1024 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO.

2, APRIL 2013

Current Status and Future Trends in

Frequency-Response Analysis With a
Transformer in Service
Eduardo Gomez-Luna, Student Member, IEEE, Guillermo Aponte Mayor, Carlos Gonzalez-Garcia, Member, IEEE,
and Jorge Pleite Guerra, Senior Member, IEEE

Abstract—This paper presents the current status and future transformer structure and, therefore, are measured as an inter-
trends in the application of the frequency-response analysis (FRA) esting output to evaluate.
technique with the transformer in service (online) through bib-
The frequency spectrum of the injected signal (input) and
liographic review and analysis. As a result, three basic stages of
the online FRA test have been identified and defined: 1) injec- the measured signal (output) are obtained through mathematical
tion and excitation signal measurement; 2) recording, filtering procedures, usually the fast Fourier transform (FFT). Finally,
and processing of measured signals; and 3) curve analysis and the ratio between the two frequency spectra is obtained.
interpretation. This work presents an overview of the online FRA In the SFRA measuring method, the excitation signal or input
technique, useful for subsequent research in this area.
is a sinusoidal signal with a low-voltage (LV) amplitude (in the
Index Terms—Frequency response, power transformer, remote 1–20 range), which is applied to a transformer terminal
monitoring, review.
in a frequency sweep (in the hertz to millihertz range); again, a
transfer function (TF) is obtained from the output/input ratio.
I. INTRODUCTION So far, the frequency-response technique for transformers has
two major limitations: 1) interpretation of measured responses

T HE frequency-response analysis (FRA) is a technique to

assess a power transformer’s condition. It is currently ac-
cepted worldwide as a complementary support to other diag-
to obtain a reliable diagnosis is not clear, and 2) it is necessary
to disconnect the transformer from the system to test it (offline
measurement).
nostic techniques. It is especially appreciated for detecting po-
Regarding the first point, progress has been made to inter-
tential mechanical problems, like displacements or deforma-
pret curves obtained offline by using transformer models, which
tions in the windings and the core sheets because these kinds
are not only able to reproduce the same frequency response of
of faults are very difficult to locate through other methods.
the actual measured system (transformer), but also represent the
The principles of the FRA method consist of measuring the
transformer’s actual electromagnetic phenomena [1]–[3]. Some
transformer response in a wide frequency bandwidth. Currently,
studies also define different coefficients to quantify differences
the test is usually performed on a no-load and de-energized
in FRA curves [4]–[9].
transformer. Different types of input signals and various stages
The second point refers to the need to disconnect the trans-
can be considered, as shown in Fig. 1.
former to test it, generating high costs and decreasing supply
According to the input signal nature, two measurement
reliability. This implies that the test is usually not scheduled
methods exist: impulse frequency-response analysis (IFRA)
as frequently as desired (predictive approach), but only when
and sweep frequency-response analysis (SFRA).
a failure is suspected (corrective approach).
The IFRA method uses a single nonperiodic signal as exci-
Consequently, online performance of the frequency-response
tation or input, injected into any of the available transformer
technique offers substantial benefits for a scheduled-based di-
terminals. Its maximum value may reach hundreds of volts and
agnosis in nonstop service and, moreover, it could even lead to
the wide frequency content is suitably ensured.
condition-based maintenance. This is the main reason why this
This excitation causes induced voltages in the remaining ends
approach has generated remarkable interest among utility com-
of the same transformer. These reflected signals depend on the
panies and the scientific community [10]–[12].
Taking into account a former proposal [13] and a biblio-
Manuscript received June 07, 2012; revised September 25, 2012; accepted graphic review and analysis whose sources are summed up
December 08, 2012. Date of publication January 15, 2013; date of current ver-
sion March 21, 2013. This work was supported in part by the Colombian Depart- in Fig. 2, this paper presents some contributions and future
ment of Administrative Science, Technology and Innovation (COLCIENCIAS trends in the application of FRA with the transformer in service
494-2009) and in part by the Spanish Government (Ministry of Science and In-
novation DPI2008-05890). Paper no. TPWRD-00574-2012.
(online FRA) to be a reference for future research in this field.
E. Gomez-Luna and G. Aponte Mayor are with the Universidad del Valle, Cali
25360, Colombia (e-mail: [email protected]; guillermo.
[email protected]).
II. CURRENT STATE OF THE ART OF
C. Gonzalez-Garcia and J. Pleite Guerra are with the Department of Elec- THE ONLINE FRA TECHNIQUE
tronic Technology, Universidad Carlos III de Madrid, Madrid 28911, Spain
(e-mail: [email protected]; [email protected]). The FRA technique with the transformer in service (online
Digital Object Identifier 10.1109/TPWRD.2012.2234141 measurement) implies using either the IFRA or the SFRA

0885-8977/$31.00 © 2013 IEEE

GOMEZ-LUNA et al.: CURRENT STATUS AND FUTURE TRENDS IN FRA 1025

Fig. 1. Scheme for the measurement of offline/online FRAs in the transformers.

niques discussed ahead) that uniquely relates faults in the trans-

former to the variation in the curves measured.

