1816 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO.

3, JULY 2010

A Generalized Delayed Signal Cancellation

Method for Detecting Fundamental-Frequency
Positive-Sequence Three-Phase Signals
Francisco A. S. Neves, Member, IEEE, Marcelo Cabral Cavalcanti, Member, IEEE, Helber Elias Paz de Souza,
Fabrício Bradaschia, Emilio J. Bueno, Member, IEEE, and Mario Rizo

Abstract—A novel scheme for obtaining the fundamental-fre- a phase-locked loop (PLL). The PLL must be capable of fol-
quency positive-sequence grid voltage vector based on a general- lowing the FFPS voltage vector as fast as possible for elimi-
ization of the delayed signal cancellation method is proposed in nating the impacts of the grid signal imperfections [1]–[3].
this paper. The technique is implemented by sampling and storing
the instantaneous voltage vector. A mathematical transfor- The synchronous reference frame PLL (SRF-PLL) is fre-
mation is then proposed through which the current and delayed quently used in three-phase systems [2] and for balanced
voltage vectors are combined. It is shown that the proposed undistorted grid conditions, good results can be reached [4].
transformation has unity gain for the fundamental-frequency The SRF-PLL can also operate correctly if only high-order
positive-sequence voltage vector, while its gain is equal to zero harmonics are present in the grid voltages, by reducing its
for some chosen components. Cascaded transformations can then
be used for eliminating the fundamental-frequency negative-se- bandwidth to attenuate these harmonics [5]. However, under
quence vector, as well as chosen positive- and negative-sequence unbalanced conditions, the second harmonic content of the
harmonic vector components and, thus, for accurately obtaining voltage vector, caused by the negative-sequence components,
the fundamental-frequency positive-sequence voltage vector. The makes the reduction of the bandwidth an inefficient solution,
output of the last transformation block is input to a synchronous since the dynamics of the PLL would become very slow for
reference frame phase-locked loop for detecting frequency and
position of the positive-sequence vector. A proposal for making such a narrow bandwidth. Better performance under unbal-
the scheme frequency adaptive is also presented. The good per- anced conditions can be achieved by separating the positive-
formance of the proposed method is verified with simulations and and negative-sequence components of the voltage vector. This
experiments by using distorted and unbalanced signals, containing is made by the double synchronous reference frame PLL
fundamental-frequency as well as positive- and negative-sequence (DSRF-PLL) that uses a decoupling cell to isolate the positive-
harmonic components. The proposed method frequency adapta-
tion capability is also verified. and negative-sequence fundamental-frequency components [6].
A cell for obtaining any other harmonic component may be
Index Terms—Converters, power quality (PQ), voltage control. implemented and then that component may be subtracted from
the original voltages. However, the computational cost greatly
increases if many harmonic components should be attenuated.
I. INTRODUCTION An alternative technique uses an enhanced PLL (EPLL) [7]
for each phase. The phase voltages and their respective values
delayed by 90 are used for obtaining the positive-sequence
voltages of the three-phase system using the instantaneous sym-

T HE CORRECT value of the fundamental-frequency pos-

itive-sequence (FFPS) grid voltage vector is essential for
achieving good control performance in grid-connected systems.
metrical components (ISC) method [8], [9]. Finally, a fourth
EPLL is applied to the output of the ISC method to estimate the
positive-sequence voltage phase angle. The fundamental-fre-
Among the many applications of FFPS detection algorithms quency negative-sequence components of the grid voltages are
are series and shunt active filters, uninterruptible power sup- eliminated by the EPLL calculator. However, some harmonics
plies, adjustable speed drives, and renewable energy genera- will pass through the EPLL and may be present in its output.
tion systems connected to the grid through voltage-source con- In [8], an improved version of the method presented in [7] is
verters (VSC). This information is usually obtained by using obtained by omitting the fourth EPLL and combining the other
three. Using this structure, a scheme for removing selected har-
Manuscript received June 22, 2009; revised September 11, 2009 and monic components is proposed in [10] at the expense of increase
November 27, 2009. First published April 05, 2010; current version published
June 23, 2010. Paper no. TPWRD-00463-2009.
in complexity.
F. A. S. Neves, M. C. Cavalcanti, and F. Bradaschia are with the Depart- In the so-called delayed signal cancelation (DSC) method
ment of Electrical Engineering and Power Systems, Federal University of Per- [11]–[14], the positive- and negative-sequence components of
nambuco, Recife 50740-530, Brazil (e-mail: [email protected]; marcelo.caval-
[email protected]; [email protected]).
the grid voltage are separated based on the voltage vector in a
H. E. P. de Souza is with the Department of Industry, Federal Institute of stationary reference frame and on the voltage vector de-
Education, Pesqueira 55200-000, Brazil (e-mail: [email protected]). layed by a quarter of cycle. After separation, the obtained pos-
E. J. Bueno and M. Rizo are with the Department of Electronics, University of itive-sequence and components are input to an SRF-PLL
Alcala, Madrid 28870, Spain (e-mail: [email protected]; mario.rizo@de-
peca.uah.es). to calculate its amplitude and angular position. The method is
Digital Object Identifier 10.1109/TPWRD.2010.2044196 suitable for unbalanced undistorted grid voltages, but the calcu-
0885-8977/$26.00 © 2010 IEEE
NEVES et al.: GENERALIZED DELAYED SIGNAL CANCELATION METHOD 1817

