Electric Power Systems Research 143 (2017) 376–388

Contents lists available at ScienceDirect

Electric Power Systems Research

journal homepage: www.elsevier.com/locate/epsr

Review

High impedance fault detection: A review

Amin Ghaderi a,∗ , Herbert L. Ginn III a , Hossein Ali Mohammadpour b
a
The University of South Carolina, SC, USA
b
POWER Engineers, Inc., CO, USA

a r t i c l e i n f o a b s t r a c t

Article history: Protection from the high impedance fault (HIF) has been one of the biggest challenges in the power
Received 22 July 2016 distribution network. HIF typically occurs when the conductors in distribution network break and touch
Received in revised form 7 October 2016 the ground surface; or lean and touch a tree branch. This fault, with current magnitude close to the load
Accepted 9 October 2016
current level, is not detectable by over-current relays. This paper aims to review the literature related to
Available online 27 October 2016
the HIF phenomenon. In this work, the HIF detection techniques are categorized, evaluated, and compared
with one another. Furthermore, the existing HIF models and HIF location techniques are reviewed. Finally,
Keywords:
the shortcoming of the existing perspective toward the HIF phenomenon and the possible road to the
High impedance fault
Fault detection techniques
future for HIF detection is discussed.
Modeling
© 2016 Elsevier B.V. All rights reserved.
Fault location techniques
Literature review

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
1.1. HIF definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
1.2. HIF hazard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
1.3. HIF characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
2. Fault detection as a pattern classification task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3. High impedance fault detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3.1. Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3.1.1. Current measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3.1.2. Voltage measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.1.3. Current and voltage measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.1.4. Magnetic field intensity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.2. Feature extraction (analysis domain) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.2.1. Time domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.2.2. Frequency domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.2.3. Time-scale domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
3.2.4. Time-frequency domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.2.5. Hybrid domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.3. Classifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.3.1. Simple threshold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.3.2. Threshold and counter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.3.3. Artificial neural network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
3.3.4. Other classifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
4. Criteria to evaluate fault detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
5. High impedance fault location techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
6. High impedance fault modeling techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

∗ Corresponding author.
E-mail address: [email protected] (A. Ghaderi).

http://dx.doi.org/10.1016/j.epsr.2016.10.021
0378-7796/© 2016 Elsevier B.V. All rights reserved.
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 377

7. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
7.1. Popularity of the HIF field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
7.2. Distinguishing HIF from other power system disturbances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
7.3. Comparison of the proposed HIF detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
8. Future of the high impedance fault detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
8.1. Increase in the processing power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
8.2. Growth of the information technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
8.3. Publishing HIF detection techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
8.4. Practical data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
9. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385

1. Introduction

Understanding high impedance fault (HIF) phenomenon,

including its physical and electrical aspects, would help the power
system protection engineers develop more intuitive detection tech-
niques. There have been several partial review articles on different
aspects of the HIF [1–3]. Prasad et al. compared the efficiency of
four low impedance fault detection techniques in detecting HIF [4].
Moreover, Li et al. reviewed some of the HIF detection techniques
before 2000 [5]. The purpose of this article, on the other hand, is
to provide a systematic, up-to-date, and comprehensive review of
the HIF phenomenon. The remainder of this paper is organized as
follows:
In the rest of this section, HIF’s definition and characteristics are
introduced. Section 2 is dedicated to discuss how to divide each
HIF detection technique into three separate components which
enables the analysis and evaluation of the existing techniques. In
Section 3, different aspects of HIF detection techniques are stud-
ied in detail, based on the analysis tools provided in Section 2. In
Section 4, nine criteria that are capable of evaluating and compar-
ing the HIF detection techniques are discussed. Sections 5 and 6
are dedicated to discussing the HIF location technique and model-
ing techniques, respectively. Detailed discussion of the existing HIF
detection techniques is presented in Section 7. The possible future
of HIF detection technique is demonstrated in Section 8. Section 9
draws the conclusion of this work.

1.1. HIF definition

HIF is a group of power system disturbances that prevents the

generation of sufficient current that is required to trip the over-
current relays, due to high grounding impedance. This term is
normally applied to faults in distribution networks (15–25 kV [6]).
Fig. 1. High impedance fault on different surfaces [22].
However, detection techniques are also proposed for HIFs in trans-
mission networks [7], low voltage power systems [8], and low
voltage 400 Hz airplane circuits [9,10]. 1.2. HIF hazard
HIFs are influenced by multiple factors, namely, the ground
surface material, surface humidity, feeder configuration, voltage Unlike most power system faults, which primarily endanger
levels, weather conditions, and load type [11]. The two most influ- electrical equipment, HIFs jeopardize human safety and could also
ential factors on the HIF characteristics are: (a) surface humidity result in legal issues [23]. Therefore, since over-current relays are
and (b) surface material. It is shown that higher surface humidity unable to detect HIFs [24], it is a crucial task to detect these types of
results in higher fault current magnitude [12]. Moreover, HIF occurs fault using algorithms that are specially designed to address such
on various materials and each results in different voltage–current issues. Followed by HIF inception, the generation and accumulation
characteristics [13,14]. The materials that are mostly touched by of flammable gas, or the evolution of the arc to close flammable
down-conductors are: tree branches [15,16], lawns, gravel, stout material could result in fire and explosion [25]. In addition to these
gravel, thin gravel [17], asphalt, concrete, sand [18], crushed stone hazards, there is a possibility that the line crewmen or members
[19], board blocks [20], and cement [13,21]. Fig. 1 shows HIF occur- of the public would have unintentional contact with the energized
ring on concrete and grass surfaces [22]. exposed conductor [18].
378 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

Certain and comprehensive reports about the statistics of the

HIFs are missing since only the faults that result in relay alarm or
trip are reported. However, it is estimated that between 5% and 10%
of the distribution faults are HIF [26]. In another study, it is reported
that 25% of the down-conductors are not detected using the over-
current relays, which are considered as HIFs [27]. Another study
increases this number up to 32% of the down-conductors [23]. All
of these reports point to the fact that the HIFs are a non-negligible
part of faults in distribution network. So it is required that some
measures are taken for isolating these faults.

1.3. HIF characteristics

The most prevalent physical signature of the HIF is the fact

that it is often accompanied with the AC electric arc [28], which
is shown in Fig. 1 for HIF on grass and concrete. For this reason, HIF
is sometimes called high impedance arcing fault [16,29,30]. The arc
ignition occurs once the magnitude of the voltage of the conductor
touching the surface exceeds a specific voltage threshold, known
as break down voltage Vbr . Consequently, the arc extinction occurs
once voltage magnitude becomes lower than Vbr . The value of Vbr
stays at different constant values for each half cycle of voltage [31].
Thus, in each cycle of the voltage, the HIF results in at least two Fig. 2. (a) Structure of a simplified pattern classification task. (b) Structure of the
arc re-ignitions and two arc extinctions [32]. In addition to the low corresponding fault detection technique.
current magnitude and accompanied arc, the HIF results in several
physical and electrical characteristics [19,33–35], including: Therefore, using the pattern classification concept, any fault detec-
tion task corresponds to the classification of the power system
• Intermittence of the arc: the arc does not generate current in a states.
steady state pattern, instead it generates a few conduction cycles In the most simplified manner, a PCT consists of three steps,
followed by several non-conduction cycles [25]. namely, (a) measurement (performed by sensors), (b) analysis
• Asymmetry in the current waveform: due to difference in the domain (feature extraction), and (c) classifier (boundary setting).
break-down voltage against positive and negative voltage values This is demonstrated in Fig. 2. Here, an example of applying these
[32], there is a difference between the peak value and shape of steps to detecting low impedance faults in distribution network is
positive and negative half cycles of HIF current [36]. discussed. The first step is to measure the line current using cur-
• Current buildup and shoulder: current magnitude gradually rent transformers. The second step is to find meaningful features
escalates (builds up) until it reaches and maintains a constant from these measurements that are faithful representation of the
value for several cycles (shoulder) [37]. fault occurrence. For this example, the current magnitude is cho-
• Non-stationary Current: the current frequency spectrum varies sen as a feature (signature) that represents the fault occurrence.
with time [22]. And finally, the third step is to find a threshold that separates the
• Randomness: both the current magnitude, and its faulty and healthy states.
conduction/non-conduction intervals are random values [38].
• Non-linearity: voltage–current characteristic of the HIF is non- 3. High impedance fault detection techniques
linear due to the existence of the arc [39,40].
• Low frequency components in current and voltage wave- The HIF does not manifest clear variations in the quantities typi-
forms: due to non-linearity of HIF the waveforms contain cally used for fault detection, namely voltage and current. However,
harmonics up to 600 Hz for the current and up to 300 Hz for meaningful variations can be identified if proper signal processing
voltage [41,42]. (measurement, feature extraction, boundary setting) is performed.
• High frequency components in current waveforms: arc results Therefore, in order to evaluate the existing detection techniques
in high frequency components in the current waveform [43]. and develop more reliable ones, it is necessary to separately study
each step of the fault detection techniques, introduced in Section
2. Fault detection as a pattern classification task 2. In this section, the existing approaches to each step are studied
and their shortcomings are discussed.
This section briefly discusses how fault detection techniques
could correspond to the pattern classification tasks (PCT) [13]. Pat- 3.1. Measurement
tern classification is defined as assigning an object or an event
to one of a number of classes based on the features derived to In order to detect the HIF, various measurements are used,
emphasize the commonalities [44]. Viewing fault detection from including, current, voltage, and magnetic field intensity. Each of
the pattern classification perspective has two important benefits: these signals will be discussed thoroughly in this section.
(1) using the pattern classification structure, it is possible to divide
the protection task into essential building blocks and address each 3.1.1. Current measurement
part separately; and (2) pattern classification provides a set of sys- The HIF can be seen as a stochastic non-linear load. Therefore,
tematic evaluation criteria to compare and categorize the detection it draws a non-stationary current with non-linear characteristics.
techniques. As a result, the current waveform demonstrates high- and low-
Generally, any power network has two operating states, namely, frequency components that are used to detect the HIF [45–48].
the healthy state (normal) and the faulty state (abnormal). Moreover, a combination of current measurements in a differential
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 379