A. Current Problems to Consider in the Implementation of the

Online FRA Technique
Applying the online FRA technique mainly involves three
drawbacks associated with the transformer connection to the
power system:
1) The measurement is performed in the presence of sinu-
soidal high voltage, involving:
— the measurement itself becomes much more complex
considering the high-voltage levels, personnel, and
Fig. 2. Sources of the literature survey. equipment safety concerns, and electric noise.[34];
— a possible impact on the measured response, given the
nonlinear effects of some transformers, particularly
method with the transformer still connected to the power core magnetization; this effect could lead to different
system. responses with no actual change inside the transformer,
Conducting an online test would lead to a more productive depending only on the trigger timing of the measure-
exploitation of the power system, as proposed in [14], which ex- ment equipment.
presses the importance of developing diagnostic tests on power 2) The transformer is electrically connected to the remaining
transformers in service, improving compliance with the elec- elements in the power system (source, load, switches, etc.)
tricity market’s requirements, and ensuring high reliability and and, therefore, the measurement is actually the response of
low-cost maintenance. the whole system (not just the transformer), complicating
Similarly, other authors [15] have stated that it is necessary the interpretation of the results [30], [31]. In addition, ac-
to implement online monitoring tools that continuously assess cessibility to the bushings is limited with respect to offline
the changing conditions of the transformer in the electricity measurements [21]. For these two reasons, the connection
network (due to atmospheric discharges, input/output burdens, configurations for online FRA cannot be as freely imple-
random electrical noise, and/or system faults) to avoid unex- mented as is in the offline approach.
pected failure and ensure power delivery. 3) The possible inclusion of elements in the transformer for
Different efforts and contributions to carry out the frequency- FRA measurement in the network could lead to vulnerabil-
response test with the transformer in service have been made ities, which may affect the power system operation.
[16]–[43]. However, as almost unanimously mentioned by dif- These three problems lead to particular difficulties in the on-
ferent authors [32], research results are still required to have a line technique feasibility, such as: How to control the wide fre-
reliable online FRA diagnosis system. quency input signal injection into the 50/60-Hz power wave.
At least two prerequisites must be fulfilled by a feasible on- What is the impact of using different input signals (shapes) in
line FRA measurement system: First, a repeatable measure, not the obtained transfer function? What is the impact of the sam-
influenced by external factors is inherent in the operation of the pling frequency on the results? How to avoid damage of mea-
transformer in the power system; second, it is necessary to pro- surement equipment and ensure personnel safety due to high
vide and implement an analysis method (based on different tech- voltages in the electrical grid? Are there any noise problems
1026 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 2, APRIL 2013