lated positive- and negative-sequence voltages are sensitive to instantaneous voltage vector and this vector delayed in time.
harmonics in the grid voltages. A mathematical transformation is proposed through which the
Some other authors proposed different ways to combine original and delayed voltage vectors are combined. It is shown
resonant filters in order to prefilter the and input signals that with the proposed generalized delayed signal cancelation
for determining the positive-sequence vector components. (GDSC) technique, the transformation may be designed so that
These components are then used in an output SRF-PLL for the FFPS voltage vector passes through it with unity gain. On the
obtaining the positive-sequence vector amplitude and position other hand, chosen positive- and negative-sequence harmonics
[15]–[18]. In the so-called dual second-order generalized inte- components are eliminated. Cascaded transformations can then
grator-quadrature signal generator (DSOGI-QSG), the resonant be used for accurately obtaining the FFPS voltage vector. The
filter parameters depend on the frequency detected by an output extracted FFPS is the input to an SRF-PLL for detecting fre-
SRF-PLL in order to achieve a frequency-adaptive strategy. quency and position of the desired voltage vector. Any of these
However, the robustness and stability to frequency variation cascaded transformations is unconditionally stable and no trans-
have not yet been analyzed due to difficulties introduced by the formation is inside the output SRF-PLL control loop.
nonlinearities involved. A frequency-locked loop (FLL) was This paper is organized as follows. The theoretical basis of
proposed in [19] for substituting the output SRF-PLL of the the proposed GDSC technique is presented in Section II. This
DSOGI-QSG. In this case, the stability for small frequency section also presents a way of cascading the proposed trans-
deviations was demonstrated. formations for extracting the FFPS vector. A frequency adap-
Usually, the discrete Kalman filter (DKF) is applied only in tation scheme is then proposed. The performances of the GDSC
single-phase power systems [20]. Recent works propose the use method and other usual methods for obtaining the FFPS voltage
of DKF for detecting the fundamental-frequency positive-se- vector are compared in Section III. The simulation cases were
quence voltage vector in three-phase systems [21]–[23]. The chosen based on the recommendations of usual standards and
KF gains are adjusted for rejecting the effects of low-order har- grid codes of some countries. Simulation and experimental re-
monics or negative-sequence components, but the response is sults are shown in Section IV. Some implementation issues are
slow compared to the previously mentioned techniques. also given. Conclusions are presented and discussed in Sec-
The Fourier transform has been used for many years for sep- tion V.
arating fundamental and harmonic components in single-phase
signals. However, for three-phase signals, there is not an algo- II. PROPOSED METHOD
rithm based on the Fourier transform to directly determine the
A. Theoretical Generalization
FFPS and/or harmonic component vectors.
A frequency-domain detection method for sequence extrac- Consider a three-phase set of unbalanced and distorted sig-
tion from a three-phase signal was proposed in [24]. The method nals described by
is based on different ways of applying the Fortescue transfor-
mation for obtaining the positive- and negative-sequence com-
ponents of the grid voltage in order to eliminate specific har-
monics from the original signals. The signal is passed through
some cascaded mathematical transformations in order to cancel
odd harmonics. However, the even harmonic components are
not eliminated, although some of them may be attenuated. (1)
Four mathematical transformations were used in [25] for The th harmonic component of the signals above can be
eliminating odd and even harmonics. Two of these transforma- written as follows:
tions are made in the stationary reference frame and they
are able to remove odd harmonics from the original signals. The
other two transformations are implemented in a synchronous
reference frame, in order to cancel even harmonic components.
This coordinate transformation requires the angular position of
the FFPS vector, which is obtained from an output SRF-PLL. (2)
Demonstrating the stability of the angular position detection then
scheme in this case becomes a difficult nonlinear problem,
since two mathematical transformation blocks are inside the
output SRF-PLL control loop. As a consequence, the control
parameters of the output SRF-PLL are difficult to be tuned
for optimal performance and a frequency adaptation scheme
was not proposed. This technique is referred to in this paper as
EDSC-PLL (extended DSC-PLL). (3)
This paper presents an original scheme that employs only
simple calculations in the stationary reference frame and
symmetrical components in the time domain [8] for extracting
the FFPS grid voltages. The method is based on computing the
1818 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010

These three-phase signals may be represented by a vector in From (10) and (11), it can be easily verified that the trans-
a stationary reference frame, eliminating the effect of the formed voltage vector is equal to the original signal
homopolar components multiplied by complex gain

(4)

or
(12)
(13)

