IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO.
9, SEPTEMBER 2011
Frequency-Adaptive Stationary-Reference-Frame
Grid Voltage Sequence Detector for Distributed
Generation Systems
Eider Robles, Josep Pou, Member, IEEE, Salvador Ceballos, Jordi Zaragoza, Student Member, IEEE,
José Luis Martín, Member, IEEE, and Pedro Ibañez, Member, IEEE
Abstract—This paper proposes a stationary-frame sequence de-
tector (SFSD) structure for determining the positive sequence in
three-phase systems. The structure includes the use of the Clarke
transformation and moving average filters (MAFs). The positive-
sequence phase angle can be obtained from the alpha–beta com-
ponents; however, such a detection becomes inaccurate if the grid
voltages are unbalanced and/or distorted. The MAF is used to fil-
ter the nonideal components. Performance of the MAF is analyzed
mathematically for a proper selection of the window width of the
optimal filter in this application. The time delay introduced by
MAFs is constant and known; hence, this may be compensated.
The proposed detector structure allows fast detection of the grid
voltage positive sequence (within one grid voltage cycle). The MAF
completely eliminates any oscillation multiple of the frequency
for which it is designed; thus, this algorithm can overcome the
presence of imbalances or harmonics in the electrical grid. Fur-
thermore, it includes a simple frequency estimator that makes
the SFSD frequency adaptive and is capable of operating under
large frequency changes. The entire SFSD is verified through
simulation and experiment, showing very good performance even
under several extreme grid voltage conditions.
Index Terms—Frequency estimation, grid-connected convert-
ers, moving average filter (MAF), sequence detector, stationary
reference frame.
I. I NTRODUCTION
D ETECTION of the fundamental-frequency positive-
sequence component of the utility voltage under un-
balanced and distorted conditions has been a major issue
nowadays. This information is needed for the proper operation
of grid-connected power electronic systems such as power
quality conditioners [1], [2], distributed generation and stor-
age systems, uninterruptible power supplies, and flexible ac
Manuscript received August 14, 2009; revised April 21, 2010 and August 31,
2010; accepted October 23, 2010. Date of publication December 10, 2010; date
of current version August 12, 2011. This work was supported by the Ministerio
de Ciencia y Tecnologia of Spain under Project ENE2007-67033-C03-01, by
the Basque Country Government, and by the Torres Quevedo Program.
E. Robles, S. Ceballos, and P. Ibañez are with the Energy Unit,
TECNALIA, 48170 Zamudio, Spain (e-mail: eider.robles@tecnalia.com;
salvador.ceballos@tecnalia.com; pedro.ibanez@tecnalia.com).
J. Pou and J. Zaragoza are with the Terrassa Industrial Electronics Group,
Department of Electronic Engineering, Technical University of Catalonia,
08222 Terrassa, Spain (e-mail: pou@eel.upc.edu; zaragoza@eel.upc.edu).
J. L. Martin is with the Department of Electronics and Telecommuni-
cations, University of the Basque Country, 48013 Bilbao, Spain (e-mail:
joseluis.martin@ehu.es).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIE.2010.2098352
0278-0046/$26.00 © 2010 IEEE
4275
transmission systems. Many synchronizing methods have been
presented over recent years. Some of these are based on a phase-
locked loop (PLL), including a feedback loop, with the purpose
of controlling a magnitude which is dependent on the voltage
phase angle [3]–[6].
PLLs are possibly the most widely used synchronization
methods in the case of grid-connected three-phase systems [7],
[8]. The dynamic response of some methods, such as the one
described in [9], is good under balanced grid voltages but very
slow when these are designed to operate under unbalanced or
distorted grid voltages. The fast PLL method presented in [10]
can cancel out the influence of certain harmonics; however, the
method has to be implemented to each harmonic separately
and should be limited to canceling out the effect of only a few
harmonics. The performance of the PLL-based synchronization
method used in [11] is good, even under voltage imbalances and
distortion. Nevertheless, in order to implement this, a number of
practical problems relating to the normalization of the detected
angle within the moving average filter (MAF) stage must be
resolved in order to avoid overflow. Additionally, this solution
was not designed to operate under variable frequency.
Other kinds of synchronization algorithms are based on the
instantaneous symmetrical component theory [12]–[14]. Some
of these have quick dynamic response and can work with
variable frequency, but grid harmonics are only attenuated and
not completely canceled out.
Alternative methods [15], [16] also apply the symmetric
component theory and extract the fundamental-frequency pos-
itive sequence by performing simple calculations. This com-
pletely eliminates several harmonics and employs finite impulse
response (FIR) filters to attenuate the rest.
Some of these methods are designed to work at a nominal
frequency and allow little deviation from this. To overcome this
drawback, frequency detectors are included in their structures to
make them frequency adaptive [17]. A number of other struc-
tures have inherently frequency-adaptive capability [18]–[20].
Nevertheless, these approaches usually increase the complexity,
worsen the dynamic response, and decrease the capability to
work in the presence of distorted voltages.
The use of MAFs in the detection of the positive sequence
is relatively recent [21]–[24]. They act as ideal filters when
the window width is properly selected for this application.
However, the inclusion of MAFs in the closed loop of a PLL
slows down the response.
4276 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011
The synchronization method presented in this paper is based
on a stationary reference frame and MAFs. Since the MAFs
are not within any closed loop, the system is always stable.
The variables are properly processed to obtain the phase of the
grid voltage positive sequence. The Clarke transformation is
used to obtain the positive- and negative-sequence components.
This method allows rapid detection of grid sequences and is no
longer affected by the presence of imbalances or harmonics in
the electrical grid. Furthermore, the stationary-frame sequence
detector (SFSD) can also operate very well under frequency
changes due to the inclusion of a frequency detector.
This paper is organized as follows. Grid voltages are char-
acterized in Section II. Section III reviews the principles of
operation of MAFs. Sections IV and V show the general
structure of the proposed SFSD and a number of simulation
results. Section VI is devoted to the frequency detector and
variable-frequency SFSD. Section VII analyzes the maximum
error due to discrete-time implementation. The experimental
results are described in Section VIII, and Section IX sets out
the conclusions.
II. C HARACTERIZATION OF G RID VOLTAGES
IN αβ C OORDINATES
Grid voltages can be considered a positive sequence with
the addition of negative and zero sequences (unbalanced volt-
ages). They can also include harmonic components which can
be either balanced or unbalanced (harmonic distortion). The
objective of a positive-sequence detector is to determine the
magnitude and phase of the positive-sequence fundamental
component, disregarding all unwanted components in the grid
voltages.
The grid voltages containing the fundamentals, harmonics,
which can be positive or negative sequences, and the zero-
sequence components are represented as follows:
⎡ ⎤ ⎡ ⎤
va cos(ωt)
⎣ vb ⎦ = V+1 ⎣ cos(ωt − 2π/3) ⎦
vc cos(ωt + 2π/3)
⎡ ⎤ ⎡ ⎤
∞ cos(nωt + θn ) v0
+ Vn ⎣ cos(nωt − 2π/3 + θn ) ⎦ + ⎣ v0 ⎦
n=−∞ cos(nωt + 2π/3 + θn ) v0
n=+1
(1)
in which V+1 and Vn are the amplitudes of the fundamentals
and harmonics, respectively. The zero sequence is represented
by v0 . Note that, for n = −1, the second term in (1) corre-
sponds to a fundamental negative sequence.
Fig. 1. Voltage vector on the αβ plane.
Under ideal grid voltages, i.e., including only the first term
in (1), the vα and vβ variables are sinusoidal waveforms with a
π/2 phase shift, whose amplitude is V+1 3/2.
Fig. 1 shows a vector representation on the αβ plane, where
V is the grid voltage vector, vα and vβ are the components of
the vector, and θ is the position angle on the plane. This angle
can be obtained by performing a simple arctangent operation
to the transformed voltage variables (arctan(vβ /vα )). In this
case, since the grid voltages are ideal (balanced and undis-
torted), this angle corresponds to the positive-sequence angle.
On the other hand, according to (2), imbalances and distor-
tion will cause oscillations in both components (vα and vβ ) and,
consequently, in the arctan(vβ /vα ) result. In this case, it will
not represent the positive-sequence angle, and the oscillations
should be removed. With n being the oscillation order in the
original abc frame and also in the transformed αβ variables,
the main order of the oscillations after the arctangent operator
is n − 1, although other frequency components appear. This is
shown in the following.
To simplify the mathematical analysis, the arctangent is
performed on the vd and vq components (Park-transformed
variables) of the grid voltages instead of on the vα and vβ
variables. As a consequence, the angle obtained, which corre-
sponds to the fundamental positive sequence, is constant instead
of continuously increasing.
Assuming a fundamental positive sequence of amplitude A
and an n-order harmonic set of amplitude B  , if the fundamental
sequence is synchronized with the d-axis, the dq components
will be
vd = A + B cos α
vq = B sin α (3)
where α = (n − 1)ωt for a positive-sequenceharmonic, α =
+ 1)ωt for a negative sequence, A = A 3/2, and B =
(n 
B  3/2.
The angle is calculated as follows:
The voltage vector can be represented in αβ coordinates 