and environmental effects affecting the online test? What is the of high-voltage connection because it is a nonintrusive
most appropriate mathematical tool to analyze the online FRA galvanic-isolated device and has good performance over
measurement? How to consider the three-phase structure in the a wide bandwidth.
measurement implementation? During the last two decades, measurement of uncontrolled
Different research addresses these concerns and they are ana- transients was a procedure proposed to determine possible de-
lyzed in the following paragraphs. The first conclusion to extract formations of the windings, as presented in [16] and [17].
is that a definitive solution for the technique application is not Automatic digital recorders were used to measure the tran-
yet available. sients produced by switching and lightning events. Using the
same approach, the five-year experience results of in-service
B. Current Alternatives for the Online FRA Technique transformer diagnosis were also published [18]. Similar ex-
To carry out the online FRA test in the transformer, first, the periences were also presented in [19]–[21], where different
most viable option (IFRA or SFRA) must be determined. Most transformer switching transients were compared in terms of
investigations, so far, have opted for the first option [16]–[20], their advantages and disadvantages. The references conclude
[25], [26], [28]–[35], [37]–[39], with much fewer online SFRA that the frequency content of signals caused by atmospheric
contributions [26], [40]–[43]. events (lightning) is concentrated in the high-frequency
A summarized analysis of these contributions is presented as bandwidth, while maneuver operations (opening and closing
follows. switches) have the highest frequency content in the low- and
1) Online IFRA Method: Application of the online IFRA medium-frequency bandwidth. Finally, the main advantage
methodology follows the three stages shown in Fig. 1, where of these signals is that an external injection system is not
the transformer is connected to the grid, and the input signal is necessary.
not aperiodic voltage. Reference [29] shows the technique used to obtain an on-
Stage one refers to the excitation signal injection (input) and line FRA response by means of pulses created from an open/
to the measurement of its reflected signal (output). This work close switching operation applied to the energized transformer.
proposes the classification of input signals in terms of their con- A similar approach was proposed in [25], [29], [30], and [37].
trolling capability. Finally, in [31], U.S. Patent No. 6 549 017 was presented,
Controlled signals: are those injected on purpose into the which describes how to obtain the frequency response of the
power system line from the measurement instrumentation to transformer in service from the time-domain response obtained
generate transformer excitation. The controlled signals have from switching operations and maneuvers. However, it high-
some advantages: on one hand, the trigger timing is controlled lights the need for further research, continued in [32], [36], [38],
and, on the other hand, the shape may also be controlled so [39], and [46].
that a wide frequency content to obtain an adequate bandwidth Once injecting and measuring input and output signals, re-
for a subsequent suitable FRA curve interpretation (from Hz spectively, are addressed, stage two comes up, which consists
to MHz) can be granted. In contrast, complex instrumentation of recording, filtering, and mathematical processing of signals.
would be required. Currently, no references of such a signal The measurement record is commonly performed through an
injecting system were found in the literature reviewed. analog-to-digital converter (ADC). High resolution, high-speed
Uncontrolled signals: are those disturbances inherent to the sampling, and sufficient memory size are essential system fea-
normal operation of the electrical network, such as wave-type tures to obtain a suitably wide bandwidth [20], [25].
switching pulse (opening and closing switches) or those con- In [30], a detailed analysis of measurement influence of dif-
sidered signal type, which arise from atmospheric events (light- ferent recorder configurations was presented, showing the im-
ning), where high-frequency components predominate. portance of choosing an appropriate sampling frequency to ob-
Both types have advantages, which is that an external injec- tain proper results for an online FRA response.
tion system is not necessary, but both are under random occur- Not many contributions are available related to filtering elec-
rences (in quantitative and qualitative terms). What is more, the trical signals coming from the power system. Particularly, [31]
shape of the disturbance is not controlled; therefore, a suffi- presents a mathematical filtering technique based on the coher-
cient spectral content is not ensured, as shown specifically in ence function and [21] and [30] merely mention the need for
[16]–[20], [23]–[35], [37]–[39], and [41]. eliminating the aliasing effect.
Considering the measurement process in the uncontrolled sig- Signal processing is a mathematical procedure applied to sig-
nals approach, some alternatives have been found: nals in the time domain to determine their frequency content.
— Measuring the voltage line through the bushing tap cou- For the online IFRA method, the tool traditionally used has been
pler (BTC), which is also a natural voltage divider with the fast Fourier transform (FFT), as presented in [16]–[20], [25],
a voltage ratio, remaining constant over a wide band- [26], [28]–[35], [37]–[39], and [41]. However, its use is ques-
width—as presented in [35], [44], and [45]. In [26], an tioned in [47] because of its limitations. The FFT requires peri-
alternative is shown for when the transformer does not odic and infinite signals, while transient signals are neither pe-
have any BTC, which consists of a noninvasive capacitive riodic nor infinite.
sensor (NICS) installed on the surface of the bushings. A possible alternative could be the window (also called short
— Measuring the current through the coil is peforned by time) Fourier transform (STFT), allowing simultaneous treat-
using wide bandwidth current transformers [21], [31], ment of the low- and high-frequency components of an uncon-
such as the Rogowski coils, which limit the problem trolled signal, improving FFT processing. However, even with
GOMEZ-LUNA et al.: CURRENT STATUS AND FUTURE TRENDS IN FRA 1027

TABLE I
CLASSIFICATION OF KEY PARAMETERS RELATED TO ONLINE FRA
1028 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 2, APRIL 2013

