWM control.
where x[k] = x(kT
u
) and the pulse width is regarded as
control input u[k] = T[k]. In this paper, T
u
=50s with
20kHz carrier frequency. [1] proposed more precise model
which can evaluate the instantaneous value than the zero-
order hold. When the pulse is allocated is the center of
control period T
u
, the PWM holder can be modeled as
follows.
A
s
= e
ApTu
, b
s
= e
Ap
Tu
2
b
p
V
DC
, c
s
= c
p
(4)
B. Single-rate Deadbeat Control
In this section, the conventional deadbeat controller is
designed to see the problem of single-rate method. The
single-rate feedback deadbeat (SR-FBDB) control law is
given by
u[k] = fx[k] + gr[k], (5)
where f is the feedback gain, g is feedforward gain, and
r[k] is the reference signal.
1) Single-rate Feedback Deadbeat Controller 1: [6]
In this section, rst single-rate feedback deadbeat controller
(SR-FBDB-1) is designed. The feedback gain f places
the closed-loop poles to origin of z-plane for the discrete-
time plant (3) with period T
u
. The feedforward gain g is
determined to make the DC gain from r[k] to y[k] unity.
Then, this controller can track the step-type reference signal
with 2 steps delay since the plant is second order system.
2) Single-rate Feedback Deadbeat Controller 2: [1][2]
Next, second single-rate feedback deadbeat controller (SR-
FBDB-2) is designed. In [1] and [2], it is proposed the
control law is given by
u[k] = fx[k] + gr[k] (6)
f = (
A
s11
b
s1

A
s12
b
s1
), g =
1
b
s1
,
from rst row of (3). This law has deadbeat characteristic to
track with 1 step as v
c
[k +1] = r[k]. Then, transfer function
from r[k] to the control input u[k] corresponds to the inverse
of the discrete-time plant model P[z]. Hence, this controller
must be able to assure perfect tracking.
However, P[z] discretized by zero-order hold has unstable
zeros when the relative degree of P(s) is greater than 2 [10].
Thus, feedforward controller becomes unstable because the
closed-loop system has the unstable zeros.
In the case of example Fig. 1, the transfer function of (4)
can be calculated as
P[z] = c
s
(zI A
s
)
1
b
s
=
4.77 10
5
(z + 0.95)
z
2
1.45z + 0.90
. (7)
Even though the relative degree of (2) is 2, P[z] has zero at
z = 0.95. Although it is in the stable region (| z |< 1), this
C=10F
R=
23.8
v vinv
VDC=
60V
L1=1mH L2=2mH
Fig. 3. Inverter system with 3rd order load.
-150
-100
-50
0
50
100
150
0 5 10 15 20
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
r[k]
Fig. 4. Simulation result for 3rd order plant(SR-FBDB-2).
zero is very oscillatory as it is closed to z = 1. Thus, the
output signal of feedforward controller oscillates with high
frequency as the sign of u[k] alternates at every T
u
.
Moreover, as described above, the discrete-time plant with
zero-order hold has unstable zeros when relative degree of
plant is greater than 2. Here, this method is applied to the
inverter system of Fig. 3. The transfer function where from
v
inv
to v can be calculated as
V
V
inv
=
R
L
1
L
2
Cs
3
+ L
1
CRs
2
+ (L
1
+ L
2
)s + R
. (8)
Fig. 4 shows the simulation result by SR-FBDB-2. The
output becomes unstable because the discrete-time plant
model of (8) that discretized of (4) has zero at z = 2.92.
Therefore, in conventional single-rate control systems, per-
fect tracking control is generally impossible.
III. MULTIRATE CONTROL
A. PTC by multirate control
A digital tracking control system generally has two sam-
plers S for the reference signal r(t) and the output y(t), and
one holder H on the input u(t). Therefore, there exist three
time periods T
r
, T
y
, and T
u
which represent the periods
of r(t), y(t), and u(t), respectively. The input period T
u
is generally decided by the speed of the actuator, the D/A
converter, or the calculations on the CPU. In case of inverter
systems, T
u
is determined by career frequency. On the other
hand, the output period T
y
is determined by the speed of
the sensor or the A/D converter. In conventional single-rate
systems, these three periods are set to equal value for the
simplication both in controller design and implementation.
In the multirate feedforward control [4], the control input
u(t) is changed n times during one sampling period (T
r
) of
reference input r(t). In other words, the reference sampling
period T
r
is set to n times longer than the control period T
u
.
Again, n is the plant order. The advantage of this method is
that the control laws to generate u
1
[i], , u
n
[i] can be set
independently. By using these extra degrees of freedom, this
method can guarantee perfect tracking at every T
r
. Note that
the single-rate systems with T
y
= T
u
cannot achieve perfect
tracking even at T
r
= nT
u
because it has one control law.
-25
-20
-15
-10
-5
0
5
10
15
20
25
0 5 10 15 20
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
r[k]
Fig. 5. Simulation result for 3rd order plant(PTC).
B. Plant Formulation with Multirate PWM
The case of T
y
= T
u
is considered. We need to obtain
the discrete-time plant formulation with multirate input. The
state transition from kT
u
to (k + 1)T
u
is given by (3) and
(4) for single-rate PWM. Thus, the state-space model from
iT
r
= kT
u
to (i + 1)T
r
= (k + n)T
u
is represented as
x[i + 1] = Ax[i] +Bu[i] (9)
y[i] = Cx[i] +Du[i] (10)
where x[i] = x(iT
r
), and multirate input and output vectors
u, y are dened as
u[i] = [u
1
[i], , u
n
[i]]
T
= [T[k], , T[k + n 1]]
T
, (11)
y[i] = [y
1
[i], , y
n
[i]]
T
= [y(kT
y
), , y((k + n 1)T
y
)]
T
. (12)
The coefcient matrices are given by