through the Clarke transformation, using a stationary reference vq
θ = θ0 + ωt + arctan (4)
frame. If this transformation is applied to the voltages in (1), vd
the following vα and vβ components are obtained:


 where θ0 is the initial angle between the d- and α-axes, which
∞
vα  3 cos(ωt)  3 cos(nωt + θn ) is henceforth assumed to be zero. From (3)
= V+1 + Vn .
vβ 2 sin(ωt) n=−∞
2 sin(nωt + θn ) vq B sin α
n=+1 = (5)
(2) vd A + B cos α
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS 4277

Fig. 2. Performance example of the MAF.

which results in The fundamental negative sequence caused by voltage dips

may be considered to be odd-order harmonics; therefore, it will
vq B B2 1 B3 1 generate even-order components on the detected angle.
= sin α − 2 sin(2α) + 3 [sin α + sin(3α)]
vd A A 2 A 4 The influence of several harmonics has also been analyzed.
4
 In this case, the arctangent is performed on the following dq
B 1 1
− 4 sin(2α) + sin(4α) . . . . (6) components:
A 4 8
vd = A + B cos(α1 ) + C cos(α2 )
In general terms, representation of arctan(x) by means of
the Taylor series leads to the following equations: vq = B sin(α1 ) + C sin(α2 ). (11)

∞
(−1)i 2i+1 Using the approach described in (8), the angle is shown to
arctan(x) = x , for |x| ≤ 1 (7)
2i + 1 be dependent on a complex sum of sinusoidal components, the
i=0
frequency of which is dependent on an arithmetic combination
or of the input harmonic frequencies. In the worst case, it may
happen that the sum of the angles in a sinusoidal component
x3 x5 x7 is constant, thus representing a continuous component in the
arctan(x) = x − + − + ···, for |x| ≤ 1.
3 5 7 detected angle. This will produce an error even after low-pass
(8) filtering the angle. Fortunately, in practice, this error is very
Assuming harmonic amplitudes to be smaller than the fun- small and infrequent.
damental one (B  < A or B < A), (6) can be truncated to the
third component because the following components are usually
very small. For the same reason, the Taylor series in (8) can be III. MAF
truncated to the second term. By replacing (6) in (8), a complex To obtain the real angle of a positive sequence, the oscillating
sum of sinusoidal terms is obtained. The general expression can components have to be removed. A MAF structure may be very
be written as useful because, under certain conditions, it can perform as an

∞
ideal low-pass filter. Performance of the MAF is analyzed as
B sin α  follows.
θ = ωt + arctan ≈ ωt + ki sin(i · α)
A + B cos α i=1
Application of the MAF operator to an input signal x(τ ) is
(9) given by
where
t
∞
 
j x̄(t) =
B Tw
ki = Ci and j increases in twos.
j=i
A t−Tw
From (9) and (10), one can deduce that an n-order harmonic
set in the grid voltages will cause infinite harmonic components
in the performed arctan(vq /vd ) and, therefore, in the detected
angle θ. However, the frequencies of these oscillations are
known and correspond to the following.
1) Odd-order harmonics generate even-order frequency os-
cillations in the dq components; therefore, the angle will
only include even-order sinusoidal components.
2) Even-order harmonics generate odd-order frequency os-
cillations in dq components; therefore, the angle will
include both even- and odd-order sinusoidal components.
1
x(τ )dτ. (12)
(10)
An example of the performance of the MAF is shown in
Fig. 2. The filter provides the mean value of the input sig-
nal in the time period from t − Tw to t, where Tw is the
window width. There is always an intrinsic delay associated
with the window width; therefore, the larger Tw is, the slower
the detection dynamic. If the input signal contains sinusoidal
components of an integer multiple of the equivalent frequency
of the MAF (fw = 1/Tw ), the output signal will be free of
oscillations, containing only the mean value of the input signal.
The MAF will need a period of time Tw to gather the data and
obtain the correct output; hence, this is the maximum delay
produced by the MAF.
4278 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011

Fig. 3. Bode diagram of the MAF.

