You are on page 1of 9


4, OCTOBER 2002 921

A Novel Fault-Detection Technique of

High-Impedance Arcing Faults in Transmission
Lines Using the Wavelet Transform
Chul-Hwan Kim, Member, IEEE, Hyun Kim, Young-Hun Ko, Sung-Hyun Byun,
Raj K. Aggarwal, Senior Member, IEEE, and Allan T. Johns, Senior Member, IEEE

Abstract—This paper describes a novel fault-detection so in view of the fact that apart from threatening the reliability
technique of high-impedance faults (HIFs) in high-voltage trans- of the electric power supply, these faults pose a risk of fires and
mission lines using the wavelet transform. Recently, the wavelet endanger life through the possibility of electric shock.
transform (WT) has been successfully applied in many fields. The
technique is based on using the absolute sum value of coefficients Most conventional fault-detection techniques for HIF mainly
in multiresolution signal decomposition (MSD) based on the involve processing information based on the feature extraction
discrete wavelet transform (DWT). A fault indicator and fault of post-HIF current and voltage. Several researchers in recent
criteria are then used to detect the HIF in the transmission line. years have presented results aimed at detecting HIF more effec-
In order to discriminate between HIF and nonfault transient tively. Hitherto, the algorithms developed include the current
phenomena, such as capacitor and line switching and arc furnace
loads, the concept of duration time (i.e., the transient time period), ratio method [3], the high-frequency method [4], the off-har-
is presented. On the basis of extensive investigations, optimal monic current method [5], the neural network and Kalman fil-
mother wavelets for the detection of HIF are chosen. It is shown tering method [1], [6]–[8]. S.J. Huang [9] has proposed an HIF
that the technique developed is robust to fault type, fault inception detection technique [10], [11] for distribution systems, which
angle, fault resistance, and fault location. The paper demonstrates uses a Morlet wavelet transform approach [12]. However, each
a new concept and methodology in HIF in transmission lines. The
performance of the proposed technique is tested under a variety of these techniques improves fault detection to a certain extent,
of fault conditions on a typical 154-kV Korean transmission-line but each has its drawbacks as well. Hitherto, a few techniques
system. (some available as commercial products) have been subjected
Index Terms—Fault detection, high-impedance arcing fault, to quite extensive testing in order to ascertain their effective-
transmission lines, wavelet transform. ness under different system and fault conditions.
It is well known that conventional Fourier transform-based
techniques (i.e., those which rely totally on spectrum analysis
I. INTRODUCTION of Fourier transform) do not possess the inherent time informa-
tion associated with fault initiation. The wavelet transform, on
H IGH-IMPEDANCE faults (HIFs) are, in general, difficult
to detect through conventional protection such as distance
or overcurrent relays. This is principally due to relay insensi-
the other hand, is useful in analyzing the transient phenomena
associated with transmission-line faults and/or switching op-
tivity to the very low level fault currents and/or limitations on erations. Unlike Fourier analysis, it provides time information,
other relay settings imposed by HIFs. This type of fault usually has the attribute of very effectively realizing nonstationary sig-
occurs when a conductor touches the branches of a tree having a nals comprising of low- and high-frequency components (such
high impedance or when a broken conductor touches the ground. as those commonly encountered in power systems networks)
In the case of an overcurrent relay, the low levels of current through the use of a variable windows length of a signal [13].
associated with HIF are below the sensitivity settings of the The advantages (particularly in terms of increased reliability
relay. In the case of a distance relay, which relies on an estima- and dependability) of a wavelet transform, which possesses
tion of impedance to fault based on the measured voltages and time and frequency information unlike the Fourier transform,
currents, the accuracy of the estimation (particularly in terms particularly in the detection of HIFs, are thus apparent, and
of relay overreach/underreach) can be significantly affected by one of these techniques is the subject of this paper. It should be
the high-impedance fault [1], [2]. HIFs, albeit uncommon, must noted that although wavelet analysis is more complex than other
nonetheless be accurately detected and removed. This is more signal-processing techniques, it is ideally suited for dealing
with nonstationary signals such as those encountered under
HIF arcing faults. This, in turn, enhances accuracy and reli-
Manuscript received April 24, 1999; revised February 14, 2002. This work ability in fault detection and the features can be applied to
was supported by the Korea Ministry of Science and Technology and Korea affecting fault-location techniques [14].
Science and Engineering Foundation.
C.-H. Kim, H. Kim, Y.-H. Ko and S.-H. Byun are with the School of Electrical This paper describes a new fault-detection technique which
and Computer Engineering, Sungkyunkwan University, Suwon-city, 440-746, involves capturing the current signals generated in a transmis-
Korea and NPT Center, Korea (e-mail: sion line under HIFs. It is shown that the technique improves the
R. K. Aggarwal and A. T. Johns are with the Department of Electronic and
Electrical Engineering, University of Bath, Bath BA2 7AY, U.K. performance of HIF detection by employing the absolute sum
Digital Object Identifier 10.1109/TPWRD.2002.803780 value based on the DWT. The detection process is performed
0885-8977/02$17.00 © 2002 IEEE