A B
C D

A
n
s
A
n1
s
b
s
A
n2
s
b
s
A
s
b
s
b
s
c
s
0 0 0 0
c
s
A
s
c
s
b
s
0 0 0
.
.
.
.
.
.
.
.
.
c
s
A
n1
s
c
s
A
n2
s
b
s
c
s
A
n3
s
b
s
c
s
b
s
0

. (13)
C. Design of Perfect Tracking Controller
In the ideal tracking control system, the transfer charac-
teristic from the command y
d
to the output y is unity. In
this section, the feedforward controller is designed so that
the transfer characteristic from the desired state x
d
[i] to the
plant state x[i] can be identity matrix I.
From (9) and (10), the transfer function from x[i + 1] to
u[i] and y[i] is described by
u[i] = B
1
(I z
1
A)x[i + 1]
=

0 I
B
1
A B
1

x[i + 1] (14)
y[i] = z
1
Cx[i + 1] +Du[i], (15)
where z = e
sTr
. In (14), the nonsingularity of matrix B is
assured for controllable plant, because B in (12) coincides
with the controllability matrix. Because all poles of the
transfer function (14) are zero, it is found that (14) is a
stable inverse system. Thus, if the control input is calculated
B (I - z A)
-1 -1
+
- +
+
y
o
u
ff u y(t) r(t)
(Tr)
(Ty)
Pn(s)
Cfb
u
fb
P(s)
(Ty)
(PWM)
(Tu)
(PWM)
(Tu)
Fig. 6. Block diagram of PTC with inter sampling.
VDC
T[k] T[k+1]
kTu (k+1)Tu (k+2)Tu
iTr (i+1)Tr
Tu Tu
Tr
(j+1)Ty
Ty Ty Ty Ty
jTy (j+2)Ty (j+3)Ty (j+4)Ty
y[j+2]
y[j]
r[i]
y[j+4]
r[i+1]
y[j+1]
y[j+3]
Fig. 7. Multirate PWM control with inter sampling.
by (16), perfect tracking is guaranteed since (16) is an exact
inverse plant.
u
0
[i] = B
1
(I z
1
A)x
d
[i + 1] (16)
Here, x
d
[i +1] is previewed desired trajectory of plant state.
The output of the nominal plant model can be calculated by
y
0
[i] = z
1
Cx
d
[i + 1] +Du
0
[i], (17)
When the tracking error e[k] = y[k] y
0
[k] is caused by
modeling error or disturbance, it can be eliminated by the
robust feedback controller as
u[k] = u
0
[k] + C
2
[z
y
]e[k]. (18)
The inverter system of Fig. 3 is applied to PTC as in the
case of SR-FBDB-2. Fig. 5 shows the simulation result. Fig.
5 shows that perfect tracking is achieved even though the
relative degree of plant is greater than 2 as (8).
D. Perfect Tracking Control with Inter Sampling
In this section, the output samples both at the top and
bottom vertices of the triangle career, although in foregoing
section it samples only at the bottom vertices. We proposed
the error suppression technique by increasing and decreasing
the pulse width every half cycle of the career. Fig. 6
shows block diagram of PTC with inter sampling, where
H
(PWM)
represents PWM holder. The nominal output y
0
[j]
is calculated every half cycle of the career in real-time. Thus,
the sampling period of PTC with inter sampling becomes T
u
=
Tr
2
= 2T
y
. The solution x(t) of (1) is represented as
x(t) = e
Ap(tt0)
x(t
0
) +