In Fig. 2, the window width is 0.01 s, exactly the same as

the input signal period that might correspond to a negative
sequence produced by a voltage dip. The MAF can eliminate
the input oscillation after gathering the data. When the input
signal changes, it contains a slight error during Tw until it has
an integer number of oscillations again and obtains the mean
value.
Selection of the window width is an important issue. Depend-
ing on this, the MAF is able to perform as an ideal low-pass
filter for some input frequency components.
For a better understanding of the MAF, the transfer function
has been obtained and analyzed. Given an input signal

x(t) = A sin(ωt + θ0 ) (13)

with ω = 2π/T . The output of the MAF will be

A
x(t) = {cos [ω(t − Tw ) + θ0 ] − cos(ωt + θ0 )} . (14)
ωTw Fig. 4. Performance of the MAF with slope inputs.

Applying the Laplace transform to (13) and (14), the transfer Fig. 3 shows the Bode diagram of this transfer function
function for θ0 = 0 becomes in which the frequency axis has been normalized (f /fw =
Tw /T ). Observe that the frequency components that comply
X(s) s [cos(ωTw ) − 1] + ω sin(ωTw ) with (20) are completely canceled out.
GMAF (s) = = . (15)
X(s) ω 2 Tw If the average value of the input signal remains constant
and Tw is properly adjusted, the output of the MAF does not
Substituting s = jω produce any steady-state error. However, if the input signal
X(jω) j [cos(ωTw ) − 1] + sin(ωTw ) increases or decreases with a constant slope x(t) = m · t, the
GMAF (jω) = = . output of the MAF includes a constant error given by d =
X(jω) ωTw
(16) mTw /2 (Fig. 4). Consequently, this value has to be added to
The magnitude and phase expressions are, respectively the output of the MAF to compensate for this error.
√
2
|GMAF (jω)| = 1 − cos(ωTw ) (17) IV. SFSD
ωTw
 Fig. 5 shows the structure of the proposed positive-sequence
cos(ωTw ) − 1 detector. By means of the Clarke transformation, the input
ϕ [GMAF (jω)] = arctan
sin(ωTw ) voltage vector is transformed into αβ components. These trans-
ωTw formed components will be sinusoidal and may contain oscil-
=− . (18) lations due to harmonics and a negative voltage sequence. A
2
four-quadrant arctangent operation is performed, and the input
The MAF is an FIR filter that, as shown in (18), has a linear voltage phase angle is detected. As well as the αβ components,
phase, with the time of the delay being constant and known. this phase angle may contain oscillations. It is thus filtered
From (17), the output magnitude of the MAF is zero when- through a MAF. Since the angle increases with a constant slope,
ever [1 − cos(ωTw )] = 0, i.e., the error produced by the MAF has to be compensated.
If the window width is Tw = Tg , with Tg being the grid
ωTw = k2π, for k = 1, 2, 3, . . . (19)
voltage period, the correction constant to compensate for the
or angle error is π. Similarly, the compensation angle is π/2 if the
selected window width of the MAF is Tw = Tg /2. Once the er-
Tw f ror is compensated, the positive-sequence angle θ+ is obtained,
= = k, for k = 1, 2, 3, . . . . (20)
T fw free of oscillations due to disturbances and imbalances.
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS
Fig. 5. Structure of the proposed PLL.
Fig. 6. Performance example: (a) Grid voltages, (b) phase angle, and (c) error
in the detected phase angle.
Fig. 6 shows a performance example of the proposed struc-
ture. The rms input voltage is 220 V at 50 Hz. At t = 30 ms,
the following odd harmonic set is included: 3rd (20%), 5th
(50%), and 7th (20%). The window width of the MAF is chosen
as Tw = Tg /2. Fig. 6(b) shows the four-quadrant arctangent
obtained (blue). Observe that it contains oscillations due to
the harmonic components. The output signal of the arctangent
operator is then filtered through a MAF, obtaining an angle
(green dotted line) which includes a steady-state error. The
angle error is properly compensated adding a constant value of
π/2, and the positive-sequence angle is obtained (red).
Fig. 6(c) shows the error between the real positive-sequence
angle and the detected one. Note that, although some transitory
errors are produced due to the intrinsic delay of the MAF (Tw ),
they are completely canceled out in the steady state.
Once the positive angle is detected, it is possible to obtain
both the positive and the negative sequences by applying the
Park transformation to the original abc input voltages with θ+
for the positive sequence and θ− = −θ+ for the negative se-
quence. The dq components obtained may contain oscillations
that can be removed using a MAF once more. In this case, it is
not necessary to compensate for steady-state errors because the
average voltage of the variables remains constant in the steady
state.
4279
Fig. 7. Constant-frequency SFSD simulation results.
V. S IMULATION R ESULTS
Fig. 7 shows a simulation test performed in the SFSD
structure. The rms line-to-neutral grid voltage is 220 V at
50 Hz, and all the MAFs have a window width of Tw = Tg /2,
where Tg is the grid voltage period. The main variables of the
proposed SFSD are displayed: (a) the grid voltages, (b) the real
and detected angles, (c) the error in the detected angle, (d) the
positive-sequence dq components, (e) the negative-sequence dq
components, and (f) the input voltage positive sequence. Fig. 8
shows a zoom of the error in the detected angle. The MAF needs
a Tw period of time to collect all the data (10 ms). After that, the
SFSD starts detecting the angle without error. At t = 30 ms, an
asymmetrical voltage dip and a π/2 phase jump are introduced.
Observe that the performance of the SFSD is very good, rapidly
eliminating the error in the detected angle within half a grid
period (10 ms) when the disturbance appears. At t = 60 ms,
an odd harmonic component set is added. The error is again
corrected within 10 ms; however, the zoom in Fig. 8 reveals
the small permanent error due to the harmonic combination, as
explained in Section II.
4280
Fig. 8. Zoom of the detected angle error of the simulation in Fig. 7.
From the input voltages and the detected angle θ, the
positive- and negative-sequence dq components are obtained.
These components may contain oscillations produced due, in
the first place, to the voltage dip process and, next, due to
the harmonic distortion. Therefore, they are filtered through
MAFs. Notice how a fundamental negative sequence appears
due to the voltage dip process and that this is zero when there
is no voltage dip. The effect of the transitory lasts longer in
these components because they are affected by the delay of two
consecutive MAFs and, hence, the longer error of one input
signal period (Tg = 2Tw ).
From the results, one can conclude that the proposed SFSD
can overcome phase jumps, voltage dips, and high harmonic
distortion. At t = 100 ms, the grid frequency rises to 55 Hz.
Now, the detected angle shows an error of 0.2 rad approxi-
mately, even when disturbances in the grid voltages disappear.
This is due to the fact that the window width of the MAF is
set to a different frequency; thus, it will not contain an integer
number of oscillations and will not be able to filter this properly.
As shown in this example, the performance of the proposed
positive-sequence detector is very good in the presence of grid
disturbances. However, a grid frequency other than the nominal
value will produce an error in the detected angle.
VI. VARIABLE -F REQUENCY SFSD
A. Frequency Detector
To make the SFSD frequency adaptive, the real frequency
must be measured to update the frequency-dependent blocks.
Several methods for the frequency measurement have been
proposed in the literature. There are the methods based on zero-
crossing detection [25]–[32], Fourier transform [29], [33], [34],
Kalman filtering [35], PLL based [36], and adaptive notch filter
based [37].
Among them, the ones based on zero-crossing detection are
the simplest. The main idea of these methods is to measure
the time elapsed between two consecutive zero crossings of the
grid voltage and, as a function of this time, estimate the fre-
quency. These methods work quite well under undistorted grid
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011
several algorithms have been proposed [25]–[32]. Almost all
of them are based on the attenuation of the undesired harmonic
components and noise as a prior stage. For example, the authors
in [25] and [26] propose a prior filter stage based on an FIR
filter. A more complicated multistage filter is introduced in [27]
and [28]. In both cases, the aim of including the filters is to
attenuate the grid voltage distortion.
In [29], the required filter stage is implemented making use
of a discrete Fourier transformation (DFT) working with a fixed
window width. In this case, if the width of the window is equal
to the period of the grid voltage, the DFT algorithm works as
an ideal low-pass filter. On the other hand, if the width of the
window does not match exactly the period of the voltage, the
DFT does not work as an ideal low-pass filter, but its behavior
is still good enough to estimate the frequency. Finally, in a
number of studies [30]–[32], grid voltages are approximated
by low-order polynomial, coefficients of which are obtained
making use of the least square technique. In those cases, due
to the low order of the chosen polynomial, the high-frequency
components of the grid voltages are attenuated.
The positive-sequence detector proposed in this paper does
not need a very accurate frequency measurement to operate
properly. A slight frequency change affects the MAF output
signal because the window width will not correspond exactly
to the input signal period. Consequently, imbalances and har-
monic disturbances will not be completely canceled out by the
MAF. This error is analyzed in the following. Given a distorted
angle
∞