By simple interchange of the variables , , and rearrange-

ment of the DWT (1) gives


Upon closer observation of this equation, it can be noticed

that there is a remarkable similarity to the convolution equation
Fig. 1. Korean transmission system (154 kV) studied. for the finite-impulse-response (FIR) digital filters, therefore


where is the impulse response of the FIR filter.

By comparing (2) with (3), it is evident that the impulse
response of the filter in the DWT equation is


By selecting or ( ) and
, the DWT can be implemented by using a multistage
filter with the mother wavelet as the lowpass filter and its
dual as the highpass filter . Also, downsampling the output
of the lowpass filter by a factor of effectively scales
the wavelet by a factor of 2 for the next stage, thereby simpli-
fying the process of dilation.
The implementation of the DWT with a filter bank is com-
Fig. 2. Typical fault current waveform at relaying point (“a”-earth HIF, fault
, 10 km, Z = 300
at V
putationally efficient. The output of the highpass filter gives the
detailed version of the high-frequency component of the signal.
Also, the low-frequency component is further split to get the
through signal decomposition, thresholding of the wavelet
other details of the input signal. By using this technique, any
transform coefficients, and duration time. Threshold value is wavelet can be implemented [13].
determined by weighting the absolute sum value for one period
in a moving window scheme and this forms the basis of a
sophisticated decision logic for the limitation of a trip decision.
The results presented in this paper relate to a typical 154-kV A. Typical Waveforms Through Wavelet Transform Realization
Korean transmission line, the faulted signals of which are Fig. 1 shows a typical 154-kV Korean Transmission System
attained using the well known Electromagnetic Transients used in the simulation studies presented herein. It consists
Program (EMTP) software. The simulation also includes an of a 26-km line length terminated in two sources of 240 and
embodiment of a realistic nonlinear HIF model [15], [16]. The 180 MVA at ends P and Q, respectively; the nominal power
relay performance is then examined for HIF signals under a frequency is 60 Hz. The simulation of the power system has
variety of different system and fault conditions encountered been carried out using the well known EMTP. Within the sim-
in practice. ulation, an emulation of the nonlinear high-impedance arcing
faults has also been embodied [16]. Fig. 2 typifies the actual
II. DISCRETE WAVELET TRANSFORM current waveform (measured at end P but which has been scaled
Analogous to the relationship between continuous Fourier down through a CT) and is for an “a”-earth HIF at 10 km from
transform (FT) and discrete Fourier transform (DFT), the con- line-end P and is for a fault near , the distortion observed
tinuous wavelet transform (CWT) has a digitally implementable in the faulted “a” phase can be directly attributed to the highly
counterpart known as the DWT, and is defined as complex and nonlinear characteristics of the HIF arc path.
To aid the development of the fault-detection technique using
the DWT, wavelet transform realization has been employed
which determines a coefficient of d1 (detail one) using dif-
ferent mother wavelets from an actual current waveform. The
where is the mother wavelet, is the input signal, and mother wavelet considered is Daubechies (db)4, biorthogonal
the scaling and translation parameters “ ” and “ ” are functions (bior)3.1, coiflets (coif)4, and symlets (sym)5. Fig. 3 depicts
of integer parameter . The result is geometric scaling (i.e., the coefficient of d1 for different mother wavelets with the
) and translation by . DWT realization. The behavior of the DWT for this actual fault
This scaling gives the DWT a logarithmic frequency coverage current waveform is illustrated in Fig. 3(a)—(d) as expected.
and this is in marked contrast to the uniform frequency coverage All of the coefficients of d1 increase on fault inception and
of, for example, the short-time Fourier transform (STFT) [13]. there are small discernible differences in the DWT outputs

