94 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO.

1, JANUARY 2009

A Generic Open-Loop Algorithm for Three-Phase

Grid Voltage/Current Synchronization With
Particular Reference to Phase, Frequency,
and Amplitude Estimation
Francisco D. Freijedo, Member, IEEE, Jesús Doval-Gandoy, Member, IEEE, Óscar López, Member, IEEE,
and Enrique Acha, Senior Member, IEEE

Abstract—This paper presents a new open-loop architecture the more efficient the control actions. Examples of applications
for three-phase grid synchronization based on moving average where faultless synchronization is crucial are in the fields of
and predictive filters, where accurate measurements of phase, fre- active power filters and power quality conditioners [1]–[13],
quency, and amplitude are carried out in real time. Previous works
establish that the fundamental positive sequence vector of a set distributed power generation systems [14], flexible ac transmis-
of utility voltage/current vectors can be decoupled using Park’s sion systems (FACTS) and custom power [15]–[17], and HVdc
transformation and low-pass filters. However, the filtering pro- transmission [18], [19].
cess introduces delays that impair the system performance. More Arguably, phase-locked loops (PLLs) are the most
specifically, when the input signal frequency is shifted above the widespread synchronization algorithms employed in grid-
nominal, a nonzero average steady-state phase error appears in the
measurements. To overcome such limitations, a suitable combina- connected converters [20]–[25]. A PLL is a nonlinear feedback
tion of predictive and moving average finite impulse response (FIR) control system that tracks the phase and frequency of the input
filters is used by the authors to achieve a robust synchronization signal fundamental component and is able to retrack the phase,
system for all input frequencies. Moving average filters are linear with no steady-state error, following a transient event such as
phase FIR filters that have a constant time delay at low frequen- phase and frequency jumps [22], [26]. PLLs have been employed
cies, a characteristic that is exploited to good effect to design a
predictive filter that compensates such time delays, enabling zero successfully in systems where an accurate phase measurement
steady-state phase errors for shifted input frequencies. In sum- is required in high- and medium-voltage grid-connected con-
mary, the main attributes of the new system are its good frequency verters [14]–[17], and in systems with self-adapting frequency
adaptation, good filtering/transient response tradeoff, and the fact variations characteristics [2], [4], [8], [12], [13]. However, one
that its dynamics is independent of the input vector amplitude. important drawback exhibited by PLLs is that their dynam-
Comprehensive experimental results validate the theoretical ap-
proach and the high performance of the proposed synchronization ics are highly dependent on the amplitude of the fundamen-
algorithm. tal positive sequence vectors. They are well suited to appli-
cations where the amplitude varies little but their tunning has
Index Terms—Grid-connected converters, phase-locked loop
(PLL), predictive control, synchronization. reported to be incased where the input signals vary over a wide
range [20], [23], [27].
I. INTRODUCTION Other schemes, classified as open-loop methods, have also
been proposed [28]. Among these are the weighted least-square
YNCHRONIZATION is an issue of paramount importance
S in the control of grid-connected power electronics equip-
ment. Most control algorithms use some form of systems volt-
estimation (WLSE) [29], the low-pass filtering transformation
angle detector (LP-TAD) [30], those that use Kalman filter [4],
[30], and the normalized positive sequence synchronous frame
age/current information, such as the phase, amplitude, and fre- (NPFS) [28]. These algorithms perform better than PLLs under
quency of the fundamental component. It is expected that the certain conditions, particularly when the input frequency is close
speedier and the more accurate these measurements are, the to the nominal. However, as the input frequency shifts, their per-
better the generation of reference signals will be and therefore formance tends to worsen. Frequency control algorithms have
been added to open-loop schemes in order to make them fre-
Manuscript received May 23, 2008; revised July 29, 2008; accepted quency adaptive [4], [28]–[30]. However, this is at the expense
August 25, 2008. Current version published February 6, 2009. This work
was supported by the Spanish Ministry of Education and Science under of increasing the complexity of their dynamics and implementa-
Project ENE2006-02930. Recommended for publication by Associate Editor R. tion by a considerable margin. So, in practical implementations,
Teodorescu. the PLL continues to be the most employed synchronization
F. D. Freijedo, J. Doval-Gandoy, and Ó. López are with the Department of
Electronic Technology, University of Vigo, Vigo 36310, Spain (e-mail: fdfrei@ algorithm, mainly due to its simplicity, frequency adaptation
uvigo.es; jdoval@uvigo.es; olopez@uvigo.es). features, and versatility.
E. Acha is with the Department of Electronics and Electrical Engineering, An open-loop single-phase system based on moving average
University of Glasgow, Glasgow G12 8QQ, U.K. (e-mail: e.acha@elec.gla.
ac.uk). finite impulse response (FIR) filters has been put forward in [31]
Digital Object Identifier 10.1109/TPEL.2008.2005580 in order to achieve amplitude-independent dynamics and good

0885-8993/$25.00 © 2009 IEEE

FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 95

filtering/transient response. Moving average FIR filters have

also been employed in PLL implementations such as in [31], and
[32]. The main drawback in using this kind of filters inside the
closed loop of a PLL is their slow transient response arising from
their phase versus frequency characteristic. It is well known in
the control theory that, in a feedback system, the phase margin
of the open-loop transfer function should be greater than 45◦ .
Otherwise, this leads to a low-bandwidth PLL and hence to a
system with a slow transient response [26]. The system in [31]
takes advantage of both frequency and transient (step) responses
of moving average filters. However, similar to other open-loop
systems, it lacks frequency adaptation features.
In this paper, the scheme proposed in [31] is extended to in-
clude three-phase system capabilities by implementing Park’s
transformation in which the rotating coordinate phase angle ro-
tates at the nominal frequency. From the Park variables, achieved
through filtering, all the needed measurements for synchroniza-
tion are obtained. In order to achieve frequency adaptation, pre-
dictive filters for compensating the delay through the moving
average filters are proposed: moving average FIR filters have
linear-phase. This implies that the time of delay is constant Fig. 1. Example of decomposition of an unbalanced three-phase set of fun-
and known for low frequencies [33]. The oscillation frequency damental voltages/currents. (a) Time-domain decomposition. (b) Equivalent
of Park variables is set by the input frequency deviation from diagram.
the nominal. Under normal operation, this deviation should not
exceed ± 1 Hz according to the International Electrotechnical
h = [Vah Vah Vah ]T ) se-
[Van Vbn Vcn ]T ), and homopolar (V
Commission standard IEC 61000-2-2, but in isolated and emer- quences (e.g., in Fig. 1)
gency systems, this limit can be exceeded resulting in a range
abc1 = V

p + V

n + V

h .
V (1)
of about ± 2 Hz or more [28]. The time of delay through the
moving average filters is always much smaller than the oscil-
p , V
|Vp | being the amplitude of V
abc can be expressed as
lation period of the Park variables. Therefore, the trajectory in
the time domain of the Park variables can be approximated to a
abc = V
V
p + V
n + V
h + f
abc (3ω1 , 5ω1 , . . .)
straight line, where only the time interval between the measured    
value and the actual value is considered [34]–[40]. Using this |Vp | sin(θ1 ) fa (ω1 , 3ω1 , 5ω1 , . . .)
   
