Download Comparison of Controllers for a Three

Comparison of Controllers for a Three-phase Phase
Locked Loop System under Distorted Conditions
V. Miñambres1,2, M.I. Milanés1,2, B. Vinagre2, E. Romero1,2
1
2
Power Electrical & Electronic Systems (PE&ES)
School of Industrial Engineering. University of Extremadura. Badajoz, Spain
Abstract—The analysis and design of a three-phase Phase
Locked Loop system to synchronize to the electrical grid is
presented. The whole phase locked loop model is obtained and a
Proportional Integral, a Proportional Integral Derivative and a
novel Fractional Proportional Integral controllers in continuous
time domain are proposed to solve the phase tracking by
improving time response, phase error and overshoot. The
performance of the three controllers is analyzed under distorted
utility conditions, such as imbalance, harmonics and phase and/or
frequency variations. The phase locked loop is totally
implemented by software, and tested by simulation via
SIMULINK and experimentally via xPCTarget, in a three phase
Autoadjustable Synchronous Reference Frame application. A
comparison in the dynamic and steady state of the phase locked
loop parameters like accurate, robustness and fastness is carried
out for the three different controllers.
Index Terms—Phase locked loops, modeling, PI control, PID
control, fractional control, power systems control.
I. INTRODUCTION
Nowadays, the electric power quality is hardly affected by
the increase of non-linear loads connected to the grid. To
assure minimum specifications in the quality, reliability and
use of the electric magnitudes, the utility interface operation of
power converters such us ac/dc converters, active filters,
cycloconverters, etc. is very common.
The positive-sequence fundamental phase angle of the utility
voltage is a critical piece of information for this kind of devices
in the operation of most power conversion and conditioning
applications such as active power filters, quality meters of
electrical magnitudes, uninterrupted power supplies and grid
connected photovoltaic or wind power generation systems, so
they need an accurate phase tracking system to work properly.
The phase obtained is used to construct a reference carrier
wave to synchronize the on/off commutations of power
devices, calculate and control the flow of active/reactive power
or transform the feedback variables to a reference frame
suitable for control purposes.
Some methods have been developed to track the phase,
which in a general way can be divided into two groups [1]:
open-loop methods, such as low pass filters, space vector filters
or extended Kalman filters, which directly estimate the phase
angle of the PCC voltage from its alfa-beta coordinates,
obtained by the Clarke transformation [2][3]; and closed-loop
methods, mainly Phase Locked Loop (PLL) methods, in which
the phase angle estimation is adaptively updated by a closedloop mechanism whose objective is to track the real value of
the angle [4]-[11]. These methods have, however, the
shortcomings of not operating properly under imbalance,
harmonic distortion or frequency variations. Most recently,
Adaptive PLL [12] and Enhanced PLL [13]-[14] have been
developed with the objective of getting frequency adaptation
and unaffected robust response under harmonics or imbalance
in the input signals. However, these methods propose threephase structures formed by four single-phase modules, which
complicate the control stage and require heavy computation.
In electric power systems, the instantaneous angle and
frequency information are typically recovered using a time
basis PLL technique. The quality of the PLL is the key to
obtain the maximum performance of the application, so it must
lock the phase of the utility voltage at the Point of Common
Coupling (PCC), which is distorted due to imbalance,
harmonic, and phase or frequency variations, as quickly and
accurately as possible. Because of that, it is essential to develop
a controller which provides fast time response, zero error in the
steady state and validity under any input signal for the PLL.
The simplest controller that fits these specifications is the
Proportional Integral (PI). It could be designed in continuoustime or discrete-time domain [8]. Then, the regulator gains
could be calculated through different control strategies such as
symmetrical optimum [7], Wiener optimization [8], second
order system pattern [7][16], etc.
This paper focuses on the design and operation comparison
of three types of different controllers for a PLL: Proportional
Integral (PI), a Proportional Integral Derivative (PID), and a
Fractional Proportional Integral (FPI), in order to find the one
with best performance. The paper is structured as follows.