the STFT, results have imperfections that can be explained by The first reference related to the online SFRA application
Heisenberg’s uncertainty principle [48], which indicates that it [26] corresponds to a chirp-type signal (signal generated in
is impossible to know the accurate time and frequency of one of a sweep frequency), which was injected to the high side of a
the components of the processed signal [47]. 765-kV/66-kV/550-V transformer and the output was moni-
The difficulties and the need for suitable signal processing tored at the lower side of the grid voltage.
have been treated in [25], where the low-voltage impulse (LVI) The second reference presents an online assessment of
method was analyzed to detect deformations in transformer winding deformation by using optimized multi-sine excitation
windings in service. compared to a chirp signal [42].
Stage three focuses on curve analysis and interpretation in In [40], a practical and preliminary online monitoring system
which the operator avails of having a suitable modeling tool. is presented, where a high-frequency signal is injected to a
A useful interpretation of the online IFRA results requires 650-kV transformer through the capacitive bushing tap.
overcoming two main difficulties inherent to the fact that the In [43], noninvasive capacitive sensors (NICs) are installed
transformer is a grid-connected device as follows. on the surface of the bushings to measure line voltage, and Ro-
• A random bias point exists when measuring the FRA re- gowski coils are also used to measure phase currents. Both de-
sponse because of the triggering moment related to the vices are suitable to handle a wide bandwidth response. But
high-voltage wave and to the load. As long as noninear again, problems are shown related to signal injection with the
effects are present in the transformer, this could lead to transformer in service.
different responses for the same unit in the same state. In Analysis of each reference in this study shows the progress
[12], [21], [30], and [33], some of these effects are identi- made so far and points to difficulties in the online FRA tech-
fied, such as: magnetic core hysteresis, temperature depen- nique’s implementation at its different stages.
dence, or an onload tap changer position. For the sake of simplicity,
• The online FRA response includes the transformer be- Table I shows a relation to the main references studied and
havior as well as the rest of the power system to which it their content.
is connected.
The diagnosis based on the online IFRA technique is sup- III. FUTURE TRENDS AND PROSPECTS OF THE
ported in [19] and [22], describing the method and the system ONLINE FRA TECHNIQUE
components; this work was extended in [28]. Some authors have
particularized the use of the transfer function to identify shifts A. Online IFRA
in the transformer windings in service [24], [27]. In [30], a com- No significant contributions were found about a signal in-
parative study between the offline and online measurements is jecting system. Nevertheless, an IFRA-controlled signal system
presented, and the impact of this leading to the high-frequency would significantly increase the measurement quality and,
response is also pointed out. therefore, interesting work remains to be done on this issue.
In [38] and [46], the experiences obtained from an actual on- The signal injection process to excite the transformer is
line FRA prototype, applying switching operations in a 345/ mainly proposed in terms of controlled signals, instead of using
140-kV, 448-MVA autotransformer are presented, showing the uncontrolled signals. This allows avoiding the randomness of
feasibility of the technique. In [49], a prototype measurement power system events and ensures monitoring the transformer
system is also presented, concluding the persistent need for fur- status when requested.
ther developments. Currently, suitable data-recording devices are available,
The advantages of an online IFRA diagnosis have also been which comply with the monitoring needs of high-frequency
defended by various institutions, as highlighted in [36]. signals in terms of resolution, sampling, and memory length.
2) Online SFRA Method: Obtaining online FRA responses Also, apart from the classical methods, new alternatives
through the SFRA method follows the same stages as the online exist to explore signal filtering and processing, such as the
IFRA case presented in Fig. 1, with the following peculiarities: wavelet transform (WT). Particularly, its multiresolution anal-
• The injected signals must be exclusively controlled signals, ysis (MRA) approach may be an interesting option, given that
defined in terms of a sinusoidal wave in a frequency sweep. it works with filter banks applied to discrete signals in sub-band
In this case, it is necessary to filter the signal from the coding, which constitutes a natural filtering. This method
power grid voltage (50/60 Hz) [41]. allows simultaneously filtering the grid frequency (50/60 Hz)
Furthermore, uncertainty remains regarding possible inter- and the electrical background noise [50], [51], some of the
ferences between the injected signals and the wide band- unwanted effects in the online FRA test.
width communication signals transmitted by the same ca- The WT allows filtering the unwanted effects, as well as ex-
bles, as described in [13], which mentions that injected tracting the transformer information by means of signal pro-
signal frequency can reach high kilohertz or even mega- cessing, in a more versatile manner than the Fourier Transform
hertz values that can interact with the control signals trav- approach. The WT allows a detailed study of time and frequency
elling along the power lines. domains, by using small time windows for high frequencies and
• The required mathematical procedure is much lighter as large windows for low frequencies. The WT has been tested
long as the response is directly obtained in the frequency to treat power system transients, partial discharge analysis, and
domain. study of harmonic signals, as presented in [52]–[56].
GOMEZ-LUNA et al.: CURRENT STATUS AND FUTURE TRENDS IN FRA 1029