approach, the transfer function of a predictive filter that predicts = |Vp | sin(θ1 − 2π/3)+ fb (ω1 , 3ω1 , 5ω1 , . . .)  (2)
the current value of Park variables is obtained. |Vp | sin(θ1 + 2π/3) fc (ω1 , 3ω1 , 5ω1 , . . .)
In summary, the use of the moving average and predic-  

tive filters enables a purely open-loop system with frequency V
p

adaptation, amplitude-independent dynamics, and good filter-
where 
ing/transient response tradeoff. The dynamics of this open-loop
system is mainly stated by the frequency and transient responses θ1 = ω1 dt + θ1 0 . (3)
at the filtering stage. Moreover, more than one moving average
filter can be used in order to improve the cancellation pattern, Here, θ1 is V
p instantaneous phase angle and θ1 is an offset
0
but this at the expense of making the transient response slower. constant phase.
Two design examples are presented. Their high performance As said, the goal of synchronization is to track V
p both in
is shown in Section VI. Simulation and real-time implementa- phase and frequency, i.e., to measure θ1 (ω1 = dθ1 /dt). From
tion results are provided and amply discussed.
p ; therefore, the influence
(2), it is clear that θ1 only depends on V
II. BACKGROUND


of Vn , Vh , and odd harmonics in θ1 measurement should be

abc = [Va Vb Vc ]T ) for which canceled.
A three-phase voltage system (V
The Park transformation for the positive sequence using an
fundamental component (V
abc1 = [Va1 Vb1 Vc1 ]T ) is unbal-
arbitrary phase angle can be expressed as
anced and also has odd harmonic components is considered. 
The input angular frequency is ω1 and its value could oscil- 2 sin(θdq ) sin(θdq − 2π 3 ) sin(θdq + 2π3 )
Tdq = (4)
late around ±2π × 2 rad/s around the nominal value ω1n = 3 cos(θdq ) cos(θdq − 2π 2π
3 ) cos(θdq + 3 )
2π × f1n rad/s, where f1n = 50 Hz is the nominal frequency
expressed in hertz. where

According to the Fortescue theorem [41], Vabc1 can be de-
composed into positive (V
n =

p = [Vap Vbp Vcp ]T ), negative (V θdq = ω1n dt + θdq 0 . (5)
96 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

−1
and the inverse of the Park transformation (Tdq )
 
sin(θdq ) cos(θdq )
    
 2π 2π  
−1  sin θ dq − cos θ dq − 
Tdq = 3 3 . (9)
    
 2π 2π 
sin θdq + cos θdq +
3 3
−1
dq results in V

p
The vector multiplication of Tdq by V
 
|Vp | sin(θ1 )
  
 2π  
−1
 |Vp | sin θ1 − 

Tdq · Vdq =  3 =V p. (10)
Fig. 2. Representation of V
p on the d–q axis defined by T d q . (a) V
p in SRF   
defined by T d q for ω 1 = ω 1 n . (b) V
p in SRF defined by T d q for ω 1 = ω 1 n
 2π 
and ω d > 0.
|Vp | sin θ1 +
3
The numerical value of θ1 , ω1 , and |Vp | can also be obtained as
Here, θdq is a phase angle rotating at the fundamental angular follows:
frequency and has an arbitrary offset constant value (θdq 0 ). 

abc gives rise to
The vector multiplication of Tdq by V |Vp | = Vd2 + Vq2 . (11)
 For the measurement of θ1 and ω1 , the angle transformation

abc = |Vp | cos(ωd t + φ) = Vd
Tdq · V Tdq →α β can be employed to represent V
p by means of Clarke
|Vp | sin(ωd t + φ) = Vq

 
T
variables Vα β = [Vα Vβ ] [42]
T d q ·V
p 
 sin(θdq ) cos(θdq )
fd (ω1 +ω1n , 3ω1n +ω1 , 3ω1 +ω1n , . . .) Tdq →α β = . (12)
+ (6) cos(θdq ) − sin(θdq )
fq (ω1 +ω1n , 3ω1n +ω1 , 3ω1 +ω1n , . . .)

dq results in
The vector multiplication of Tdq →α β by V
where  


Vα |Vp | sin(θ1 )
Vα β = Tdq →α β · Vdq = = . (13)
ωd = ω1 − ω1n (7) Vβ |Vp | cos(θ1 )

and From (13)

θ1 = atan2(Vα , Vβ ) ∈ [−π, π] (14)

φ = θ1 0 − θdq 0 . (8)
and by definition
Here, Vd and Vq are the so-called direct and quadrature Park
variables and V
dq = [Vd Vq ]T is the dq vector. dθ1
ω1 = . (15)
From (6) and (7), Vd and Vq are dc values when the input dt
frequency is the nominal one. Fig. 2(a) depicts V
p in the syn- It is important to note that this approach is independent of the
chronous reference frame (SRF) defined by Tdq and ω1 = ω1n ;
p .
concrete θdq 0 value; of course, Vd and Vq depend on it but not V

p on the axis are Vd and Vq . Another impor- −1
the projections of V The use of Tdq and Tdq →α β allows to know all the parameters
tant result from (6) is that the effect of the nonnegative sequence of V
p without knowing the exact value of θdq .
0
and harmonic components results in even harmonic components

h = 0).
(Tdq · V
III. ALGORITHM IMPLEMENTATION
Usually, in normal operation, ω1 = ω1n . In such a situation,

p is rotating in the SRF at ωd : clockwise if ωd < 0 and anti-
V As proved in the previous section, it is possible to synchro-
clockwise if ωd > 0. Fig. 2(b) depicts such a situation: the dot- nize V
p by means of angle transformations and low-pass filters.
ted vector represents V
p in a previous point of time. As ωd > 0, Fig. 3 shows the proposed scheme for grid synchronization im-


the current Vp (solid line) has been rotated anticlockwise with plemented in a digital device at sampling frequency fs . The
angular frequency ωd . matrix elements generation block generates the components of
−1
Since ωd  ω1n is considered, Vd and Vq are very low- Tdq , Tdq , and Tdq →α β . The filtering block obtains Vd and Vq in
frequency components in (6). Even harmonic components are order to decouple V
p from V
abc . In order not to have delay in Vd
also a bit shifted from integer multiples of ω1n . The main and and Vq measurements, predictive filters are linked to the mov-
more interesting implication of this approach is the suitability ing average filters. Next sections explain the design approach of
of using a low-pass filter for decoupling V
p through Vd and Vq these blocks in depth.
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 97