θ(t) = mt + An sin(nmt) (21)
n=1
where m is the slope of the angle, n is the harmonic order of
the angle, and An is the harmonic amplitude. The output of the
MAF will be
∞
Tw  An 1
θ(t) = mt−m − [cos(nmt)−cos(nm(t−Tw ))] .
2 n=1 Tw nm
(22)
When the frequency is detected properly, the window width
Tw will correspond to the input signal period Tw = T (or
half period T /2). The slope of the angle will be m = 2π/Tw .
Substituting in (22) the filtered angle will result in
2π
θ(t)ideal = t−π (23)
T
in which π (or π/2) is the constant delay that requires
compensation.
When there is a slight difference between the real and de-
tected frequencies, Tw will be different from T . Substituting
m = 2π/T in (22) results in
2π Tw
θ(t)real = t−π
T T
conditions. However, in the presence of harmonics or other grid ∞  


An T 2π 2π
disturbances, they may lead to erroneous frequency estimations − cos n t −cos n (t−Tw ) .
T n2π
n=1 w
T T
due to the nonreal crossing points caused by the distortion.
In order to improve its behavior under these circumstances, (24)
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS 4281

Fig. 10. Zero crossing in the accepted band.

Fig. 9. Block diagram of the frequency detector.

Fig. 11. Zero crossing outside the accepted band.

The error produced because the MAF window width is errors due to transients when disturbances appear is to use both
not properly adjusted is determined by subtracting (23) from crossing points and calculate the frequency separately.
(24). This error can be separated into two terms: a constant The following situations may occur depending on the loca-
component (e1 ) and an oscillating component (e2 ) tion of the next zero-crossing point.

1) A zero-crossing point is found in the 25-Hz/s band. It
Tw
e1 = π 1 − (25) is considered to be a valid value, and the frequency is
T
updated. This situation is shown in Fig. 10. Grid voltages
∞
An T start undisturbed, and suddenly, one harmonic component
e2 = is added. In this case, the transition is not strong enough,
n=1 w
T n2π
 

and the zero-crossing point is found within the band. Any
2π 2π
× cos n t − cos n (t − Tw ) . (26)
T T
From (25) and (26), one can deduce that if the difference
between Tw and T is small, the total error in the detected angle
will be small and hence acceptable.
Therefore, in this paper, a relatively simple but efficient
adaptive zero-crossing method is proposed. Fig. 9 shows the
flowchart of this detection method. At the beginning of the
process, two zero-crossing points are needed to calculate
the initial frequency. It has been designed to allow frequency
changes of up to 25 Hz/s. Hence, every time a new frequency is
measured, the upper and lower frequency limits are calculated
to determine a 25-Hz/s frequency band for the next zero-
crossing point. Then, it will wait for the following zero-crossing
point and proceed differently depending on whether it is within
the band or not. Both the positive- and negative-slope zero-
crossing points can be used for detecting the frequency in order
to improve the detection dynamics. The best way to avoid
other zero-crossing point before or after the band may not
be considered if there has been a crossing point within
the band.
2) There are several zero-crossing points within the band.
When the first zero-crossing point is found, the frequency
is updated, and the new band limits are calculated. The
algorithm will look for a new zero-crossing point much
further from the second zero-crossing point. Hence, this
will be outside the new band and therefore will not be
considered.
3) There is a zero-crossing point before the band. This is
shown in Fig. 11. This zero-crossing point would repre-
sent an important frequency change and may not be real,
but it may be caused by a transitory disturbance, in this
case the sudden addition of a harmonic component set.
This uncertain crossing point is saved, but the frequency
is not updated. If the next zero-crossing point is found in
the 25-Hz/s band of the previous one, the frequency will
then be updated.
4282 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011