t
t0
e
Ap(t)
b
p
u()d, (19)
by giving initial value x(t
0
) and piecewise continuous input
u(t). From (19), the nominal plant output y
0
[j](= x
0
[j]) can
be calculated
x
0
[j + 1] = A
0
x
0
[j] + b
0
(T
on
, T
off
) (20)
A
0
= e
ApTy
0
2
4
6
8
2 2.2 2.4 2.6 2.8 3
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
r[k]
Vc*(t)
Fig. 8. Simulation result for sinusoidal wave with 3rd harmonics (SR-
FBDB-1).
0
2
4
6
8
2 2.2 2.4 2.6 2.8 3
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
r[i]
Vc*(t)
Fig. 9. Simulation result for sinusoidal wave with 3rd harmonics (PTC).
10
1
10
2
10
3
10
4
-40
-30
-20
-10
0
10
Frequency [Hz]
G
a
i
n

[
d
B
]
S1[z]
S2[z]
Fig. 10. Sensitivity function.
b
0
=

TyTon(k)
0
e
Apt
b
p
(E)dt (off on)

Ty
TyT
off
(k)
e
Apt
b
p
(E)dt (on off)
,
where T
on
(k) is the time PWM pulse becomes on from off,
T
off
(k) is the time PWM pulse becomes off from on.
The PWM pulse is increased and decreased by feedback
controller C
fb
when the tracking error is caused by modeling
error or disturbance. The nominal plant output y
0
from (20)
and detection value y are compared every sampling period
of the output T
y
. Therefore, as Fig. 7, the PWM pulse can
become an asymmetry type when the tracking error is caused
by the modeling error and disturbance.
Analytical solution of (20) can be obtained for low degree
of the plant like this paper. However, it is necessary to
calculate (20) numerically for higher-order plant by using
the Pade approximation etc. Moreover, if the reference signal
is already known, the computation cost can be reduced by
computing (20) off-line.
IV. SIMULATION RESULTS
A. Output for Arbitrary Waveform
As [11], the reference signal is given
V