The TF considered for the transformer diagnosis, obtained REFERENCES

after measuring and processing the signals, should contain only
the transformer information in a unique response set. It should [1] J. A. Martinez and B. A. Mork, “Transformer modeling for low- and
not depend on the measurement (offline or online, IFRA or mid-frequency transients—A review,” IEEE Trans. Power Del., vol.
SFRA) or on the operating point conditions. To achieve this, 20, no. 2, pt. 2, pp. 1625–1632, Apr. 2005.
[2] F. de León, “Discussion of “transformer modeling for low- and mid-
some interpretation tools must be considered, such as: artificial frequency transients—A review”,” IEEE Trans. Power Del., vol. 23,
intelligence, statistical methods, or Transformer models, which no. 3, pp. 1696–1697, Jul. 2008.
could be particularly useful for online measurement curves. [3] C. González and J. Pleite, “Transformer modeling approaches for Fre-
quency Response Analysis,” in Proc. XIX Int. Conf. Elect. Mach., Sep.
6–8, 2010, pp. 1–6.
[4] Standard of People’s Republic of China “Frequency Response Analysis
on Winding Deformation of Power Transformers”, Document 15182-
B. Online SFRA Method 2005, Elect. Power Ind.
[5] Power Transformers. Part 18: Measurement of Frequency Response,
The stages to follow for the online SFRA method are the same IEC 60076-18 Int. Standard, Int. Electrotech. Comm., Jul. 2012, Ed. 1.
[6] G. M. Kennedy, “Using cross-correlation coefficients to analyze
as those implemented for online IFRA (Fig. 1). Nonetheless, dif- transformer Sweep Frequency Response Analysis (SFRA) traces,”
ferences exist in how each stage is applied, as will be discussed: presented at the IEEE Power Eng. Soc. Power Africa Conf. Expo.,
Some systems with a chirp signal injection through the BTC Johannesburg, South Africa, Jul. 16–20, 2007.
[7] S. Ryder, “Methods for comparing frequency response analysis mea-
have been proposed by means of blocking filters [33], [35], [44], surements,” in Proc. IEEE Conf. Rec. Int. Symp. Elect. Insul., Boston,
and [45]. These systems care not only about the measurement MA, 2002, pp. 187–190.
itself but also about preventing interferences from/to the grid. [8] P. M. Nirgude, D. Ashokraju, A. D. Rajkumar, and B. P. Singh, “Appli-
There is a common application of these filters in power-line cation of numerical evaluation techniques for interpreting frequency re-
sponse measurements in power transformers,” IET Sci. Meas. Technol.,
communication (PLC), which is used for the grid’s remote-con- vol. 2, no. 5, pp. 275–285, 2008.
trol system. The latest technologies and trends on PLCs are de- [9] H. Firoozi and M. Shishehchian, “Frequency response analysis-condi-
scribed in [57] and [58]. Furthermore, in [59], the role of the tion assessment of power transformers using mathematical and statis-
tical criteria,” presented at the 9th Int. Conf. Properties Appl. Dielect.
PLC in the smart grid concept was presented, showing the great Mater., Harbin, China, Jul. 19–23, 2009.
advantages and future trends in communication and control sys- [10] M. Wang, A. J. Vandermaar, and K. D. Srivastava, “Review of condi-
tems for power grids because it supports a timely decisions ap- tion assessment of power transformers in service,” IEEE Elect. Insul.
proach based on monitoring the current state of the transformer Mag., vol. 18, no. 6, pp. 12–25, Nov. 2002.
[11] S. Tenbohlen, T. Stirl, and G. Bastos, “Experience based evaluation
in operation, as shown in [34]. The SFRA systems may avail of of economic benefits of online monitoring systems for power trans-
the PLC approach and adapt it to the FRA needs. formers,” in CIGRE, Paris, France, 2002.
From the diagnosis point of view, modeling may be a good [12] T. Krieg and M. Napolitano, “Techniques and experience in on-line
transformer condition monitoring and fault diagnosis in ElectraNet
approach [1], but other approaches are worth exploring like the SA,” in Proc. Int. Conf. Power Syst. Technol., 2000, pp. 1019–1024.
wavelet transform, which may also be a good tool to interpret [13] C. Gonzalez, J. Pleite, V. Valdivia, and J. Sanz, “An overview of the
the results [60], [61]. on line application of Frequency Response Analysis (FRA),” in Proc.
IEEE Int. Symp. Ind. Electron., 2007, pp. 1294–1299.
[14] C. Bengtsson, “Status and trends in transformer monitoring,” IEEE
Trans. Power Del., vol. 11, no. 3, pp. 1379–1384, Jul. 1995.
[15] T. Anderson and A. Golder, “Validation of on-line power system
IV. CONCLUSIONS dynamics measurements,” presented at the IEEE Power Eng. Soc.
Summer Meeting. Conf., Edmonton, AB, Canada, 1999.
[16] R. Malewski, J. Douville, and L. Lavallée, “Measurement of switching
The main conclusions regarding the current state of the art on transients in 735 kV substations and assessment of their severity for
the online FRA for transformer diagnosis are as follows: transformer insulation,” IEEE Trans. Power Del., vol. 3, no. 4, pp.
Firstly, different issues still need to be overcome, as pointed 1380–1390, Oct. 1988.
in this paper. [17] R. Malewski and B. Poulin, “Impulse testing of power transformers
using the transfer function method,” IEEE Trans. Power Del., vol. 3,
So far, the online IFRA method seems to offer greater no. 2, pp. 476–489, Apr. 1988.
possibilities to obtain satisfactory results related to the SFRA [18] R. Malewski, A. Claudi, E. Gockenbach, R. Maier, and K. H. Fellmann,
approach or, at least, more contributions proposing the IFRA “Five years of monitoring the impulse test of power transformers with
digital recorders and the transfer function method,” presented at the
system are available. CIGRE, Paris, France, 1992, paper no. 12-201, unpublished.
Three stages have been identified in this work to sort the [19] T. Leibfried and K. Feser, “On-line Monitoring of Transformers by
main issues when implementing the online FRA measurement. means of the transfer function method,” presented at the Conf. Rec.
The particularization of these stages to the SFRA and IFRA ap- IEEE Int. Symp. Elect. Insul., Pittsburgh, PA, Jun. 5–8, 1994.
[20] T. Leibfried and K. Feser, “Off line and on-line monitoring of power
proaches has also been presented. transformers using the transfer function method,” in Proc. IEEE Conf.
On one hand, the signal injection process was discussed. The Rec. Int. Symp. Elect. Insul., Montreal, QC, Canada, Jun. 16–19, 1996,
use of controlled signals is preferred to uncontrolled signals, pp. 34–37.
[21] T. Leibfried and K. Feser, “Monitoring of power transformers using
given their greater controllability and repeatability; however, the transfer function method,” IEEE Trans. Power Del., vol. 14, no. 4,
they are much more complex to implement. On the other hand, pp. 1333–1341, Oct. 1999.
signal filtering and processing was also considered and the [22] T. Leibfried, “On-line monitors keep transformers in service,” IEEE
Wavelet Transform was identified as a useful tool with even Comput. Appl. Power, vol. 11, no. 3, pp. 36–42, Jul. 1998.
[23] T. Leibfried, Siemens, “Online monitoring of power trans-
more potential than the standardized Fourier Transform and its formers—System technology and data evaluation,” in Proc. High
derivatives. Voltage Eng. Symp., Conf. Publ., Germany, Aug. 1999, pp. 22–27.
1030 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 28, NO. 2, APRIL 2013