4) There is a zero-crossing point after the band. If no zero-

crossing point is found before or within the band, this
uncertain crossing point is saved, but the frequency is not
updated. If the next zero-crossing point is found in the
25-Hz/s band of the previous one, the frequency will then
be updated.
5) There are zero-crossing points before and after the band.
In this case, the nearest point to the band is taken and
saved, but again, the frequency is not updated until the
next zero-crossing point is within the band. This could
represent nonreal crossing points due to any kind of
disturbance, and the frequency will be updated properly
within a grid voltage cycle.
When no zero-crossing point is found in the band, the nearest Fig. 12. Simulation result of the proposed frequency detector.
point to the band is saved. This is usually due to a transitory in
the grid voltage. The waveform will be periodical, and thus, in up to 2-Hz/s maximum variations). Observe that the error in the
the following period, a new zero-crossing point will be found frequency detected is small (maximum of 0.5 Hz) during the
within the band of the saved one, and the frequency will be variation until it stabilizes. The SFSD hardly notices this error.
updated. At t = 340 ms, a phase jump of π is introduced. A phase jump
There are several ways of performing the method. In this is usually the worst disturbance for a frequency detector. Notice
PLL application, a high level of precision is not required. For that the proposed frequency detector is not affected by such an
example, for a frequency difference of 0.5 Hz, with Tw = abrupt phase jump.
0.02 s, the error in the detected angle given by (25) will be
0.0314 rad. The influence of a fifth-order harmonic given by B. Variable-Frequency SFSD
(26) would be 0.0098An rad, with An being the amplitude of
the harmonic in the angle which is usually much lower than In the structure shown in Fig. 5, several blocks are designed
the unity. This shows that small frequency deviations provoke to perform at a constant frequency. The simulation test shown
slight errors in the positive-sequence detection. However, it in Fig. 7 proves that a grid frequency that is different from
is preferable to apply the method to more than one phase. It the one selected will provoke errors in the SFSD variables.
can be applied to line-to-neutral phases, line-to-line phases, or To solve this problem, the frequency-dependent blocks have
alpha–beta transformed phases. to be updated with the calculated frequency. Then, all the
MAF window widths have to be adapted to the new frequency.
1) Line-to-neutral voltages. Three frequency detectors are
Consequently, the MAF window will always contain an integer
needed. The average of the detected three frequencies is
number of unwanted oscillations produced by harmonic dis-
accepted as the frequency value. Any frequency change
tortion or voltage dips. Therefore, it will be able to remove
will be detected quickly; up to three measurements are
them without any error. Both Park transformations used for
made within a period. When a phase fails, the average of
the detection of positive- and negative-sequence components
the remaining phases will be used. If two phases fail, the
make use of a synchronous reference frame rotating at the
frequency will be calculated with the remaining phase.
detected frequency. When the input frequency is stable, the
2) Line-to-line voltages. This is similar to the previous case.
response time corresponds exactly to the MAF window width.
The only difference is that if one phase fails, there will
When the frequency is changing, the response time will vary
still be three line-to-line phases to detect the frequency.
depending on the moment the variation starts; in the worst case,
Only when two phases fail will one line-to-line phase go
half a period is added to the window width. After the required
to zero.
maximum time, the value of the angle theta is obtained with
3) α−β transformed components. This only needs two fre-
practically no error. Only a small error can be produced due to
quency detectors, but the frequency is updated less often.
the sampling.
The α−β components can still be used when a phase
fails. Furthermore, if two phases fail, one variable (α or
β) can still be used for the frequency update. VII. D ISCRETE -T IME I MPLEMENTATION
Fig. 12 shows the performance of the frequency detector Discrete-time implementation can produce errors if the num-
in the presence of distorted grid voltages when the frequency ber of samples needed in the frequency estimation or in the
varies. The simulation starts with 220 V at 50 Hz. At t = MAF integration is not integer. In both cases, the maximum
60 ms, the frequency starts increasing at 20 Hz/s, and it reaches error of one sample would be determined by (25). Considering
54 Hz at the end of the simulation. At t = 180 ms, third and fifth a sampling frequency of fs = 10 kHz, Fig. 13 shows a repre-
harmonics are included with amplitudes of 50% and 30% of the sentation of the detected angle error in (25) for grid frequencies
fundamental, respectively. This simulation represents the worst from 25 to 75 Hz. Fig. 14 shows a zoom in the proximities of
case since the detector is only applied to one phase, and the 50 Hz. It can be observed how frequencies matching an integer
frequency variation is very fast (the Spanish grid code considers number of samples will have zero error.
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS
Fig. 13. Error due to discrete-time implementation for different frequencies.
Fig. 14. Zoom of the error in Fig. 13 in the proximities of a grid frequency
of 50 Hz.
VIII. C OMPARISON
The proposed variable-frequency SFSD scheme has been
compared to the traditional PLL and the DSOGI-FLL [13]
found in the literature. The former has been tuned with a high
bandwidth, imposing a rising time of 2 ms and a natural fre-
quency of 557 rad/s. The PI parameters result in the following:
Kp = 2.06 and Ki = 814. The latter has been tuned with the
optimum parameters given in [13]: k = γ = 1.41. Fig. 15(a)
and (b) shows the grid voltages and the angle errors, respec-
tively. The rms line-to-neutral grid voltage is 220 V at 50 Hz,
and the MAFs have a window width of Tw = Tg /2, where Tg is
the grid voltage period. At the beginning, all the methods detect
the angle without error. At t = 60 ms, the frequency changes
up to 50.5 Hz. At t = 80 ms, a fifth-order harmonic component
(20%) is added until t = 150 ms. The SFSD needs a Tw time to
cancel the error, while both the traditional PLL and the DSOGI-
FLL cannot completely cancel the effects of the harmonics. At
t = 200 ms, an asymmetric voltage dip (20%, 30%, and 40%) is
introduced until t = 250 ms. Again, the SFSD corrects the error
in a Tw time. The traditional PLL maintains an oscillation due
to the negative sequence, and the DSOGI-FLL shows a slower
dynamic performance.
IX. E XPERIMENTAL R ESULTS
The proposed algorithm was programmed in a DSP board
TMS320F2812 operating at 100 MHz. It only needs 15 μs to
obtain the frequency and the positive sequence. A number of
experimental results were obtained. An Omicron three-phase
generator CMC-156 was used to provide the grid voltages.
4283
Fig. 15. Comparative simulation: (a) Grid voltages and (b) angle errors for
the SFSD, DSOGI-FLL, and traditional PLL.
All the tests were performed with a line-to-neutral voltage of
50 Vrms.
Fig. 16 shows the main variables of the proposed positive-
sequence detector. The grid frequency is maintained constant at
50 Hz throughout the process. The figure shows the grid volt-
ages, the detected angle, the detected frequency, the positive-
sequence filtered dq components, and the positive sequence. An
abrupt short circuit occurs at t = 30 ms (a 100% single-phase
voltage dip). At t = 110 ms, the previous voltage dip recovers,
and a three-phase 70% voltage dip and odd harmonic sets are
introduced as follows: up to the 25th harmonic according to
Vh = Vfundamental /h for h = 3, 5, . . . , 25. At t = 190 ms, the
input voltage is completely restored. Observe the fast detection
of the fundamental amplitude change and how the positive
sequence is perfectly characterized. There is no change in the
positive-sequence phase during this process, as the detected
angle shows. Thus, the frequency and angle are detected ac-
curately, showing good performance in the presence of voltage
dips and high harmonic distortion.
The experimental test shown in Fig. 17 shows the perfor-
mance of the variable-frequency SFSD when the input voltage
has a time-varying frequency. The input frequency starts at
50 Hz. At t = 200 ms, it starts to vary at 10 Hz/s up to 59 Hz.
In addition, a single-phase voltage dip of 30% occurs at t =
500 ms until the end. Fig. 18 shows a zoom of the transitory.
Note that, even in the presence of such adverse operating condi-
tions, the proposed positive-sequence detector is able to detect
the angle correctly and that the dq components are perfectly
synchronized.
The SFSD has also been tested in a real application to
connect a neutral point clamped (NPC) converter to a 400-V
grid as shown in Fig. 19. It is part of a laboratory wind-turbine
testbench consisting of a permanent-magnet synchronous gen-
4284 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011

