You are on page 1of 8

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO.

1, JANUARY 2000

247

Characterization of Distribution Power Quality


Events with Fourier and Wavelet Transforms
Surya Santoso, Member, IEEE, W. Mack Grady, Senior Member, IEEE, Edward J. Powers, Fellow, IEEE,
Jeff Lamoree, Member, IEEE, and Siddharth C. Bhatt, Member, IEEE

AbstractIt is the objective of this paper to present unique features that characterize power quality events and methodologies
to extract them from recorded voltage and/or current waveforms
using Fourier and wavelet transforms. Examples of unique features
include peak amplitudes, RMS, frequency, and statistics of wavelet
transform coefficients. These features are derived from well documented theories, power engineers heuristics gained through long
years of experience, and power quality data collected in recent
years. Converter operation, transformer energization, and capacitor energization (which includes normal, back-to-back, and restrike on opening energization), representing three common power
quality events at the distribution level, are presented. These examples provide the basis for further characterization of other power
quality events.
Index TermsFourier and wavelet transforms, identifying features, power quality.

I. INTRODUCTION

TUDIES of power quality phenomena have emerged as an


important subject in recent years due to renewed interest in
improving the quality of the electricity supply. As sensitive electronic equipment continues to proliferate, the studies of power
quality will be further emphasized. According to the Electric
Power Research Institute, in 1985 20% of the total electric load
in the US was electronic, however this number will jump to approximately 5060% by the year 2000 [1].
In recent years, power quality phenomena have been investigated directly from actual recorded disturbance waveforms
thanks to widely available power monitor equipment. These
disturbance recordings are stored as three-phase voltage and
current time-series which bring a wealth of information about
the characteristics of the associated power quality events.
Since these characteristics are unique to each power quality
event, they are useful to portray the process of power quality
disturbances and to pinpoint their causes.
In this paper, we present unique features that characterize
power quality events and methodologies in order to extract
them from recorded waveforms. We unify the use of one of the
most important mathematical transforms in power engineering,

Manuscript received July 31, 1997; revised September 18, 1998.


S. Santoso was with the Department of Electrical and Computer Engineering,
The University of Texas at Austin, Austin, TX 78712-1084 and is now with
Electrotek Concepts, Inc., Knoxville, TN 37923.
W. M. Grady and E. J. Powers are with the Department of Electrical and
Computer Engineering, The University of Texas at Austin, Austin, TX 787121084.
J. Lamoree is with Electrotek Concepts, Inc., Knoxville, TN 37923.
S. C. Bhatt is with EPRI Customer Systems Group, Palo Alto, CA 94523.
Publisher Item Identifier S 0885-8977(00)00564-1.

the Fourier transform, with a newer transform, the wavelet


transform, to extract characterizing power quality features.
The Fourier transform is used to characterize steady state phenomena, whereas the wavelet transform is applied to transient
phenomena. We also make use of other mathematical analysis
techniques such as curve-fitting techniques to estimate sudden
voltage step changes. While theories of Fourier and wavelet
transforms are not presented here due to space limitation, [2],
[3], [7] provide excellent overviews in these subjects.
Power quality events are characterized by their maximum amplitudes, crest voltages, RMS, frequency, statistics of wavelet
transform coefficients, instantaneous voltage drops, number of
notches, duration of transients, etc. These characteristics are different for each power quality event, thus they are unique identifying features.
In Section II, characteristics of three common power quality
events at the distribution level are presented. They are converter
operation, transformer energization, and capacitor energization
(which is further broken down into normal, back-to-back, magnification, and restrike on opening energization). These examples provide the basis for further characterization of other power
quality events. Utilizing characteristics described in Section II,
Section III concludes this paper with an application to build an
event identification module.
II. PQ EVENTS AND THEIR FEATURES
Unique features characterizing the three events mentioned
above are presented in this section. These identifying features
are derived from well documented theories [3][6], power engineers heuristics gained through long years of experience, and
power quality data collected in recent years.
In performing wavelet transformations, a given signal is decomposed into four decomposition scales with a Daub4 wavelet
as a mother wavelet. A Daub4 wavelet is a Daubechies wavelet
with 2 vanishing moments [2], [8], [9].
We shall now begin our discussion starting with transformer
energization, followed by converter operation and capacitor energizing events.
A. Transformer Energizing
Transformer energizing events are caused by magnetizing inrush current when a transformer is energized. The inrush current can be significantly higher than the rated load current. Due
to saturation effects in the transforme iron core, inrush currents
are rich in both odd and even harmonics, particularly the 2nd
through 5th harmonics. These harmonics are also reflected in
the voltage waveforms.