B. Filtering Block
1) Moving Average Filters as Low-Pass Filters: As shown,
low-pass filters could be employed to decouple Vd and Vq . The
most important feature of a low-pass filter is its amplitude ver-
sus frequency response. In this particular case, unitary gain for
low-frequency values and high attenuation at higher frequen-
cies, specially at ω1 harmonics, are required. A filter with this
frequency response is relatively easy to design and implement
in current digital devices. However, another important feature of
the filter is its step response. The step response sets the system
Fig. 3. Block diagram of the synchronization algorithm. response to a transient. In filtering, there is a tradeoff between
the cancellation pattern and the step response [33].
Another and more problematic issue in the design of this
system is that ω1 is unknown. As shown, when ω1 = ω1n , Vd
and Vq are low-frequency components rotating at ωd . In such a
situation, Vd and Vq are obtained with a time delay (td ) set by
the phase versus frequency response of the low-pass filters. If it
is not compensated, this time delay would provoke a constant
average steady-state phase error (θe ) proportional to td and ωd
θe = ωd td = θ1 − θ1m eas (17)
where θ1m eas is the measurement of θ1 through the system. The
ω1 measurement would not be affected since the time derivative
of a constant is zero. Amplitude measurements would also be
Fig. 4. Electronic oscillator model. (a) Poles/zeros map of the oscillator (in
radians per second, continuous model). (b) Block diagram of the proposed obtained with delay.
digital oscillator (implemented at fs = 10 kHz). The only way to know td in open loop, without knowing
ωd , is by using a linear-phase FIR filter. Linear-phase refers to
the condition where the phase response of the filter is a linear
A. Matrix Elements Generation function of frequency, excluding phase wraps at ±180◦ . This
1) Electronic Oscillator: Fig. 4(a) shows the closed-loop results in td through the filter being the same at all frequencies
poles of an analog RC oscillator. Following Barkhausen criteria, [33].
this system oscillates at ω1n . From the control theory, this system The difference equation that defines how the input signal (x)
tends to instability since its poles are on the imaginary axis. is related with the output signal (y) of an FIR filter is
Fig. 4(b) shows the block diagram of this electronic oscillator 1 −1
N
model implemented digitally. The integrators are saturated to set y(k) = b0 xk + b1 xk −1 + · · · + bN 1 −1 xk −N 1 −1 = bi xk −i
the amplitude of the output signals (±1 p.u.); this nonlinearity i=0
also avoids instability [43]. The output of one of the integrator (18)
should be different from zero (initial condition). where k represents the current number of sample, bi are the
This oscillator generates sin(θdq ) and cos(θdq ) rotating at filter coefficients, and N1 − 1 is the filter order, so N1 must be
ω1n . θdq 0 is not measured since it is not necessary. A small an integer number. The window length or time length (tw ) of an
drawback of the oscillator implementation is that the generated FIR filter is set by the number of samples and the sampling time
waves sin(θdq ) and cos(θdq ) are not pure sinusoidal waveforms (1/fs )
due to the nonlinear behavior of the saturation and sampling 1
process. For this reason, the saturation limits of the integra- t w = N1 . (19)
fs
tors should be adjusted accordingly, as shown in Fig. 4(b), so
An FIR filter is linear-phase if and only if its coefficients are
the amplitude of the components rotating at ω1n is 1 p.u. The
symmetrical around the center coefficient; under such a condi-
total harmonic distortion (THD) of sin(θdq ) and cos(θdq ) im-
tion, td is also a function of N1
plemented at fs high enough is negligible: e.g., 0.70% when
implemented at fs = 10 kHz. N1 1
td = . (20)
2) Trigonometric Rules Net: All the elements necessary 2 fs
−1 −1 
for implementing Tdq , Tdq , and Tdq can be generated A particular case of linear-phase FIR filters is moving average
from the oscillator block outputs sin(θdq ) and cos(θdq ). Dis- filters that implement the average function over the last N1
placed components are obtained in this block using the samples. The filter coefficients of the moving average FIR filters
relations are found via the following equation:
 √ 1
2π  3 b(i)i=0, ..., N 1 −1 = . (21)
sin θdq ± = −0.5 sin(θdq ) ∓ cos(θdq ). (16) N1
3 2
98 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

Fig. 6. Filtering blocks of the design examples. (a) Filtering stage of S1.
(b) Filtering stage of S2.

TABLE I
DESIGN EXAMPLE 1 VALUES

Equation (18) for the moving average FIR filter can be expressed
in the Z domain as
Y (z) = H1 (z)X(z)
1  
= X(z) + X(z)z −1 + · · · + X(z)z −(N 1 −1)
N1
 N 1 −1

1  −i
= z X(z) (22)
N1 i=0

where H1 (z) is the Z-domain transfer function relating the

input and output of the filter. From (22) and through the basic
properties of the Z transform, a more suitable implementation
form of H1 (z) is obtained
1 1 − z −N 1
H1 (z) = . (23)
N1 1 − z −1
The magnitude versus frequency response of H1 (z) is mathe-
matically described by the Fourier transform of a rectangular
pulse of amplitude 1/N1 and duration N1 samples in the time
domain
 
 sin(πf N1 /fs ) 
Mag H1 (f ) =   (24)
N1 sin(πf /fs ) 
where f is defined in the range [0, fs /2] [44].
In this application, a cancellation of the even harmonics of f1n
is expected (100, 200, . . . Hz). From (24), N1 value canceling
all the even harmonics is derived
Fig. 5. Time and frequency responses of H (z)1 implemented at fs = 10 kHz
and N 1 = 100. (a) Frequency response (in hertz). (b) Part (a) zoomed around dc fs
(in hertz). (c) Impulse response (tw = 0.01 s). (d) Step response (tw = 0.01 s). N1 = (25)
f1n
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 99

Fig. 7. Design Example 1: time and frequency responses of H (z)1 H (z)2 . (a) Frequency response (in hertz). (b) Part (a) zoomed around low frequency with
ε = 0. (c) Part (a) zoomed around low frequency with ε = 0.0095. (d) Impulse response (tw = 0.01 s). (e) Step response (tw = 0.01 s).