c
= 10sin(2500)t + 2sin(21500)t. (21)
The three periods of SR-FBDB-1 are set to T
y
= T
u
=
T
r
=50s. The career frequencies of SR-FBDB-1 and PTC
are selected to be equal with T
u
=50s for fair comparison.
Thus, the three periods of PTC are set to T
y
= T
u
=50s
T
r
=100s. Fig. 8 shows that SR-FBDB-1 has tracking error
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
0 5 10 15 20 25 30 35 40 45 50
v
o
l
t
a
g
e
[
V
]
time[ms]
Inter Sampling
with Inter Sampling
(A)
(B)
without
Fig. 11. Simulation result(Error comparison).
-30
-20
-10
0
10
20
30
40
0 2 4 6 8 10
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*(t)
(A)without Inter Sampling
-30
-20
-10
0
10
20
30
40
0 2 4 6 8 10
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*(t)
(B)with Inter Sampling
Fig. 12. Simulation results with plant variation.
0 500 1000 1500 2000 2500 3000
0
0.5
1
1.5
2
2.5
3
3.5
frequency[Hz]
v
o
l
t
a
g
e
[
V
]
(A)without inter sampling
(B)with inter sampling
Fig. 13. Simulation results with plant variation(FFT).
even at the sampling points. On the other hand, Fig. 9 shows
that PTC has zero tracking error at every sampling period.
Thus, we nd that the multirate control has better tracking
performance than the single-rate system even though these
two methods have same T
u
.
B. Effectiveness of Inter Sampling
In this section, we design the feedback controller C
fb
and
consider the stability of the closed-loop system about each of
(A) conventional PTC (T
y
=50s) and (B) PTC with inter
sampling (T
y
=25s). We use the lead-lag controller as C
fb
.
We determined that the cutoff frequency of the closed-loop
transfer function is equal to
1
100
of the nyquist frequency for
(A) and (B). Then, it becomes possible to achieve higher
bandwidth because (A) is discretized by sampling frequency
T
y
=50s, but (B) is discretized by T
y
=25s. In Fig. 10,
S
1
(z) and S
2
(z) are the sensitivity function of the closed-
loop system of (A) and (B), respectively. From Fig. 10, we
nd that it makes bandwidth twice by using inter sampling.
The gain of sensitivity can be attenuated about 6dB up to
300Hz.
We verify characteristic of disturbance rejection to com-
pare the feedback characteristic of (A) and (B). The sinu-
soidal wave with amplitude 2V and frequency 100Hz is given
at the plant output as disturbance in Fig. 6 when the reference
r(t) is assumed to be 0. From Fig. 11, the characteristics of
disturbance rejection of (B) is better than (A).
-4
-2
0
2
4
6
8
10
12
0 0.2 0.4 0.6 0.8 1
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*[i]
-2
0
2
4
6
8
10
12
0 0.2 0.4 0.6 0.8 1
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*[i]
(a) for sinusoidal wave with (b) for sinusoidal wave with
3rd harmonics(SR-FBDB-1) 3rd harmonics(PTC)
Fig. 14. Experimental result with nominal plant.
-1
-0.5
0
0.5
1
0 5 10 15 20 25 30 35 40
v
o
l
t
a
g
e
[
V
]
time[ms]
Inter Sampling
with Inter Sampling
(A)
(B)
without
Fig. 15. Experimental result(Error comparison).
0 2 4 6 8 10
- 30
- 20
- 10
0
10
20
30
40
ti me[ms]
v
o
l
t
a
g
e
[
V
]
Vc(t)
Vc*[i ]
0 2 4 6 8 10
- 30
- 20
- 10
0
10
20
30
40
ti me[ms]
v
o
l
t
a
g
e
[
V
]
Vc(t)
Vc*[i ]
O 6OO 1OOO 16OO 2OOO 26OO 8OOO
O
2
4
6
8
1O
froquonoy|Hz|
v
o
!
t
a
_
o
|
V
|
A)w1thout !ntor Samp!1n_
B)w1th !ntor Samp!1n_
(a) without Inter Sampling (b) with Inter Sampling (c) FFT analysis
Fig. 16. Experimental result with plant variation.
Next, we examine that the plant has parameter varia-
tion from the nominal model. We consider the inductance
variation L = 0.4L
n
where L
n
is the nominal inductance
value, and the reference signal is given the sinusoidal wave
with 3rd harmonics that consists of a combination rst
harmonic with amplitude 30V and frequency 300Hz with 3rd
harmonics with amplitude 6V and frequency 900Hz with this
L(= 0.4L
n
), the closed-loop system is still stable. Fig. 12
shows the simulation results with plant variation, and Fig.
13 shows the FFT analysis of error both (A) and (B). From
Fig. 13, (B) can attenuate the error tracking than (A) in low-
frequency region where the sensitivity improved in Fig. 10.
Thus, we nd the effectiveness of inter sampling method.
V. EXPERIMENTAL RESULTS
Fig. 14 shows the experimental results. In the case of the
reference is sinusoidal wave with 3rd harmonics, SR-FBDB-
1 has tracking error even at the sampling points, as shown in
Fig. 14(a). On the other hand, Fig. 14(b) shows that perfect
tracking is achieved every sampling periods by PTC.
Fig. 15 shows the feedback characteristic of disturbance
rejection. As well as the simulation, the reference r(t) is
assumed to be 0 and the sinusoidal wave of amplitude 2V,
frequency 100Hz is given to the plant output as disturbance
in software. From Fig. 15, the disturbance has been sup-
pressed in (B) more than (A).
Fig. 16 shows the experimental results with plant variation.
As well as the simulation, L = 0.4L
n
and the reference is
sinusoidal wave with 3rd harmonics. From Fig. 16(c), (B)
can attenuate the tracking error than (A) in low-frequency
region. Thus, we nd the effectiveness of inter sampling
method even in experiments.
L=1mH
C=10F
vc vinv VDC=
60V
DSP
Gate Signal
vc
LL=1.6mH
CL=220F
RL=
15.8
iL
iL
Fig. 17. Inverter system with rectier load.
VI. APPLICATION TO NONLINEAR LOAD
In this section, the proposed method is applied to a
nonlinear load. Here, we think about the inverter system of
Fig. 17. The plant is represented as
x(t) = A
p
x(t) +b
p
(v
inv
(t) L