08858977/00$10.00 2000 IEEE

248

Fig. 1. Typical waveforms of a transformer energizing event in phases (a) A,


(b) B, and (c) C.

Fig. 1 shows the three-phase voltage for a transformer energizing event. The energization event occurs on all phases simultaneously. In the following, characteristics of transformer energizing are presented.
1) Waveshape Before Transformer Energization: The waveshape before the energization is very smooth and sinusoidal. The
wavelet transform variables in terms of squared wavelet transform coefficients (WTC) are utilized to determine the smoothness of the waveshape before transformer energization. As a
case study, we analyze the phase A voltage waveform shown in
Fig. 1(a) which is also shown in Fig. 2(a). The squared wavelet
transform coefficients at scales 1, 2, 3, and 4 of the transform
are shown in Fig. 2(b)(e), respectively.
can be located by
The starting time of energizing event
finding the maximum value of the squared WTC1 or squared
WTC2, or both depending on the tiny transient that occurs at the
energizing instant. Fig. 2(b) and (c) indicates that the starting
time is found by locating the maximum value of the squared
WTC1 because it occurs earlier than that of squared WTC2. The
ms.
energizing instant is at
The smoothness of the waveform prior to energizing (014
ms) can be investigated through the statistics of the squared
WTC3 and squared WTC4. For phase A voltage, the maximum value of squared WTC3 before the energizing time
is
, and the average value
of the largest 30 coefficients of the squared WTC3 before
is
.
the energizing time
for
The wavelet statistics for squared WTC4 are
and
,
respectively.
2) Waveshape During Transformer Energization: Unlike
the waveshape before energization where there was no significant activity detected by the wavelet transform, the waveshape
during the energization is loaded with disturbance activity,
which is picked up by the wavelet transform. There are
additional high values of squared WTCs scattered over the

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

Fig. 2. The wavelet transform of a transformer energizing waveform. (a) The


time domain signal. (b)(e), squared WTCs at scales 1, 2, 3, and 4, respectively.

energization period which are associated with the harmonic


distortion.
The statistics of squared WTC3 and squared WTC4 during
the energization are quite different compared to those before
energization. The maximum squared WTTC3 during the
is 0.0108, which is
energization
. The average of the largest
54 times
30 coefficients of squared WTC3 during the energization
is 0.004, which is 40 times
. For squared WTC4, the values
and
for
are 0.0609 and 0.0202, which are 122 and 200 times the before
energization values, respectively.
3) Harmonic Voltage: During the transformer energization,
the voltage waveforms are contaminated with harmonics, in particular the 2nd through 5th harmonics. A Fourier transform is
performed to find the harmonic voltage for the waveform during
the energization. The second harmonic voltage is the predorninant one, as shown in Fig. 3.
B. Converter Operation
Notching in a voltage waveform is characteristic of converter
operation. These notches are the result of commutation during
a multiphase converter operation. During this period, a momentary phase-to-phase short circuit occurs and pulls the voltage
phases as close as possible to zero. The phase-to-phase short circuit causes the notches in the voltage waveforms. Fig. 4(a)(c)
show three phase voltage notching waveforms for phases A, B,
and C, respectively, in which there are 6 notches in every cycle.
The following are the unique identifying features for voltage
notching waveforms.
1) Number of Notches per Cycle: Notches appear periodically throughout the waveform as a result of commutation
during multiphase converter operation. In a conventional
six-pulse rectifier, the current is commutated six times from
one phase to another in one cycle to produce a dc voltage on its

SANTOSO et al.: CHARACTERIZATION OF DISTRIBUTION POWER QUALITY EVENTS WITH FOURIER AND WAVELET TRANSFORMS

Fig. 3.