TABLE II monics of f1n are canceled, and the gain is also very low around
DESIGN EXAMPLE 2 VALUES
these frequencies. The linear-phase feature can also be noticed.
Fig. 5(c) and (d) shows the impulse response and step response
set by (19), in this case 0.01 s.
In sum, moving average FIR filters offer a very good harmonic
cancellation pattern, a very good transient response, and a fixed
where N1 is an integer number, since it sets the order of a digital and known delay through them (linear phase).
filter; hence, fs should be multiple of f1n . 2) Predictive Filter: This block is the key in the high per-
Fig. 5 shows the features of a moving average filter with formance of the system. As shown, the moving average FIR
fs = 10 kHz and N1 = 100. Fig. 5(a) shows that the even har- filters introduce a fixed and known delay. If moving average
100 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

Fig. 9. S1: test to show predictive filters action. (a) Phase jump tran-
sient in an unbalanced (|V n | = 0.1|V p |) set of input waves rotating at ω 1 =
2π × 51 rad/s. (b) Phase error: transient and steady state with predictive filters
Fig. 8. Design Example 2: frequency and step responses of [H 2 (z)]. Zero average phase error is achieved. (c) Phase error: transient and
 steady state without predictive filters [H 2 (z)]. There is no zero average phase
H 1 (z)H 1 (z)H 2 (z). (a) Frequency response (in hertz). (b) Step re-
sponse (tw = 0.02 s). (c) Part (a) zoomed around low frequency with error.
ε = 0.009.
well to a straight line. Therefore, it is possible to predict future
samples from current samples into this trajectory [34]–[40].
FIR filters are used to decouple Vd and Vq , and td is not com- y2 (k + 1) being the current sample of a straight line trajectory,
pensated, there is a steady-state error in the measurements when it can be expressed as
ω1 = ω1n , which is the main limitation of the system proposed
in [31]. However, as td is known from (20), it can be compen- 1
y2 (k + 1) = y2 (k) + m (26)
sated by an open loop with a predictive filter. This approach is fs
detailed next.
where m is the slope. In the same way
Equation (20) is used to calculate td of the moving average
filter. While td  2π/ωd , it is correct to say that in any td 1
interval, both Vd and Vq trajectories in the time domain fit very y2 (k + 2) = y2 (k + 1) + m . (27)
fs
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 101

Fig. 10. S1 response to a big frequency step at 0.2 s. Unbalanced (|V n | =

0.1|V p |) input wave. Frequency step (up), phase error (center), and frequency er-
ror (down). Zero steady-state error and transient duration of 0.01 s are achieved.

From (26) and (27)

y2 (k + 2) = 2y2 (k + 1) − y2 (k). (28)
N2 being the number of samples of delay through a moving
average FIR filter, it is set by td (20) and fs as
N1
N2 = td fs =
. (29)
2
Knowing N2 from the moving average FIR filter parameters,
through (28), the future N2 sample can be predicted from the
current and past samples as follows:
y2 (k + N2 ) = 2y2 (k + N2 − 1) − y2 (k + N2 − 2)
= 3y2 (k + N2 − 2) − 2y2 (k + N2 − 3)
= (N2 + 1)y2 (k) − N2 y2 (k − 1). (30)
Equation (30) can be expressed as the Z-domain transfer
function
H2 (z) = (N2 + 1) − N2 z −1 . (31)
The predictive filter defined by (31) compensates the delay
through the moving average FIR filter: for low frequencies, the
phase versus frequency response of H2 (z) has a linear positive
phase that cancels the negative linear phase of H1 (z).
By inspecting the phase versus frequency response of the
whole filtering stage [H1 (z)H2 (z)], H2 (z) can be optimized
accordingly in order not to have almost any phase delay without
introducing amplitude error. H2 (z) is redefined as
(N2 + 1) − (N2 − )z −1
H2 (z) = (32)
1+
where is an optimization parameter (  1). Fig. 11. S1 tested under a distorted set of input waves rotating at 49.5 Hz.
At 1 s, a 1–0.2 p.u. sag with +45◦ phase jump has been programmed.
(a) S1: set of distorted input waves in time domain. Unbalance (|V n | = 0.1|V p |)
IV. DESIGN EXAMPLES and harmonics (≈ 10% fifth, 7% seventh, 4% eleventh). (b) Phase error: the
transient lasts 0.01 s and zero average steady-state error. (c) |V p |, amplitude
Following the theoretical approach of the previous section, error, V d , and V q measurements. The transient lasts 0.01 s and zero average
two design examples are contributed. steady-state error. (d) V α and V β measurements. They are not distorted.
102 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

Fig. 13. S1: steady state error for an unbalanced (|V n | = 0.1|V p |) set of
Fig. 12. S1: steady state error for a balanced set of inputs oscillating at 48 Hz.
inputs oscillating at 51 Hz. (a) Steady-state measurements for a balanced set of
(a) Balanced set of voltages oscillating at 48 Hz and V α measurement. M1 is
voltages oscillating at 51 Hz. M1 depicts V a − V α as the error signal tends to
V a − V α that tends to zero. (b) Zoom of (a) around the zero cross. The average
zero in steady state. (b) Zoom of (a) around the zero cross. The average error is
error is slightly positive due to the sampling process.
slightly positive due to the sampling process.

A. Design Example 1
Fig. 6(a) shows the filtering block that employs two pairs of minimized by the optimization of the value [see Fig. 7(c)].
linked moving average and predictive filters. The values of the Fig. 7(d) and (e) shows that the transient response is kept in half
parameters are summarized in Table I. a fundamental cycle.
The whole system implementing this filtering block is named As said, in this approach, only odd harmonics in V
abc are
S1. Fig. 7(a) shows H1 (z)H2 (z) frequency response. The can- considered since nonlinear electrical loads causing high even
cellation pattern is optimized for even harmonics in Vd and Vq harmonic disturbances (in V
abc ) are not usual and their use must
caused by odd harmonics in the input signal V
abc [31]. Fig. 7(b) be precluded in ac networks [45]–[47]. To deal with the presence
proves the feasibility of the predictive filter approach: the phase of even harmonics in V
abc , N1 = 200 and N2 = 100 could be
delay is almost zero for low frequencies. This error is even considered.
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 103

Fig. 15. S1: steady-state phase measurement at 48 Hz. Ch1 is θ1 , Ch2 is V a ,

Ch3 is V b , and Ch4 is V c (phase in ≈ 2 rad, voltages in p.u.). The measurement
does not have ripple.

in this section [see Fig. 6(b)]. The predictive filter should be

recalculated according to (29), since the time of delay through
the two filters is doubled. The whole system implementing this
filtering block is named S2, and its values are summarized in
Table II.
Fig. 8 shows the frequency and step responses of design ex-
ample 2. The cancellation pattern is improved at cost of doubling
the transient time. As in other systems such as PLLs, there is a
tradeoff between filtering and transient response.