Fig. 16. Experimental results. Constant frequency input, voltage dip, and high harmonic distortion.

Fig. 17. Experimental results. Varying frequency input and voltage dip.

erator and a back-to-back NPC power conversion stage. The
system has been subject to hard distortion in the test of Fig. 20.
Observe that, even under such extreme grid voltage conditions,
the detected angle and the positive-sequence dq components are
detected correctly.
X. C ONCLUSION
In this paper, a new PLL has been presented for positive-
sequence detection in a three-phase system. The structure is
based on a MAF that guarantees the complete cancellation
of the effects produced by grid voltage imbalances and har-
monics on the detected variables, except in certain infrequent
conditions when the error would be very small. Simulation
and experimental results have been presented, which verify
good performance of the system in the presence of harmonics,
imbalances, and phase changes in the grid voltages.
Performance of the presented PLL has proved considerably
better than the basic PLL design. In addition, its performance
shows clear advantages in the presence of harmonics over
the newest and more sophisticated positive-sequence detectors.
On the other hand, it has a similar dynamic response and
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS 4285

behavior in the presence of voltage dips. However, a remarkable

advantage of the proposed system is its simplicity compared
to those with a similar performance. Furthermore, the PLL
includes a simple frequency detector that makes it frequency
adaptive. This guarantees its proper operation under large fre-
quency changes.
The proposed positive-sequence detector has been applied
to a laboratory wind-turbine platform. The grid-connected
converter maintains synchronization with the voltage positive
sequence even under extreme grid voltage distortion. The pro-
posed positive-sequence detector can be applied to any other
grid-connected converter application, such as in photovoltaic
generation, active filters, static compensators, dynamic voltage
restorers, etc.
R EFERENCES
[1] M. Cirrincione, M. Pucci, G. Vitale, and A. Miraoui, “Current harmonic
compensation by a single-phase shunt active power filter controlled by
adaptive neural filtering,” IEEE Trans. Ind. Electron., vol. 56, no. 8,
pp. 3128–3143, Aug. 2009.
[2] K. H. Kwan, Y. C. Chu, and P. L. So, “Model-based H-infinity control of
Fig. 18. Experimental results. Zoom of the transitory in Fig. 17. a unified power quality conditioner,” IEEE Trans. Ind. Electron., vol. 56,
no. 7, pp. 2493–2504, Jul. 2009.
[3] M. Karimi-Ghartemani and M. R. Iravani, “A nonlinear adaptive filter
for online signal analysis in power systems: Applications,” IEEE Trans.
Power Del., vol. 17, no. 2, pp. 617–622, Apr. 2002.
[4] M. Karimi-Ghartemani, H. Karimi, and M. R. Iravani, “A
magnitude/phase-locked loop system based on estimation of frequency
and in-phase/quadrature-phase amplitudes,” IEEE Trans. Ind. Electron.,
vol. 51, no. 2, pp. 511–517, Apr. 2004.
[5] R. M. Santos Filho, P. F. Seixas, P. C. Cortizo, L. A. B. Torres, and
A. F. Souza, “Comparison of three single-phase PLL algorithms for UPS
applications,” IEEE Trans. Ind. Electron., vol. 55, no. 8, pp. 2923–2932,
Aug. 2008.
[6] C. H. da Silva, R. Rodrigues, L. E. Borges, G. Lambert-Torres,
B. K. Bose, and S. U. Ahn, “A digital PLL scheme for three-phase system
using modified synchronous reference frame using modified synchronous
reference frame,” IEEE Trans. Ind. Electron., vol. 57, no. 11, pp. 3814–
3821, Nov. 2010.
[7] D. Jovcic, “Phase locked loop system for FACTS,” IEEE Trans. Power
Syst., vol. 18, no. 3, pp. 1116–1124, Aug. 2003.
[8] L. G. B. Rolim, D. R. da Costa, Jr., and M. Aredes, “Analysis and soft-
ware implementation of a robust synchronizing PLL circuit based on the
pq theory,” IEEE Trans. Ind. Electron., vol. 53, no. 6, pp. 1919–1926,
Dec. 2006.
[9] S.-K. Chung, “A phase tracking system for three phase utility interface
inverters,” IEEE Trans. Power Electron., vol. 15, no. 3, pp. 431–438,
Fig. 19. Grid-connected NPC converter. May 2000.
[10] P. Rodríguez, J. Pou, J. Bergas, J. I. Candela, R. Burgos, and
D. Boroyevich, “Decoupled double synchronous reference frame PLL for
power converters control,” IEEE Trans. Power Electron., vol. 22, no. 2,
pp. 584–592, Mar. 2007.
[11] J. Pou, E. Robles, S. Ceballos, J. Zaragoza, A. Arias, and P. Ibáñez, “Con-
trol of back-to-back-connected neutral-point-clamped converters in wind
mill applications,” in Proc. EPE, Dresden, Denmark, Sep. 2–5, 2007.
[12] M. Karimi-Ghartemani and M. R. Iravani, “A method for synchroniza-
tion of power electronic converters in polluted and variable-frequency
environments,” IEEE Trans. Power Syst., vol. 19, no. 3, pp. 1263–1270,
Aug. 2004.
[13] P. Rodríguez, A. Luna, M. Ciobotaru, R. Teodorescu, and F. Blaabjerg,
“Advanced grid synchronization system for power converters under unbal-
anced and distorted operating conditions,” in Proc. IEEE IECON, Paris,
France, Nov. 7–10, 2006, pp. 5173–5178.
[14] S. Alepuz, S. Busquets-Monge, J. Bordonau, A. Martinez-Velasco,