manner is used to detect the HIF [49]. It should be noted, however, fractal geometry is used to reveal the level of chaos (random-
that since the distribution lines and current transformers act as low ness) in the current waveform as the HIF signature. In [57,63], time
pass filters, the informative part of the HIF current would be lost domain mathematical morphology technique, which reveals the
before reaching the processor. This makes current measurement irregularity in the waveform is used to extract HIF features.
less reliable. It should also be noted that the current measurement is Although time domain techniques are generally simple, their
more efficient in strong distribution networks where the frequency assumption of infinite frequency window, which result in the loss of
components of the current waveform are not strongly attenuated. frequency domain information, decreases their ability in accurately
extracting the HIF signature.
3.1.2. Voltage measurement
Voltage measurement has also been used to detect the HIF 3.2.2. Frequency domain
[50–52]. However, this measurement is only efficient when there is As discussed in Section 1.3, the arc associated with an HIF
enough voltage drop from the fault location to the relay. Therefore, results in the generation of low- and high-frequency components
since the HIF current magnitude is fairly low, such measurement is in the current spectrum. Frequency domain techniques, which take
only reliable in very weak distribution systems. advantage of these generated frequency components, fall into two
classes: low-frequency based techniques and high-frequency based
3.1.3. Current and voltage measurement techniques.
Increasing the reliability of the detection technique is also pos- The former are based on the fact that arc nonlinearity results
sible through measuring both voltage and current. Traveling waves in low order harmonics. Third harmonics of the current and volt-
of the voltage and current have been utilized for HIF detection age are widely used as the frequency domain signature of the HIF
in [53,54]. In other studies, the synchrophasors of current and [56,64]. In addition, other harmonics such as 2nd and 5th harmon-
voltage are used to detect the HIF [55,56]. It should be noted, ics are used to increase the reliability of the 3rd harmonic based
however, that such reliability comes at the price of adding addi- techniques [33,65]. In [66], low order even, odd, and interharmon-
tional measurement devices, high bandwidth sensors, or GPS based ics are used as HIF feature. In [67], the magnitude of the frequency
communication, which increase the overall cost of the detection components in the vicinity of low order harmonics (such as 30 Hz
technique. and 150 Hz for 60 Hz power system) are used as features of the HIF.
In other techniques several low order harmonics are used together
3.1.4. Magnetic field intensity to detect the HIF [68,69].
In [57], HIF indicators, installed in the vicinity of the sending The high frequency based techniques take advantage of the ran-
end of feeders, are utilized to measure the magnetic field intensity. dom and sharp variation of the HIF current [43]. They use the Fast
Such measurement, though highly reliable, increases the cost and Fourier Transform (FFT) to extract the harmonics. However, due to
complexity of the detection techniques. the fact that HIF current is a non-stationary signal, the frequency
components of the current are time-dependent. Therefore, several
3.2. Feature extraction (analysis domain) techniques use more complex methods, such as Kalman filtering,
to estimate the frequency components of the HIF current [70,71].
The second step of any detection technique, i.e. the analysis Although the frequency domain techniques are capable of
domain, is responsible for revealing the hidden characteristics (fea- extracting some salient properties of the HIF, their inability to
tures) of the HIF by transforming the measured data into more temporally localize the frequency components decreases their
informative analysis domains. It is important to study the anal- detection efficiency. The reason is that several disturbances in dis-
ysis domains that are capable of mathematically manifesting the tribution network, such as capacitor switching and transformer
characteristics of the HIF, such randomness, non-linearity, non- energization, share similar frequency domain features with the HIF.
stationary nature, asymmetry. The analysis domains utilized for
HIF detection fall into four classes, namely, time domain, frequency 3.2.3. Time-scale domain
domain, time-scale domain, and time-frequency domain. These Time-scale analysis, also known as wavelet transform (WT),
domains will be discussed in this section. maps the time-domain signals into time-scale domain. It catches
It should be noted that in the case of large number of features both the frequency information and the instant (location) in time
derived from the analysis domain, which consequently increases where these frequency components occur. This capability makes
the complexity of the classifier, a feature selection step should be the wavelet based methods powerful tools for analyzing signals
added to the design. The function of feature selection is to decrease with time-varying spectrum, such as HIF current [58,72–78]. There-
the feature set dimension by selecting most informative features. fore, this unique ability results in the popularity of WT in HIF
For example, the Principal Component Analysis is used to decrease detection (about 40% of all the HIF detection techniques have been
the feature set dimensionality and therefore the classifier complex- wavelet based).
ity in [22,58]. In [79], two factors are shown to be influential on the WT based
techniques, namely, sampling frequency and choice of mother
3.2.1. Time domain wavelet. It is illustrated in [13] that the HIF results in frequency
Time domain analysis extracts the temporal irregularity of the components up to 10 kHz, which necessitates the sampling fre-
HIF waveforms. In [59], the magnitude of zero-sequence of voltage quency of at least 20 kHz, based on Nyquist–Shannon sampling
and current are used as features of the HIF. However, simple time- theorem. Several articles have explored the choice of mother
domain indices, such as magnitude, RMS, peak, and mean values wavelet by comparing the detection method efficiency [80]. Several
could not efficiently demonstrate hidden irregularities available in mother wavelets such as daubechies [74,81], Morlet [82], symlet
HIF. [83], and rbior [13] have been utilized to decompose the current
Some more computationally complex time domain based tech- and voltage signals into different details and approximations. In
niques have been proposed to detect the HIF. In [60], the chaotic [76], daubechies mother wavelet is suggested to be superior to
functions, which look for the variation of system linearity states, other mother wavelets due to its capability in accurate detection of
are used for HIF detection. In [61], the summation of autocorre- signal with low amplitudes, its short duration, and its fast decay.
lation and partial autocorrelation of power quantities and their In spite of the advantages of the WT based techniques, it is a
derivatives are used as features of the HIF. In [62], time-domain challenge to design a systematic detection technique based on WT.
380 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