Firstly, the explanation of the PLL applied to an
Autoadjustable Synchronous Reference Frame (ASRF) [14]
application is exposed. Then, a PLL model is obtained. With
the model, an open loop transfer function is attained and the PI
and PID controllers are designed by using the root locus
method. The FPI is designed in the frequency domain by
improving the frequency response of the previous two designed
controllers. Afterwards, the closed loop model is simulated and
the PLL is implemented by software to carry out an ASRF
application, which consists on the extraction of the positivesequence fundamental three phase component of an unbalanced
three-phase signal with harmonics. The ASRF is checked by
simulation using SIMULINK, and then, it is used in
experimental environment with the xPCTarget platform.
Finally, the results of the three controllers are summarized in a
comparison table.
II. THEORETICAL EXPLANATION OF THE PLL AND THE ASRF
APPLICATION
The PLL and the ASRF application used to compare the
controllers proposed in this paper are shown in Fig. 1. In the
following sections, a theoretical explanation of the PLL under
sinusoidal input voltages is firstly presented. Afterwards, the
adaptation of the model necessary in case of distorted input
voltages is detailed, and finally the ASRF application is
explained.
A. Tracking Method. PLL under sinusoidal input voltage
The instantaneous three-phase voltage at the PCC uPCC(a,b,c)
could be converted to the 0-d-q dynamic reference frame
uPCC(0,d,q) using the Park transformation:
uPCC ( 0, d , q ) = P ⋅ uPCC ( a, b, c ) ,
(1)
where P is the transformation matrix, defined as:
1
1
⎛ 1 2
⎞
2
2
⎟
2⎜
2π
2π
cos θ
cos (θ − 3 ) cos (θ + 3 ) ⎟ .
P=
(2)
3 ⎜⎜
2π
2π ⎟
θ
θ
θ
sin
sin
sin
−
−
−
−
+
(
)
(
)
3
3 ⎠
⎝
If the input voltage is sinusoidal with amplitude U, then:
⎛ uPCC , a ⎞ ⎛ U sin (ωin t + ϕ ) ⎞
⎟
⎜
⎟ ⎜
uPCC ( a, b, c ) = ⎜ uPCC ,b ⎟ = ⎜ U sin (ωin t + ϕ − 23π ) ⎟ .
(3)
⎜u
⎟ ⎜⎜ U sin ω t + ϕ + 2π ⎟⎟
( in
⎝ PCC , c ⎠ ⎝
3 )⎠
Denoting (ωin t + ϕ ) by θin , (3) and (1) could be substituted
in (2) and, after applying some trigonometric relations and
rejecting the homopolar component (since it will not be used),
the following expression is obtained:
⎛ uPCC , d ⎞
3 ⎛ sin (θin − θ ) ⎞
uPCC ( d , q ) = ⎜
U⎜
(4)
⎟.
⎟=
2 ⎝⎜ − cos (θin − θ ) ⎠⎟
⎝ uPCC , q ⎠
When the difference angle Δθ = θin − θ is big, the angle θ
is locked soon due to the feedback. Then, Δθ will be small,
and the Taylor approximation for the sine and cosine around
zero could be applied [16]:
⎛ uPCC , d ⎞
3 ⎛ Δθ ⎞
uPCC ( d , q ) = ⎜
U ⎜ ⎟.
(5)
⎟=
2 ⎝ −1 ⎠
⎝ uPCC , q ⎠
Finally, the voltage division is calculated to work only with
the angle, that is, ud is normalized:
ud uq = −Δθ .
(6)
With this approach, the PLL does not depend of the level of
the voltage input.
B. Adaptation of the PLL to be used under distorted input
voltage
Taking into account that the input voltage is distorted, the
expressions exposed in the previous section are only valid if
the positive sequence fundamental component is obtained from
the input. This is carried out by a Butterworth Low Pass Filter
(LPF) in the d and q components. With a cut-off frequency of
f co = 13Hz the LPF gets their mean values U d and U q , which
assures no harmonics in each one. The first one is used for the
PLL control, and both of them for the ASRF reconstruction.
Fig. 1. PLL with ASRF application.
C. PLL and ASRF application
The aim is to make zero U d / U q through the PLL controller
C, which is possible when Δθ = 0 , as can be extracted from (6)
. The division achieves the controller to be valid for any level
of the input voltage. On the one side, the angle θ will
correspond to the angle of the q axis and, on the other side, U q
will carry on with the positive-sequence fundamental signal.