Fig. 3. Coefficient of d1 under “a”-earth HIF using DWT (a) db4 mother wavelet. (b) sym5 mother wavelet. (c) bior3.1 mother wavelet (d) coif4 mother wavelet.

for the four different mother wavelets that are considered. 2) the classification ability between the faulted phase and the
The performance of the DWT realization was evaluated under healthy phase.
different fault types, fault inception angle, and fault location, In order to select the most suitable mother wavelet, the
and some of the results will be shown. It should be mentioned maximum sum value (this is over a 1-cycle period at power
that due to a limitation of space, only the coefficient associated frequency) of d1 coefficients based on wavelet analysis is
with “a”-earth HIF at 10 km is shown. adopted for this work. Consider, for example, the waveforms
shown in Fig. 4 which illustrate the maximum sum value of d1
coefficients of the three-phase current signals (as measured at
B. Selection of Mother Wavelet for HIF Detection the relaying point) for an “a”-earth HIF at a distance every 1
km from end P in Fig. 1. First of all, considering Fig. 4(a) (this
As a second step in fault-detection technique, selection of the
is based on the db4 mother wavelet), it is clearly evident that
mother wavelet is essential to enhance the performance of HIF
the maximum sum value of d1 coefficients is significantly
detection technique to extract the useful information rapidly.
larger for the faulted “a”-phase than for the two healthy
For the technique considered here, this process leads to an ac-
phases “b” and “c.” This is also true when employing sym5
curate classification between the faulted phase and the healthy
and bior3.1 mother wavelets [as evident from Fig. 4(b) and
phase in the first instance, thereby significantly improving the
4(c)], although the levels are somewhat smaller in the case
performance and speed of the HIF detection process. As men-
of the former. However, when employing the coif4 mother
tioned in the foregoing section, the mother wavelets considered
wavelet, although there is a discernible difference between the
are db4, bior3.1, coif4, and sym5. For comparison of the perfor-
levels attained for the faulted “a” phase and the two healthy
mances attained using different mother wavelets, two conditions
phases, in comparison to the previous three mother wavelets
are compared as follows.
considered, they are significantly lower. Also, equally important
1) a significant magnitude of d1 coefficient for detecting the is that the differences in magnitudes between the faulted and
fault; healthy phases in the case of coif4 is much smaller than the

Fig. 4. Variation in the maximum sum value of d1 coefficients for three-phase current signals (a) db4 mother wavelet (b) sym5 mother wavelet (c) bior3.1 mother
wavelet (d) coif4 mother wavelet.

corresponding other types of mother wavelets as apparent from sample number that signifies the duration time for which a
the graphs shown in Fig. 4, and this is true for all fault positions. transient event (such as HIF) has to persist continuously. This
Thus, with regard to the selection of the mother wavelet, the is to discriminate between HIF and nonfault transient events
simple criterion adopted herein is based on the magnitudes of such as capacitor and line switching, arc furnace loads, etc. As
the summated coefficients d1 and the differences in magnitudes can be seen, when is greater than or equal to FC, the
between the faulted and healthy phases. In this respect, after value of FI is incremented and as soon as it attains the level ,
a series of studies employing the foregoing d1 coefficients this indicates an internal fault and a trip signal is initiated. As
distribution approach, the db4 and sym5 are appropriate for shown in Fig. 5, the absolute sum value is based on
detection of HIF in transmission line. The db4 mother wavelet summating the d1 coefficients over a 1-cycle period and the
is chosen for this study. sampling rate employed is 3840 Hz (i.e., 64 samples/cycle at
60 Hz). The summated values associated the three phases are
C. Fault-Detection Algorithm compared with a preset threshold level FC. The whole process
Fig. 5 shows the fault-detection procedure of the proposed is based on a moving window approach where the 1-cycle
technique, where FI is a counter that signifies the sample window is moved continuously by one sample.
number (and, therefore, the time period) for which useful infor- It is apparent from the foregoing decision logic that the
mation through DWT realization under HIF persists. criteria for the protection relay to initiate a trip signal is
is the sum value of the detailed output (d1 component) for a such that must stay above the threshold level FC
1-cycle period and is represented as an absolute value, fault continuously for samples (after fault inception). In this
criterion (FC) is the signal magnitude threshold as the lower respect, an extensive series of studies has revealed that in
limit of that is used to detect the HIF, and is the order to maintain relay stability for external faults (i.e., faults

Fig. 5. Block diagram of fault-detection technique.