[24] J. Christian, K. Feser, U. Sundermann, and T. Leibfried, “Diagnostics [45] A. Setayeshmehr, A. Akbari, H. Borsi, and E. Gockenbach, “On-line
of power transformers by using the transfer function method,” in Proc. monitoring and diagnoses of power transformer bushings,” IEEE
IEEE High Voltage Symp., Aug. 1999, pp. 37–40. Trans. Dielect. Elect. Insul., vol. 13, no. 3, pp. 608–615, Jun. 2006.
[25] M. Wang, A. J. Vandermaar, and K. D. Srivastava, “Condition moni- [46] L. Coffeen, J. Mcbride, N. Woldemariam, J. Benach, and L. V. Zel, “A
toring of transformers in service by the low voltage impulse method,” summary of NEETRAC on-line Frequency Response Analysis (FRA)
presented at the Inst. Elect. Eng. High Voltage Eng. Symp., Vancouver, and a new EPRI commercial prototype FRA installation at first energy,”
BC, Canada, Aug. 1999. presented at the NEETRAC and EPRI Substation Equipment Maint.
[26] S. Birlasekaran and F. Fetherston, “Off/on-line FRA condition moni- Optim. Diagnos. Conf., San Antonio, TX, Mar. 2009.
toring technique for power transformer,” IEEE Power Eng. Rev., vol. [47] D. C. Robertson, O. I. Camps, J. S. Mayer, and W. B. Gish, “Wavelets
19, no. 8, pp. 54–56, Aug. 1999. and electromagnetic power system transients,” IEEE Trans. Power
[27] N. Pinhas, S. Islam, and J. Hullett, “On the development of transfer Del., vol. 11, no. 2, pp. 1050–1058, Apr. 1996.
function method for fault identification in large power transformers [48] M. Doroslovacki, “Discrete-time signals and uncertainty relations in-
on load,” in Proc. Annu. Rep. Conf. Elect. Insul. Dielect. Phenomena, volving ordinary second moments in time and frequency,” in Proc. Int.
2000, vol. 2, pp. 747–751. Symp. Time-Frequency Time-Scale Analysis, Oct. 1994, pp. 186–189.
[28] K. Feser, J. Christian, C. Neumann, U. Sundermann, T. Leibfried, A. [49] Elect. Power Res. Inst. (EPRI), On-line frequency response anal-
Kachler, and M. Loppacher, “The transfer function method for detec- ysis–report on commercialization of system, Tech Rep. 1012345,
tion of winding displacements on power transformers after transport, Dec. 12, 2006. [Online]. Available: http://www.epri.com/ab-
short circuit or 30 years of service,” in CIGRE, Paris, France, 2000. stracts/Pages/ProductAbstract.aspx?ProductId=000000000001012345
[29] A. Vandermaar, “On-line frequency response analysis system,” Devel- [50] M. B. Figueiredo, A. D. Almeida, and B. Ribeiro, “Wavelet decompo-
opment of Specifications. EPRI Project Manager. Technical Progress, sition and singular spectrum analysis for electrical signal denoising,”
Develop. Spec. No. 1001942, Elect. Power Res. Inst. Palo Alto, CA, in Proc. IEEE Int. Conf. Syst., Man, Cybern., 2011, pp. 3329–3334.
Dec. 2001. [51] V. L. Pham and K. P. Wong, “Wavelet-transform-based algorithm for
[30] M. Wang, “Winding movement and condition monitoring of power harmonic analysis of power system waveforms,” Proc. Inst. Elect.
transformers in service,” Ph.D. dissertation, Dept. Elect. Comput. Eng., Eng., Gen., Transm. Distrib., vol. 146, no. 3, pp. 249–254, May 1999.
Univ. British Columbia, Vancouver, BC, Canada, 2003. [52] M. Karimi, H. Mokhtari, and M. R. Iravani, “Wavelet based on-line dis-
[31] L. Coffeen, “System and method for on-line impulse frequency re- turbance detection for power quality applications,” IEEE Trans. Power
sponse analysis,” U.S. Patent 654 901 7B2, Apr. 15, 2003. Del., vol. 15, no. 4, pp. 1212–1220, Oct. 2000.
[32] L. Coffen, “On-line frequency response analysis system,” Develop- [53] P. Purkait and S. Chakravorti, “Pattern classification of impulse faults
ment of specifications, EPRI Project Manager, Technical Progress. in transformers by wavelet analysis,” IEEE Trans. Dielect. Elect. Insul.,
Palo Alto, CA, Research Program No. 1001770, Mar. 2003. vol. 9, no. 4, pp. 555–561, Aug. 2002.
[33] R. Wimmer and K. Feser, “Calculation of the transfer function of a [54] K. L. Butler and M. Bagriyanik, “Characterization of transients in
power transformer with online measuring data,” in Proc. II Int. Conf. transformers using discrete wavelet transformer,” IEEE Trans. Power
Advances Process., Testing Appl. Dielect. Mater., Sep. 2004, pp. 1–5. Syst., vol. 18, no. 2, pp. 648–656, May 2003.
[34] M. Bagheri, M. S. Naderi, T. Blackburn, and T. Phung, “Practical [55] E. Zheng, Z. Liu, and L. Ma, “Study on harmonic detection method