The PLL control loop consists on a C controller which takes
its error signal U d / U q and gives the amount of pulsation Δω
necessary to add or subtract to a reference pulsation ω0 to
catch the input voltage pulsation ω in = 2π f in . The LPF makes
the controller to be immune to the input harmonics and
imbalance. Obviously, ω 0 is calculated as the nominal value of
the input voltage frequency (50Hz in electric systems) which
helps the algorithm to improve the convergence. The result is
the fundamental pulsation ω = 2π f , so the positive sequence
fundamental angle θ is obtained by integrating it. The phase
will be locked to the q axis rotation when the error signal
Ud / Uq = 0 , that is, ω = ωin and θ = θin , so Δθ = 0 .
The ASRF reconstruction of the positive-sequence
fundamental voltage uPCC,1¨+(a,b,c) is possible by applying the
inverse Park transformation P-1 to U q with the angle θ. The
component U d does not affect in the steady state, but using it a
softer dynamic state is achieved, and the homopolar is rejected.
III. MODEL OF THE PLL
The model of the designed PLL is presented in Fig. 2. It is
obtained from the PLL system after being linearized (5) and
normalized (6). Obviously, the LPF is taking into account to
design properly the controller C. An approximation delay is
modeled too, to simulate the sample rate Ts of the algorithm.
From the model, the open loop transfer function without C is
derived ( Ts = 1e − 4s ):
1 1
PLL TF ( s ) = LPF ( s )
=
s Ts s + 1
(7)
2
( 2π f co )
1 1
=
.
2
s 2 + 2 2 f co s + ( 2π f co ) s Ts s + 1
6672
.
0.0001 s + 1.012 s 3 + 116.2 s 2 + 6672 s
The root locus is shown in Fig. 3.
PLL TF ( s ) =
4
(8)
Now, by tuning the controller C, the poles of the system
could be located where the system response is stable and fast.
Fig. 2. Model of the PLL.
IV. DESIGN OF THE CONTROLLER FOR THE PLL
A. PI controller
The PI controller is designed by using the root locus tool
from the MATLAB toolbox. The pole introduced by the PI is
located at the origin, and the zero is located where the root
locus shows the best stable situations (-14.33). The criteria
selected then to locate the four poles is based on the closeness
to the x axis. Fig. 4a shows the result: Two pairs of complex
conjugate poles at the same position and one real pole with a
high negative value.
The PI controller responds to the next expression:
K
412.18
.
CPI ( s ) = K p + i = 28.8526 +
(9)
s
s
B. PID controller
The PID controller is designed like the PI. The difference is
that the PID introduces two zeros (both at -57.59 in the design),
that is, an added degree of freedom, so the root locus has more
stable situations, plus a faster transitory. Fig. 4b displays the
result: Two pairs of complex conjugates poles at the same
position and one real pole with a high negative value.
The PID controller responds to the next expression:
K
1658.3
CPID ( s ) = K p + i + K d s = 56.382 +
+ 0.479 s. (10)
s
s
C. FPI controller
Fractional calculus is a generalization of the integration and
differentiation to the non-integer (fractional) order fundamental
operator a Dtα , where a and t are the limits and α (α ∈ \) is the
order of the operation. Among many different definitions, one
of the most commonly used for the general fractional integrodifferential operation is the Grünwald–Letnikov (GL):
[(t − a ) h]
⎛α ⎞
−α
α
(−1) j ⎜ ⎟ f (t − jh),
(11)
∑
a Dt f (t ) = lim h
h→0
j =0
⎝ j⎠
where [⋅] means the integer part.
Podlubny [17] proposed a generalization of the PID
controller, namely the PIλDμ controller, involving an integrator
of order λ and a differentiator of order μ. He also demonstrated
the better response of this type of controller, in comparison
with the classical PID controller, when used for the control of
fractional order systems. A frequency domain approach by
using fractional order PID controllers was also studied [18].
x 10
4
Root Locus of the open loop PLL model without C
Root Locus of the open loop PLL model without C
200
2
150
1.5
100
Imaginary Axis
1
Imaginary Axis
The controller block C for the PLL must be, at least, a PI, to
obtain a zero steady state error, because the model is similar to
the position control of a servo with a ramp input. In this paper
it will be analyzed the classical PI and PID controllers and, as a
novelty, a FPI controller is designed and compared with the
usual ones.