249

Harmonic voltages of a transformer energizing waveform.

Fig. 5. Wavelet transform for phase A voltage. (a) The voltage waveform,
(b)(e) are the squared WTCs at scales 14, respectively.

where
and
, are the number of notches and the
, respectively.
number of cycles in the voltage waveform,
The number of notches in the waveform can be found by simply
counting the number of impulses in the squared WTC1. The
number of cycles in the waveform can be estimated simply using
the following relationship,
(2)

Fig. 4. Typical waveform of a voltage notching event in phases (a) A, (b) B,


(c) C.

output. This operation gives rise to six notches per cycle which
appear periodically throughout the waveform. Likewise, in a
twelve-pulse rectifier, there will be twelve notches per cycle.
We shall now discuss the procedure to count the number of
notches per cycle. For this purpose, the phase A voltage waveform shown in Fig. 4(a) [also shown in Fig. 5(a)] is chosen
as a study case. The wavelet transform is utilized to count the
number of notches per cycle. The wavelet transform of the phase
is shown in Fig. 5(b)(e) for scales 14, respecA voltage
tively.
High magnitudes of squared WTC1 in Fig. 5(b) clearly indicate the occurrence of the notches. Squared WTC1 associated
with notches at the peaks of the waveform are much higher than
those of associated with notches between peaks because of latter
are less severe. They all appear as perodic impulses. However,
squared WTC2, squared WTC3, and squared WTC4 are not so
useful to estimate the number of notches per cycle because the
impulse pattern does not clearly stand out. From our experience,
it is sufficient to estimate the number of notches from squared
WTC1 only.
is found using
The number of notches per cycle
the following relation,
(1)

is the time duration of the waveform in ms.


where
ms
For the voltage waveform shown in Fig. 5(a),
. Therefore using (1),
and the number of notches is
is
cycles. Finally, using
the number of cycles in
(2), the number of notches per cycles can be computed which
. This number indicates that the voltage
is
notching waveform is due to a six-pulse rectifier operation.
2) Harmonic Voltages: Voltage notching waveforms are
rich in harmonics. In a six-pulse rectifier, the voltage waveforms
harmonics, where
. In a twelve-pulse
are rich in
harmonics,
rectifier, the voltage waveforms are rich in
. Voltage waveforms in a twelve-pulse rectifier
where
contain 12 notches in every cycle.
Fig. 6 shows the harmonic voltages up to the 25th harmonic.
The predominant harmonic voltages are the 5th, 7th, 11th,
13th, 17th, 19th, 23rd, and 25th. The 13th harmonic, however,
is rather small. This harmonic voltage spectrum provides
further evidence about the nature of the notching waveform.
)th
If the waveform is due to a six-pulse rectifier, the (
harmonics should be present. However, due to imperfections
in the converter systems, some of these harmonics may not
)th harmonics
really stand out, but the majority of the (
do appear.
C. Capacitor Energizing
Capacitor energizing transient events are one of the most
common transient events present in power systems. The
transient events occur when a capacitor is switched on.
At the switching instant, a fast change in the bus voltage

250

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

Fig. 6. Harmonic voltages for phase A voltage waveform.