Fig. 7. Variation in the sum value for “a”-earth HIF at 10 km from end P (a)
fault inception angle 90 (b) fault inception angle 0 .

and are 0.085 and 128, respectively. The setting values of

these thresholds are dependent on the system environments.
Note that the latter corresponds to a two-cycle period at power


In this section, some typical results illustrate the performance
of the protection technique being developed. It should be noted
that although not shown herein, responses/limitations due to
CTs, relay hardware (such as current interface module com-
prising anti-aliasing filters and analog-to-digital converters),
etc. have been taken into account in the simulation so that the
relay performance attained pertains closely to that expected in

Fig. 6. Variation in the sum value for “a”-earth HIF as a function of fault A. Single-Phase-Earth Fault
distance (a) fault inception angle 90 (b) fault inception angle 0 . Fig. 6 depicts the absolute sum value of d1 associated with the
three-phase currents (measured at end P of the system shown
behind busbar P and beyond busbar Q in Fig. 1) and also in Fig. 1) using the db4 mother wavelet; the graphs shown in
restrain under no-fault conditions, the optimal settings for FC Fig. 6(a) and 6(b) are for an “a”-phase-earth HIF, as a function

Fig. 9. Variation in the sum value for “a”- “b” earth HIF at 10 km from end P
(a) fault inception angle 90 (b) fault inception angle 0 .

Fig. 7 shows the behavior of absolute sum value of d1 for

one particular fault position (i.e., 10 km from end P, as a func-
Fig. 8. Variation in the sum value for “a”- “b” earth HIF as a function of fault tion of time, the graphs also depict the moving-window regime
distance (a) fault inception angle 90 (b) fault inception angle 0 .
adopted). It is apparent that for both fault inception angles con-
sidered, the faulted “a” phase stays above the set threshold level
of fault distance and are for fault inception angles of and ( ) for more than two cycles; this corresponds to a
, respectively. As expected, the magnitudes of the faulted sample number of approximately 192 and is well above the set
“a” phase are much higher than the healthy phases and this is level of . The protective relay will thus initiate a trip
true for both fault inception angles considered. Also important, signal. For comparison purposes, the healthy phase currents “b”
although the magnitudes reduce in size as the fault moves away and “c” are also shown and, as expected, their magnitudes are
from end P, the very significant difference between the faulted well below the threshold level .
and healthy phases is retained for an appreciable time period. It
should be mentioned that in all HIF cases, the fault impedance
B. Double-Phase-Earth Fault
has been represented by a realistic nonlinear arc model, but
for reference purposes only, this can be considered as approxi- Fig. 8 typifies the variations in the absolute sum values of the
mately equivalent to about 300 . coefficients d1 for a double-phase-earth HIF fault. The graphs

Fig. 11. Coefficient of d1 under arc furnace phenomenon.

