This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON POWER DELIVERY

A Method for Averting Saturation From Series

Transformers of Dynamic Voltage Restorers
Darlan A. Fernandes, Member, IEEE, Fabiano F. Costa, Member, IEEE, and Monti A. Vitorino, Member, IEEE

AbstractThis paper proposes a technique for preventing

saturation in series transformers from dynamic voltage restorer
(DVR) systems. The method consists in correcting the voltages
which are injected through the transformers into the power system
to compensate voltage sags. It restricts the compensating voltages
during the sag whenever it predicts that a maximum limit for
the flux linkage is about to be exceeded. The prediction is carried
out at the beginning of a stabilized voltage sag. Moreover, the
technique allows a certain level of sag compensation even when
the estimated flux is expected to exceed the saturation limit. The
voltage sag level and phase are computed through an adaptive
recursive least squares (RLS). The RLS estimation incorporates
a transient period before it achieves a stable state whenever
there is a sag event. The DVR is not supposed to operate in this
period. Therefore, this paper also outlines a simple procedure for
detecting the RLS estimation stable level. Simulation of different
scenarios of voltage sags and the results from the experimental
implementation of a DVR show the effectiveness of the method.
Index TermsDynamic voltage restorer (DVR), flux linkage,
power quality (PQ), recursive least squares, saturation, transformer.

I. INTRODUCTION

OLTAGE sags are the most incident disturbances inflicting the power system. Surveys indicated that 92% of
interruptions in industrial facilities may occur due to voltage
sags [1]. The economic impact to the industries and utilities is
severe due to equipment damage and loss of production [2].
To mitigate this problem, the utilities can invest in the power
system design in order to reduce the faults incidence and the
time for their clearance. Also, redundant lines can be installed
to feed critical loads [3]. Unfortunately, these solutions are
complex and costly to implement. This enables local-based
alternatives where some equipment is fixed in the system-load
Manuscript received August 05, 2013; revised January 28, 2014 and March
07, 2014; accepted March 09, 2014. This work was supported by the Brazilian
Research Council (CNPq) through project processes nos. 482736/2011-9,
454638/2012-4, and 420442/2013-8. Paper no. TPWRD-00876-2013.
D. A. Fernandes is with the Department of Electrical Engineering, Federal
University of Paraba (UFPB), Joo Pessoa PB, 58.051-900, Brazil (e-mail:
darlan@cear.ufpb.br).
F. F. Costa is with the Department of Electrical Engineering, Federal University of Bahia (UFBA), Salvador BA, 40210630, Brazil (e-mail: fabiano.
costa@ufba.br).
M. A. Vitorino is with the Department of Electrical Engineering, Center of
Electrical Engineering and Informatics, Federal University of Campina Grande
(UFCG), Campina Grande- PB, 58.429-900, Brazil (e-mail: vitorino@dee.ufcg.
edu.br).
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/TPWRD.2014.2311577

Fig. 1. System configuration with a DVR.

interface. One example of this approach is the usage of dynamic

voltage restorers (DVRs).
A DVR is one of the most effective custom power devices for
voltage sag and swell compensation and it has been attracting
growing attention in recent years [1], [4][8]. A typical test
system, incorporating a DVR, is depicted in Fig. 1. The DVR
injects compensating voltages to the power lines through a
three-phase series transformer or three single-phase series
transformers. A problem may arise when the DVR system corrects a severe sag. In this situation, the compensating voltages
can cause flux linkage in the core to exceed the transformer
nominal limit [9]. The exceeding flux is caused by a dc component whose amplitude depends on the initial phase angle of
the compensating voltage. This, in turn, leads to overcurrent
and overheating, reducing the useful life of the transformer. To
overcome this problem, one alternative is to enlarge the series
transformers. This solution brings an increase in physical size,
weight, and cost to the transformer [10]. Another approach is
to apply DVR systems without transformers [11], [12]. This
increases the number of switches and their associated control
circuits needed to apply the compensating voltage. Finally,
there is the strategy of controlling the flux linkage by limiting
the voltage injected to compensate the sag. This approach has
a compromising nature. It deals with the necessity of reducing
the sag and the demand of not changing the transformers flux
limit.
In [13], the flux linkage in the DVRs transformers is kept
under a maximum limit by shutting off the reference voltages
while the currents are over a specified limit, or the reference
voltages are reversed. The drawback is to impose a full sag to the