(5) The real parameters and are determined so that chosen

harmonic components of order , , are
eliminated, where and are chosen as constant integers. This
where the complex quantities indicated by the symbol are is achieved by making
phasors, containing the information about the initial values of
magnitude and phase angle of sinusoidal components
(14)
(6)
The complex constant parameter is determined by imposing
It can also be observed from (5) that the terms inside the the desired transformation gain for a specific frequency vector.
brackets are the Fortescue’s positive- and negative-sequence For example, in order to ensure unity gain for the FFPS voltage
vector
phasors, denoted here by and . Then

(7) (15)

Without loss of generality, the positive- or the negative-se-

B. Implementation of the Proposed Method
quence vector harmonic component can be represented by
An ideal FFPS voltage vector detector should be able to elimi-
(8) nate the fundamental-frequency negative-sequence vector along
with all positive- and negative-sequence harmonic vectors. On
where the other hand, the gain for the FFPS vector should be equal to
for positive-sequence harmonic components one. It is not possible to cancel all harmonic component vec-
for negative-sequence harmonic components. tors using only one transformation of the type described in (11).
(9) However, some transformations may be chosen to be put in cas-
Consider a positive- or negative-sequence vector harmonic cade for eliminating harmonic component vectors around the
component delayed in time by an angle corresponding to ra- FFPS vector.
dians of the fundamental-frequency (or radians with respect As shown in (14), all even harmonic component vectors can
to the harmonic component) be eliminated by a transformation in which and
, where is chosen to keep the unitary gain for
the FFPS voltage vector (15). Thus, the gain for this first
(10) transformation is
It should be noted that if , then the voltage vectors
and rotate counterclockwise and vector is (16)
radians behind . Conversely, if , then the voltage
Four other transformations ( , , , and ) were chosen in
vectors and rotate clockwise and vector is order to eliminate the harmonic vectors of order ,
radians in advance. In both cases, counterclockwise was consid- , , and , respectively.
ered to be the positive direction. The parameters for these transformations are ,
A mathematical transformation in which the original and de- , and ; , and
layed voltage vectors are combined is now proposed ; , and ;
, and . The corresponding gains are
(11)

where the complex gain and the angle are constants. (17)
The aforementioned transformation allows choosing some
specific harmonic components of the original signal to be (18)
eliminated. It is desirable, however, that the transformation
gain for the FFPS voltage vector (or any other chosen harmonic (19)
component to be detected, with positive or negative sequence)
could be made equal to one. (20)
NEVES et al.: GENERALIZED DELAYED SIGNAL CANCELATION METHOD 1819

Fig. 1. Block diagram of the proposed solution without the frequency adaptation scheme.

In order to exemplify the proposed method, transformations It must be noted that the output PLL does not influence trans-
, , , , and are cascaded, as shown in Fig. 1. Using this formations , since they are all implemented in (i.e.,
proposed solution, only the positive- and negative-sequence har- transformations do not use the angular position estimated
monics of order are not canceled by the mathematical by the output SRF-PLL), as occurred in the EDSC-PLL method
operations. These harmonic components can be attenuated by [25]. The output SRF-PLL transfer function may be obtained
adequately choosing the bandwidth of the output SRF-PLL. The by considering that the estimated output vector position is
decision about the bandwidth of the output PLL must take into approximately equal to the correct value [5]. The vector
consideration the increase in the overall response time of the obtained from the to transformation was normalized for
FFPS vector detection scheme versus the total harmonic distor- making the PLL transfer function independent from the voltage
tion (THD) reduction achieved. vector magnitude. For achieving an overshoot of 20% indepen-
The first step for implementing each transformation is mea- dently from the bandwidth, the parameters were calculated by
suring the phase quantities and calculating vector . The considering a relative damping factor [26]. However,
vector components must be stored in order to determine . the natural frequency choice depends on the lowest order har-
The transformed vector is obtained by summing the current and monic component that must be filtered by the output PLL. The
the delayed vectors multiplied by and , respectively. lower the natural frequency is, the lower the bandwidth of the
This is done in a matrix form by performing the following PLL is and the slower its response will be. Using the SRF-PLL,
operation: a low natural frequency must be chosen in order to filter out the
effects of low-order harmonics on the output voltage. For the
proposed method, a much higher bandwidth may be specified,
(21) since most low-order harmonics are eliminated by the cascaded
transformations.
As indicated in (11), the transformed output vector is
where , , , and calculated by using the input vector and from this vector,
. delayed in time by a value corresponding to radians of the
Consequently, transformations , , , , and are imple- fundamental frequency signal. Thus, the samples of the and
mented through very low computational effort operations components of a transformation input vector must be stored
by the time delay period. For this reason, when a transformation
input vector changes, the correct filtering action is achieved only
(22) after the time necessary for the delayed input signal become af-
fected. The maximum overall response time of transformations
(23) corresponds to 345 of the fundamental frequency cycle
(i.e., 19.17 ms if the grid frequency is 50 Hz). However, it should
be pointed out that the proposed transformations are finite-im-
pulse-response (FIR) filters and, therefore, do not change the
input signals frequency is equal to the rated value. Furthermore,
as shown in (27) and (28), if the input signal’s frequency is equal
(24) to the rated value ( , where is the ratio between the ac-
tual and rated grid frequencies), then there will be no error in the
(25) detected magnitude or phase angle. Therefore, the ability of the
output PLL stage to synchronize with the input fundamental fre-
quency and detect the correct angle is maintained. The response
time of the fundamental-frequency positive-sequence vector de-
tection is the sum of transformations delays with that of the
(26) output PLL. On the other hand, a much faster output SRF-PLL
may be used, since it will not need to filter low-order harmonics.
1820 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010

Fig. 3. Block diagram of the proposed solution with the frequency adaptation
scheme.