2.5
0.5
0
-0.5
-1
50
0
-50
-100
-1.5
-150
-2
-2.5
-2.5
-200
-2
-1.5
-1
-0.5
0
Real Axis
0.5
1
1.5
2
x 10
4
-100
-50
0
Real Axis
50
100
(a)
(b)
Fig. 3. Root locus of the PLL model without controller. (a) Full view.
(b) Detail at the origin.
(a)
(b)
Fig. 4. Root locus with poles fixed. Detail at the origin. (a) PI controller
(b) PID controller.
In power electronics, fractional order controllers have been
successfully applied in [19] and [20]. In the first one, a Power
Electronic Buck Converter is controlled by using different
strategies: linear fractional controllers (Smith predictor and
phase - lag compensation) and fractional sliding. In the second
one, a fractional PI controller is used for the control of a Shunt
Active Power Filter outperforming the results obtained with a
traditional PI.
The generalized PIλDμ controller has the next form (see [21]
and references therein):
K
CFPI ( s ) = K p + λi + K d s μ .
(12)
s
The FPI controller is designed in the frequency domain for
improving the phase margin obtained with the other two
controllers: with an order of integration less than one (PI), the
FPI could maintain zero error on steady state by introducing
less phase lag.
The final three Bode plots of the open loop model systems
for each controller are shown in Fig. 5. As it can be seen, the
stable conditions are met in a very narrow margin of
frequencies for the PI and PID, so there is another possible
improvement here for the FPI: The phase margin of the model
with the FPI is larger than the others models (see the values
listed in TABLE. I), so a more stable and robust system is
Gain (dB)
0
-50
-100
-1
10
0
10
1
10
2
Phase (º)
-150
-200
-250
-300
-1
10
10
0
10
1
10
2
Frequency (rad/s)
Fig. 5. Bode plots of the closed loop system model with different controllers.
PI (—blue). PID (--red). FPI (..magenta).
x 10
Ud/Uq (V)
Ud/Uq (V)
0.1
0
-0.1
-0.2
-0.3
0
0.05
0.1
0.15
0.2
0.25
0.3
ef (Hz)
0
-1
0
0.05
0.1
0.15
0.2
0.25
-2
-4
19.98
19.99
20
19.99
20
19.99
Time (s)
20
-6
0
-2
-4
-6
-8
19.98
0.3
x 10
0.3
-4
4
e (rad)
0.2
0.1
2
θ
0
-0.1
-4
0
x 10
1
VI. ASRF ALGORITHM SIMULATION
0
0.05
0.1
0.15
Time (s)
0.2
0.25
0
19.98
0.3
(a)
(b)
Fig. 6. PLL`s model simulation with different controllers. PI (—blue).
PID (--red). FPI (..magenta). (a) Transient state. (b) Steady state.
5
Ud/Uq (V)
0.1
0.15
0.2
0.25
-3
-5
19.98 19.99
0.3
20
0.2
0.05
0.1
0.15
0.2
0.25
0
-0.2
19.98 19.99
0.3
0.3
0.2
0.1
0
-0.1
2
x 10
20
-3
0
θ
e (rad)
θ
4
2
0
-2
-4
0
0.05
x 10
0
ef (Hz)
0
-0.1
-0.2
-0.3
0
ef (Hz)
Ud/Uq (V)
0.1
e (rad)
The ASRF algorithm (see Fig. 1) has been simulated with a
three-phase PCC voltage input with harmonics and imbalance.
In Fig. 7 the PLL variables are shown, while Fig. 8 and Fig. 9
displays the ASRF inputs and outputs in transient and steady
state, with the next input voltage parameters: phase a 230 V,
phase b 180 V, phase c 280 V, all the phases with 1/7 of 7th
harmonic and 51 Hz. The ASRF obtains the positive sequence
fundamental component of the input voltage, at 51 Hz. The
sample rate was 1e-4 s, that is, a frequency of 10 kHz.