0885-8977 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2

IEEE TRANSACTIONS ON POWER DELIVERY

load during a certain period. In [14], the controller uses the magnitude of the positive-sequence component of the line voltages
to identify voltage sags. The flux is estimated by means of the
integral of the voltage and whenever it reaches a given limit, the
voltage is set to zero. The method proposed in [15] also makes
use of the flux estimation in order to limit the compensating voltages. The voltage injection action is divided into three intervals.
Between the sag detection instant and one-sixth of the fundamental period , after the sag detection, the injected compensating voltage is fully applied to compensate the sag. Between
and
, the injected voltage is adjusted to zero. From the
instant
on, after the detection, again the full compensating
voltage is applied. In [16], two methods for flux-linkage control are presented. In the first one, the compensating voltage be
multiplied by half during the first half fundamental cycle after
the sag detection instant and, after this period, the full compensating voltage is used. This ensures that the dc flux is wiped off.
The second method predicts, at the detection instant, whether
the flux will surpass the maximum limit within half a cycle past
the zero cross. If the flux is exceeded, then it is introduced a form
factor to limit the compensating voltage. This method has the
advantage of allowing a level of voltage to be injected through
the transformers during all periods of the sag.
The inrush control systems for all of the aforementioned
works rely on the estimation of the load voltage phasors to
compute the compensating voltages. Some works employ the
standard least-squares error to estimate these phasors [17],
[18], notwithstanding, in general, that the techniques for estimating them are not exploited or discussed by the authors.
This paper expands the ideas developed in [16] by dealing with
the possibility of more restrictive limits for the saturation in
the transformers core. In addition, it proposes the use of an
adaptive RLS-based technique for the estimation of phasors,
suitable to be incorporated to the compensation voltage control
as well as the inrush control. The voltage phasor amplitudes are
employed to verify whether the flux surpasses the transformers
flux limit. The RLS algorithm necessarily includes a transition
time before its estimation achieves a constant level. In the
proposed method, it is assumed that the DVR system should
not operate before it has a stabilized reference for the sag.
Therefore, this paper also proposes a simple procedure to detect
whether the RLS estimation reaches a constant value for the
estimation of phasors.
II. METHOD FOR CONTROLLING SATURATION
This section devises the method of controlling saturation proposed in this paper. The fundamental idea is to constrain the
compensating voltage
by multiplying it by a form factor .
In order to accomplish such a goal, one must predict, at the moment of the sag detection, the value for the form factor to be
applied up to the end of the next half cycle (or the next whole
cycle) of the compensating voltage after the sag detection and
keep the flux at its limit value. In general, can be described as

the linked flux in the transformers core at a given instant

can be expressed by

(2)
Solving (2) and assuming that the transformer is demagnetized,
0 at the
0, the following expression for the flux
that is,
is obtained:
(3)
The first part of (3) represents the ac component of the flux,
while the second one is its dc component. Whenever the injected
voltage started at a zero cross, that is,
, the peak of
the flux reaches its maximum value. For instance, if
,
the expression for the flux is given by
(4)
The technique proposed in this paper is inspired by the one
described in [16]. Consider Fig. 2, where the injected voltage
starts at angle . It is possible to predict the maximum excursion
for the flux linkage through the following integration:

(5)
where is a form factor which is first set to unity. Note that between and
, the injected voltage contributes positively to
the flux. Between
and
, the voltage contributes negatively to the flux. Therefore, in the situation depicted in Fig. 2,
at the angle
, the flux reaches its minimum value. If the
module of the prediction provides a value higher than the allowed limit for the transformer, then the parameter must be
adjusted to a value which restricts the amplitude to be equal to
the lower limit. Thus, if
, make
in (5)
and find as
(6)
Applying the factor , computed through (6), to the compensating voltage during its negative semicycle, ensures that the
flux will not surpass the minimum limit. When the injected
voltage starts within a negative semicycle, at the point
,
is predicted through

(7)
, the injected voltage is required to be scaled by
If
the form factor computed by

(1)
where and are, respectively, the fundamental frequency and
the initial phase of the compensating voltage. By Faradays law,

(8)

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
FERNANDES et al.: METHOD FOR AVERTING SATURATION FROM SERIES TRANSFORMERS

A. RLS Estimator
To model voltages acquired from power systems, it is usual
to describe them as a sum of sinusoids, with one being the
fundamental, and the others being harmonics. If the voltage is
corrupted by harmonics, this representation ensures that the dynamics of the harmonics do not contaminate the parameters estimation related to the fundamental sinusoid. Hence, denoting
the data of voltages by , the model is a sum of sinusoids
provided by

Fig. 2. Compensating voltage for one of the phases.

(11)
It must be noted that the procedure described before only
shifts the flux curve so that, up to end of the semicycle, subsequently after the start of the voltage injection, its value is not
higher than the transformers flux limit. It still remains a dc component which can cause the flux value to surpass the allowed
limit within the subsequent opposite semicycle. Therefore, in
the proposed method, the condition
(9)

where
and
are, respectively, the amplitude and the
phase of the sinusoid of
frequency, and is the time index.
The time interval
is the sampling period. Its selection does
not interfere with the RLS performance once the Nyquist criterion is observed [19]. The first of the sinusoids is related to the
fundamental phasor. The model described by (11) is not applicable for the RLS algorithm. The parameters
are not linear
with respect to the model . Thus, the model is rewritten as

must be verified. Note that is the peak value for the compensating voltage. If the condition (9) is not observed, the compensating voltage must be computed as
(10)
where
.
The method implementation is carried out in such a manner
that, whenever the initial phase is detected within the interval
from
up to , it must be subtracted by . This is necessary because the RLS algorithm computes ranging from 0 up
and leaves out the interval
up to 0.
In the proposed method, there is a need to compute the amplitude and phase of the compensating voltage. This task is carried
out by a recursive least squares method [19] which, for each
instant, updates the amplitude and phase estimation. This is discussed in the next section. Furthermore, the DVR voltage correction only takes place in the moment where the estimation for
the parameters is stabilized, that is, during the estimation transient when there is no compensating voltage injected into the
grid. The proposed method is applied for each one of the three
phases, accordingly to the flowchart depicted in Fig. 3.
III. COMPENSATING VOLTAGE CONSTRUCTION
The compensating voltage construction applied in this paper
makes use of an RLS algorithm which computes the amplitude
and the phase for each sample of the grid voltage
. The
RLS algorithm is applied for each one of the three grid phases.
It is worth emphasizing that the DVR is not meant to compensate the voltage while the RLS estimation is in its transient period. Therefore, in this paper, a method is proposed for flagging
whether the RLS estimation is stable. The next subsection is
dedicated to explain the RLS algorithm used in this paper. The
following one outlines the procedure in which the compensating
voltage is only injected when the RLS estimation is constant.

(12)
where
tions

and

are related to the model (11) through equa-

(13)
(14)
Equation (12) can compactly be written as
(15)
where

is a vector of regressors given by

(16)

and
is a vector of parameters to be determined and whose
elements are given by
(17)
It should be noted that the subscript in
refers to the
estimation of the parameters carried out for the instant
.
For instance, the element
is the estimation of
for the
instant
.
The discordance between the data signal and its model
at a given instant is the prediction error
, provided by
(18)

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4

IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 3. Flowchart for the proposed method.