The detected frequency can be used to determine the number

of samples in one fundamental cycle and the number of samples
corresponding to each delay in transformations .
Fig. 2. Frequency response of the cascaded A 0 E transformations. In order to reduce the GDSC mentioned errors due to fre-
quency variations, a frequency-adaptive scheme using two
GDSC-PLL blocks is proposed, as shown in Fig. 3. The
C. Frequency Adaptation Scheme GDSC-PLL-1 transformations use the rated grid fre-
quency, but as discussed before, give the correct estimate of the
If the grid frequency varies, the delays used in the proposed actual system frequency. This estimated frequency is used to
transformations will not correspond to 180 , 60 , 30 , or 15 . adapt the number of samples of GDSC-PLL-2 transformations
The detection algorithm takes 50 Hz as the fundamental fre- .
quency and the actual frequency will be filtered as an interhar- Using a constant sampling frequency, the number of samples
monic component near the fundamental. In this case, the cas- corresponding to the delays in transformations should
caded transformation gains and phase angles will be different be adjusted when the input frequency varies, for keeping the
than those for 50-Hz signals. transformation gains constant. However, the delay of a trans-
The transformations can be viewed as vector FIR fil- formation may not correspond to an integer number of samples.
ters applied to three-phase systems. An evaluation of the pro- The nearest integer is then used in the GDSC-PLL-2, causing
posed transformations filtering performance can be made by an error in the transformation gain.
computing the absolute value and phase angle of the cascaded In order to compute the effect of the rounding error over op-
gains given in (16)–(20) for different harmonic components, as erations , let be the number of samples that
shown in Fig. 2. In this figure, the negative frequency is used correspond to the delay in (11) and the integer
to indicate negative-sequence vector signals. Thus, in the de- number nearest to . Thus, the delayed signal actually used
scribed nonadaptive GDSC method, a fundamental frequency in the transformation is
variation in the input signals causes magnitude and phase-angle
errors in the positive-sequence detected vector. If transforma- (29)
tions are used in cascade, then the per-unit magnitude
error and phase-angle error in degrees can where .
be obtained by The frequency-adaptive transformation gain becomes equal
to

(27) (30)

(28) Considering a fixed sampling frequency of 18 kHz and using

(30), the magnitude and phase angle of the cascaded
For example, the magnitude and phase-angle errors for a 10% transformations were computed and plotted in Fig. 4. As can be
increased input frequency are equal to 0.016 p.u. and , observed for input signals frequency varying from 40 to 60 Hz,
respectively. the maximum phase-angle error caused by rounding the delays
It can be easily verified from (28) that a frequency devia- to the nearest number of integer sampling periods is around one
tion causes a constant steady-state error in the angular posi- degree. The magnitude error is negligible.
tion of the FFPS estimated vector (output from ). Thus, the A low-pass filter (LPF) is recommended because the GDSC-
transformation output and the input signals have the same PLL-1 output frequency oscillates during the first milliseconds
frequency. As a consequence, the frequency obtained from the after a severe disturbance in the grid signals. Since the frequency
output SRF-PLL is influenced only by components of frequency in a power system does not experiment fast transients, an LPF
, where and is the fundamental frequency. with a low cutoff frequency should be used.
However, the components are strongly attenuated by The computational burden for implementing transformations
the output SRF-PLL. The output frequency is then a very good is very low (16 real multiplications and 18 real sums).
estimate of the input signals’ fundamental frequency. For this reason, the use of two GDSC-PLL blocks for achieving
NEVES et al.: GENERALIZED DELAYED SIGNAL CANCELATION METHOD 1821

TABLE I
INDIVIDUAL HARMONIC VOLTAGE LIMITS IN LOW- AND MEDIUM-VOLTAGE NETWORKS (PERCENT OF FUNDAMENTAL COMPONENTS) [27], [28]

The corresponding compatibility level for the total harmonic distortion is THD = 8%.
0
v (h) = (2:27(17=h) 0:27)% and v (h) = (0:25(10=h) + 0:25)%.