The PLL algorithm response is similar to the PLL model
simulation. The differences are mainly in case of the PID
controller, which applies an excess effort in the beginning of
the transient state and has some oscillations around zero in the
steady state. These unexpected effects do not happen in the
model simulation because of the linearization.
The phase error, which has been measured in the steady
state, is close to zero (see the values listed in TABLE. I). The
rapidness under the described situation (like a step frequency
input) is also summarized in TABLE. I.
The input margin has been measured by changing the
frequency of a sinusoidal and balanced input while the phase
was not locked. Obviously, the PID is the best one with regards
to this parameter because of its higher efforts, and
consequently its rapidness. The PI and FPI are similar at this
point. The values are in TABLE. I.
10
-100
θ
The PLL model (see Fig. 2) has been simulated with an input
angle θin modeled by a ramp with a slope ωin = 2π f in . If the
input voltage is sinusoidal and balanced, the plot of the voltage
vector will rotate with constant frequency and the phase
changes linearly.
Fig. 6 shows the results Ud/Uq, ef = fin – f and eθ = Δθ, in a
simulation test with f in = 51Hz . One observes that the FPI
controller is the slowest one with a small steady state error that
is going con closer to zero. The PID is the fastest and the PI
stays as an intermediate solution. The PI and PID offer quasizero steady state error. The controllers are well designed.
50
ef (Hz)
V. SIMULATION ANALYSIS OF THE MODEL
100
e (rad)
expected. The disadvantage is the DC gain response, which is
smaller too, and the steady state will be reached slowly. The
phase margin reveals that the PID is able to adapt to higher
changes.
Finally, the FPI controller responds to the next expression:
K
K
100 0.3
CFPI ( s) = K p + λi = K p + i s1− λ = 20 +
s . (13)
s
s
s
By defining an integrator, and then, the complementary
fractional term, the error in the steady state is forced to zero.
The fractional controller is implemented by the “fractional
PID” block from the “ninteger” [22] MATLAB toolbox with
the fractional parameters crone, mcltime, frac n = 10 and
bandwidth = 0.1-100 [22].
0
0.05
0.1
0.15
Time (s)
0.2
0.25
0.3
-2
19.98 19.99
20
Time (s)
(a)
(b)
Fig. 7.
PLL`s algorithm simulation with different controllers.
PI (—blue). PID (--red). FPI (..magenta). (a) Transient state. (b) Steady state.
The steady state has been analyzed to calculate the Total
Harmonic Distortion (THD) of the output voltage for the phase
a. One can observe from TABLE. I that the THD is closed to
zero and equal in all the cases. This is because the LPF block is
responsible for filtering the harmonic content and it has been
used the same filter for all the cases.
PI θ in and θ (rad)
0
-500
0
0.05
0.1
PID θ in and θ (rad)
500
0
-500
0
0.05
0.1
500
0
-500
board. The Advantech embedded PC PPT-154T with the
National Instruments PCI-6071E presented in Fig. 10 is the
equipment used as Target.
The HP-6834B AC power source, which can be seen in Fig.
10, has been utilized to generate the PCC voltage. These
voltages have been measured using HALL effect sensors.
8
FPI θ in and θ (rad)
PI uPCC and u +PCC,1 (V)
PID uPCC and u +PCC,1 (V)
FPI uPCC and u +PCC,1 (V)
500
0
0.05
Time (s)
0.1
6
4
2
0
0
0.05
0.1
0
0.05
0.1
0
0.05
Time (s)
0.1
8
6
4
2
0
8
6
4
2
0
8
PI θ in and θ (rad)
500
0
-500
19.96
19.97
19.98
19.99
20
PID θ in and θ (rad)
500
0
-500
19.96
19.97
19.98
19.99
20
500
FPI θ in and θ (rad)
FPI uPCC and u +PCC,1 (V)
PID uPCC and u +PCC,1 (V)
PI uPCC and u +PCC,1 (V)
Fig. 8. Simulation of the ASRF algorithm in the transient state with different
controllers. Input (distorted) and output (sinusoidal and balanced) signals.