The RLS algorithm updates the estimation for the parameters

according to the following equation:

In this technique, the covariance matrix is updated according to

the following rule:
(22)

(19)
where

is a gain given by
(20)

is the so-called covariance matrix which is updated by

and
the following recursive equation:
(21)
Before the algorithm is triggered, the initial covariance matrix must be adjusted [20]. Usually, this initial matrix is set to
be diagonal with the elements that have high values in comparison with the values of the parameters to be estimated. Then,
this matrix has a higher norm in a sense that it has greater capability to modify a norm of a vector which is multiplied by it.
During the RLS application, as the estimative converges for the
true values of the parameters, the
norm is reduced. Hence,
this algorithm is not intrinsically adaptive. In order to provide
adaptability for the RLS, it must be added to some mechanism
where the covariance matrix is updated so that its norm value
is always within an adequate range. The technique selected in
this paper is designated modified random walking (MRW) [19].

where and are arbitrarily adjusted. This algorithms structure is suitable for the proposed flux control application. The
monitoring of the error
can be used not only to provide
adaptability for the RLS, but also to detect the voltage sag.
The DVR action must be performed only after the transient
of the parameters estimation. Thus, the next subsection outlines
a simple manner of detecting when the transient is finished, that
is, a mean of detecting the constant level for the parameters
estimation.
B. Constant Level Detection
In order to detect a constant level for the parameters estimation, one can average an -length moving window for the estimation of the amplitude
through the equation
(23)
where is the last sample of the amplitude estimation. This
average can be used to compute a sum given by
(24)

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
FERNANDES et al.: METHOD FOR AVERTING SATURATION FROM SERIES TRANSFORMERS

Fig. 4. Simulation platform of the DVR system.

This sum tends to be zero whenever the parameters estimation

is a constant level. Hence, if is less or equal to a given limit
immediately after a sag detection, one can ensure that the RLS
estimation for the parameters, that is, the amplitude and phase
of the sag, has passed its transition time.
For each voltage sample, the RLS algorithm computes the
phase
and the amplitude
which is provided to the constant level detector. This detector sets a flag signal to 1 whenever
the parameters estimation is stable. On the instant that the parameters start changing, the RLS sets the flag signal to 0. Therefore, the flag signal can be used as an enable signal to the DVR
action. The compensating voltage is injected by the DVR as
if flag
if flag

(25)

where is the difference between a reference value

and
the estimated
.
The constant level detection procedure imposes the sampling
rate to be fast enough so that a given constant level is detected
with no critical delay in comparison with the fundamental cycle.
In the simulations and experiments, the sampling frequency has
been set in 10 kHz and the moving average
has been carried
out with five samples.
IV. SIMULATED RESULTS
In order to test the performance of the proposed algorithm,
a DVR system has been simulated by using the packages SimPowerSystems and Simulink from Matlab. Thus, the simulation
is comprised of a three-phase load which is fed by a three-phase
voltage source and is protected by the DVR, as depicted in
Fig. 4. Table I shows the parameters used for the simulation.
The filter parameters are computed so that the cutoff frequency
is around 890 Hz.
In the figure, it can be observed that the compensating voltage
is injected by means of a four-leg voltage-source inverter (VSI)
controlled by a pulsewidth-modulation (PWM) strategy. The
four-leg VSI allows the compensating voltages to be unbalanced
[21]. The VSI voltage output is passed through an LC filter. The
RLS block incorporates the procedure for detecting whether the