equal to , , and ,
respectively. In the phase-to-phase case, these phase voltages
are , , and ,
respectively. The same fifth and seventh harmonic voltage
vector components of case 1 are considered during the fault.
In the fourth simulation case, highly distorted input signals
are considered. The harmonic components of the three-phase
test voltages are the maximum-allowed values according to IEC
standards [27], [28] as indicated in Table I. These standards do
not specify the angle of each harmonic component vector. In
the simulations presented, each angle is equal to the respective
harmonic order, in degrees.
According to some grid codes [27], wind turbines connected
Fig. 4. Magnitude and phase angle of the detected FFPS signal by using the to the distribution system must remain connected during a rate
0
cascaded adaptive A E transformations. of change in frequency up to and including 0.5 Hz/s. Another
requirement is that a single-phase jump of 20 in the point of
common coupling (PCC) should not initiate a regulation of the
a frequency-adaptive scheme can be executed in usual digital
frequency, which occurs typically if the detected frequency goes
processors.
outside the range Hz Hz. These frequency re-
III. COMPARISON OF SYNCHRONIZATION METHODS quirements are verified through two more simulation tests. In
the fifth test, after the first second, a fundamental-frequency re-
In order to verify the proposed algorithm effectiveness, sim- duction of 0.5 Hz/s is simulated until the frequency reaches 47
ulations were carried out for obtaining the positive-sequence Hz and after that, the frequency remains equal to 47 Hz. In the
fundamental-frequency voltages from unbalanced and distorted last test, a 20 phase jump in the FFPS voltage vector is consid-
input signals. The grid fundamental frequency is 50 Hz and ered in order to verify whether the detected frequency reaches
the sampling frequency is 18 kHz. In all situations, the prefault the limit for initiating the frequency regulation scheme.
voltage is After that, different disturbances are Table II provides a quantitative comparison of the proposed
applied during 120 ms and then the initial balanced undistorted methods with five existing synchronization methods, based
input voltages are restored. These simulation cases are used to on the output total harmonic distortion (THD) and response
compare the performance of the proposed technique with usual time. The methods used for comparison are the SRF-PLL,
synchronization methods. DSRF-PLL, DSOGI-PLL, DKF, EDSC-PLL, and the proposed
The grid codes of most countries require wind farms to re- methods, without (GDSC-PLL) and with the frequency adapta-
main connected to the grid during severe voltage sags. In some tion scheme (GDSC-A-PLL), as shown in Fig. 3.
cases, restrictions are also demanded in terms of the renewable The DKF method used in comparisons with the proposed
energy source ability to control the active and reactive power method is the one presented in [22], where the DKF is modeled
during and after the fault [27]. For attending these requirements, by using four state variables: the three positive-sequence phase
the fast detection of the FFPS voltage vector is very important. components plus a fourth state for allowing the phase-angle
In the first simulation case, a three-phase balanced voltage sag detection without an external PLL. The average values of the
with a phase jump of 20 is considered. During the fault, the re- process and measurement noises are assumed to be zero, with
maining voltage is 15%. Further, 6% of 5th negative-sequence covariance matrices defined as
and 5% of 7th positive-sequence harmonic voltage vectors are
added to the fundamental-frequency components.
The second and third tests consider the application of (31)
phase-to-neutral and phase-to-phase voltage sags, respectively.
In both cases, the test recommendations of IEC 61000 standards The measurement noise covariance is a diagonal matrix
for a system with a remaining voltage of 40% are followed obtained prior to the filter operation. The determination of the
[29]. The phase voltages in the phase-to-neutral voltage dip are process noise covariance is generally more difficult because
1822 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010

TABLE II
COMPARISON OF THE SYNCHRONIZATION METHODS

typically it is not possible to directly observe the estimated output PLL and output filter may have a high bandwidth and,
process. Therefore, it has been selected as a diagonal matrix, thus, a fast response. For each transformation , the output
guaranteeing uncorrelated state variables, with a small value to values of and from the previous cascaded transformation
assign a little uncertainty to the process model. must be stored in order to determine the delayed voltage vector.
In Table II, the numbers inside parentheses indicate the case Thus, the total amount of real variables to be stored for the cas-
simulated as already described. For example, (1) means a three- caded transformations is equal to twice the total number of sam-
phase balanced voltage sag with phase jump and fifth negative- ples necessary for the delays in the cascaded transformations.
sequence and seventh positive-sequence harmonics and (2) and This number for EDSC-PLL and GDSC-PLL methods is also
(3) correspond to the described phase-to-neutral and phase-to- included in Table II.
phase voltage sags, respectively, and so on. It is also shown in
Table II whether the method is capable of compensating signals IV. SIMULATION AND EXPERIMENTAL RESULTS
with dc offset.
The THD presented in Table II is the highest among those Experiments using a OMICRON-CMC256 programmable
calculated for the three phases. In practice, the higher harmonic source [30] to generate the same simulated input voltages of
component of the THD that can be calculated is equal to half of cases (1) to (4) are presented. These signals were acquired
the sample frequency of the input signals. In all simulations and using 12-b converters and processed in a floating point
experiments, the sample frequency is 18 kHz. Then, all THD Texas TMS320C6713 DSP. A sampling rate of 18 kHz was
results have harmonic components up to 9 kHz. For simulation chosen so that the number of samples corresponding to any
cases 1 to 4, the THD was measured during the sixth cycle after of the delays in transformations is an integer [11]. In
the fault so that the positive-sequence detection method had al- order to plot the output quantities, the DSP variables to be
ready been stabilized. For the other cases, the THD was com- examined were sent to digital outputs and saved for plotting
puted right after the frequency reaches its final value of 47 Hz. using MATLAB. Figs. 5–8 show the experimental results. Each
The THD was calculated from one presents three graphics (from top to bottom): input phase
voltages , output phase voltages , and the angle
of the FFPS voltage vector . The detected FFPS voltage
(32) vector magnitude is plotted together with the voltages.
In all cases, this amplitude becomes constant after approxi-
where is the rms value of the output voltage , or mately one fundamental period after the disturbance, which
and is each frequency component. is in accordance with the simulation results. This constant
The response time in Table II is the time necessary for the amplitude indicates that the output signals become balanced
detected angle absolute error to remain inside a tolerance range three-phase signals. Further, the detected angle rate of change
of 1.5 (0.0262 rad). In some cases, the detection method was becomes constant. To recover correctly, the three-phase signal,
not able to guarantee the error to remain in the tolerance range. the estimated , , and are input to the inverse Park
In those cases, the response time was represented by “ .” In (1), transformation.
the detected angle error remained inside the tolerance range and Experimental results showing the error of the detected FFPS
the response time was indicated as 0.00 ms. In all situations, the voltage vector position cannot be plotted since the actual posi-
proposed method led to zero steady-state error. tion is not available. These errors obtained by simulations are
As can be seen, the performance of the proposed technique is presented in Fig. 9. The response times, close to one funda-
better than any of the previous ones for the criteria considered in mental period, are in accordance with the theoretical predic-
the comparison. The response delay in the proposed technique tions. However, a small remaining error is observed in the re-
is caused mainly by the time necessary for obtaining the de- sponses for cases (1) and (3), in which the system experiences
layed signals for the cascaded GDSC transformations, since the a phase jump. This error is caused by the slow dynamics of the
NEVES et al.: GENERALIZED DELAYED SIGNAL CANCELATION METHOD 1823