i
L
(t)), y(t) = c
p
x(t) (22)
A
p
=

0 1

1
LC
0

, b
p
=

0
1
LC

, c
p
=

1 0

that is coincide with the model which has the resistance load
R = in (1).

i
L
(t) is regarded as disturbance, which can be
suppressed by feedback controller. The system of multirate
PWM control with inter sampling is designed by using the
discrete-time state space model of (22) with PWM holder.
We apply the lead-lag controller as the feedback controller
C
fb
, which is determined that the cutoff frequency of the
closed-loop transfer function is equal to
1
50
of the nyquist
frequency for (A) and (B). Fig. 18(a) shows the sensitivity
function of the closed-loop system. The experiments are
carried out with the reference signal is given the sinusoidal
wave with amplitude 30V and frequency 50Hz. Fig. 18(b)
shows the FFT analysis of tracking error both (A) without
inter sampling and (B) with inter sampling. From Fig.
18(b), the tracking error of (B) is much smaller than that
of (A) in low frequency band around 50Hz. However, the
sensitivity function is amplied in high frequency band over
the resonance frequency. The tracking error of (B) is larger
10
1
10
2
10
3
10
4
-40
-30
-20
-10
0
10
20
30
Frequency [Hz]
G
a
i
n

[
d
B
]
(A)without Inter Sampling
(B)with Inter Sampling
(a)Sensitivity function
0 500 1000 1500 2000 2500 3000
0
0.2
0.4
0.6
0.8
1
frequency[Hz]
v
o
l
t
a
g
e
[
V
]
(A)without inter sampling
(B)with inter sampling
(b)FFT analysis
Fig. 18. Application to the rectier load.
+ -
+
Vinv
u
d(t)
P(s)
d*[k]
Plant
(PWM)
(Tu)
Tu
VDC
Fig. 19. Block diagram with disturbance compensator.
than that of (A) in high frequency band. When the nominal
plant model is designed as the resistance load R = , it is
difcult to suppress the error in all frequency band only by
making closed-loop bandwidth higher with inter sampling.
Therefore we proposed the disturbance compensation as
feedforward manner with detection i
L
(t). Here, the rectier
load can be regarded as the current source in Fig. 17. Thus,
the plant model is represented as
V
c
=
1
LC
s
2
+
1
LC
(V
inv
LsI
L
) := P(s)(V
inv
+ d), (23)
which consists of LC lter and the current source sI
L
as
disturbance d. The disturbance compensation is proposed
as feedforward in Fig. 19. Because the control input u[k]
is on-time T[s], the disturbance compensation d