TABLE I
SIMULATED SYSTEM PARAMETERS

RLS estimation for the amplitude and phase is stabilized. It also

encloses the logic to enable the RLS through the flag signal.
Three different scenarios of voltage sags have been simulated. For the first case, consider that a three-phase grid is under
a phase-to-phase sag during 58 ms, as illustrated in Fig. 5(a).
Fig. 5(b) shows the least-squares amplitude estimation for
phase-A. Besides the amplitude, the least-squares algorithm
also estimates the angle . Moreover, the estimation enables
detection where the changing in the estimation starts and where
it is finished by means of a flag signal. In Fig. 5(b), these
instants are indicated by the dashed line. We can highlight four
instants. The first two are related to the beginning of the sag and
the last two are associated with the end of the sag. Recall that
the compensating voltage is not triggered while the amplitude
and phase are varying. Fig. 5(c) shows the injected voltage by
the DVR for the two sagged phases. It is worth noting that the
injected voltage for phase-A is ignited at the instant where the
estimation for the phase-A amplitude is stabilized. Although
not shown in this figure, the same is true for the other phase.
Fig. 5(d) shows the flux linkage associated with each phase.
The fluxes for the two phases do not exceed the limit of 0.38
Wbturn, given in Table I. Fig. 5(e) shows the corrected voltages
applied to the load.
In the second case, the voltage sag is depicted by Fig. 6(a)
where it is shown that a phase-to-phase fault for an angle
is different from the previous one. Fig. 6(b) shows the amplitude estimation carried out by the least-squares algorithm for
phase-A. Again, the instants where the sag is initiated and finished are highlighted by the dashed line. Fig. 6(c) shows a set
of four curves of voltages injected by the DVR into the grid.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6

Fig. 5. Simulated results for the DVR system: Case I. (a) Voltage sags on
phases A and B. (b) Amplitude estimation of phase-A. (c) Compensating voltages injected by the DVR. (d) Flux linkages in the transformers. (e) Corrected
voltages applied to the load.

The dashed curves are the compensating voltages that would be

injected without any control for the saturation. The solid lines
represent the compensating voltages restrained by the control of
saturation proposed in this paper. Fig. 6(d) shows four curves
for the fluxes within the transformers. The dashed curves are
the ones obtained without control of flux saturation. The solid
ones are obtained through the usage of the proposed algorithm.
One verifies in this case that the solid lines do not surpass the exceeding limits opposed to the situation depicted with the dashed
lines. Nevertheless, they still include a dc component. Fig. 6(e)
show the voltages applied to the load.
The third case simulates a sag for single phase, as depicted in
Fig. 7(a). Similar to the other cases, Fig. 7(b) shows the RLS amplitude estimation for the sagged phase voltage. Fig. 7(c) shows
the dashed curve for the voltage without any constraint with the
control of saturation, and the solid line represents the compensating voltage constrained by the proposed control of saturation. We verify that the difference between the curves is confined to the first quarter of the cycle. The flux curves associated
with these voltages are shown in Fig. 7(d). One can note that
the flux associated with the constrained voltage and represented
by the solid line does not include any dc component. This happens because unlike the second case, the condition described in

IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 6. Simulated results for the DVR system: Case II. (a) Voltage sags on
phases A and B. (b) Amplitude estimation of phase-A. (c) Compensating voltages injected by the DVR. (d) Flux linkages in the transformers. (e) Corrected
voltages applied to the load.

(9) is not verified and the compensating voltage must be given

according to (10). This ensures that the dc component is completely removed. The voltages applied to the load are sketched
in Fig. 7(e).
V. EXPERIMENTAL RESULTS
In order to test the performance of the proposed method in
practice, a DVR system has been set up in the laboratory. Fig. 8
shows the photographs of the system. The experimental apparatus is composed of a stage of power and the other stage of
signal and data acquisition (DAQ) is shown, respectively, in
Figs. 8(a) and (b). With regards to the stage of power, the main
component is the voltage-source inverter (VSI), composed of
four legs. The dc link is made out of four capacitors of 2200
F/450 V equivalent to 2200 F/900 V. The modulated voltages are filtered by an LC filter for each phase. The values for the
components are the same as those presented in Table I. The loads
are made of power resistors. The DAQ system is formed of Hall
sensors for voltages and currents. The signal processing is carried out by a Texas Instruments digital signal controller (DSC)
TMS320F28335. The core of the series transformers used by the
DVR system is made of ferromagnetic grain-oriented material

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
FERNANDES et al.: METHOD FOR AVERTING SATURATION FROM SERIES TRANSFORMERS

Fig. 8. Experimental dynamic voltage restorer. (a) Power stage. (b) Control
and acquisition system.