V. SIMULATION RESULTS
The S1 system proposed in the previous section has been
simulated in order to test its performance. Simulation results
have been obtained through Matlab/Simulink, using the fixed-
step discrete-time solver at fs = 10 kHz.
The most important feature of time-domain simulation with
respect to real-time implementation is the possibility of obtain-
ing curves of phase error, frequency error, and amplitude error
in the time domain. These errors are defined as the difference
Fig. 14. S1: transient response for a sag with −45◦ phase jump. (a) S1: between the actual value and the measurement value.
transients response to a sag with −45◦ phase jump. M1 depicts V a − V α as
the error signal, which is zero in steady-state. As expected, the transients last The amplitudes of the input waves are in per unit (p.u.), so
0.01 s. Time scale at 5 ms. (b) S1: transients response to a sag with −45◦ phase they can represent both voltages or currents. The magnitude
jump. M1 depicts V a − V α as the error signal, which is zero in steady state. As displayed to indicate the amount of unbalance in V
abc is the
expected, the transients last 0.01 s. Time scale at 2 ms.
negative sequence vector magnitude (|Vn |).
Fig. 9 shows the phase error for an unbalanced (|Vn | =
0.1|Vp |) input wave rotating at 51 Hz (ω1 = 2π × 51 rad/s).
B. Design Example 2
A −45◦ phase jump has been programmed to show both steady-
Even though S1 is a very good practical solution dealing with state and transient responses. The result is also displayed for the
unbalance and harmonics, the presence of other factors such as system without predictive filters. As expected, the system with-
interharmonics, noise, and notching (also noise) could degrade out predictive filters has the nonzero average steady-state phase
the performance of the measurements. In fact, the limitations of error set by (17). With the predictive filter, steady-state zero
the proposed system are mainly set by its cancellation pattern. In error is achieved. As also expected, the transient lasts 0.01 s.
order to improve the cancellation pattern, more than one moving Fig. 10 shows the phase error when the input wave of the

average filters [H1 (z) and H1 (z)] can be linked, as proposed previous test has a big frequency step in 0.2 s: from 48 to 52 Hz.
104 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

Fig. 17. S1 versus S2 in terms of transient response. Ch1 is V α , Ch2 is V a ,

Fig. 16. Comparison between S1 and S2 in terms of noise rejection. Ch1 Ch3 is V b , and Ch4 is V c (voltages in p.u., ω 1 = 2π × 49 rad/s). (a) Transient
is V α , Ch2 is V a , Ch3 is V b , and Ch4 is V c (voltages in p.u., ω 1 = response of S1. The steady state is reached in 0.01 s. (b) Transient response of
2π × 49.75 rad/s). (a) Measurement with S1. V α also contains notches. S2. The steady state is reached in 0.02 s. The settling time is doubled.
(b) Measurement with S2. V α is clean.

VI. EXPERIMENTAL RESULTS

Before and after the transient of duration 0.01 s, the average The proposed algorithm has been implemented in the dSpace
phase and frequency errors are zero. platform DS1103 at fs = 10 kHz and tested in the laboratory.
Fig. 11 shows the results for an unbalanced The execution time of S1 and S2 was 11.2 and 15.8 µs, respec-
(|Vn | = 0.1 · |Vp |) and harmonic contaminated (≈ 10% fifth, tively. The three-phase voltage system has been generated with
7% seventh, 4% eleventh) input wave rotating at 49.5 Hz. In a three-phase programmable ac source (three Chroma 61501
1 s, a deep magnitude change (sag) from 1 to 0.2 p.u. with a modules). In order to obtain the test signals containing sags,
+45◦ phase jump has been programmed. Fig. 11(b) shows how unbalance, harmonics, and notching signals, the three-phase ac
steady-state phase error is achieved; the transient lasts 0.01 s. source has been used as an arbitrary power amplifier connected
Fig. 11(c) shows |Vp |, amplitude error, Vd , and Vq ; as expected, to a three-phase arbitrary waveform generator based on a DSP
Vd and Vq rotate at 0.5 Hz. The average amplitude error is card [48]. The input signals have been acquired through the
nonzero only during the transient (0.01 s). Fig. 11(d) shows the dSpace I/O interface by means of LEM LV25-P voltage trans-
Clarke variables employed in the calculation of θ1 (14). ducers. As in previous section, p.u. units have been employed.
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 105