[k] is
approximately converted to T
d
[s] as
T
d
=
T
u
V
DC
d

. (24)
Fig. 20(a) and (b) show the experiment results. Fig. 21
shows the FFT analysis of error both of Fig. 20(a) and (b).
From Fig. 21, the tracking error with disturbance compen-
sation is much smaller than that without disturbance com-
pensation. Thus, we nd the effectiveness of the disturbance
compensation.
VII. CONCLUSION
A novel perfect tracking control (PTC) method was pro-
posed for inverter systems based on multirate PWM control.
The advantage of this method is that the feedforward con-
troller can be designed without considering the unstable zero
problem. Moreover, by combining the proposed feedforward
controller with a robust feedback controller, robust tracking
performance is obtained.
In PTC, the output samples both at the top and bottom
vertices of the triangle career, although in conventional it
samples only at the top or bottom vertices. Therefore, it is
possible to achieve higher bandwidth of feedback system.
Then, the PWM pulse can become an asymmetry type when
the tracking error is caused by the modeling error and distur-
bance. The advantages of this approach were demonstrated
-40
-30
-20
-10
0
10
20
30
40
-25 -20 -15 -10 -5 0 5 10 15 20 25
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*(t)
Id
(a)without disturbance compensation.
-40
-30
-20
-10
0
10
20
30
40
-25 -20 -15 -10 -5 0 5 10 15 20 25
v
o
l
t
a
g
e
[
V
]
time[ms]
Vc(t)
Vc*(t)
Id
(b)with disturbance compensation.
Fig. 20. Experimental Results with rectier load.
0 500 1000 1500 2000 2500 3000
0
0.5
1
1.5
2
frequency[Hz]
v
o
l
t
a
g
e
[
V
]
(a)without disturbance compensation
(b)with disturbance compensation
Fig. 21. FFT analysis with rectier load.
through simulations and experiments on the voltage control
of single-phase inverter.
We also applied the proposed method to nonlinear load.
In the proposed method, the error is suppressed in the low
frequency region, but the error has increased in the high
frequency region. Therefore we proposed the disturbance
compensation as feedforward and conrmed the effectiveness
in experiments.
REFERENCES
[1] K. P. Gokhale, A. Kawamura, and R. G. Hoft: Dead beat micro-
processor control of PWM inverter for sinusoidal output waveform
synthesis, IEEE Trans. Industry Applications, vol. 23, No. 5, pp.
901-910, 1987
[2] K. Ishihara and A. Kawamura: DSP based real time output waveform
synthesis for three phase PWM inverter, T.IEE Japan, Vol. 110-D, No.
6, pp. 627636, 1990 (in Japanese)
[3] Y. Igarashi, T. Izumi, T. Yokoyama, and T. Haneyoshi: A study of
instantaneous value control with voltage variation compensation for
UPS inverters, PCC2002, Vol. 2, pp. 653658, Osaka, 2002
[4] H. Fujimoto, Y. Hori, and A. Kawamura: Perfect tracking control
based on multirate feedforward control with generalized sampling
periods, IEEE Trans, Industrial Electronics, Vol. 48, No. 3, pp. 766
772, 2001
[5] H. Fujimoto, Y. Hori, and S. Kondo: Perfect Tracking Control Based
on Multirate Feedforward Control and Applications to Motion Control
and Power Electronics, PCC2002, Vol. 1, pp. 196201, Osaka, 2002
[6] D. Ogino, H. Fujimoto and S. Kondo: Multirate Two-degree-of-
freedom Deadbeat Control of Single-phase Inverter, T.IEE Japan, Vol.
125-D, No. 7, pp. 751757, 2005 (in Japanese)
[7] H. Abe and H. Fujimoto: Multirate Perfect Tracking Control of
Inverter for Arbitrary Waveform, IEE of Japan Technical Meeting
Record, SPC-06-35, pp. 6772, January 2006 (in Japanese)
[8] M. Tomizuka: Zero phase error tracking algorithm for digital con-
trol, ASME, J. Dynam. Syst., Measur., and Contr., vol. 109, pp. 65
68, March 1998
[9] E. Shimada and T. Yokoyama: Current Control for Utility Interactive
Inverter Using Multi Sampling Method based on FPGA, JIASC2005,
pp.405410, August 2005 (in Japanese)
[10] K. J.

Astrom, P. Hangander, and J. Sternby: Zeros of Sampled
Systems, Automatica, Vol. 20, No. 1, pp. 3138, 1984
[11] S. Hashino, K. Wada and T. Shimizu: A Generation Control of
Arbitrary AC Waveform for the Single-phase Voltage Source PWM
Inverter Utilizing an Adaptive Frequency Loss-less Resonator, T.IEE
Japan, Vol. 127-D, No. 2, pp. 103111, 2007 (in Japanese)

S-ar putea să vă placă și