Fig. 7. Simulated results for the DVR system: Case III. (a) Voltage sag on
phase-A. (b) Amplitude estimation of phase-A. (c) Compensating voltage injected by the DVR. (d) Flux linkage in the transformer. (e) Corrected voltages
applied to the load.

in toroidal shape. These characteristics minimize the leakage inductance. The transformer is rated 1 kVA and the ratio of turns
of the secondary to the primary circuit is 1:1.
The first result shown in this section is in regards to the determination of the saturation value for the DVRs transformers.
Toward this purpose, it has been set a measurement circuit with
capacitors and resistors. Fig. 9 shows the experimental result
for the hysteresis curve for the transformer. The extreme points
of this curve represent the beginning of saturation. To the right
of this curve, the extreme point is (5.94;37.90) V. The second
coordinate is relative to the maximum value of the flux
,
which is 0.38 Wbturn.
Fig. 10 summarizes three different cases of voltage sags corrected by the DVR system. In the first case, the compensating
voltages constrained (solid) and nonconstrained (dashed) start
at 15 ms as shown in Fig. 10(a). The associated flux curves
are shown in Fig. 10(d). It can be noticed that the dashed flux
curve exceeds the minimum limit. By means of (5), the proposed
method predicts, at 15 ms, whether the flux will exceed the imposed limit. Thus, through (6), a form factor to restrain the
compensating voltage is computed and its related flux linkage
is shown by the solid line.

Fig. 9. Experimental hysteresis curve of the transformer (H. ch.

ch. 20 V/div).

5 V/div, V.

Fig. 10(b) depicts a situation where the compensating voltage

that is not constrained leads the flux to exceed the maximum
limit. Again, an adequate form factor is computed and applied
to the voltage during the half fundamental cycle. This procedure
shifts down the flux to its allowed limit, as shown in Fig. 10(e)
(solid line).
Fig. 10(c) outlines two curves. The dashed curve represents a
full voltage compensation which is associated with the dashed
curve shown in Fig. 10(f). Constraining the compensating
voltage by the form factor would result in the dashed-dotted
flux curve also shown in Fig. 10(f), which is out of the allowed lower limit. The solid curve in Fig. 10(c) is the voltage
constraint according to (10). This voltage result in a flux is
represented by the solid line in Fig. 10(f), which is within the
allowed limits.
VI. CONCLUSION
This paper has proposed a method for controlling flux saturation in transformers used by a DVR system. The DVR system
makes use of an RLS algotithm to compute the compensating
voltage. The method relies on the correct computation of the
compensating voltage phasor which is constrained whenever
it can provoke saturation. The compensation is never rendered

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8

IEEE TRANSACTIONS ON POWER DELIVERY

Fig. 10. Experimental DVR compensations and their associated flux-linkage curves. (a) (c) Compensating voltages with and without the control of saturation.
(d)(f) Associated flux-linkage curves.