This is due to its (1) narrow high frequency support, (2) subjectivity advantages, the ANNs make the total design of the detection algo-
to the choice of mother wavelet, and (3) loss of feature resolution rithm more complex. Moreover, choosing the number of layers and
(low ability for the interpretation). number of perceptrons in each layer are trial-and-error procedures
which decrease the objectivity of the ANN based techniques.
3.2.4. Time-frequency domain
In order to maintain the time-frequency localization of the 3.3.4. Other classifiers
wavelet transform, while avoiding some of its shortcomings, the Numerous classifiers have been utilized less frequently in HIF
time-frequency domain analysis is proposed for HIF detection. detection. Fuzzy inference [112,113], and support vector machine
Time-frequency domain analysis estimates the energy of the signal [114,115] have been more attractive than other classifiers, such
in each point in time and frequency coordinates. In [84], a time- as decision trees [116], the genetic algorithm [21], the knowledge
frequency based technique (S-transform) is used to detect the HIF, based classifier [117], Bayes-based classifier [13], z-test classifier
and its reliability is shown to be superior to wavelet based tech- [118], and rule-based classifier [7]. Though less studied and utilized,
niques. In [85] the ridges of time-frequency distribution are used these classifiers could still demonstrate high level of efficiency in
as a powerful method in detecting the HIF. Moreover, in [22,86] detecting the HIF, if accompanied by appropriate measurement and
the strength of the time-frequency based techniques in detecting analysis domain.
non-stationary signal, such as HIF, is demonstrated.
Time-frequency domain analysis has the following advantages: 4. Criteria to evaluate fault detection techniques
(1) time- and frequency-localization, (2) coherent time-frequency
support, and (3) features with the ability of interpretation [87]. Several indices have been introduced for the evaluation of fault
However, the calculation of time-frequency distribution requires detection techniques, namely, selectivity, sensitivity, speed [119],
high computational capability compared to the other domains. dependability, and security [120]. However, the existing criteria do
not completely evaluate the detection algorithms for more com-
3.2.5. Hybrid domain plex faults, such as HIF. For example, for high impedance faults
Several articles have utilized a combination of the four domains occurring on distribution lines that supply sensitive loads, such
discussed above. A combination of time-domain and frequency- as traffic lights or hospitals, disconnecting the line would not be
domain based techniques is used in [88,89]. Moreover, in [90] the best decision. Therefore, the dependability criteria would not
wavelet and frequency domain are combined to extract signatures be enough to evaluate the correct decisions of the relay. Another
of the HIF. The combination time domain, frequency domain, and reason for the necessity of new evaluation criteria arises from the
wavelet domain is also reported in [91]. Although such hybridiza- fact that, in general, the HIF detection techniques are computation-
tion increases the detection method reliability, the complexity of ally complex. Therefore, it is desirable that each method is able to
the hybrid techniques might impeded their practicality. potentially detect (identify) other types of faults in the system, in
order to minimize the overall computation on the microprocessors.
3.3. Classifier However, the existing criteria could not address such necessity.
These unique characteristics of the HIF and its detection tech-
The third step of each fault detection technique is finding the niques necessitate a more comprehensive set of criteria for their
boundary which separates a faulty state from a healthy one. For evaluation. In this article the fault detection evaluation criteria
HIF detection, various classifiers with different levels of complexity proposed in [22,121] are utilized to evaluate and compare the HIF
have been developed. In this section, each of these classifiers is detection techniques.
discussed in more detail. The proposed set of criteria is extracted from mapping the fault
detection techniques to the pattern classification tasks (introduced
3.3.1. Simple threshold in Section 2). This allows the utilization of the existing pat-
Similar to the traditional protection techniques, such as over- tern classification evaluation techniques. Confusion matrix and its
current relays, several techniques utilize a simple threshold to derivations are used for such purpose. For a two-class classifiers
separate HIF from non-HIF conditions [92–96]. Though such clas- (such as detection algorithms), the confusion matrix (CM ) is a two-
sifiers highly simplify the techniques, the complicated nature of by-two matrix defined as:
the HIF often necessitates more complex classifiers that are able to
 
True Positive False Positive
distinguish the HIF from other disturbances. CM = (1)
False Negative True Negative

3.3.2. Threshold and counter where in the case of detection algorithms, true positive (TP) shows
One approach for increasing the reliability of the simple thresh- the number of right detection of the faults, true negative (TN)
old classifiers is to accumulate their confidence by adding a counter demonstrates number of right decision of healthy conditions, false
[96–100]. In other words, if the algorithm detects the HIF for a positive (FP) shows the number of faults that are not detected,
consecutive number of times, the classifier decides that a fault has and finally false negative (FN) is number of healthy condition that
occurred. Such a design has several drawbacks, such as decreasing mistakenly classified as faults.
the technique speed, a requirement for larger memory, and missing Based on this matrix, a new set of reliability criteria are proposed
less steady HIFs. which are summarized in Table 1. More detailed explanation can
be found in [121]. The proposed criteria are used in Section 7 to
3.3.3. Artificial neural network compare the existing high impedance fault detection techniques.
Artificial neural networks (ANNs) represent “computational
paradigms based on a biological metaphor” [101]. They are known 5. High impedance fault location techniques
to have high accuracy in pattern classification and generaliza-
tion, fast response, noise removal ability, and prediction capability An accurate and fast fault location algorithm increases the
[102,103]. They have also been extensively used in HIF detection. power system reliability and the restoration speed, decreases the
About 23% of HIF detection techniques, such as [104–111], utilize economic loss due to the faults, and pinpoints the weak part of the
some topologies of artificial neural networks. Despite their reputed power system [122–126]. There are three classes of techniques used
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 381

Table 1
Summary of the fault detection evaluation criteria [22,121].

Criterion Acronym Formula Functionality Class

TP+TN
Accuracy A TP+FP+FN+TN
% Overall precision
TP
Dependability D TP+FP
% Faulty state detection precision
TN
Security S TN+FN
% Healthy state detection precision
TN Reliability criteria
Safety SF TN+FP
% Safety related criterion
TP
Sensibility SN TP+FN
% Sensitive load related criterion
Tone−cycle
Speed v Tdetection
Detection speed

Cost C Software and hardware complexity

Overall performance
Objectivity OBJ Objectivity to fault type and network
criteria
Completeness COM Ability to hold important information

for HIF location, namely, impedance based, traveling wave based,

and communication based.
Impedance based techniques, which estimate the fault location
based on the measured impedance magnitude, are used in [127].
The randomness and high magnitude of the HIF impedance would
decrease the reliability of this class of location techniques.
The second class of fault location techniques, traveling wave
(TW) based, falls into two subclasses, namely, active and pas-
sive techniques. Passive TW techniques estimate the fault location
based on TW transients received from the fault [115,128]. Their
disadvantage is that, due to very small variation in current and volt-
age after the HIF, large enough TWs might not be generated. This
makes these techniques less reliable. Active TW techniques are also
discussed in [129–131]. They inject a TW to the distribution sys-
tem and extract its reflection to locate the fault. Their disadvantage
is that the injected TW should have high enough energy to travel
to the end of the line and return without remarkable attenuation.
Moreover, the shunt capacitors in the network could largely atten-
uate the TWs. It should be noted that the TW based techniques Fig. 3. HIF model proposed in [63].
require very high bandwidth transducers and signal generators
which make them economically less attractive.
In addition, there have been several works on the location of HIF where g is the instantaneous conductance of the arc,  is a time
using communication based techniques [100,132,133]. It is noted constant, and G is stationary conductance of the arc calculated as
that the realization of the communication based techniques results follows:
in additional hardware and software complexity.
|i(t)|
G(t) = (4)
(u0 + R|i(t)|)l
6. High impedance fault modeling techniques
where i(t) is the arc instantaneous current, L is the arc length, R is
resistive component per length of the arc, and u0 is the constant
There have been decades of experience on the accurate model-
voltage parameter per length of the arc.
ing of the arc, which is required for the simulation of the HIF and
The methods in this class are capable of modeling the ran-
development of new detection techniques [134–136]. Arc modeling
dom nature of the HIF, through random variables, such as u0 for
falls into three categories; first group uses empirical data to find the
the aforementioned example. However, their weak point is the
equivalent resistance of the arc, as a function of several parameters
assumption of constant arc length which could be erroneous in the
such as arc length and arc current [137,138]. For example, in [138]
case of HIFs from bouncing down conductors.
the resistance of the arc is calculated using the following equation:
The third group of models uses a combination of electrical ele-
√ ments, such as voltage sources, diodes, resistances, and inductances
L to mimic the HIF characteristics [140]. For example, in [63] the HIF
Rarc = 25 (2)
Iarc 0.85 is modeled using the circuit diagram shown in Fig. 3. In this circuit,
the value of voltage sources (Vp , Vn ) and series resistances (Rp , Rn )
are random variables used to model the random nature and asym-
where Rarc is the arc equivalent resistance, L is the arc length, and
metry of the HIF. The disadvantage of these models is that they
Iarc is the arc current.
should be empirically verified for different types of HIF.
Although the techniques in this class are simple and easy to
solve, they cannot model the random nature of the HIF.
In the second group of the HIF models, the differential equa- 7. Discussion
tions are utilized to calculate the equivalent conductance of the
arc [16,105,139]. For example, in [105], the following differential High impedance fault and its definition, characteristics, detec-
equation is used: tion techniques, location technique, and modeling methods have
been reviewed. In this section, other aspects of the literatures on
dg G(t) − g(t) the HIF are discussed, namely, the popularity of the HIF field, HIF
= (3)
dt  security, and comparison of the existing HIF detection techniques.
382 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

Fig. 4. (a) normalized yearly average citations, (b) number of articles published in the periodicals.