0
-500
19.96
19.97
19.98 19.99
Time (s)
20
6
4
2
0
19.96
19.97
19.98
19.99
20
19.97
19.98
19.99
20
19.98 19.99
Time (s)
20
8
6
4
B. Software
The ASRF algorithm (Fig. 1) has been implemented with the
xPCTarget tool from MATLAB toolbox via SIMULINK. The
code is obtained by compiling the ASRF blocks scheme and
saved into a “dlm” format archive. To do that, a third party
compiler is required (the selected was VISUAL BASIC).
The “dlm” archive is loaded in the Target through the
“xpcexplorer” application in the Host via Ethernet. With this
application, all the options and parameters of the Target loaded
code could be controlled. The sample rate was 10 kHz.
The Target has its own real time OS. It is created through
“xpcexplorer” by setting the proper configuration.
The block used to get the measurements has been the
xPCTarget A/D of the DAQ board and then, the measured
values are corrected according to the voltage sensor
attenuation. The outputs are registered through the xPCScope
block which allows saving the data into the Target hard disk.
C. Results
The experimental test has been carried out by setting a threephase voltage input with the next parameters: Phase a 230 V,
phase b 200 V, phase c 260 V, and all the phases with 1/5 of 5th
and 1/7 of 7th harmonics and 51 Hz, which represents a
situation where the PCC voltage has imbalance and harmonics.
After making the experiment, the saved data are analyzed. The
results in the transient and in the steady state are displayed in
Fig. 11 and Fig. 12, demonstrating that the ASRF works with
the three controllers.
2
0
19.96
8
6
4
2
0
19.96
19.97
Fig. 9. Simulation of the ASRF algorithm in the steady state with different
controllers. Input (distorted) and output (sinusoidal and balanced) signals.
VII. ASRF EXPERIMENTAL TESTS
A. Hardware
The hardware to implement the ASRF is based on the
xPCTarget from MATLAB. It proposes a x86 based PC, called
Target, to run the algorithm, which is controlled by another PC,
called Host, via Ethernet. Obviously, to measure the PCC
voltages it is necessary to install a xPCTarget compatible DAQ
Fig. 10. AC power source and Target PC running the ASRF algorithm.
Harmonic espectrum
8
-200
-400
0
0.05
0.1
6
4
2
0
0
0.05
0
-200
-400
0
0.05
0.1
400
200
0
-200
-400
0
0.05
Time (s)
0.1
8
200
0
-200
-400
19.96
19.97
19.98
19.99
20
PID θ in and θ (rad)
400
200
0
-200
-400
19.96
19.97
19.98
19.99
20
400
200
0
-200
-400
19.96
19.97
19.98 19.99
Time (s)
20
0
5
10
15
0
20
h: Harmonic order
0
5
10
4
80
0
0.05
0.1
8
6
Harmonic espectrum
100
Voltage. THD: 0.0%
2
20
(b)
Harmonic espectrum
100
15
h: Harmonic order
(a)
Voltage. THD: 0.0%
80
60
40
20
60
40
20
4
0
2
0
0
0.05
Time (s)
0.1
0
5
10
15
h: Harmonic order
20
0
0
5
10
15
h: Harmonic order
20
(c)
(d)
Fig. 13. Spectrum analysis of the experimental tests. Individual harmonic
distortion ratio (taking as base value the fundamental component of the phase a
input voltage) of the: (a) Input voltage, (b) PI output voltage, (c) PID output
voltage and (d) FPI output voltage.
According to the amplitudes of the three outputs, one can
conclude from Fig. 13 that the most accurate controller is the
PID.
8
PI θ in and θ (rad)
400
40
20
6
0
60
0.1
IHDh = |Ah/A1|%
PID θ in and θ (rad)
200
FPI θ in and θ (rad)
PI uPCC and u +PCC,1 (V)
PID uPCC and u +PCC,1 (V)
40
20
Fig. 11. Experimental tests of the ASRF algorithm in the transient state with
different controllers. Input (distorted) and output (sinusoidal and balanced)
signals.