challenges in online transformer winding deformation diagnostics,” based on FFT and wavelet transform,” in Proc. IEEE 2nd Int.Conf.
in Proc. 2nd Int. Conf. Elect. Power Energy Convers. Syst. (EPECS), Signal Process. Syst., 2010, pp. 1–4.
2011, pp. 1–6. [56] S. R. Paraskar, M. A. Beg, and G. M. Dhole, “Continuous wavelet
[35] C. González, J. Pleite, J. Sanz, and V. Valdivia, “On-line measurements transform for discrimination between inrush and fault current transients
of the frequency response of the transformer by means of the bushing in transformer,” Int. J. Sci., Spirituality, Bus. Technol., vol. 1, no. 1,
tap coupler,” in Proc. Advanced Res. Workshop Transformers, Baiona, Mar. 2012.
Spain, Oct. 2007, pp. 1–6. [57] S. Ramseier and H. Spiess, Feb. 2006, “El canto de las líneas
[36] L. Coffeen, D. Cantrelle, and J. Mango, Transm. Distrib. World, eléctricas,” Revista ABB No. 2, Aug. 2006. Baden, Switzerland.
vol. 59, no. 10, p. 62, Oct. 2007. [Online]. Available: http://connec- [Online]. Available: http://www05.abb.com/global/scot/scot221.
tion.ebscohost.com/c/articles/27325791/on-line-monitor-tracks-trans- nsf/veritydisplay/cd6491bf4accb5cfc1257264004903aa/$file/re-
former-health view_s.indd_1.pdf
[37] A. Setayeshmehr, H. Borsi, E. Gockenbach, and I. Fofana, “On-line [58] D. D. Giustina, L. Andersson, C. Casirati, S. Zanini, and L. Cremas-
monitoring of transformer via transfer function,” in Proc. IEEE Elect. chini, “Testing the broadband power line communication for the dis-
Insul. Conf., Canada, Jun. 2009, pp. 278–282. tribution grid management in a real operation environment,” in Proc.
[38] L. Coffeen, J. Mcbride, N. Woldemariam, and J. Benach, “A summary Int. Symp. Power Electron., Elect. Drives, Autom. Motion, 2012, pp.
of NEETRAC on-line frequency response analysis (FRA) a new EPRI 785–789.
commercial prototype FRA installation at first energy,” presented at [59] S. Galli, A. Scaglione, and Z. Wang, “For the grid and through the grid:
the EPRI Substation Equipment Maintenance Optimizaton Diagnostics The role of power line communications in the smart grid,” Proc. IEEE,
Conf., San Antonio, TX, Mar. 2009. vol. 99, no. 6, pp. 998–1027, Jun. 2011.
[39] L. Coffeen, J. McBride, N. Woldemariam, J. Benach, and L. V. D. Zel, [60] J. C. Gonzales, E. E Mombello, and G. D. Guidi, “Diagnosis of power
“An on-line Frequency Response Analysis (FRA) update,” TechCon transformers through Frequency Response Analysis applying multires-
pp. 129–151, 2009. olution analysis and pattern classification,” in Proc. IEEE/Power Eng.
[40] T. De Rybel, A. Singh, J. A. Vandermaar, M. Wang, J. R. Marti, and Soc. Transm. Distrib. Conf. Expo.: Latin America, Nov. 8–10, 2010,
K. D. Srivastava, “Apparatus for online power transformer winding pp. 270–277.
monitoring using bushing tap injection,” IEEE Trans. Power Del., vol. [61] E. Gómez-Luna, D. Silva, G. Aponte, J. Pleite, and D. Hinestroza, “Ob-
24, no. 3, pp. 996–1003, Jul. 2009. taining the electrical impedance using wavelet transform from the time
[41] T. De Rybel, A. Singh, P. Pak, and J. R. Marti, “online signal injec- response,” IEEE Power Eng. Lett., Dec. 04, 2012, accepted for publi-
tion through a bus-referenced current transformer,” IEEE Trans. Power cation.
Del., vol. 25, no. 1, pp. 27–34, Jan. 2010.
[42] S. Gopalakrishna, V. Jayashankar, K. V. Jagadeesh, and N. M. Mohan,
“Online assessment of winding deformation based on optimised exci- Eduardo Gomez-Luna (S’11) was born in Pradera
tation,” in Proc. IEEE Int. Workshop Appl. Meas. Power Syst., 2010, Valle, Colombia, in 1982. He is currently pursuing
pp. 84–89. the Ph.D. degree in electrical engineering from the
[43] V. Behjat, A. Vahedi, A. Setayeshmehr, H. Borsi, and E. Gockenbach, Universidad del Valle, Cali, Colombia.
“Diagnosing shorted turns on the windings of power transformers He is part of the Research Group on High Voltage
based upon online FRA using capacitive and inductive couplings,” (GRALTA), Universidad del Valle. His research
IEEE Trans. Power Del., vol. 26, no. 4, pp. 2123–2133, Oct. 2011. interests focus on power transformer diagnosis,
[44] M. Melo and M. Alves, “Experience with on-line monitoring of ca- power electronics, applied mathematics, and signal
pacitance and tangent delta of condensive bushings,” in Proc. IEEE processing.
Transm. Distrib. Conf. Expo., 2008, pp. 1–6, Latin America.
GOMEZ-LUNA et al.: CURRENT STATUS AND FUTURE TRENDS IN FRA 1031