7.1. Popularity of the HIF field 8. Future of the high impedance fault detection techniques

Two indices are utilized to demonstrate the popularity of the HIF It is desirable to shed light on the possible road ahead of the HIF
field; (1) the average number of citation normalized to the num- field. The possible future of the field is outlined in this section based
ber of years the papers have been published, which is shown in on existing experiences and emerging technologies.
Fig. 4(a), and (2) the number of articles published in the periodi-
cals, shown in Fig. 4(b). Note that the articles before and in 1994 8.1. Increase in the processing power
are lumped at 1994. These figures show that the HIF, as an ongo-
ing problem still draws remarkable attention from power system The advancement in the information technology and the
researchers. increase in the computing power have been shifting the power
system protection field toward a modern era [205]. High speed
microprocessors pave the road toward fast and reliable digital
7.2. Distinguishing HIF from other power system disturbances relays. With such high computational power, it is possible to imple-
ment HIF detection techniques with higher complexity. Therefore,
HIF detection techniques are only responsible for the protec- it is possible for the detection techniques and move toward more
tion from the HIF, and therefore, they should stay inactive when complex analysis domains and classifiers [206].
any other power system disturbances occur (the security crite-
rion in Table 1). For any new proposed technique, this security 8.2. Growth of the information technology
should be verified against the disturbances that can be confused
with the HIF. These events share at least one characteristic with In addition, the blossoming of the information technology in
the HIF: (a) non-linear loads share the non-linearity characteristic power systems with the introduction of GPS-coordinated phasor
[83,90], (b) line switching, load switching, and capacitor switching measurement unit (PMUs) has led to a higher degree of free-
share the high-frequency characteristic [45,46,83], and (c) trans- dom in monitoring, protection, and control of the power systems
former inrush current and CT saturation share the low-frequency [207,208]. Utilizing this technology could significantly improve the
and asymmetry characteristic [46,90]. decision-making for complex events, such as HIF, by providing
redundant insights from different part of the network.
7.3. Comparison of the proposed HIF detection techniques
8.3. Publishing HIF detection techniques
In Tables 2 and 3, the existing techniques are compared using
some of the criteria introduced in Section 4. Unfortunately, large Most of the published HIF detection techniques miss an essential
part of the literature did not provide the information required to part: quantitative and qualitative evaluation. It is highly suggested
extract all of the criteria. Some observation could be made from that the new publications on HIF detection – and in general any
these tables as follows: fault detection – report their reliability using criteria specifically
designed for their evaluation (such as criteria mentioned in Table 1).
Moreover, it is desired for HIF detection techniques to report
1. Wavelet Transform dominates the analysis domain. However that against which disturbances they verified their security (as
other domains such as time-frequency domain have been gain- discussed in Section 7.2). And finally, for the articles that use real-
ing some interest in recent years. world data to evaluate and verify their methods, it is desirable to
2. The methods that utilize simulation, in general, demonstrate report the surfaces and conditions that are used to perform the
higher reliability compared to methods based on real-world tests.
data. This reveals the necessity for more practical models, so that
the results from real-world- and simulation-based techniques 8.4. Practical data acquisition
are more compatible.
3. In general, the security of the methods is smaller than other crite- It is discussed in [121] that the detection techniques that are
ria. This suggests that, although the existing techniques detect either based on simulation or the real-world data acquired with the
most of the HIFs, the rate of unnecessary trips decreases their Hall-effect transformers lose their efficiency in real-world imple-
overall reliability. mentation. This is because of two erroneous assumptions: (1) the
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 383

Table 2
Classification and comparison of the existing techniques: time domain and frequency domain techniques.

Ref. Analysis Measurement Classifier A (%) D (%) S (%) v Year Data Network Article
domain
[104] Artificial neural network 98.3 96.3 2011 Real-world Distribution Journal
[141] 1982 Real-world Distribution Journal
[142] Threshold 1985 Real-world Distribution Journal
[60] Current 2010 Simulation Distribution Conference
[94] 2014 Real-world Distribution Conference
[143] Time domain 1992 Real-world Distribution Journal
[36] Threshold and counter 0.50 1994 Real-world Distribution Journal
[62] 1995 Real-world Distribution Conference
[144] Current and 2003 Simulation Low voltage Journal
Threshold
voltage
[61] 2012 Simulation Distribution Journal
[59] Threshold and counter 2007 Real-world Distribution Conference
[57] Magnetic field Genetic algorithm 2013 Real-world Distribution Journal
[63] Voltage Threshold and counter 2013 Simulation Distribution Journal
[109] 99.8 99.0 100 0.01 2001 Simulation Distribution Conference
Artificial neural network
[145] 98.0 2.00 2009 Simulation Distribution Journal
[66] 0.50 2006 Simulation Distribution Conference
Decision tree
[71] 2010 Simulation Distribution Journal
[70] 99.7 99.5 0.50 2012 Simulation Distribution Journal
[68] 81.0 1999 Real-world Distribution Journal
[146] 80.0 83.0 77.0 2008 Simulation Distribution Conference
[147] 2011 Simulation Distribution Conference
Fuzzy classifier
[148] 2011 Simulation Distribution Conference
[47] 2012 Simulation Distribution Journal
[113] 2012 Simulation Distribution Journal
[149] Genetic algorithm 2007 Simulation Distribution Conference
[116] Support vector machine 99.0 98.0 100 2010 Simulation Distribution Journal
Current
[150] 1982 Real-world Distribution Journal
[67] 1987 Real-world Distribution Journal
[43] 1989 Real-world Distribution Journal
[12] 1990 Real-world Distribution Journal
[151] 1990 Real-world Distribution Journal
[152] 1990 Simulation Distribution Journal
[153] 1991 Real-world Distribution Conference
[154] 1993 Real-world Low voltage Journal
Frequency
[155] Threshold 1994 Real-world Distribution Journal
domain
[27] 1995 Real-world Distribution Journal
[156] 95.0 1996 Simulation Distribution Conference
[157] 2005 Simulation Distribution Conference
[158] 2007 Real-world Distribution Journal
[159] 2009 Real-world Low voltage Conference
[95] 2010 Real-world Distribution Journal
[93] 2014 Real-world Distribution Journal
[160] 1991 Real-world Distribution Journal
[17] Threshold and counter 84.4 2000 Real-world Distribution Conference
[96] 0.50 2014 Real-world Distribution Conference
[118] Z-test classifier 0.02 1986 Simulation Distribution Journal
[161] 1993 Simulation Distribution Conference
[162] 95.3 2003 Simulation Distribution Conference
[106] 2005 Simulation Distribution Journal
[163] 2006 Simulation Distribution Conference
[164] 1999 Simulation Distribution Conference
Current and
[65] Artificial neural network 2012 Simulation Distribution Conference
voltage
[165] 1995 Simulation Distribution Conference
[110] 1996 Simulation Distribution Journal
[69] 1998 Simulation Distribution Journal
[166] 1991 Real-world Distribution Journal
[167] 1994 Real-world Distribution Conference
Threshold
[168] 2004 Simulation Distribution Conference
[169] Threshold and counter 1989 Simulation Distribution Journal
[112] Fuzzy classifier 98.8 2012 Simulation Distribution Journal
[129] 2010 Simulation Transmission Journal
[55] 2013 Simulation Distribution Conference
Multiple Threshold
[56] 2006 Simulation Transmission Journal
measurements
[170] 2013 Simulation Distribution Conference
[132] 99.9 99.5 99.5 2004 Simulation Transmission Journal
Threshold and counter
[100] 99.9 99.5 99.5 2004 Simulation Transmission Journal
[171] Threshold 2008 Simulation Distribution Conference
Voltage
[172] Threshold and counter 2013 Simulation Distribution Conference
384 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

Table 3
Classification and comparison of the existing techniques continued: wavelet domain, hybrid domain, and time-frequency domain techniques.

Ref. Analysis Measurement Classifier A (%) D (%) S (%) v Year Data Network Article
domain
[173] 1990 Simulation Distribution Journal
[174] 1997 Simulation Distribution Conference
[107] 2005 Real-world Distribution Journal
[175] 2005 Simulation Distribution Conference
Artificial neural network
[176] 2005 Simulation Distribution Conference
[177] 100 2006 Real-world Distribution Conference
[76] 99.7 99.5 0.50 2011 Simulation Distribution Journal
[178] 2011 Simulation Distribution Journal
[48] 97.5 95.0 100 2012 Simulation Distribution Journal
[46] 96.1 98.5 93.5 0.40 2014 Simulation Distribution Journal
[13] Bayes classifier 2008 Real-world Distribution Journal
[72] Fuzzy classifier 2014 Simulation Distribution Journal
[21] 98.3 95.0 100 2005 Real-world Distribution Journal
Genetic algorithm
[45] Current 98.4 2014 Simulation Distribution Journal
[179] 2008 Simulation Distribution Conference
[58] Support vector machine 99.6 99.8 0.66 2011 Real-world Distribution Journal
[114] 94.9 90.0 90.9 0.11 2014 Simulation Distribution Conference
[9] 2005 Simulation Low-voltage Conference
[74] 93.2 2007 Simulation Distribution Journal
[180] 1.00 2011 Real-world Distribution Conference
[181] 1.00 2011 Real-world Distribution Conference
[182] Wavelet domain Threshold 2012 Simulation Distribution Conference
[183] 1.00 2013 Simulation Distribution Conference
[184] 2014 Real-world Distribution Conference
[185] 100 100 100 2014 Simulation Distribution Conference
[92] 2014 Simulation Distribution Conference
[19] 2014 Simulation Distribution Conference
[186] 2014 Real-world Distribution Conference
[187] 1998 Simulation Distribution Journal
[188] 2001 Simulation Distribution Journal
[189] 2002 Simulation Distribution Journal
Threshold and counter
[81] 0.03 2004 Real-world Distribution Journal
[99] 2009 Simulation High-voltage Journal
[190] 0.50 2010 Simulation Distribution Journal
[97] 2011 Real-world Distribution Conference
[191] 2004 Simulation Distribution Conference
[105] Artificial neural network 2006 Simulation Distribution Journal
[108] 2007 Simulation Distribution Journal
[82] Current and voltage 1999 Real-world Distribution Journal
[192] Improved data conversion 2006 Simulation Distribution Journal
[7] Rule based classifier 2008 Simulation Transmission Journal
[75] 2011 Real-world Transmission Journal
[193] Threshold 2011 Simulation Distribution Conference
[98] 2012 Simulation Transmission Conference
[194] 2007 Real-world Distribution Conference
Multiple
[195] Threshold 2007 Real-world Distribution Journal
measurement
[15] 1.00 2008 Real-world Distribution Journal
[37] 2013 Simulation Distribution Conference
Traveling wave
[196] Threshold 2013 Simulation Distribution Conference
[197] 2007 Real-world Distribution Conference
[51] Threshold 2014 Simulation Distribution Journal
[198] Voltage 2.00 2008 Simulation Transmission Conference
[199] Threshold and counter 2.00 2009 Simulation Transmission Conference
[200] 2.00 2010 Simulation Transmission Journal
[50] 0.50 2014 Simulation Distribution Journal
[90] Fuzzy classifier 99.5 99.0 99.8 0.20 2008 Simulation Distribution Journal
[201] Current Threshold 2008 Real-world Distribution Conference
Hybrid
[202] Artificial neural network 1.00 1990 Simulation Distribution Journal
domain
[203] Threshold and counter 2007 Real-world Distribution Conference
[117] Knowledge based classifier 1990 Real-world Distribution Journal
Current and voltage
[11] Threshold 1990 Simulation Distribution Journal
[204] 99.2 99.1 99.1 1.00 2013 Real-world Distribution Conference
Artificial neural network
[84] Time-frequency 98.0 2.00 2008 Simulation Distribution Journal
Current
domain
[22] Support vector machine 93.6 100 81.5 1.00 2015 Real-world Distribution Journal
[85] Threshold 2013 Simulation Distribution Conference