FPI uPCC and u +PCC,1 (V)
60
0
400
IHDh = |Ah/A1|%
0
Voltage. THD: 0.8%
80
IHDh = |Ah/A1|%
200
Harmonic espectrum
100
Voltage. THD: 24.1%
80
IHDh = |Ah/A1|%
PI θ in and θ (rad)
400
FPI θ in and θ (rad)
FPI uPCC and u +PCC,1 (V)
PID uPCC and u +PCC,1 (V)
PI uPCC and u +PCC,1 (V)
100
6
4
2
0
19.96
19.97
19.98
19.99
20
VIII. COMPARISON BETWEEN THE CONTROLLERS
8
6
4
2
0
19.96
19.97
19.98
19.99
20
19.98 19.99
Time (s)
20
8
6
4
2
0
19.96
19.97
The results obtained in the comparison of the controllers
carried out in section IV. C. and section VI. under specific
simulation conditions, are summarized in TABLE. I. From this
table, one can conclude that the selection of the controller
depends on the kind of application and its necessities, because
each one has their own pros and cons.
If the robustness is the priority, the FPI is at the top. The
fastest is the PID. In the search of the widest input frequency
range, the PID must be the selected again. The THD reveals
that every controller works equally well under distorted
conditions. The minimum phase error is reached by the PI.
Fig. 12. Experimental test of the ASRF algorithm in the steady state with
different controllers. Input (distorted) and output (sinusoidal and balanced)
signals.
The phase is tracked by avoiding the harmonic and
imbalance of the input, so the output is the positive sequence
fundamental three-phase voltage. This is demonstrated in Fig.
13 by showing the Individual Harmonic Distortion (IHD) ratio
of the phase a for the three controllers (taking as base value the
fundamental component of the phase a input voltage). As it can
be seen, only the fundamental voltage remains so the THD is
0%, except in the PI case, which has a THD of 0.8%.
TABLE. I
COMPARISON OF CONTROLLERS
Robustness.
Phase
margin (º)
PI
32.8
PID
26.1
FPI
42.9
Settling
time (s)
0.22
0.16
2.5 (eθ
0.01rad)
Input
frequency
margin (Hz)
39-61
12-103
(oscillating)
40-59
THD
Phase error
(%)
(rad)
0.8
3e-6
8.6e-3
8.6e-3
0
(oscillating)
3,59e-4
IX. CONCLUSIONS
A full working three-phase PLL has been designed. Three
different controllers have been tested to lock the phase: PI, PID
and a novel FPI. Due to the use of LPFs and the normalization
in the PLL model, each controller is immune to harmonics and
imbalance, respectively, so all the controllers fit under any
distorted condition. Every controller has been designed using
well known control strategies, despite of the complex PLL
system model.
The performance of each controller has been analyzed and
summarized in a table in order to compare them. The result is a
wide variety, because each controller has its own advantages
and drawbacks, so one concludes that the selection must be
done depending on the application specifications.
The PLL has been also tested experimentally for an ASRF
application, confirming the performance with every controller.
X. REFERENCES
[1]
M. Karimi-Ghatemani and M. Reza Iravani, "A Method for
Synchronization of Power Electronic Converters in Polluted and
Variable-Frecuency Environments,” IEEE Transactions on Power
Systems, Vol. 19, No. 3, pp. 1263-1270. Aug. 2004.
[2]
[3]
J.W. Dixon, J.J. García and L. Morán, "Control System for Three-Phase
on Circuits and System, pp. 421-424. 2001.
[14] M. Karimi-Ghatemani, H. Moktari and M. Reza Iravani, M. Sedighy, "A
Signal Processing System for Extraction of Harmonics and Reactive
Current of Single-Phase Systems,” IEEE Transactions on Power
Delivery, Vol. 19, No. 3, pp. 979-986. Jul. 2004.
[15] M. I. Milanés, E. Romero, A. Rico, V. M. Miñambres and F. Barrero,
"Novel method for synchronization to disturbed three-phase and singlephase systems," in IEEE International Symposium on Industrial
Electronics (ISIE 2007), pp. 860-865.
[16] G. Zhou, X. Shi, C. Fu and Y. Wang, "Operation of a Three-phase Soft
Phase Locked Loop Under Distorted Voltage Conditions Using
Intelligent PI Controller," in Proc. 2006 IEEE Region 10 Conf.
(TENCON 2006), pp. 315-320.
[17] I.
Podlubny,
“Fractional-order
systems
and
PIλDμ controllers,”
IEEETransactions on Automatic Control, Vol. 44 No. 1, pp 208-214.