Fig. 5. Experimental results for case test (1). From top to bottom: the input Fig. 7. Experimental results for case test (3). From top to bottom: the input
voltages, the output voltages, and the FFPS angular position  . voltages, the output voltages, and the FFPS angular position  .

Fig. 6. Experimental results for case test (2). From top to bottom: the input Fig. 8. Experimental results for case test (4). From top to bottom: the input
voltages, the output voltages, and the FFPS angular position  . voltages, the output voltages, and the FFPS angular position  .

detected frequency (Fig. 3) and longer simulations showed fundamental-frequency negative-sequence components and all
that it converges to zero in approximately ten fundamental fre- positive- and negative-sequence harmonic voltage vectors up
quency periods. to 24th order and the offsets. The last transformation output
signals are input to a SRF-PLL for obtaining the synchronous
V. CONCLUSION and components, the fundamental frequency vector, and the
In this paper, a method for obtaining the FFPS voltage position of the FFPS vector. Since the components input to
vector in three-phase grid-connected systems is proposed and the SRF-PLL do not have any low-order harmonic components,
analyzed. The method is based on a generalized delayed signal a high bandwidth PLL may be used. A way of combining two
cancelation process and uses only simple mathematical trans- GDSC-PLL blocks for making the strategy frequency adaptive
formations in the stationary reference frame. It is proved that is also proposed. The presented method performance was
the current and delayed voltage vectors may be combined verified through simulation and experimental tests according to
in a mathematical transformation in order to cancel specific IEC standards. The test cases include balanced and unbalanced
harmonic positive- and negative-sequence voltage vectors from voltage sags, harmonic distortion, and frequency variation.
the original signal. Further, it is shown that cascading five of Compared to five other FFPS detection schemes, the proposed
the proposed transformations makes it possible to cancel the method presented the best performance in terms of detected
1824 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 25, NO. 3, JULY 2010