while the RLS amplitude phasor estimation is varying. Hence,

the RLS algorithm is combined with a technique for detecting
whether the estimation for the amplitude reached a constant
value. This ensures that the compensating voltage is always at
a proper level. In some cases, this is performed at a cost of not
completely compensating the sag for a certain period of time.
Yet, this is a compromise solution to be applied to loads which
can withstand some level of sag within a limited period. The
method has been put on test by simulations of different scenarios
of voltage sags. In addition, a DVR prototype, including the proposed method, has been set up. For the studied cases, the results
show that the load voltages are properly corrected in less than
one fundamental cycle, which is a quite satisfactory time for a
large class of loads.
REFERENCES
[1] Y.-H. Chen, C.-Y. Lin, J.-M. Chen, and P.-T. Cheng, An inrush mitigation technique of load transformers for the series voltage sag compensator, IEEE Trans. Power Electron., vol. 25, no. 8, pp. 22112221,
Aug. 2012.
[2] R. H. G. Tan and V. K. Ramachandaramurthy, Voltage sag acceptability assessment using multiple magnitude-duration function, IEEE
Trans. Power Del., vol. 27, no. 4, pp. 19841990, Oct. 2012.
[3] M. A. Mora and J. V. Milanovic, Monitor placement for reliable estimation of voltage sags in power networks, IEEE Trans. Power Del.,
vol. 27, no. 2, pp. 936944, Apr. 2012.
[4] J. Roldn-Perz, A. Garca-Cerrada, J. L. Zamora-Macho, P. RonceroSnchez, and E. Acha, Troubleshooting a digital repetitive controller
for a versatile dynamic voltage restorer, Elect. Power Energy Syst.,
vol. 57, pp. 105115, May 2014.
[5] P. Kanjiya, B. Singh, A. Chandra, and K. A.-. Haddad, SRF theory
revisited to control self supported dynamic voltage restorer (DVR) for
unbalanced and nonlinear loads, IEEE Trans. Ind. Appl., vol. 49, no.
6, pp. 23302340, Dec. 2013.

[6] S. R. Naidu and D. A. Fernandes, Dynamic voltage restorer based on

4-leg voltage source converter, IET Gen. Transm. Distrib., vol. 3, no.
5, pp. 437447, May 2009.
[7] T. Jimichi, H. Fujita, and H. Akagi, A dynamic voltage restorer
equipped with a high-frequency isolated dc-dc converter, IEEE
Trans. Ind. Appl., vol. 47, no. 1, pp. 169175, Jan. 2011.
[8] F. B. Ajaei, S. Farhangi, and R. Iravani, Fault current interruption by
the dynamic voltage restorer, IEEE Trans. Power Del., vol. 28, no. 2,
pp. 903910, Apr. 2013.
[9] D. I. Taylor, J. D. Law, B. K. Johnson, and N. Fischer, Single-phase
transformer inrush current reduction using preuxing, IEEE Trans.
Power Electron., vol. 27, no. 1, pp. 245252, Jan. 2012.
[10] S. W. Middlekauff and E. R. Collins, System and customer impact:
Considerations for series custom power devices, IEEE Trans. Power
Del., vol. 13, no. 1, pp. 278282, Jan. 1998.
[11] A. Y. Goharrizi, S. H. Hosseini, M. Sabahi, and G. B. Gharehpetian,
Three-phase HFL-DVR with independently controlled phases, IEEE
Trans. Power Electron., vol. 27, no. 4, pp. 17061718, Apr. 2012.
[12] M. I. Mareia, A. B. Eltantawyb, and A. A. El-Sattar, An energy optimized control scheme for a transformerless DVR, Elect. Power Syst.
Res., vol. 83, no. 1, pp. 110118, Feb. 2012.
[13] J. G. Nielsen, M. Newman, H. Nielsen, and F. Blaabjerg, Control and
testing of a dynamic voltage restorer (DVR) at medium voltage level,
IEEE Trans. Power Electron., vol. 19, no. 3, pp. 806813, May 2004.
[14] P. Cheng, W. Chen, Y. Chen, C. Ni, and J. Lin, A transformer inrush
mitigation method for series voltage sag compensators, IEEE Trans.
Power Electron., vol. 22, no. 5, pp. 890899, Sep. 2007.
[15] T. Jimichi, H. Fujita, and H. Akagi, An approach to eliminating dc
magnetic flux from the series transformer of a dynamic voltage restorer, IEEE Trans. Ind. Appl., vol. 44, no. 3, pp. 809816, May/Jun.
2008.
[16] C. Fitzer, A. Arulampalam, M. Barnes, and R. Zurowski, Mitigation of
saturation in dynamic voltage restorer connection transformers, IEEE
Trans. Power Electron., vol. 17, no. 6, pp. 10581066, Nov. 2002.
[17] H. Abdollahzadeh, M. Jazaeri, and A. Tavighi, A new fast-converged
estimation approach for dynamic voltage restorer (DVR) to compensate voltage sags in waveform distortion conditions, Elect. Power Energy Syst., vol. 54, pp. 598609, Jan. 2014.
[18] F. B. Ajaei, S. Afsharnia, A. Kahrobaeian, and S. Farhangi, A fast
and effective control scheme for the dynamic voltage restorer, IEEE
Trans. Power Del., vol. 26, no. 4, pp. 23982406, Oct. 2011.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
FERNANDES et al.: METHOD FOR AVERTING SATURATION FROM SERIES TRANSFORMERS