Guillermo Aponte Mayor was born in Palmira, Jorge Pleite Guerra (SM’12) is an Associate Pro-
Colombia, in 1955. He received the B.Sc. and Ph.D. fessor at Carlos III University of Madrid, Madrid,
degrees in electrical engineering from Universidad Spain, where he is also the Head of the Master’s
del Valle, Cali, Colombia, in 1978 and 2011, re- Program in Advanced Electronics Systems. His main
spectively, and the M.Sc. degree in high-voltage research line is the modeling of electric and analog
engineering from the University of Manchester electronic devices and, particularly, their application
Institute of Science and Technology (UMIST), in diagnosing power transformers.
Manchester, U.K., in 1985.
Currenlty, he is a Professor at Universidad del
Valle and the Director of the Research Group on
High Voltage GRALTA. His research areas are focus
on power transformer diagnosis and substations.

Carlos Gonzalez-Garcia (S’06–M’12) was born

in Madrid, Spain, in 1980. He received the B.Sc.,
M.Sc., and Ph.D. degrees in electrical engineering
from Carlos III University of Madrid, Madrid, Spain,
in 2004, 2009, and 2012, respectively.
His research interests focus on power transformer
diagnosis, power electronics, and modeling of mag-
netic devices. Currently, he is an Assistant Professor
at Carlos III University of Madrid in the Electronics
Technology Department, Power Electronics Group.