TABLE III
BRIEF COMPARATIVE AMONG SIGNIFICANT SYSTEMS WITH GOOD UNBALANCE REJECTION

As the system is an open-loop system, no feedback error S2 as expected from Figs. 7(e) and 8(b). Fig. 17 emphasizes the
signals can be presented. However, considering only the fun- fact that the system gain is independent of the input amplitude.
damental component, Va = Vα for balanced conditions and A. Brief Comparative
unbalalanced systems when Van = −Vah , i.e., V
n and V
h
cancel each other for the a phase [see Section II, (2), and Table III shows a brief comparative with other significant
(13)]. Therefore, testing the systems under such conditions, algorithms that successfully deal with unbalance [25], [28], [29].
Va − Vα represents the error, i.e., Va − Vα = 0 implies zero The settling time (Ts ) is defined as the time required for the
error in phase, frequency, and amplitude. The subtraction func- system to settle within a certain percentage of the steady-state
tion of the oscilloscope has been used to obtain this error value [50].
(channel M1). As proved, for S1 and S2, Ts is accurately set by the step
Fig. 12 shows the error signal of S1 in steady state when response of the FIR filtering stage (19): 0.01 and 0.02 s, re-
the set of inputs is balanced and oscillates at 48 Hz. As shown, spectively. In relation to device implementation, it is clear that
the error is negligible. In the zoomed figure, the small positive the moving average filters are a part of resource consuming,
delay between the input (Va ) and the output or processed signal as shown in the increment of execution time from S1 to S2.
(Vα ) could be noticed. This error is due to the sampling process However, this and other implementations [51]–[53] prove that
(execution and conversion times). current digital devices are powerful enough to implement them
Fig. 13 shows the error signal of S1 in steady state for an un- in real time. With regard to amplitude dependence, due to the lin-
balanced set of inputs oscillating at 51 Hz. As explained before, earity of all the employed digital filters, the system is amplitude
Van = −Vah and |Vn | = 0.1|Vp | in order to test the error signal. independent.
These figures also prove the absence of the steady-state error The settling time Ts (within 2% of the final value) of the
of the algorithm. Fig. 13(b) again shows the small delay in the double SRF (DSRF)-PLL has been estimated using the data
visualization due to the sampling process. (ωc = 2π × 25 rad/s, ξ = 0.707) of the equivalent linearized
Fig. 14 shows the error signal of S1 before, during, and after systems studied in [25] through the formula [50]
a transient. The programmed transient is a sag with postfault
4
component [49]. During the fault, |Vp | = 0.8 p.u. and Van = Vh Ts = . (33)
with |Vn | = 0.05 p.u., and a −45◦ phase jump with respect to ξωc
the balanced wave of |Vp | = 1 p.u. The accuracy of this estimation can be checked in [25, Figs. 11
Once proved the theoretical approach in terms of zero steady- and 12]. The execution time of the DSRF-PLL is not shown
state error and transient response, other interesting figures are in [25], but it is smaller than the implementation sampling time
shown. Because of the features of the input waves, the error (50 µs).
signal is not available for these tests. The nonlinearity introduced by the phase detector in PLLs and
Fig. 15 shows how the measurement of θ1 through S1 for a also in the DSRF-PLL has the drawback of amplitude depen-
set of unbalanced (|Vn | = 0.09|Vp |) and harmonic contaminated dence, even though this can be reduced through a normalization
(≈ 9% fifth, 5% seventh, 3% eleventh) input waves rotating at stage [20], [23], [27].
48 Hz is rippleless. Ts of NPFS presents a high variation depending on the kind
Figs. 16 and 17 compare S1 with S2 in terms of the tradeoff of transient, as shown in [28, Fig. 12], which is a nonlinear
between filtering and transient response. feature. This time is always higher than 0.02 s.
Fig. 16 shows the test made for a balanced input wave contain- The WLSE algorithm, tested when the input frequency is
ing notches, which, under the frequency domain point of view, tracked [29, Figs. 6 and 7], presents a very quick response and
is noise. Fig. 16(a) shows that Vα measured with S1 contains good harmonic/noise rejection. However, its frequency adapta-
notches, even though smoother than in input waves. Fig. 16(b) tion algorithm is very slow when compared with the NPFS one.
shows that Vα measured with S2 is clean. These results are The weak point of the WLSE algorithm is its slow frequency
expected from Figs. 7(a) and 8(a). adaptation dynamics [25].
Fig. 17 shows how S2 improves the frequency rejection of In general, it could be said that adaptive-filtering-based al-
S1 at the cost of incrementing the resources (time of execution) gorithms such as Kalman based and WLSE have a tradeoff
and transient response. This test has been realized with an un- between transient response and frequency adaptation, as proved
balanced (|Vn | = 0.09|Vp |) input wave that presents a very deep in [4]. The WLSE approach of [29] is in the extreme of very fast
sag: from 1 to 0.1 p.u. with −45◦ phase jump. S1 is faster than transient response but with a very slow frequency adaptation.
106 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 24, NO. 1, JANUARY 2009