occurs because the voltage in the capacitor cannot change


instantaneously. The transient frequency is determined by the
combination of the capacitance of the capacitor bank and the
system inductance. The oscillation frequency in the voltage
waveform is typically between 3001000 Hz and lasts for less
than half a cycle of the power frequency.
There are several kinds of capacitor energizing events. We
shall present normal energizing, back-to-back energizing, and
restrike on capacitor opening events. We shall discuss each of
these categories and formulate a procedure to extract these identifying features.
1) Normal Energizing of Utility Capacitors: Energization
of utility capacitors is a daily operation in the utility system.
They are switched into the system in anticipation of load increase at a customer site, to correct power factor, to support
voltage on the system, and so on. These energization of utility
capacitors are considered normal energizing. The following are
the identifying features for normal capacitor switching.
a) Overvoltage: At the switching instant, the voltage in
the capacitor cannot change instantaneously. The bus voltage is
pulled down, and then rises as the capacitor begins to charge.
During the process, the capacitor voltage may overshoot and
ring at the natural frequency. The overvoltage in normal energizing is usually between 1.11.4 p.u. Figs. 7 and 8 show
two typical normal energization of utility capacitors. The overvoltage is not larger than 1.1 p.u in either case.
b) Polarity and Magnitude of the Step Voltage: One of
the most common identifying features for normal energizing of
utility capacitors is the polarity of the step voltage. The behavior of this feature and how it relates to normal energizing of
capacitors axe illustrated using the above two normal switching
cases.
If the power quality monitor is located at or near capacitors
that have no series reactor, a fast initial voltage step will be observed. The voltage step at the instant of closing cannot go beyond zero if the capacitor has no initial net charge at the closing
instant or if the capacitor is grounded. If the power quality monitor is located farther away from the capacitor, the voltage step
change may not be observed, or at least it is not as prominent.
These situations are depicted in Figs. 7 and 8, respectively. In
any event, sudden changes of voltages never go across the zero
line, i.e., they do not change polarity. This behavior is exhibited
in nearly all normal energizing of utility capacitors.
Now let us consider phase C voltage shown in Fig. 7(c) as
a study case. The first 22 ms of this voltage are also shown in
Fig. 9 for the sake of clarity. When the switch is closed on this

Fig. 7. Typical waveforms of normal energizing of utility capacitors in phases


(a) A, (b) B, and (c) C. PQ monitor is located near or at the capacitors.

Fig. 8. Typical waveforms of normal energizing of utility capacitors in phases


(a) A, (b) B, and (c) C. PQ monitor is located faraway from the capacitors.

phase, the sudden step voltage change occurs. The voltage in


phase C at the closing instant is 1.5 at point . The utility capacitor pulls the system voltage down to point , but the voltage
does not pass zero. The polarity of the voltage at point is the
and are
same as that of point . The voltages at points
p.u and
p.u, respectively. They are
, which is 0.325 ms.
separated by
can be computed by taking the absoThe voltage change
to . The slope of the step voltage
lute difference of
is then computed by taking the ratio of
to
. For the above
and its slope
are 0.929
example, the voltage change
p.u and 2.85 103 p.u/s, respectively.
c) Oscillation Frequency of the Energizing Event in the
Phase Voltage: The oscillation frequency of any kind of capacitor energizing in the phase voltage is generally between
3001000 Hz. Thus, the frequency of oscillation is helpful in
identifying capacitor energizing in general, but it cannot be used

SANTOSO et al.: CHARACTERIZATION OF DISTRIBUTION POWER QUALITY EVENTS WITH FOURIER AND WAVELET TRANSFORMS

251

Fig. 9. The first 22 ms of phase C voltage shown earlier in Fig. 7(c).

to discriminate normal energizing versus other types of capacitor energizing.


Extracting the oscillation frequency from the capacitor energizing transients is difficult. First, the transient event from a
given voltage waveform has to be localized using the squared
wavelet transform coefficients. Once it is localized, the sinusoidal part has to be removed. The solid line waveform shown
in Fig. 10(a) is the first 22 ms of phase C voltage shown earlier in
Fig. 7(c). The dotted line waveform in Fig. 10(a) is a curve-fitted
which is found using a polynomial curve-fitting
waveform
technique. The transient event can now be removed from
by subtracting
from it. The removed waveform
is now
shown in Fig. 10(b).
The frequency of the transient event in the voltage phase is
now estimated by counting the number of crest voltages that
are larger than 0.1 p.u and the duration of the corresponding
crest voltages. This method does not always work if the removed
is very small in magnitude. In such a case, the
waveform
from the phase voltage
oscillation frequency feature
cannot be extracted. In this example, the estimated oscillation
Hz. Whereas the estimated oscilfrequency is
lation frequency for the waveform shown in Fig. 8(c) using the
Hz.
above technique is
d) Wavelet Transform Coefficients: The wavelet transform is utilized to provide additional features in identifying
normal capacitor switching. Consider the phase C voltage
shown in Fig. 7(c). We shall perform the wavelet transforms
at scales 14. The WTCs for these scales are now shown
in Fig. 11(b)(e), respectively. Fig. 11(a) shows the phase C
voltage waveform which was shown earlier in Fig. 7(c).
In identifying normal energization of utility capacitors, we
utilize only results from the wavelet transform at scales 1
and 2. The maximum squared magnitude at scales 1 and 2
and
, respectively, indicate
the onset of the disturbance and its disturbance activity. The
duration of the transient is estimated using WTC1 or WTC2.
The small oscillating wave following the spike-like event in
WTCI and/or WTC2 provides an estimate of the energizing
transient duration.
In this example, the maximum values of the squared WTC1
and
and squared WTC2 are
, respectively. The duration of the
transient event estimated using WTC1 is
cycle.