C. Silva, J. Pontt, and J. Rodriguez, “Control strategies based on sym-
metrical components for grid-connected converters under voltage dips,”
IEEE Trans. Ind. Electron., vol. 56, no. 6, pp. 2162–2173, Jun. 2009.
[15] H. E. P. Souza, F. Bradaschia, F. A. S. Neves, M. C. Cavalcanti,
G. M. S. Azevedo, and J. P. Arruda, “A method for extracting the fun-
Fig. 20. Experimental results. Real grid-connected distortion test: (a) Grid damental frequency positive-sequence voltage vector based on simple
voltages, (b) αβ components, (c) detected angle, and (d) dq components of the mathematical transformations,” IEEE Trans. Ind. Electron., vol. 56, no. 5,
positive sequence. pp. 1539–1547, May 2009.
4286 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 9, SEPTEMBER 2011
[16] F. A. S. Neves, M. C. Cavalcanti, H. E. P. de Souza, F. Bradaschia,
E. J. Bueno, and M. Rizo, “A generalized delayed signal cancellation
method for detecting fundamental-frequency positive-sequence three-
phase signals,” IEEE Trans. Ind. Electron., vol. 25, no. 3, pp. 1816–1825,
Jul. 2010.
[17] R. F. de Camargo and H. Pinheiro, “Synchronisation method for three-
phase PWM converters under unbalanced and distorted grid,” Proc.
Inst. Elect. Eng.—Elect. Power Appl., vol. 153, no. 5, pp. 763–772,
Sep. 2006.
[18] H. S. Timorabadi and F. P. Dawson, “A three-phase frequency adaptive
digital phase locked loop for measurement, control, and protection in
power systems,” in Proc. IEEE PCC, Nagoya, Japan, Apr. 2–5, 2007,
pp. 183–190.
[19] G. Escobar, M. Martinez-Montejano, A. Valdez, P. Martinez, and
M. Hernandez-Gomez, “Fixed reference frame phase-locked loop (FRF-
PLL) for grid synchronization under unbalanced operation,” IEEE Trans.
Ind. Electron., vol. 58, no. 5, pp. 1943–1951, May 2011.
[20] F. A. S. Neves, H. E. P. de Souza, F. Bradaschia, M. C. Cavalcanti,
M. Rizo, and F. J. Rodriguez, “A space-vector discrete Fourier transform
for unbalanced and distorted three-phase signals,” IEEE Trans. Ind. Elec-
tron., vol. 25, no. 3, pp. 1816–1825, Jul. 2010.
[21] A. M. Salamah, S. J. Finney, and B. W. Williams, “Three-phase phase-
lock loop for distorted utilities,” IET Elect. Power Appl., vol. 1, no. 6,
pp. 937–945, Nov. 2007.
[22] A. Ghoshal and V. John, “A method to improve PLL performance
under abnormal grid conditions,” in Proc. NPEC, Bangalore, India,
Dec. 17–19, 2007.
[23] E. Robles, S. Ceballos, J. Pou, J. Zaragoza, and I. Gabiola, “Grid syn-
chronization method based on a quasi-ideal low-pass filter stage and a
phase-locked loop,” in Proc. IEEE Power Electron. Spec. Conf., Rhodes,
Greece, Jun. 15–19, 2008, pp. 4056–4061.
[24] F. D. Freijedo, J. Doval-Gandoy, O. Lopez, and E. Acha, “A generic open-
loop algorithm for three-phase grid voltage/current synchronization with
particular reference to phase, frequency, and amplitude estimation,” IEEE
Trans. Power Electron., vol. 24, no. 1, pp. 94–107, Jan. 2009.
[25] L. Asnin, V. Backmutsky, and M. Gankin, “Comparative characteristics
of main methods for dynamic estimation of frequency and magnitude
parameters in power systems,” in Proc. 22nd Conv. Elect. Electron. Eng.,
Dec. 1, 2002, pp. 35–38.
[26] L. Asnin, V. Backmutsky, M. Gankin, J. Blashka, and M. Sedlachek, “DSP
methods for dynamic estimation of frequency and magnitude parame-
ters in power system transients,” in Proc. Power Tech, Porto, Portugal,
Sep. 10–13, 2008, vol. 4.
[27] O. Vainio and S. J. Ovaska, “Digital filtering for robust 50/60 Hz zero-
crossing detectors,” IEEE Trans. Instrum. Meas., vol. 45, no. 2, pp. 426–
430, Apr. 1996.
[28] O. Vainio and S. J. Ovaska, “Noise reduction in zero crossing detection
by predictive digital filtering,” IEEE Trans. Ind. Electron., vol. 42, no. 1,
pp. 58–62, Feb. 1995.
[29] M. B. Duric and Z. R. Durisic, “Frequency measurement in power net-
works in the presence of harmonics using Fourier and zero crossing
technique,” in Proc. Power Tech, St. Petersburg, Russia, Jun. 27–30, 2005,
pp. 1–6.
[30] M. M. Begovic, P. M. Djuric, S. Dunlap, and A. G. Phadke, “Frequency
tracking in power networks in the presence of harmonics,” IEEE Trans.
Power Del., vol. 8, no. 2, pp. 480–486, Apr. 1993.
[31] D. W. P. Thomas and M. S. Woolfson, “Evaluation of frequency track-
ing methods,” IEEE Trans. Power Del., vol. 16, no. 3, pp. 367–371,
Jul. 2001.
[32] M. Sanaye-Pasand and V. J. Marandi, “Frequency estimation of distorted
signals for control and protection of power system,” in Proc. 8th IEE
Int. Conf. Develop. Power Syst. Protection, Amsterdam, The Netherlands,
Apr. 5–8, 2004, pp. 631–634.
[33] H. K. Kwok and D. L. Jones, “Improved instantaneous frequency estima-
tion using an adaptive short-time Fourier transform,” IEEE Trans. Signal
Process., vol. 48, no. 10, pp. 2964–2972, Oct. 2000.
[34] J. Yang and C. Liu, “A precise calculation of power system frequency,”
IEEE Trans. Power Del., vol. 16, no. 3, pp. 361–366, Jul. 2001.
[35] A. Routray, A. K. Pradhan, and K. P. Rao, “A novel Kalman filter for
frequency estimation of distorted signals in power systems,” IEEE Trans.
Instrum. Meas., vol. 51, no. 3, pp. 469–479, Jun. 2002.