real-world installed measurement devices can pass high frequency high frequency components of the current in a way that it would
components (similar to the Hall-effect transformers), (2) there is not be measured even with Hall-effect transformers. Two solutions
no environmental noise in the vicinity of measurement devices are possible for this problem: first, development of HIF models
[209,210]. Moreover, in [67], it is mentioned that the capacitor that take into account the existence of noise and non-ideality of
bank throughout the distribution network would attenuate the the measurement devices, and second, development of techniques
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 385

based on practical measurement devices with magnetic cores and [22] A. Ghaderi, H. Mohammadpour, H. Ginn III, Y. Shin, High impedance fault
existence of environmental noise. detection in distribution network using time-frequency based algorithm,
IEEE Trans. Power Deliv. PP (99) (2014) 1.
[23] M. Aucoin, Status of high impedance fault detection, IEEE Trans. Power
9. Conclusion Appar. Syst. PAS-104 (3) (1985) 637–644.
[24] J. Barnard, A. Pahwa, Determination of the impacts of high impedance faults
on protection of power distribution systems using a probabilistic model,
This paper performs a comprehensive survey on the literature Electr. Power Syst. Res. 28 (1) (1993) 11–18.
of the high impedance fault, with the emphasis on its detection [25] A. Hamel, A. Gaudreau, M. Cote, Intermittent arcing fault on underground
low-voltage cables, IEEE Trans. Power Deliv. 19 (4) (2004) 1862–1868.
techniques, and including HIF physics, modeling, and location tech-
[26] M. Adamiak, C. Wester, M. Thakur, C. Jensen, High impedance fault detection
niques. Moreover, HIF detection techniques are categorized based on distribution feeders, GE Industrial Solution (2006).
on dividing each techniques into three steps. Additionally, a com- [27] B. Russell, C. Benner, Arcing fault detection for distribution feeders: security
prehensive set of criteria is utilized to compare the existing HIF assessment in long term field trials, IEEE Trans. Power Deliv. 10 (2) (1995)
676–683.
detection techniques. Finally, a possible road to the future of the [28] B. Russell, R. Chinchali, C. Kim, Behaviour of low frequency spectra during
HIF field is proposed. arcing fault and switching events, IEEE Trans. Power Deliv. 3 (4) (1988)
1485–1492.
[29] R. Paththamperuma, I. Perera, K. Perera, C. Perera, N. De Silva, U. Javatunga,
References High Impedance Arcing Fault Detection in Low Voltage Distribution
Network, Digital Library of University of Moratuwa, 2013, pp. 1–6.
[1] C.-L. Huang, H.-Y. Chu, M.-T. Chen, Algorithm comparison for high [30] N.I. Elkalashy, M. Lehtonen, H.A. Darwish, A.-M.I. Taalab, M.A. Izzularab, A
impedance fault detection based on staged fault test, IEEE Trans. Power novel selectivity technique for high impedance arcing fault detection in
Deliv. 3 (4) (1988) 1427–1435. compensated MV networks, Eur. Trans. Electr. Power 18 (4) (2008) 344–363.
[2] H.-Y. Chu, M.-T. Chen, C.-L. Huang, S. lin Chen, S.-S. Yen, High impedance [31] T. Dalstein, B. Kulicke, Neural network approach to fault classification for
fault tests on the Taipower primary distribution system, Electr. Power Syst. high speed protective relaying, IEEE Trans. Power Deliv. 10 (2) (1995)
Res. 19 (2) (1990) 105–114. 1002–1011.
[3] C.G. Wester, High impedance fault detection on distribution systems, in: [32] N. Zamanan, J. Sykulski, Modelling arcing high impedances faults in relation
Rural Electric Power Conference, 1998, pp. 1–5. to the physical processes in the electric arc, in: 6th WSEAS International
[4] C. Prasad, N. Srinivasu, D. Prasad, M. Saiveerraju, Reliability of different fault Conference on Power Systems, 22–24 September 2006, 2006, pp. 28–33.
detection algorithms under high impedance faults, in: 2013 International [33] B. Aucoin, R. Jones, High impedance fault detection implementation issues,
Conference on Advanced Computing and Communication Systems (ICACCS), IEEE Trans. Power Deliv. 11 (1) (1996) 139–148.
2013, pp. 1–5. [34] A.V. Masa, J.-C. Maun, W. Werben, Characterization of high impedance faults
[5] L. Li, M. Redfern, A review of techniques to detect downed conductors in in solidly grounded distribution networks, in: 17th Power Systems
overhead distribution systems, in: Seventh International Conference on Computation Conference, 2011, pp. 1–6.
(IEE) Developments in Power System Protection, 2001, 2001, pp. 169–172. [35] F.B. Costa, B.A. Souza, N.S.D. Brito, J.A.C.B. Silva, W.C. Santos, Real-time
[6] J. Tengli, et al., High impedance fault detection technology, Rep. PSRC Work. detection of transients induced by high-impedance faults based on the
Group D 15 (1996) 1–12. boundary wavelet transform, IEEE Trans. Ind. Appl. 51 (6) (2015)
[7] T. NengLing, C. JiaJia, Wavelet-based approach for high impedance fault 5312–5323.
detection of high voltage transmission line, Eur. Trans. Electr. Power 18 (1) [36] A. Sultan, G. Swift, D. Fedirchuk, Detecting arcing downed-wires using fault
(2008) 79–92. current flicker and half-cycle asymmetry, IEEE Trans. Power Deliv. 9 (1)
[8] A. Thakallapelli, V. Jijabai, Detection of high impedance faults by distance (1994) 461–470.
relays using prony method, Int. J. Adv. Technol. Eng. Res. 2 (2) (2012) 39–45. [37] W.C. Santos, F.V. Lopes, N.S.D. Brito, B.A.D.F. Souza Jr., W.L.A. Neves, High
[9] W.-J. Li, Y.-C. Li, Arc fault detection based on wavelet packet, in: Proceedings impedance fault detection and location based on electromagnetic transient