[10] M. Karimi-Ghartemani, “A novel three-phase magnitude-phase-locked

loop system,” IEEE Trans. Circuits Syst. I, vol. 53, no. 8, pp.
1792–1802, Aug. 2006.
[11] J. Svensson, M. Bongiorno, and A. Sannino, “Practical implementation
of delayed signal cancellation method for phase-sequence separation,”
IEEE Trans. Power Del., vol. 22, no. 1, pp. 18–26, Jan. 2007.
[12] G. Saccomando and J. Svensson, “Transient operation of grid con-
nected voltage source converter under unbalanced voltage conditions,”
in Proc. IEEE Ind. Appl. Soc., 2001, pp. 2419–2424.
[13] M. Bongiorno, J. Svensson, and A. Sannino, “Effect of sampling
frequency and harmonics on delay-based phase-sequence estimation
method,” IEEE Trans. Power Del., vol. 23, no. 3, pp. 1664–1672, Jul.
2008.
[14] Z. Yao, “Fundamental phasor calcultion with short delay,” IEEE Trans.
Power Del., vol. 23, no. 3, pp. 1280–1287, Jul. 2008.
[15] X. Yuan, W. Merk, H. Stemmler, and J. Allmeling, “Stationary-frame
generalized integrators for current control of active power filters with
zero steady-state error for current harmonics of concern under unbal-
anced and distorted operating conditions,” IEEE Trans. Ind. Appl., vol.
38, no. 2, pp. 523–532, Mar./Apr. 2002.
[16] R. I. Bojoi, G. Griva, V. Bostan, M. Guerriero, F. Farina, and F. Pro-
fumo, “Current control strategy for power conditioners using sinusoidal
signal integrators in synchronous reference frame,” IEEE Trans. Power
Electron., vol. 20, no. 6, pp. 1402–1412, Nov. 2005.
[17] P. Rodríguez, R. Teodorescu, I. Candela, A. Timbus, M. Liserre, and
F. Blaabjerg, “New positive-sequence voltage detector for grid syn-
chronization of power converters under faulty grid conditions,” in Proc.
IEEE Power Electronics Specialists Conf., Jun. 2006, pp. 1–7.
[18] L. R. Limongi, R. I. Bojoi, C. Pica, F. Profumo, and A. Tenconi, “Anal-
ysis and comparison of phase locked loop techniques for grid utility ap-
plications,” in Proc. IEEE Power Conversion Conf., 2007, pp. 674–681.
[19] P. Rodríguez, A. Luna, I. Candela, R. Teodorescu, and F. Blaabjerg,
“Grid synchronization of power converters using multiple second order
generalized integrators,” in Proc. IEEE Annu. Conf. Industrial Elec-
tronics Soc., Nov. 2008, pp. 755–760.
Fig. 9. Simulation results of the error in the detected FFPS voltage vector an- [20] V. Moreno, M. Liserre, A. Pigazo, and A. Dell’Aquila, “A compara-
gular position for cases (1) to (4), respectively. tive analysis of real-time algorithms for power signal decomposition in
multiple synchronous reference frames,” IEEE Trans. Power Electron.,
vol. 22, no. 4, pp. 1280–1289, Jul. 2007.
0
[21] A. Pigazo and V. Moreno, “ 3 3w signal model for power system
voltages THD and time response of the voltage vector-angle harmonics and unbalance identification using Kalman filtering,” IEEE
detection. Trans. Power Del., vol. 23, no. 2, pp. 1260–1261, Apr. 2008.
[22] M. S. Pádua, S. M. Deckmann, F. P. Marafão, and D. Colón, “Sim-
plified models of Kalman filter for fundamental frequency, amplitude
REFERENCES and phase angle detection,” in Proc. Brazilian Power Electronics Conf.,
Sep. 2007, pp. 175–180.
[1] J. G. Nielsen, M. Newman, H. Nielsen, and F. Blaabjerg, “Control and [23] M. S. Pádua, S. M. Deckmann, and F. P. Marafão, “Frequency-ad-
testing of a Dynamic Voltage Restorer (DVR) at medium voltage level,” justable positive sequence detector for power conditioning applica-
IEEE Trans. Power Electron., vol. 19, no. 3, pp. 806–813, May 2004. tions,” in Proc. IEEE Power Electronics Specialists Conf., 2005, pp.
[2] V. Kaura and V. Blasko, “Operation of phase locked loop system under 1928–1934.
distorted utility conditions,” IEEE Trans. Ind. Appl., vol. 33, no. 1, pp. [24] R. Cutri and L. Matakas, “A fast instantaneous method for sequence
58–63, Jan./Feb. 1997. extraction,” in Proc. Brazilian Power Electronics Conf., Oct. 2007, pp.
[3] D. Yazdani, A. Bakhshai, G. Joos, and M. Mojiri, “A nonlinear adap- 1–6.
tive synchronization technique for grid-connected distributed energy [25] H. E. P. de Souza, F. Bradaschia, F. A. S. Neves, M. C. Cavalcanti,
sources,” IEEE Trans. Power Electron., vol. 23, no. 4, pp. 2181–2186, G. M. S. Azevedo, and J. P. Arruda, “A method for extracting the fun-
Jul. 2008. damental frequency positive-sequence voltage vector based on simple
[4] L. N. Arruda, S. M. Silva, and B. J. C. Filho, “PLL structures for utility mathematical transformations,” IEEE Trans. Ind. Electron., vol. 56, no.
connected systems,” in Proc. Industry Applications Soc. Annu.Meet., 5, pp. 1539–1547, May 2009.
Sep./Oct. 2001, vol. 4, pp. 2655–2660. [26] E. J. Bueno, F. J. Rodriguez, F. Espinosa, and S. Cobreces, “SPLL
[5] S.-K. Chung, “A phase tracking system for three phase utility interfece design to flux oriented of a VSC interface for wind power applications,”
inverters,” IEEE Trans. Power Electron., vol. 15, no. 3, pp. 431–438, in Proc. IEEE Annu. Conf. Industrial Electronics Soc., Nov. 2005, pp.
May 2000. 1–6.
[6] P. Rodríguez, J. Pou, J. Bergas, I. Candela, R. Burgos, and D. Boroye- [27] F. Iov and F. Blaabjerg, “Advanced power converters for universal and
vich, “Decoupled double synchronous reference frame PLL for power flexible power management in future electricity network.” Delivery
converters control,” IEEE Trans. Power Electron., vol. 22, no. 2, pp. D2.1 of UNIFLEX-PM. [Online]. Available: www.eee.nott.ac.uk/uni-
584–592, Mar. 2007. flex/Documents/W2-AU-DV-2001-B.pdf, Feb. 2007
[7] M. Karimi-Ghartemani and M. R. Iravani, “A method for synchroniza- [28] M. McGranaghan and G. Beaulieu, “Update on IEC 61000-3-6: Har-
tion of power electronic converters in polluted and variable-frequency monic emission limits for customers connected to MV, HV and EHV,”
environment,” IEEE Trans. Power Syst., vol. 19, no. 3, pp. 1263–1270, in Proc. IEEE Transmission and Distribution Conf. Exhibit., May 2006,
Aug. 2004. pp. 1158–1161.
[8] M. Karimi-Ghartemani and H. Karimi, “Processing of symmetrical [29] Testing and measuring techniques—Voltage dips, short interruptions
components in time-domain,” IEEE Trans. Power Syst., vol. 22, no. 2, and voltage variations immunity tests for equipment with input current
pp. 572–579, May 2007. more than 16 A per phase, IEC Std. 61000-4-34 Electromagnetic Com-
[9] U. K. Rao, M. K. Mishra, and A. Ghosh, “Control strategies for load patibility (EMC) Part 4-34, 2004.
compensation using instantaneous symmetrical component theory [30] Omicron CMC256—4 Phase Voltage/6 Phase Current Test Set. OMI-
under different supply voltages,” IEEE Trans. Power Del., vol. 23, no. CRON–Testing solutions for protection and measurement systems.,
4, pp. 2310–2317, Oct. 2008. 2003.
NEVES et al.: GENERALIZED DELAYED SIGNAL CANCELATION METHOD 1825