[36] H. Karimi, M. Karimi-Ghartemani, and M. R. Iravani, “Estimation of
frequency and its rate of change for applications in power systems,” IEEE
Trans. Power Del., vol. 19, no. 2, pp. 472–480, Apr. 2004.
[37] M. Mojiri, M. Karimi-Ghartemani, and A. Bakhshai, “Estimation of
power system frequency using an adaptive notch filter,” IEEE Trans.
Instrum. Meas., vol. 56, no. 6, pp. 2470–2477, Dec. 2007.
Eider Robles was born in Bilbao, Spain, in 1980.
She received the M.S. degree in electronic and au-
tomatic engineering from the University of Deusto,
Bilbao, in 2003 and the Ph.D. degree from the
University of the Basque Country, Bilbao, in 2010.
Since 2003, she has been with Tecnalia Research
& Innovation, Zamudio, Spain, where she is cur-
rently a Research Engineer with the Energy Unit. She
is the author or coauthor of more than 20 technical
papers. Her main research activity deals with grid
connection and control of renewable energy systems.
Josep Pou (S’97–M’03) received the B.S., M.S.,
and Ph.D. degrees in electrical engineering from the
Technical University of Catalonia (UPC), Terrassa,
Spain, in 1989, 1996, and 2002, respectively.
During 1989, he was the Technical Director of
Polylux S.A. In 1990, he was with the faculty of
UPC as an Assistant Professor, where he became an
Associate Professor in 1993. From February 2001
to January 2002 and from February 2005 to January
2006, he was a Researcher with the Center for Power
Electronics Systems, Virginia Polytechnic Institute
and State University, Blacksburg. He is currently with the Terrassa Industrial
Electronics Group, Department of Electronic Engineering, UPC. He has au-
thored more than 80 published technical papers and has been involved in several
industrial projects and educational programs in the fields of power electronics
and systems. His research interests include modeling and control of power
converters, multilevel converters, power quality, renewable energy systems, and
motor drives.
Salvador Ceballos received the B.Sc. degree in
physics from the University of Cantabria, Santander,
Spain, in 2001 and the B.Eng. and Ph.D. degrees
in electronic engineering from the University of the
Basque Country, Bilbao, Spain, in 2002 and 2008,
respectively.
Since 2002, he has been with Tecnalia Research
& Innovation, Zamudio, Spain, where he is currently
a Development Engineer with the Energy Unit. From
May 2008 to May 2009, he was a Visiting Researcher
with the Hydraulic and Maritime Research Centre,
University College Cork, Cork, Ireland. He has authored more than 30 pub-
lished technical papers. His research interests include multilevel converters,
fault-tolerant power electronic topologies, and renewable energy systems.
Jordi Zaragoza (S’08) received the B.S. degree in
electronic engineering and the M.S. degree in auto-
matic and electronic industrial engineering from the
Technical University of Catalonia (UPC), Terrassa,
Spain, in 2001 and 2004, respectively, where he is
currently working toward the Ph.D. degree in the
Terrassa Industrial Electronic Group, Department of
Electronic Engineering.
In 2003, he was with the faculty of UPC as an
Assistant Professor. From September 2006 to
September 2007, he was a Researcher with the
Energy Unit, Robotiker–Tecnalia Research Centre, Zamudio, Spain. He has
authored more than 20 published technical papers. His research interests
include modeling and control of power converters, multilevel converters, wind
energy, and power quality.
ROBLES et al.: GRID VOLTAGE SEQUENCE DETECTOR FOR DISTRIBUTED GENERATION SYSTEMS
José Luis Martín (M’98) received the M.S. and
Ph.D. degrees in electrical engineering from the Uni-
versity of the Basque Country, Bilbao, Spain, in 1988
and 1992, respectively.
From 1989 to 1995, he was an Assistant Professor
in electronic technology with the Department of
Electronics and Telecommunications, University of
the Basque Country, where he became an Associate
Professor in 1995 and the Head of the Department
from 1995 to 2001. From 2001 to 2005, he was the
Vice-Dean of the Faculty of Engineering in Bilbao.
He manages the Applied Electronics Research Team of the University of the
Basque Country.
4287
Pedro Ibañez (M’04) was born in Bilbao, Spain,
in 1964. He received the M.Sc. and Ph.D. de-
grees in electrical engineering from the University
of the Basque Country, Bilbao, in 1988 and 1991,
respectively.
From 1988 to 1997, he was an Assistant Professor
in electronic technology with the Department of
Electronics and Telecommunications, University of
the Basque Country, where he became an Associate
Professor in 1997. Since 1992, he has been with
Tecnalia Research & Innovation, Zamudio, Spain,
where he is currently the Technology Director of the Energy Unit. He has
worked in many projects related to electronics systems, digital control systems,
and power converters for energy applications. He has participated in more
than 30 research projects supported by public institutions (including European
Projects) and private companies. He is the author or coauthor of more than 30
technical papers.