of 2005 International Conference on Machine Learning and Cybernetics, analysis, in: International Conference on Power Systems Transients
2005, vol. 3, 2005, pp. 1783–1788. (IPST2013), 2013, pp. 1–6.
[10] C. Wagner, L. Fountain, Arcing fault currents in low-voltage a-c circuits, [38] G. Parise, U. Grasselli, V. Di Luozzo, Arcing fault in sub-distribution
Trans. Am. Inst. Electr. Eng. 67 (1) (1948) 166–174. branch-circuits, IEEE Trans. Power Deliv. 8 (2) (1993) 580–583.
[11] C. Kim, B. Russell, K. Watson, A parameter-based process for selecting high [39] E. Acha, A. Semlyen, N. Rajakovic, A harmonic domain computational
impedance fault detection techniques using decision making under package for nonlinear problems and its application to electric arcs, IEEE
incomplete knowledge, IEEE Trans. Power Deliv. 5 (3) (1990) 1314–1320. Trans. Power Deliv. 5 (3) (1990) 1390–1397.
[12] A. Emanuel, D. Cyganski, J. Orr, S. Shiller, E. Gulachenski, High impedance [40] V.T.H.F.R.S. Maximov, J.L. Guardado, High impedance fault location
fault arcing on sandy soil in 15 kV distribution feeders: contributions to the formulation: a least square estimator based approach, Math. Problems Eng.
evaluation of the low frequency spectrum, IEEE Trans. Power Deliv. 5 (2) (2014) 1–10.
(1990) 676–686. [41] T. Lai, L. Snider, E. Lo, D. Sutanto, High-impedance fault detection using
[13] A.-R. Sedighi, M. Haghifam, O. Malik, M.-H. Ghassemian, High impedance discrete wavelet transform and frequency range and RMS conversion, IEEE
fault detection based on wavelet transform and statistical pattern Trans. Power Deliv. 20 (1) (2005) 397–407.
recognition, IEEE Trans. Power Deliv. 20 (4) (2005) 2414–2421. [42] T. Lai, L. Snider, E. Lo, Wavelet transform based relay algorithm for the
[14] J.R. Macedo, J.W. Resende, C.A. Bissochi, D. Carvalho, F.C. Castro, Proposition detection of stochastic high impedance faults, Electr. Power Syst. Res. 76 (8)
of an interharmonic-based methodology for high-impedance fault detection (2006) 626–633, selected Topics in Power System Transients.
in distribution systems, IET Gener. Transm. Distrib. 9 (16) (2015) 2593–2601. [43] C. Benner, P. Carswell, B.D. Russell, Improved algorithm for detecting arcing
[15] N. Elkalashy, M. Lehtonen, H. Darwish, A.-M. Taalab, M. Izzularab, faults using random fault behavior, Electr. Power Syst. Res. 17 (1) (1989)
DWT-based detection and transient power direction-based location of 49–56.
high-impedance faults due to leaning trees in unearthed MV networks, IEEE [44] D.G.S.R.O. Duda, P.E. Hart, Pattern Classification, Wiley-Interscience, 2000.
Trans. Power Deliv. 23 (1) (2008) 94–101. [45] Y. Hong, W.-S. Huang, Locating high-impedance fault section in electric
[16] N. Elkalashy, M. Lehtonen, H. Darwish, M. Izzularab, A.-M. Taalab, Modeling power systems using wavelet transform, k-means, genetic algorithms, and
and experimental verification of high impedance arcing fault in medium support vector machine, Math. Problems Eng. (2014) 1–9.
voltage networks, IEEE Trans. Dielectr. Electr. Insul. 14 (2) (2007) 375–383. [46] Z. Moravej, S.H. Mortazavi, S.M. Shahrtash, DT-CWT based event feature
[17] A. Lazkano, J. Ruiz, E. Aramendi, L. Leturiondo, J. Gonzalez, Study of high extraction for high impedance faults detection in distribution system, Int.
impedance fault detection in Levante area in Spain, in: Ninth International Trans. Electr. Energy Syst. 25 (12) (2015) 3288–3303.
Conference on Harmonics and Quality of Power, 2000. Proceedings, vol. 3, [47] M.S.A. Aziz, M.A.M. Hassan, E.A. Zahab, High-impedance faults analysis in
2000, pp. 1011–1016. distribution networks using an adaptive neuro fuzzy inference system,
[18] M. Carpenter, R. Hoad, T. Bruton, R. Das, S. Kunsman, J. Peterson, Electr. Power Compon. Syst. 40 (11) (2012) 1300–1318.
Staged-fault testing for high impedance fault data collection, in: 2005 58th [48] A. Abohagar, M. Wazir, N. Algeelani, Hybrid algorithm for detection of high
Annual Conference for Protective Relay Engineers, 2005, pp. 9–17. impedance arcing fault in overhead transmission system, Int. J. Electron.
[19] A.V. Masa, F.B. Costa, S. Werben, B.A. Souza, Overcoming limitations of the Electr. Eng. (2012) 43–60.
present protection technology for smart grid: innovative approaches to [49] E. Sortomme, S. Venkata, J. Mitra, Microgrid protection using
detect high impedance faults, in: T D Conference and Exposition, 2014 IEEE communication-assisted digital relays, IEEE Trans. Power Deliv. 25 (4)
PES, 2014, pp. 1–5. (2010) 2789–2796.
[20] J.-H. Ko, J.-C. Shim, C.-W. Ryu, C.-G. Park, W.-Y. Yim, Detection of high [50] A. Bakar, M. Ali, C. Tan, H. Mokhlis, H. Arof, H. Illias, High impedance fault
impedance faults using neural nets and chaotic degree, in: 1998 location in 11kv underground distribution systems using wavelet
International Conference on Energy Management and Power Delivery, 1998. transforms, Int. J. Electr. Power Energy Syst. 55 (0) (2014) 723–730.
Proceedings of EMPD ’98, vol. 2, 1998, pp. 399–404. [51] M.S. Ali, A.H. Abu Bakar, H. Mokhlis, H. Arof, H. Azil Illias, High-impedance
[21] A.-R. Sedighi, M.-R. Haghifam, O. Malik, Soft computing applications in high fault location using matching technique and wavelet transform for
impedance fault detection in distribution systems, Electr. Power Syst. Res. underground cable distribution network, IEEJ Trans. Electr. Electron. Eng. 9
76 (1–3) (2005) 136–144. (2) (2014) 176–182.
386 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