Fig. 10. Oscillation frequency of capacitor energizing is estimated with the


help of curve-fitting technique.

Fig. 11. The wavelet transform coefficients for phase C voltage waveform
shown in Fig. 8(c).

The maximum values of the squared WTC1 and squared


WTC2 can always be found; however, the duration of the
may not always be found. This
transient event
is especially true when the transient frequency is very low.
Under this circumstance, both the oscillation frequency in the
and the duration of the transient event
phase voltage
cannot be extracted. The duration of the capacitor
energizing transient for phase C voltage shown in Fig. 8 is also
0.49 cycle, whereas the maximum values of the squared WTC1
and squared WTC2
are
and
, respectively. These three features
from the wavelet transform will be utilized as identifying
features when available.
e) Oscillation Frequency of the Energizing Event in the
Phase Current: As described earlier, the oscillation frequency
is
of the capacitor transient event in the phase voltage

252

very helpful in discriminating capacitor energizing events from


other events. However, the frequency feature from the phase
cannot be used to distinguish normal capacvoltage
itor energizing from other types of capacitor energizing disturbances. Fortunately, the oscillation frequency of the transient
for the normal energizing
event at the phase current
is very distinct from other types of capacitor energizing disturbances. The frequency of the phase current during the energization is typically between 3001000 Hz.
Consider the phase C voltage shown in Fig. 8(c) and its phase
current shown in Fig. 12(c). The oscillation frequency of the
is computed using the Fourier transform
phase current
once the transient event is localized by the wavelet transform.
is 395.6 Hz.
For this example, the frequency
2) Back-to-Back Capacitor Energizing: Back-to-back
energizing of utility capacitors occurs when the capacitor
being switched into the utility system is located near an
already energized capacitor. The voltage waveforms of the
back-to-back energizing events look very much the same as
those in normal energizing. Almost all features of the normal
energizing apply to back-to-back energizing. Therefore, discriminating back-to-back energizing events against normal
ones is a very delicate task. Fig. 13 shows a three-phase voltage
of back-to-back energizing of utility capacitors. The associated
phase currents are shown in Fig. 14.
In the following we shall present two key features that may
help distinguish back-to-back from normal energizing; the step
voltage change and the oscillatory frequency of the phase current.
f) Step Voltage: Back-to-back energizing transients
involve two capacitors in close vicinity. One of them is fully
energized and the other is switching on. The resulting step
voltage is not as large as if the fully energized bank did not
exist. Furthermore, the voltage at point , defined earlier (see
Fig. 9) does not drop close to zero as in the previous case. In
other words, during the energization process, the bus voltage
cannot drop to zero (although the capacitor has no net charge)
because a nearby fully energized capacitor supports the bus
voltage. This feature may not always be able to be used to
differentiate back-to-back from normal energizing because the
step voltage in back-to-back energizing behaves exactly like
that in normal energizing when the power quality monitor is
located faraway from the capacitor.
Based on our observations, the voltage change in
back-to-back capacitor energizing is typically no larger
than 0.5 p.u. None of the step voltages in phases A, B, and C in
Fig. 13 are larger than 0.5 p.u.
g) Oscillation Frequency of the Energizing Event at the
Phase Current: The frequency of the phase current,
provides distinct evidence for the back-to-back energizing
event. The typical frequency for back-to-back in the phase
is larger than 1000 Hz, whereas that for the
current
normal energizing is less than 1000 Hz. Fig. 14 shows the
energizing transient in the phase current. The frequency in these
phase currents is obviously higher than those in normal energizing shown in Fig. 12. The frequency of the transient event
in the phase current is therefore very useful to distinguishing
back-to-back from normal energizing events.