Francisco A. S. Neves (M’00) was born in Campina Fabrício Bradaschia was born in São Paulo, Brazil,
Grande, Brazil, in 1963. He received the B.S. and in 1983. He received the B.S. and the M.S. degrees
M.Sc. degrees in electrical engineering from the in electrical engineering from the Federal University
Federal University of Pernambuco, Recife, Brazil, in of Pernambuco, Recife, Brazil, in 2006 and 2008, re-
1984 and 1992, respectively, and the Ph.D. degree in spectively, where he is currently pursuing the Ph.D.
electrical engineering from the Federal University of degree.
Minas Gerais, Belo Horizonte, Brazil, in 1999. From 2008 to 2009, he was a Visiting Scholar
He was a Visiting Scholar at the Georgia Institute at the University of Alcalá, Madrid, Spain. His
of Technology, Atlanta, in 1999, and at Alcala Uni- research interests are direct ac-ac converter topolo-
versity, Spain, from 2008 to 2009. Since 1993, he has gies with high efficiency, pulsewidth modulation
been with the Department of Electrical Engineering, for three-phase matrix converters, discrete-time
Federal University of Pernambuco, where he is currently a Professor of Elec- controls, and grid synchronization methods.
trical Engineering. His research interests include power electronics, renewable
energy systems, power quality, and grid synchronization methods.

Emilio J. Bueno (S’05–M’06) was born in Madrid,

Spain, in 1972. He received the M.S. and Ph.D.
Marcelo Cabral Cavalcanti (M’07) was born in degrees in electronics engineering from the Univer-
Recife, Brazil, in 1972. He received the B.S. degree sidad de Alcalá, Madrid, Spain, in 1999 and 2005,
in electrical engineering from the Federal University respectively.
of Pernambuco, Recife, Brazil, in 1997 and the M.S. Currently, he is an Associate Professor with
and Ph.D. degrees in electrical engineering from the the Department of Electronics, Universidad de
Federal University of Campina Grande, Campina Alcalá, where he is a member of the Research
Grande, Brazil, in 1999 and 2003, respectively. Group “Electronics Engineering Applied to the
Since 2003, he has been with the Department Renewable Energies.” His research interests include
of Electrical Engineering, Federal University of power-electronic systems, control of power-elec-
Pernambuco, where he is currently a Professor of tronic converters, power quality, and distributed power generation systems.
Electrical Engineering. His research interests are
renewable systems, power quality, and three-phase matrix converters.

Mario Rizo received the M.D. degree in telecom-

munication engineering from the Alcalá University,
Helber Elias Paz de Souza was born in Cabo de Madrid, Spain, in 2009, where he is currently pur-
Santo Agostinho, Brazil, in 1983. He received the suing the Ph.D. degree.
B.S. and M.S. degrees in electrical engineering Since 2007, he has been a member of the Elec-
from the Federal University of Pernambuco, Recife, tronic Engineering Applied to Renewable Energies
Brazil, in 2006 and 2008, respectively, where he is Research Group of Alcalá University. His research
currently pursuing the Ph.D. degree. areas include massive integration of renewable ener-
Since 2009, he has been with the Department of gies in power systems, control of power converters
Industry, Federal Institute for Education, Science and used in renewable energy systems, and power quality.
Technology of Pernambuco. His research interests
are grid synchronization methods, power quality,
and three-phase matrix converters.