From Table III, it is clear that S1 and S2 present very good [11] W. Stefanutti, P. Mattavelli, G. Spiazzi, and P. Tenti, “Digital control
dynamics and are suitable for real-time implementation with a of single-phase power factor preregulators based on current and voltage
sensing at switch terminals,” IEEE Trans. Power Electron., vol. 21, no. 5,
high performance. pp. 1356–1363, Sep. 2006.
[12] C. A. Busada, H. G. Chiacchiarini, and J. C. Balda, “Synthesis of sinu-
soidal waveform references synchronized with periodic signals,” IEEE
VII. CONCLUSION Trans. Power Electron., vol. 23, no. 2, pp. 581–590, Mar. 2008.
[13] P. Xiao, K. A. Corzine, and G. K. Venayagamoorthy, “Multiple reference
An open-loop algorithm has been presented to estimate phase, frame-based control of three-phase PWM boost rectifiers under unbal-
frequency, and amplitude of three-phase utility grids, using Park anced and distorted input conditions,” IEEE Trans. Power Electron.,
and Clarke transformations, a digital oscillator model, moving vol. 23, no. 4, pp. 2006–2017, Jul. 2008.
[14] F. Blaabjerg, R. Teodorescu, M. Liserre, and A. V. Timbus, “Overview
average filters, and predictive filters. of control and grid synchronization for distributed power generation sys-
1) It has a very good filtering/transient response tradeoff. It tems,” IEEE Trans. Ind. Electron., vol. 53, no. 5, pp. 1398–1409, Oct.
is specially advantageous in the presence of unbalances 2006.
[15] C. N. M. Ho, H. S. H. Chung, and K. T. K. Au, “Design and implementation
and harmonic distortion. of a fast dynamic control scheme for capacitor-supported dynamic voltage
2) Since it is based on the implementation of linear filters restorers,” IEEE Trans. Power Electron., vol. 23, no. 1, pp. 237–251, Jan.
(oscillator, moving average, and predictive), its implemen- 2008.
[16] Y. W. Li, F. Blaabjerg, D. M. Vilathgamuwa, and P. C. Loh, “Design and
tation is simple and low in resources. No explicit trigono- comparison of high performance stationary-frame controllers for DVR
metric functions are required. implementation,” IEEE Trans. Power Electron., vol. 22, no. 2, pp. 602–
3) Its dynamics are independent of the amplitude of the input 612, Mar. 2007.
[17] H. Akagi, S. Inoue, and T. Yoshii, “Control and performance of a trans-
waves, and therefore, the system is highly stable. formerless cascade PWM STATCOM with star configuration,” IEEE
4) It is frequency adaptive without resorting to an extra fre- Trans. Ind. Appl., vol. 43, no. 4, pp. 1041–1049, Jul./Aug. 2007.
quency loop. It yields zero steady-state average error for [18] D. Jovcic, N. Pahalawaththa, and M. Zavahir, “Analytical modelling of
HVDC-HVAC systems,” IEEE Trans. Power Del., vol. 14, no. 2, pp. 506–
all the measurements with independence of the input sig- 511, Apr. 1999.
nal frequency. [19] H. Chen, Z. Xu, and F. Zhang, “Nonlinear control for VSC based HVDC
These features bode well for the applicability of the algo- system,” in Proc. IEEE Power Eng. Soc. Gen. Meeting Jun. 18–22, 2006,
pp. 1–5.
rithm in the control of grid-connected power converters, grid [20] D. Jovcic, “Phase locked loop system for FACTS,” IEEE Trans. Power
monitoring in distributed power generation systems, and other Syst., vol. 18, no. 3, pp. 1116–1124, Aug. 2003.
applications where a fast and accurate online measurement of [21] V. Kaura and V. Blasko, “Operation of a phase locked loop system under
distorted utility conditions,” IEEE Trans. Ind. Appl., vol. 33, no. 1, pp. 58–
phase, frequency, and amplitude are mandatory. 63, Jan. 1997.
[22] S. K. Chung, “Phase-locked loop for grid-connected three phase conver-
REFERENCES sion systems,” in Proc. IEEE Electr. Power Appl., May 2000, vol. 147,
no. 3, pp. 213–219.
[1] H.-L. Jou, K.-D. Wu, J.-C. Wu, and W.-J. Chiang, “A three-phase four- [23] F. D. Freijedo, J. Doval-Gandoy, O. Lopez, D. Pineiro, C. Martinez-
wire power filter comprising a three-phase three-wire active power filter Penalver, and A. Nogueiras, “Real-time implementation of a SPLL for
and a zigzag transformer,” IEEE Trans. Power Electron., vol. 23, no. 1, FACTS,” in Proc. IECON, Paris, France, Nov. 2006, pp. 2390–2395.
pp. 252–259, Jan. 2008. [24] L. G. B. Barbosa, Rolim, D. R. Rodrigues, da Costa, Jr., and M. Aredes,
[2] C. Lascu, L. Asiminoaei, I. Boldea, and F. Blaabjerg, “High performance “Analysis and software implementation of a robust synchronizing PLL
current controller for selective harmonic compensation in active power circuit based on the pq theory,” IEEE Trans. Ind. Electron., vol. 53, no. 6,
filters,” IEEE Trans. Power Electron., vol. 22, no. 5, pp. 1826–1835, Sep. pp. 1919–1926, Dec. 2006.
2007. [25] P. Rodriguez, J. Pou, J. Bergas, J. I. Candela, R. P. Burgos, and D. Boroye-
[3] M. Routimo, M. Salo, and H. Tuusa, “Comparison of voltage-source and vich, “Decoupled double synchronous reference frame PLL for power
current-source shunt active power filters,” IEEE Trans. Power Electron., converters control,” IEEE Trans. Power Electron., vol. 22, no. 2, pp. 584–
vol. 22, no. 2, pp. 636–643, Mar. 2007. 592, Mar. 2007.
[4] V. M. Moreno, M. Liserre, A. Pigazo, and A. Dell’Aquila, “A compar- [26] R. E. Best, Phase Locked Loops Design, Simulation and Applications, 4th
ative analysis of real-time algorithms for power signal decomposition in ed. New York: McGraw-Hill, 1999.
multiple synchronous reference frames,” IEEE Trans. Power Electron., [27] S. Muyulema, E. J. Bueno, F. J. Rodriguez, S. Cobreces, and D. Diaz,
vol. 22, no. 4, pp. 1280–1289, Jul. 2007. “Response of the grid converters synchronization using p.u. magnitude
[5] L. Asiminoaei, C. Lascu, F. Blaabjerg, and I. Boldea, “Performance im- in the control loop,” in Proc. IEEE Int. Symp. Ind. Electron. 2007 (ISIE
provement of shunt active power filter with dual parallel topology,” IEEE 2007), Jun. 4–7, pp. 186–191.
Trans. Power Electron., vol. 22, no. 1, pp. 247–259, Jan. 2007. [28] R. F. de Camargo and H. Pinheiro, “Synchronisation method for three-
[6] P. T. Cheng and T. L. Lee, “Distributed active filter systems (DAFSs): phase PWM converters under unbalanced and distorted grid,” Proc. Inst.
A new approach to power system harmonics,” IEEE Trans. Ind. Appl., Electr. Eng. Electr. Power Appl., vol. 153, no. 5, pp. 763–772, Sep. 2006.
vol. 42, no. 5, pp. 1301–1309, Sep./Oct. 2006. [29] H. S. Song and K. Nam, “Instantaneous phase-angle estimation algorithm
[7] L. Kunjumuhammed and M. Mishra, “A control algorithm for single- under unbalanced voltage-sag conditions,” Proc. Inst. Electr. Eng. Gen-
phase active power filter under non-stiff voltage source,” IEEE Trans. eration, Transmiss. Distrib., vol. 147, pp. 409–415, Nov. 2000.
Power Electron., vol. 21, no. 3, pp. 822–825, May 2006. [30] J. Svensson, “Synchronisation methods for grid-connected voltage source
[8] M. Liserre, R. Teodorescu, and F. Blaabjerg, “Multiple harmonics control converters,” in Proc. Inst. Electr. Eng. Generation, Transmiss. Distrib.,
for three-phase grid converter systems with the use of PI-RES current vol. 148, no. 3, pp. 229–235, May 2001.
controller in a rotating frame,” IEEE Trans. Power Electron., vol. 21, [31] F. Freijedo, J. Doval-Gandoy, O. Lopez, and C. Martinez-Penalver, “New
no. 3, pp. 836–841, May 2006. algorithm for grid synchronization based on fourier series,” in Proc. EPE
[9] A. I. Maswood and F. Liu, “A unity-power-factor converter using the Conf. Sep. 2–5, 2007, pp. 1–6.
synchronous-reference-frame-based hysteresis current control,” IEEE [32] Power System BlockSet, User Manual. Mathwoks, Natick: MA, 2000.
Trans. Ind. Appl., vol. 43, no. 2, pp. 593–599, Mar. 2007. [33] S. K. Mitra, Digital Signal Processing. A Computer-Based Approach, S.
[10] B. Wang, G. Venkataramanan, and A. Bendre, “Unity power factor W. Director, Ed. New York: McGraw-Hill, 1998.
control for three-phase three-level rectifiers without current sensors,” [34] R. Wu, S. B. Dewan, and G. R. Slemon, “A PWM AC-to-DC converter
IEEE Trans. Ind. Appl., vol. 43, no. 5, pp. 1341–1348, Sep./Oct. with fixed switching frequency,” IEEE Trans. Ind. Appl., vol. 26, no. 5,
2007. pp. 880–885, Sep./Oct. 1990.
FREIJEDO et al.: GENERIC OPEN-LOOP ALGORITHM FOR THREE-PHASE GRID VOLTAGE/CURRENT SYNCHRONIZATION 107