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

Fig. 12. Current waveforms at phases A, B, and C of the voltage waveforms


shown in Fig. 8.

The oscillation frequency of the energizing event in the phase


current is computed using the technique described in the earfor phases A, B, and C curlier section. The frequency
rents shown in Fig. 14(a)(c), are 1.09, 2.28, and 2.40 KHz, respectively. These frequencies are certainly higher than those in
normal energizing events.
3) Capacitor Switch Restrike on Opening: Switching off a
capacitor (i.e., capacitor de-energizing) does not usually result
in any transient oscillation. However, the bus voltage may drop
slightly due to loss of voltage support from the capacitor. During
the deenergizing process, the contactor of a capacitor opens and
discontinues the flow of current. If the contactor does not open
successfully, an arc will be established between the contacts
and the capacitor will be re-energized. The voltage at the instant of restrike will oscillate at the natural frequency characterized by the capacitance of the capacitor and the system inductance. Fig. 15 shows a three-phase voltage of restrike transients
of utility capacitors. The restrikes occur on phases A and B only.
Phase C voltage shows no transient because the utility capacitor
on phase C successfully de-energized.
Features of capacitor restriking on opening are similar to
other capacitor energizing transients. However, discriminating
capacitor switch restrike on opening events from the previous
three capacitor switching events are quite straightforward.
The following are identifying features for capacitor switching
restrike transients.
h) Polarity of the Voltage Step: Consider phase A voltage
shown in Fig. 15(a). The magnified version of the transient is
shown in Fig. 16. The voltage change from point and point
is quite dramatic. The voltage at points and are
p.u and
p.u, respectively. Unlike step voltages from
the previous two energizing transients, the step voltages of the
restrike transients do change polarity. If the voltage at point is
positive then the voltage at point is negative, and vice versa.
The change of polarity is attributed to trapped charges in the
capacitor. This is the unique feature associated with capacitor
restrike transients.

SANTOSO et al.: CHARACTERIZATION OF DISTRIBUTION POWER QUALITY EVENTS WITH FOURIER AND WAVELET TRANSFORMS

Fig. 13. Typical waveforms of back-to-back energizing of utility capacitors in


phases (a) A, (b) B, and (c) C.

Fig. 15.

253

Capacitor switch restrike on opening at phases (a) A, (b) B, and (c) C.

Fig. 16. Voltage waveform of phase A in detail.

Fig. 14. Current waveforms at phases A, B, and C of the voltage waveforms


shown in Fig. 13 (only the first ee ms is shown).

i) Squared WTCs: Since the voltage step is very drais also very
matic, the slope of the voltage step change
large. This condition translates to a very high magnitude of
squared WTCs.
Fig. 17 shows squared WTCs at scales 14 for the phase A
voltage. The high magnitudes of the squared WTCs indicate
location of step voltage. The maximum value of squared
is 0.63 whereas that for scale 2
WTC1
is 1.34. Features from these two scales are
sufficient to identify the event.
III. APPPLICATION TO EVENT IDENTIFICATION MODULE
We have presented identifying features of three types of
power quality events at distribution level. These identifying
features are very useful to understand the physics underlying
the disturbance processes.

Fig. 17. The squared WTC of voltage waveform (a) at scales (b) 1, (c) 2, (d)
3, and (e) 4.