[52] M. Thomas, N. Bhaskar, A. Prakash, Voltage based detection method for high [81] A. Lazkano, J. Ruiz, E. Aramendi, L. Leturiondo, Evaluation of a new proposal
impedance fault in a distribution system, J. Inst. Eng. (India): Ser. B (2015) for arcing fault detection method based on wavelet packet analysis, in:
1–11. Power Engineering Society Summer Meeting, 2001, vol. 3, 2001, pp.
[53] A. Milioudis, G. Andreou, D. Labridis, Enhanced protection scheme for smart 1328–1333.
grids using power line communications techniques. Part II: Location of high [82] S.-J. Huang, C.-T. Hsieh, High-impedance fault detection utilizing a Morlet
impedance fault position, IEEE Trans. Smart Grid 3 (4) (2012) 1631–1640. wavelet transform approach, IEEE Trans. Power Deliv. 14 (4) (1999)
[54] A. Milioudis, G. Andreou, D. Labridis, Enhanced protection scheme for smart 1401–1410.
grids using power line communications techniques. Part I: Detection of high [83] B. Vahidi, N. Ghaffarzadeh, S. Hosseinian, S. Ahadi, An approach to detection
impedance fault occurrence, IEEE Trans. Smart Grid 3 (4) (2012) of high impedance fault using discrete wavelet transform and artificial
1621–1630. neural networks, Soc. Model. Simul. Int. 86 (4) (2010) 203–215.
[55] M. Zanjani, H. Kargar, M. Zanjani, High impedance fault detection of [84] S. Samantaray, K. Panigrahi, P. Dash, High impedance fault detection in
distribution network by phasor measurement units, in: 2012 Proceedings of power distribution networks using time-frequency transform and
17th Conference on Electrical Power Distribution Networks (EPDC), 2012, probabilistic neural network, Gener. Transm. Distrib. IET 2 (2) (2008)
pp. 1–5. 261–270.
[56] C. Lee, J. Park, J. Shin, Z. Radojevie, A new two-terminal numerical algorithm [85] J. Pan, H. Zhu, L. Chen, Time-frequency ridge analysis using on distinguish
for fault location, distance protection, and arcing fault recognition, IEEE regular signal from arc-fault signal, in: 2013 IEEE International Workshop
Trans. Power Syst. 21 (3) (2006) 1460–1462. on Intelligent Energy Systems (IWIES), 2013, pp. 203–208.
[57] M. Sarlak, S. Shahrtash, High-impedance faulted branch identification using [86] J.-Y. Cheng, S.-J. Huang, C.-T. Hsieh, Application of Gabor–Wigner transform
magnetic-field signature analysis, IEEE Trans. Power Deliv. 28 (1) (2013) to inspect high-impedance fault-generated signals, Int. J. Electr. Power
67–74. Energy Syst. 73 (2015) 192–199.
[58] M. Sarlak, S. Shahrtash, SVM based method for high impedance faults [87] L. Cohen, Time-frequency distributions—a review, Proc. IEEE 77 (7) (1989)
detection in distribution networks, COMPEL 30 (2) (2011) 431–450. 941–981.
[59] D. Hou, N. Fischer, Deterministic high-impedance fault detection and phase [88] Y. Sheng, S. Rovnyak, Decision tree-based methodology for high impedance
selection on ungrounded distribution systems, in: Industrial Commercial fault detection, IEEE Trans. Power Deliv. 19 (2) (2004) 533–536.
Power Systems Technical Conference, 2007. ICPS 2007. IEEE/IAS, 2007, pp. [89] J. Lee, C. Jung, I. Kim, Y. Baek, Protective relay testing and characteristic
1–10. analysis for high impedance faults in transmission lines, Electr. Power Syst.
[60] A. Siadatan, H. Karegar, V. Najmi, New high impedance fault detection, in: Res. 56 (2) (2000) 129–137.
2010 IEEE International Conference on Power and Energy (PECon), 2010, pp. [90] A. Etemadi, M. Sanaye-Pasand, High-impedance fault detection using
573–576. multi-resolution signal decomposition and adaptive neural fuzzy inference
[61] N. Faridnia, H. Samet, B. Doostani Dezfuli, A new approach to high system, Gener. Transm. Distrib. IET 2 (1) (2008) 110–118.
impedance fault detection based on correlation functions Artificial [91] W. Charytoniuk, W.-J. Lee, M.-S. Chen, J. Cultrera, T. Maffetone, Arcing fault
Intelligence Applications and Innovations, vol. 381, 2012, pp. 453–462. detection in underground distribution networks-feasibility study, IEEE
[62] A. Mamishev, B. Russell, C. Benner, Analysis of high impedance faults using Trans. Ind. Appl. 36 (6) (2000) 1756–1761.
fractal techniques, IEEE Trans. Power Syst. 11 (1) (1996) 435–440. [92] H. Wu, B. Phung, D. Zhang, J. Chen, Modeling and detection of high
[63] S. Gautam, S. Brahma, Detection of high impedance fault in power impedance faults, in: 2014 International Conference on Smart Green
distribution systems using mathematical morphology, IEEE Trans. Power Technology in Electrical and Information Systems (ICSGTEIS), 2014, pp.
Syst. 28 (2) (2013) 1226–1234. 88–93.
[64] H. Li, E. Southern, P. Crossley, S. Potts, S. Pickering, B. Caunce, G. Weller, A [93] V. Torres, J. Guardado, H. Ruiz, S. Maximov, Modeling and detection of high
new type of differential feeder protection relay using the global positioning impedance faults, Int. J. Electr. Power Energy Syst. 61 (0) (2014)
system for data synchronization, IEEE Trans. Power Deliv. 12 (3) (1997) 163–172.
1090–1099. [94] P. Gonzalez, E. Romero, V. Minambres, M. Guerrero, Experimental tests of
[65] B. Vahidi, N. Ghaffarzadeh, S.H. Hosseinian, S.M. Ahadi, An approach to high impedance faults in MV rural distribution network, in: 2014
detection of high impedance fault using discrete wavelet transform and International Conference on Optimization of Electrical and Electronic
artificial neural networks, Simulation 86 (4) (2010) 203–215. Equipment (OPTIM), 2014, pp. 1–10.
[66] S.M. Shahrtash, M. Sarlak, High impedance fault detection using harmonics [95] M. Sarlak, S.M. Shahrtash, D.A. Khaburi, Design and implementation of a
energy decision tree algorithm, in: 2006 International Conference on Power systematically tunable high impedance fault relay, ISA Trans. 49 (3) (2010)
System Technology, 2006, pp. 1–5. 358–368.
[67] M. Aucoin, B. Russell, Detection of distribution high impedance faults using [96] H.-G. Yeh, D.H. Tran, R. Yinger, High impedance fault detection using
burst noise signals near 60 Hz, IEEE Trans. Power Deliv. 2 (2) (1987) orthogonal transforms, in: Green Energy and Systems Conference (IGESC).
342–348. 2014 IEEE, 2014, pp. 67–72.
[68] P.R. Jota, F.G. Jota, Fuzzy detection of high impedance faults in radial [97] J. da Silva, F. Costa, G. Lira, W. Santos, W. Neves, B. Souza, N. Brito, M. Paes,
distribution feeders, Electr. Power Syst. Res. 49 (3) (1999) 169–174. High impedance fault location – case study with developed models from
[69] L. Snider, Y. Yuen, The artificial neural-networks-based relay algorithm for field experiments, in: 21st International Conference on Electricity
the detection of stochastic high impedance faults, Neurocomputing 23 (1-3) Distribution, 2011, pp. 1–4.
(1998) 243–254. [98] A.S.H. Mahari, High impedance fault detection and discrimination using a
[70] S. Samantaray, Ensemble decision trees for high impedance fault detection wavelet packet transform-based algorithm, in: 6th Power Systems
in power distribution network, Int. J. Electr. Power Energy Syst. 43 (1) Protection and Control Conference, 2012, pp. 1–6.
(2012) 1048–1055. [99] G.T.R.D.M. Sushama, A.J. Laxmi, Detection of high-impedance faults in
[71] S.R. Samantaray, P.K. Dash, High impedance fault detection in distribution transmission lines using wavelet transform, ARPN J. Eng. Appl. Sci. 4 (3)
feeders using extended Kalman filter and support vector machine, Eur. (2009) 6–12.
Trans. Electr. Power 20 (3) (2010) 382–393. [100] Y.-H. Lin, C.-W. Liu, C.-S. Chen, A new PMU-based fault detection/location
[72] M. Banejad, H. Ijadi, High impedance fault detection discrete wavelet technique for transmission lines with consideration of arcing fault
transform and fuzzy function approximation, J. Artif. Intell. Data Min. 2 (2) discrimination—Part II: Performance evaluation, IEEE Trans. Power Deliv. 19
(2014) 149–158. (4) (2004) 1594–1601.
[73] N.R. Varma, D.B.V.S. Ram, D.K.S.R. Anjaneyulu, Development of fault [101] V.S.S. Vankayala, N.D. Rao, Artificial neural networks and their applications
detection algorithm for high impedance faults in distribution network using to power systems—a bibliographical survey, Electr. Power Syst. Res. 28 (1)
multiresolution analysis, Int. J. Eng. Res. Technol. 3 (9) (2014) 573–576. (1993) 67–79.
[74] M. Yang, J.G.J. Gu, High impedance faults detection technique based on [102] A. Jain, R. Duin, J. Mao, Statistical pattern recognition: a review, IEEE Trans.
wavelet transform, Int. J. Electr. Comput. Syst. Eng. 1 (3) (2007) 143–147. Pattern Anal. Mach. Intell. 22 (1) (2000) 4–37.
[75] D.K. Ibrahim, S.M. Saleh, Unsymmetrical high-impedance earth fault central [103] A. Ghosh, D. Lubkeman, The classification of power system disturbance
relay for transmission networks, Electr. Power Compon. Syst. 39 (2011) waveforms using a neural network approach, IEEE Trans. Power Deliv. 10 (1)
1469–1492, n/a. (1995) 109–115.
[76] I. Baqui, I. Zamora, J. Mazón, G. Buigues, High impedance fault detection [104] M. Sarlak, S. Shahrtash, High impedance fault detection using combination
methodology using wavelet transform and artificial neural networks, Electr. of multi-layer perceptron neural networks based on multi-resolution
Power Syst. Res. 81 (7) (2011) 1325–1333. morphological gradient features of current waveform, Gener. Transm.
[77] P. Biradar, V.R. Sheelvant, High-impedance fault detection using wavelet Distrib. IET 5 (5) (2011) 588–595.
transform, Int. J. Eng. Res. Gen. Sci. (2015) 166–173. [105] M. Michalik, W. Rebizant, M. Lukowicz, S.-J. Lee, S.-H. Kang, High-impedance
[78] A. Mahari, H. Seyedi, High impedance fault protection in transmission lines fault detection in distribution networks with use of wavelet-based
using a WPT-based algorithm, Int. J. Electr. Power Energy Syst. 67 (2015) algorithm, IEEE Trans. Power Deliv. 21 (4) (2006) 1793–1802.
537–545. [106] H.K. Zadeh, An ANN-based high impedance fault detection scheme: design
[79] P. Qi, J. Lezama, S. Jovanovic, P. Schweitzer, Adaptive real-time DWT-based and implementation, Int. J. Emerg. Electr. Power Syst. 4 (2) (2005) 1–14.
method for arc fault detection, in: Proceedings of ICEC 2014. The 27th [107] M. Yang, J. Gu, Detecting high impedance faults utilizing combined phase
International Conference on Electrical Contacts, 2014, pp. 1–6. voltages with neutral line current, Int. J. Emerg. Electr. Power Syst. 2 (2)
[80] M. Sulaiman, A. Tawfan, Z. Ibrahim, Using probabilistic neural network for (2005) 1–21.
classification high impedance faults on power distribution feeders, World [108] M. Michalik, M. Lukowicz, W. Rebizant, S.-J. Lee, S.-H. Kang, Verification
Appl. Sci. J. 23 (10) (2013) 1274–1283. of the wavelet-based HIF detecting algorithm performance in solidly
 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388 387