1999.
digital control. Past, present and future of PID control, Terrasa, Spain,
S. Luo and Z. Hou, "An Adaptive Detecting Method for Harmonic and
M. C. Benhabib and S. Saadate, “A New Robust Experimentally
G.-C. Hsieh and J.C. Hung, "Phase-Locked Loop Techniques - A
609-615. Dec. 1996.
D Detjen, J. Jacobs and R. W. De Doncker, H.-G. Mall, "A New Hybrid
pp. 53–58.
[19] A. J. Calderón, B.M. Vinagre, V. Feliu. "Fractional order control
strategies for power electronic buck converters." Signal Processing, 86
(2006) pp. 2803-2819.
[20] A. Rico, M. I. Milanés, E. Romero, and B. M. Vinagre. "PI vs fractional
PI for the control of a shunt active power filter," in Proceedings of the
ASME IDETC/CIE, Las Vegas, USA, Sept. 4-7 2007.
[21] C. A. Monje, B. M. Vinagre, V. Feliu, Y. Chen, “Tuning and auto-tuning
of fractional order controllers for industry applications,” Control
Engineering Practice, Vol. 16, pp. 798–812. 2008.
Filter to Dampen Resonances and Compensate Harmonic Currents in
[22] D. V. Mata de Oliveira. (2005, Aug.). Ninteger v. 2.3 Fractional control
Industrial Power Systems With Power Factor Correction Equipment,”
toolbox for MatLab. Instituto Superior Técnico, Universidade Técnica de
IEEE Transactions on Power Electronics, Vol. 16, No. 6, pp. 821-827.
Lisboa. http://web.ist.utl.pt/duarte.valerio/ninteger/ninteger.htm.
Nov. 2001.
V. Kaura and V. Blasko, "Operation of a Phase Locked Loop System
Under Distorted Utility Conditions,” IEEE Transactions on Industry
Applications, Vol. 33, No. 1, pp. 58-63. Jan./Feb. 1997.
S.-K. Chung, "A Phase Tracking System for Three Phase Utility Interface
Inverters,” IEEE Transactions on Power Electronics, Vol. 15, No. 3, pp.
431-438. May 2000.
[9]
(PLL) System,” Proceedings of the 44th IEEE 2001 Midwest Symposium
Vol. 42, No. 6, pp. 636-641. Dec. 1995.
Survey,” IEEE Transactions on Industrial Electronics, Vol. 43, No. 6, pp.
[8]
on Power Systems, Vol. 18, No. 3, pp. 1116-1124. Aug. 2003.
[13] M. Karimi-Ghatemani and M.R. Iravani, "A New Phase-Locked Loop
PID controllers: A frequency domain approach,” in IFAC workshop on
Journal, Vol 15, No. 3. 2006.
[7]
Vol. 3, pp. 1033-1038. 2001.
[12] D. Jovcic, "Phase-Locked Loop System for FACTS,” IEEE Transactions
and Unbalanced Loads,” IEEE Transactions on Industrial Electronics,
Validated Phase Locked Loopfor Power Electronic Control,” EPE
[6]
Voltage Restorer (DVR),” IEEE Power Eng. Soc. Winter Meeting. Conf.,
Active Power Filter which Simultaneously Compensates Power Factor
42, No. 1, pp. 85-89. Feb. 1995.
[5]
Barnes, N. Jenkins, "Software Phase-Locked Loop applied to Dynamic
[18] B. Vinagre, I. Podlubny, L. Dorčá k and V. Feliu, (2000), “On fractional
Reactive Currents,” IEEE Transactions on Industrial Electronics, Vol.
[4]
[11] C. Zhan, C. Fitzer and V.K. Ramachandaramurthy, A. Arulampalam, M.
J. Phinney and D.J. Perreault, "Filters With Active Tuning for Power
Applications,” IEEE Transactions on Power Electronics, Vol. 18, No. 2,
pp. 636-647. Mar. 2003.
[10] P. Jintakosonwit and H. Fujita, H. Akagi, "Control and Performance of a
Fully-Digital-Controlled Shunt Active Filter for Installation on a Power
Distribution System,” IEEE Transactions on Power Electronics, Vol. 17,
No. 1, pp. 132-140. Jan. 2002.