We utilized these identifying features to build an event identification module. The identification module extracts relevant
identifying features using the above methodologies. In order
to identify the PQ events, we made use of a rule-based expert
system as a reasoning paradigm. The identifying features presented in the previous section were incorporated into expert
system knowledge bases. The module was designed to mimic
power quality engineers thought process and conversely, the

254

IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

module permits power engineers to follow its reasoning process


and how it reaches its decision. These features are important to
assist power engineers in pinpointing the cause of a disturbance
event. The event identification module is described in [8].
ACKNOWLEDGMENT
This work represents a concerted effort between The University of Texas and Electrotek Concepts under the financial
support of EPRI to develop an event identification module. We
would like to thank Electroteks power engineers, especially, M.
F. McGranaghan, R. F. Scott, and Dr. H. V. Nguyen in providing
technical assistance. Valuable discussions with A. C. Parsons of
The University of Texas at Austin are appreciated.
REFERENCES
[1] R. J. Gilleskie, Utility expertise in providing power quality and other
engineering services, in Proceedings of the National Science Foundation Conference on Unbundled Power Quality Services in the Power Industry, Nov. 1996, pp. 2127.
[2] I. Daubechies, Ten Lectures on Wavelets. Philadelphia: SIAM, 1992.
[3] J. Arrilaga, Power System Harmonics. Chichester (West Sussex)/New
York: Wiley, 1985.
[4] G. T. Heydt, Electric Power Quality, 2nd ed. West Lafayette, IN: Stars
in a Circle Publications, 1991.
[5] R. C. Dugan, M. F. McGranaghan, and H. W. Beaty, Electrical Power
Systems Quality. New York: McGraw-Hill, 1996.
[6] A. Greenwood, Electrical Transients in Power Systems, 2nd ed. New
York: John Wiley & Sons, Inc., 1991.
[7] G. P. Tolstov, Fourier Series. New York: Dover Publications, Inc.,
1976.
[8] S. Santoso, E. J. Powers, and W. M. Grady, Wavelet-based power
quality event identification system, Part 1: The wavelet transform and
feature extraction of power quality disturbances and Part 2: Design and
implementation,, EPRI Report, Aug. 1997.
[9] S. Santoso, E. J. Powers, W. M. Grady, and A. C. Parsons, Power quality
disturbance waveform recognition using wavelet-based neural classifier,
Part 1: Theoretical foundation and Part 2: Applications,, IEEE Trans. on
Power Delivery, PE-405-PWRD-0-01-1997 and PE-599-PWRD-0-011997, 1997.

Surya Santoso (M97) received the Ph.D. degree in electrical and computer
engineering from The University of Texas at Austin in 1996. From January
97August 97, he was a postdoctoral fellow under the auspices of the Electric
Power Research Institute. He is now with Electrotek Concepts, Inc, Knoxville,
TN. His primary responsibility is to research, develop, and implement advanced
algorithms for the state-of-the-art power quality and energy monitoring instrumentation. He is a Member of IEEE-PES and IEEE-SP.

W. Mack Grady (SM83) received the Ph.D. degree in electrical engineering


from Purdue University, West Lafayette, IN, in 1983. He is a Professor of electrical and computer engineering at The University of Texas at Austin. His areas
of interest include power systems analysis, power system harmonics, power
quality, and short term load forecasting.

Edward J. Powers (F83) received the Ph.D. degree in electrical engineering


from Stanford University, Stanford, CA, in 1965. He is a Professor of electrical
and computer engineering at The University of Texas at Austin. His current
interests include wavelet and higher-order statistical signal processing and their
application to transient and nonlinear phenomena.

Jeff Lamoree (M90) received the M.E. degree in electric power engineering
from Rensselaer Polytechnic Institute, Troy, NY, in 1990. He is currently the
General Manager for Electrotek R&D at Electrotek Concepts, Inc., Knoxville,
TN. He is responsible for the design, engineering, software development and
project management involved in Electroteks research and development project
area. These projects focus primarily on the development of power quality analysis tools for utility engineers and are typically funded by the Electric Power Research Institute. These tools include the PQ Database, the PQ Diagnostic System
and the PQ Toolbox.

Siddharth C. Bhatt (M72) received the Ph.D degree in physics from Rensselaer Polytechnic Institute, Troy, NY, in 1972. He was a Principal Engineer in
design, simulation, and computer products for 13 years with General Electric.
He spent two years with Litton Industries in electronic warfare simulation and
diagnostics systems. He is currently with EPRI managing various R&D projects
involving power quality systems, power quality and load signature analysis, advanced power quality monitoring projects using advanced technologies such as
wavelets, artifical neural networks, and expert systems.

You might also like