[19] F. F. Costa, D. A. Formiga, R. R. Ferreira, T. Sousa, and F. B. Costa,

A recursive least-squares aided by pre-filtering for phasor-estimation
in distance protection, presented at the IEEE Int. Conf. Elect. Power
Eng., Grenoble, France, Jun. 2013.
[20] L. Ljung, System Identification: Theory for the User, 2nd ed. Upper
Saddle River, NJ, USA: Prentice-Hall, 1999.
[21] D. A. Fernandes, F. F. Costa, and E. C. S. Jr, Digital-scalar PWM
approaches applied to four-leg voltage-source inverters, IEEE Trans.
Ind. Electron., vol. 60, no. 5, pp. 20222030, May 2013.

Darlan A. Fernandes (S07M08) received the

B.S. degree in electrical engineering from the Federal University of Paraba, Campina Grande, Brazil,
in 2002, and the M.S. and Ph.D. degrees in electrical
engineering from the Federal University of Campina
Grande, Campina Grande, Brazil, in 2004 and 2008,
respectively.
From 2007 to 2011, he was a Professor with the Industry Department in the Federal Center of Technological Education, Rio Grande do Norte, Brazil. Currently, he is an Assistant Professor with the Department of Electrical Engineering, Federal University of Paraba. His research interests are in the applications of power electronics in distribution systems, power
quality, and photovoltaic systems.

Fabiano F. Costa (M07) received the B.S. degree

in electrical engineering from the University of So
Paulo, So Paulo, Brazil, in 1997, the M.S. degree in
electrical engineering from the Federal University of
Paraba (UFPB), Campina Grande, Brazil, in 2001,
and the Ph.D. degree in electrical engineering from
the Federal University of Campina Grande (UFCG),
Campina Grande, Brazil, in 2005.
Currently, he is an Assistant Professor with the Department of Electrical Engineering, Federal University of Bahia, Salvador, Brazil. His current research
interests include applications of digital signal processing for distributed generation, islanding techniques, and real-time digital simulation of power systems.

Monti A. Vitorino (S11M13) was born in

Campina Grande, Paraiba, Brazil, in 1983. He
received the B.Sc., M.Sc., and Ph.D. degrees in
electrical engineering from the Federal University of
Campina Grande, Campina Grande, Paraiba, Brazil,
in 2007, 2008, and 2012, respectively.
From 2011 to 2012, he was a Scholar with the
Center for Power Electronics Systems, Virginia
Polytechnic Institute and State University, Blacksburg, VA, USA. From 2012 to 2013, he was with the
Department of Electrical Engineering, Center for Renewable and Alternative Energies, Federal University of Paraiba (UFPB), Joao
Pessoa, Brazil, where he was an Assistant Professor of Electrical Engineering.
Since 2013, he has been with the Department of Electrical Engineering, Center
of Electrical Engineering and Informatics, Federal University of Campina
Grande (UFCG), Campina Grande, Brazil, where he is currently an Assistant
Professor of Electrical Engineering. His research interests include electrical
motor drives, power electronics, and renewable energy.