depicts the behavior of the “a”-phase absolute sum value

for an “a”-earth fault for faults just behind the busbar
P and beyond the remote busbar Q. It is apparent that although
the levels of the signals are quite high, they are nonetheless
well below the threshold level , thereby restraining
the relay from asserting a trip decision. This is the case for
voltage maximum and voltage minimum faults. Although not
shown here, this is also true in the case of double-phase-earth
HIF faults.
It should be mentioned that when considering the behavior
of the aforementioned signals under nonfault transient condi-
tions, such as capacitor and line switching, arc furnace loads,
etc. (these signals are not shown in this paper), although the
magnitudes of the absolute sum value can momen-
tarily exceed the threshold level . In some cases,
importantly, an extensive series of studies has revealed that in all
cases, the levels are significant for a much shorter period ( 1
cycle) compared to a fault situation. This effectively means that
Fig. 10. Variation in the sum value for “a”-earth external faults (a) fault the decision logic as described in Section III-C would inhibit
inception angle 90 (b) fault inception angle 0 .
relay operation in the case of the former.
shown are for two fault inception angles, one near voltage max-
imum of the “a”-phase [Fig. 8(a)] and the other is in relation to D. Nonfault Transient Events
voltage zero of the “a”-phase [Fig. 8(b)]. Here again, there is a In the development of any new fault-detection technique,
large difference in magnitudes between the faulted phases and such as the type described in this paper, it is important to ensure
the healthy phase and, as expected, both of the faulted phases that it is secure under nonfault events such as capacitor and line
(“a” and “b” involved in the fault) attain high levels. switching, arc furnace loads, etc. In this respect, it should be
When considering the dynamic behavior of the absolute sum noted that for this fault detector, the parameter ‘ ’ (comprising
value as a function of time, for one particular fault po- of 128 samples) within the decision logic plays a crucial role in
sition (which is 10 km from end P in Fig. 1), Fig. 9 portrays char- maintaining the fault detector’s security.
acteristics which are very much in line with what is expected The HIF detector presented herein was extensively tested
(i.e., both the faulted phases “a” and “b” stay well above the set under many nonfault transient events. As an example, Fig. 11
level ) for an appreciable time after fault inception depicts the behavior of coefficient d1 (with DWT realization) to
and the healthy phase stays well below the threshold throughout arc furnace phenomenon. As can be seen, its magnitude is quite
the period of interest. large when the load is switched in but decreases rapidly within
a short duration of time thereafter. With regard to the absolute
C. Relay Performance Under External Faults sum value of d1, it is apparent from Fig. 12 that although its
As mentioned before, it is vitally important for the designed magnitude exceeds the preset threshold level ( ) for
protection to be stable under external fault conditions. Fig. 10 both the “b” and “c” phases, this is only momentary, certainly