[35] T. G. Habetler, “A space vector-based rectifier regulator for AC/DC/AC Francisco D. Freijedo (M’07) was born in Spain
converters,” IEEE Trans. Power Electron., vol. 8, no. 1, pp. 30–36, Jan. in 1978. He received the M.Sc. degree in physics
1993. from the University of Santiago de Compostela,
[36] D. Holmes and D. Matin, “Implementation of a direct digital predictive Santiago de Compostela, Spain, in 2002. He is cur-
current controller for single and three phase voltage source inverters,” in rently working toward the Ph.D. degree in the Depart-
Proc. IAS, San Diego, CA, Oct. 6–10, 1996, vol. 2, pp. 906–913. ment of Electronic Technology, University of Vigo,
[37] Y. A. R. I. Mohamed and E. F. El-Saadany, “Robust high bandwidth Vigo, Spain.
discrete-time predictive current control with predictive internal model— Since 2005, he has been an Assistant Professor
A unified approach for voltage-source PWM converters,” IEEE Trans. at the University of Vigo. His current research inter-
Power Electron., vol. 23, no. 1, pp. 126–136, Jan. 2008. ests include power quality problems, grid-connected
[38] P. Cortes, J. Rodriguez, D. E. Quevedo, and C. Silva, “Predictive current switching converters, ac power conversion, and flex-
control strategy with imposed load current spectrum,” IEEE Trans. Power ible ac transmission system (FACTS).
Electron., vol. 23, no. 2, pp. 612–618, Mar. 2008.
[39] S.-J. Jeong and S.-H. Song, “Improvement of predictive current control
performance using online parameter estimation in phase controlled recti-
fier,” IEEE Trans. Power Electron., vol. 22, no. 5, pp. 1820–1825, Sep.
2007. Jesús Doval-Gandoy (M’99) received the M.Sc.
[40] G. Abad, M. Rodriguez, and J. Poza, “Two-level VSC based predictive degree from the Polytechnic University of Madrid,
direct torque control of the doubly fed induction machine with reduced Madrid, Spain, in 1991, and the Ph.D. degree from
torque and flux ripples at low constant switching frequency,” IEEE Trans. the University of Vigo, Vigo, Spain, in 1999.
Power Electron., vol. 23, no. 3, pp. 1050–1061, May 2008. From 1991 to 1994, he was with the industry. He is
[41] C. Fortescue, “Method of symmetrical coordinates applied to the solution currently an Associate Professor at the University of
of polyphase netwotks,” Trans. AIEE, vol. 37, pp. 1027–1140, 1918. Vigo. His current research interests include the area
[42] G. C. Paap, “Symmetrical components in the time domain and their appli- of ac power conversion.
cation to power network calculations,” IEEE Trans. Power Syst., vol. 15,
no. 2, pp. 522–528, May 2000.
[43] G. Gonzalez, “Theory of oscillators,” in Foundations of Oscillator Circuit
Design, A. H. Publishers, Ed. Norwood, MA: Artech House, 2007.
[44] S. W. Smith, The Scientist and Engineer’s Guide to Digital Signal Pro-
cessing. San Diego, CA: California Technical, 1998.
[45] R. Costa-Castello, R. Grino, and E. Fossas, “Odd-harmonic digital repet- Óscar López (M’05) was born in Spain in 1975. He
itive control of a single-phase current active filter,” IEEE Trans. Power received the M.Sc. degree in 2001 from the University
Electron., vol. 19, no. 4, pp. 1060–1068, Jul. 2004. of Vigo, Vigo, Spain, where he is currently working
[46] R. Costa-Castello, R. Grino, and E. Fossas, “Reply to “Concerning ‘odd- toward the Ph.D. degree in the Department of Elec-
harmonic digital repetitive control of a single-phase current active fil- tronic Technology.
ter’”,” IEEE Trans. Power Electron., vol. 21, no. 4, pp. 1159–1160, Jul. He is currently an Assistant Professor at the Uni-
2006. versity of Vigo. His current research interests include
[47] K. Zhou, K.-S. Low, D. Wang, F.-L. Luo, B. Zhang, and Y. Wang, “Zero- the areas of power switching converters technology.
phase odd-harmonic repetitive controller for a single-phase PWM in-
verter,” IEEE Trans. Power Electron., vol. 21, no. 1, pp. 193–201, Jan.
2006.
[48] (2008, Jul. 24). Programmable ac power source model 61500 series
[Online]. Available: www.chromaate.com
[49] M. H. J. Bollen, Understanding Power Quality Problems. Voltage Sags Enrique Acha (M’90–SM’99) was born in México.
and Interrupions, R. J. Herrick, Ed. Piscataway, NJ: IEEE Press, 2000. He received the B.Sc. degree in electrical engineering
[50] R. C. Dorf and R. H. Bishop Modern Control Systems, D. A. George, Ed. from the Universidad Michoacana de San Nicolás de
Englewood Cliffs, NJ: Prentice-Hall, 2001. Hidalgo, Morelia, México, in 1979, and the Ph.D. de-
[51] H. Awad, H. Nelsen, F. Blaabjerg, and M. J. Newman, “Operation of static gree from the University of Canterbury, Christchurch,
series compensator under distorted utility conditions,” IEEE Trans. Power New Zealand, in 1988.
Syst., vol. 20, no. 1, pp. 448–457, Feb. 2005. He was a Postdoctoral Fellow at the University of
[52] B. P. McGrath, D. G. Holmes, and J. J. H. Galloway, “Power converter Toronto, Toronto, ON, Canada, and the University of
line synchronization using a discrete Fourier transform (DFT) based on Durham, Durham, U.K. He is currently the Professor
a variable sample rate,” IEEE Trans. Power Electron., vol. 20, no. 4, of electrical power systems at the University of Glas-
pp. 877–884, Jul. 2005. gow, Glasgow, U.K. He has authored or coauthored
[53] M. A. Perez, J. R. Espinoza, L. A. Moran, M. A. Torres, and E. A. Araya, three text books on various aspects of power electronic applications in electrical
“A robust phase-locked loop algorithm to synchronize static-power con- power systems, and over one hundred research papers.
verters with polluted AC systems,” IEEE Trans. Ind. Electron., vol. 55, Prof. Acha is the IEEE Power Engineering Society (PES) Summer Distin-
no. 5, pp. 2185–2192, May 2008. guished Lecturer.