[3] R. E. Lee and M. T. Bishop, “Performance testing of the ratio ground

relay on a four-wire distribution feeder,” IEEE Trans. Power App. Syst.,
vol. 102, pp. 2943–2949, Sept. 1983.
[4] B. M. Aucoin and B. D. Russell, “Detection of distribution high
impedance faults using burst noise signals near 60 Hz,” IEEE Trans.
Power Delivery, vol. 2, pp. 342–348, Apr. 1987.
[5] D. I. Jeerings and J. R. Linders, “A practical protective relay for down-
conductor faults,” IEEE Trans. Power Delivery, vol. 6, pp. 565–574, Apr.
[6] A. A. Girgis, W. Chang, and E. B. Makram, “Analysis of
high-impedance fault generated signals using a Kalman filtering
approach,” IEEE Trans. Power Delivery, vol. 5, pp. 1714–1722, Oct.
[7] S. Ebron, S. L. Lubkeman, and M. White, “A neural network approach
to the detection of incipient faults on power distribution feeders,” IEEE
Trans. Power Delivery, vol. 5, pp. 905–912, Apr. 1990.
[8] E. A. Mohamed and N. D. Rao, “Artificial neural network based fault
diagnostic system for electric power distribution feeders,” Elect. Power
Syst. Res., vol. 35, no. 1, pp. 1–10, 1995.
[9] S. J. Huang and C. T. Hsieh, “High impedance fault detection utilizing a
Morlet wavelet transform approach,” IEEE Trans. Power Delivery, vol.
14, pp. 1401–1410, Oct. 1999.
[10] A. E. Emanuel and E. M. Gulachenski, “High impedance fault arcing on
sandy soil in 15 kV distribution feeders: Contributions to the evaluation
Fig. 12. Variation in the sum for arc furnace loads. of the low frequency spectrum,” IEEE Trans. Power Delivery, vol. 5, pp.
676–686, Apr. 1990.
[11] D. C. T. Wai and X. Yibin, “A novel technique for high impedance fault
for a much short period than samples, thus ensuring identification,” IEEE Trans. Power Delivery, vol. 13, pp. 738–744, July
the stability of the fault detector. [12] O. Chaari, M. Meunier, and F. Brouaye, “Wavelets: A new tool for the
resonant grounded power distribution systems relaying,” IEEE Trans.
Power Delivery, vol. 11, pp. 1301–1308, July 1996.
V. CONCLUSION [13] G. Strang and T. Q. Nguyen, Wavelets and Filter Banks. Cambridge,
MA: Wellesley-Cambridge, 1996.
This paper has presented a novel technique for transmis- [14] F. H. Magnago and A. Abur, “Fault location using wavelets,” IEEE
sion-line fault detection under high-impedance Earth faults Trans. Power Delivery, vol. 13, pp. 1475–1480, Oct. 1998.
[15] V. L. Buchholz, M. Nagpal, J. B. Neilson, and W. Zarecki, “High
using the discrete wavelet transform for which a near optimal impedance fault detection device tester,” IEEE Trans. Power Delivery,
mother wavelet (suitable for a vast majority of different samples vol. 11, pp. 184–190, Jan. 1996.
and fault conditions) has been selected after an extensive series [16] A. T. Johns, R. K. Aggarwal, and Y. H. Song, “Improved techniques for
modeling fault arcs on faulted EHV transmission systems,” Proc. Inst.
of studies. The DWT-based technique presented herein has Elect. Eng. Gener. Transm. Distr., vol. 144, no. 2, pp. 148–154, 1994.
a number of distinct advantages over other traditional HIF
detection techniques.
For example, it is robust to a variation in different system and Chul-Hwan Kim (M’97) was born in Korea on January 10, 1961. He
fault conditions and is predominantly dependent upon the non- received the B.Sc., M.Sc., and Ph.D. degrees in electrical engineering from
linear behavior of the HIF; it is stable for external faults and has Sungkyunkwan University, Korea, in 1982, 1984, and 1990, respectively.
Currently, he is Professor in the School of Electrical and Computer
the ability to discriminate clearly between internal faults and Engineering at Sungkyunkwan University, Korea. He was a Visiting Academic
nonfault transient events such as capacitor and line switching, in the University of Bath, U.K., in 1996, 1998, and 1999. In 1990, he
arc furnace loads, etc; it has the inherent attribute of distin- became a Full-Time Lecturer at Cheju National University, Cheju, Korea. His
research interests include power system protection, artificial intelligence, the
guishing between the faulted phase(s) and healthy phase(s) and modeling/protection of underground cable, and EMTP software.
this is a significant advantage for transmission systems in which
single-pole tripping is employed, and which therefore requires
phase selection. The technique developed is based on current Hyun Kim was born in Korea on October 4, 1972. He received the B.Sc. and
signals only and, therefore, requires the use of current trans- M.Sc. degrees in electrical engineering from Sungkyunkwan University, Korea,
formers (CTs) only. in 1997 and 1999, respectively.
Currently, he is involved with condition monitoring in a nuclear power plant
Work is now in progress to extend the research for at Woori Technology, Inc., where he has been since March 1999.
ultra-high–voltage transmission systems comprising of long
lines and for fault detection/fault location in low-voltage
distribution lines. The outcome of the research will be reported Young-Hun Ko was born in Korea on July 7, 1975. He received the B.Sc. and
in due course. M.Sc. degrees in electrical engineering from Sungkyunkwan University, Korea,
in 1998 and 2000, respectively.
Currently, he is a Researcher of product R&D center in Daewoo Telecom.
[1] A. F. Sultan, G. W. Swift, and D. J. Fedirchuk, “Detection of high
impedance arcing faults using a multi-layer perceptron,” IEEE Trans. Sung-Hyun Byun was born in Korea on February 11, 1973. He received the
Power Delivery, vol. 7, pp. 1871–1877, Oct. 1992. B.Sc. and M.Sc. degrees in electrical engineering from Sungkyunkwan Univer-
[2] B. D. Russel, C. L. Benner, and A. V. Mamishev, “Analysis of high sity, Korea, in 1996 and 1998, respectively.
impedance fault using fractal techniques,” IEEE Trans. Power Syst., vol. Currently, he works in power system protection and control at the Korea
11, pp. 435–440, Feb. 1996. Power Exchange (KPX), Seoul.

Raj Aggarwal (SM’91) received the B.Sc. and Ph.D degrees in electrical engi- Allan T. Johns (SM’88) received the B.Sc. and Ph.D. degrees from the Univer-
neering from the University of Liverpool, U.K., in 1970 and 1973, respectively. sity of Bath, U.K. In 1982, he was awarded the degree of D.Sc. for an original
Currently, he is a Professor and Head of the Power and Energy Systems Group and substantial contribution to knowledge of electrical engineering.
at the University of Bath, U.K. His main areas of research interests are power Currently, he is Emeritus Professor in the Electrical Engineering Department
system modeling and the application of digital techniques and artificial intelli- at the University of Bath, U.K.
gence, protection and control, and power quality issues. Dr. Johns is is a fellow of the IEE U.K.
Dr. Aggarwal is a fellow of the IEE U.K.