grounded MV networks, IEEE Trans. Power Deliv. 22 (4) (2007) [138] L.E. Fisher, Resistance of low-voltage ac arcs, IEEE Trans. Ind. Gen. Appl.
2057–2064. IGA-6 (6) (1970) 607–616.
[109] R. Keyhani, M. Deriche, E. Palmer, A high impedance fault detector using a [139] V. Torres, S. Maximov, H.F. Ruiz, J.L. Guardado, Distributed parameters
neural network and subband decomposition, in: Sixth International model for high-impedance fault detection and localization in transmission
Symposium on Signal Processing and its Applications, 2001, vol. 2, 2001, pp. lines, Electr. Power Compon. Syst. 41 (14) (2013) 1311–1333.
458–461. [140] A. Sedighi, A new model for high impedance fault in electrical distribution
[110] F. Jota, P. Jota, High-impedance fault identification using a fuzzy reasoning systems, Int. J. Sci. Res. Comput. Sci. Eng. 2 (4) (2014) 6–12.
system, Transm. Distrib. IEE Proc. Gener. 145 (6) (1998) 656–661. [141] H. Calhoun, M.T. Bishop, C.H. Eichler, R. Lee, Development and testing of an
[111] J. Garcia, V. Vega Garcia, N. Kagan, Detection of high impedance faults in electro-mechanical relay to detect fallen distribution conductors, Power
overhead multi grounded networks, in: 2014 11th IEEE/IAS International Eng. Rev. IEEE PER-2 (6) (1982) 50–51.
Conference on Industry Applications (INDUSCON), 2014, pp. 1–6. [142] R. Lee, M. Bishop, A comparison of measured high impedance fault data to
[112] M. Sulaiman, A. Tawfan, Z. Ibrahim, Detecting high impedance fault in digital computer modeling results, Power Eng. Rev. IEEE PER-5 (10) (1985)
power distribution feeder with fuzzy subtractive clustering model, Aust. J. 35–36.
Basic Appl. Sci. 7 (2013) 81–91. [143] A. Sultan, G. Swift, D. Fedirchuk, Detection of high impedance arcing faults
[113] M. Sulaiman, A. Tawafan, Z. Ibrahim, High impedance fault detection on using a multi-layer perceptron, IEEE Trans. Power Deliv. 7 (4) (1992)
power distribution feeder, Int. Rev. Modell. Simul. 5 (5) (2012) 2197–2204. 1871–1877.
[114] S. Sahoo, M.E. Baran, A method to detect high impedance faults in [144] K. Li, Z. Xiangjun, Y. Xianggen, Novel methods for high-impedance
distribution feeders, in: T D Conference and Exposition. 2014 IEEE PES, 2014, ground-fault protection in low-voltage supply systems, Electr. Power
pp. 1–6. Compon. Syst. 31 (12) (2003) 1133–1150.
[115] H. Livani, C. Evrenosoglu, A machine learning and wavelet-based fault [145] S. Samantaray, P. Dash, S. Upadhyay, Adaptive Kalman filter and neural
location method for hybrid transmission lines, IEEE Trans. Smart Grid 5 (1) network based high impedance fault detection in power distribution
(2014) 51–59. networks, Int. J. Electr. Power Energy Syst. 31 (4) (2009) 167–172.
[116] S.R. Samantaray, P.K. Dash, High impedance fault detection in distribution [146] S. Saleem, A. Sharaf, A fuzzy artmap based high impedance arc fault
feeders using extended Kalman filter and support vector machine, Eur. detection scheme, in: Canadian Conference on Electrical and Computer
Trans. Electr. Power 20 (3) (2010) 382–393. Engineering, 2008. CCECE 2008, 2008, pp. 000871–000876.
[117] B.D. Russell, Computer relaying and expert systems: new tools for detecting [147] M. Abdel Aziz, M. Moustafa Hassan, E. Zahab, Applications of ANFIS in high
high impedance faults, Electr. Power Syst. Res. 20 (1) (1990) 31–37. impedance faults detection and classification in distribution networks, in:
[118] S. Balser, K. Clements, D. Lawrence, A microprocessor-based technique for 2011 IEEE International Symposium on Diagnostics for Electric Machines,
detection of high impedance faults, IEEE Trans. Power Deliv. 1 (3) (1986) Power Electronics Drives (SDEMPED), 2011, pp. 612–619.
252–258. [148] M. Abdel Aziz, M. Moustafa Hassan, E. Zahab, Applications of ANFIS in high
[119] W. Rebizant, J. Szafran, A. Wiszniewski, Digital signal processing in power impedance faults detection and classification in distribution networks, in:
system protection and control, Springer Science & Business Media, 2011. 2011 IEEE International Symposium on Diagnostics for Electric Machines,
[120] S.H. Horowitz, A.G. Phadke, Power System Relaying, vol. 22, John Wiley & Power Electronics Drives (SDEMPED), 2011, pp. 612–619.
Sons, 2008. [149] N. Zamanan, J. Sykulski, A. Al-Othman, Arcing high impedance fault
[121] A. Ghaderi, H.A. Mohammadpour, H. Ginn, High impedance fault detection detection using real coded genetic algorithm, in: Third IASTED Asian
method efficiency: simulation vs. real-world data acquisition, in: Power and Conference Power and Energy Systems, 2–4 April 2007, 2007,
Energy Conference at Illinois (PECI), 2015 IEEE, 2015, pp. 1–5. pp. 35–39.
[122] M.M. Saha, J.J. Izykowski, E. Rosolowski, Fault Location on Power Networks, [150] B. Aucoin, B. Russell, Distribution high impedance fault detection utilizing
Springer-Verlag London, 2010. high frequency current components, IEEE Trans. Power Appar. Syst. PAS-101
[123] A. Ahmadimanesh, S. Shahrtash, Transient-based fault-location method for (6) (1982) 1596–1606.
multiterminal lines employing s-transform, IEEE Trans. Power Deliv. 28 (3) [151] A. Girgis, W. Chang, E. Makram, Analysis of high-impedance fault generated
(2013) 1373–1380. signals using a Kalman filtering approach, IEEE Trans. Power Deliv. 5 (4)
[124] S. Azizi, M. Sanaye-Pasand, A straightforward method for wide-area fault (1990) 1714–1724.
location on transmission networks, IEEE Trans. Power Deliv. 30 (1) (2015) [152] D. Jeerings, J. Linders, Unique aspects of distribution system harmonics due
264–272. to high impedance ground faults, IEEE Trans. Power Deliv. 5 (2) (1990)
[125] M. Davoudi, Parameter-free fault location for transmission lines based on 1086–1094.
optimisation, Transm. Distrib. IET Gener. 9 (2015) 1061–1068 (7). [153] A. Sultan, G. Swift, Security testing of high impedance fault detectors, in:
[126] L. Iurinic, A.H. Orozco, R. Ferraz, A. Bretas, Distribution systems high WESCANEX ’91 ‘IEEE Western Canada Conference on Computer, Power and
impedance fault location: a parameter estimation approach, IEEE Trans. Communications Systems in a Rural Environment’, 1991, pp. 191–197.
Power Deliv. PP (99) (2015) 1. [154] R. Christie, H. Zadehgol, M. Habib, High impedance fault detection in low
[127] A. Dasco, R. Marguet, B. Raison, Fault distance estimation in distribution voltage networks, IEEE Trans. Power Deliv. 8 (4) (1993) 1829–1836.
network for high impedance faults, in: PowerTech, 2015 IEEE Eindhoven, [155] A. Sultan, G. Swift, D. Fedirchuk, Detection of high impedance arcing faults
2015, pp. 1–6. using a multi-layer perceptron, IEEE Trans. Power Deliv. 7 (4) (1992)
[128] S. Zimath, C. Dutra, R. Matos, L. de Oliveira, J. de Resende, J. Moutinho, 1871–1877.
Traveling wave fault location applied to high impedance events, in: 12th IET [156] A. Sharaf, R. El-Sharkawy, R. Al-Fatih, M. Al-Ketbi, High impedance fault
International Conference on Developments in Power System Protection detection on radial distribution and utilization systems, in: Canadian
(DPSP 2014), 2014, pp. 1–6. Conference on Electrical and Computer Engineering, 1996, vol. 2, 1996, pp.
[129] M.-y. Chen, J.-q. Zhai, Z.-q. Lang, J.-c. Liao, Z.-y. Fan, High impedance fault 1012–1015.
location in transmission line using nonlinear frequency analysis, in: Life [157] F. Uriarte, V. Centeno, High-impedance fault detection and localization in
System Modeling and Intelligent Computing, vol. 6328, 2010, distribution feeders with microprocessor based devices, in: Proceedings of
pp. 104–111. the 37th Annual North American Power Symposium, 2005, 2005, pp.
[130] N. Apostolos, T. Georgios, P. Dimitris, Detection and location of high 219–224.
impedance faults in multiconductor overhead distribution lines using power [158] N. Ramezani, M. Sarlak, S. Shahrtash, D. Khabori, Design and implementation
line communication devices, IEEE Trans. Smart Grid 2 (6) (2015) 894–902. of an adaptive high impedance fault relay, in: Power Engineering
[131] A. Ghaderi, H.A. Mohammadpour, H. Ginn, Active fault location in Conference, 2007. IPEC 2007. International, 2007, pp. 1131–1136.
distribution network using time-frequency reflectometry, in: Power and [159] T.A. Kawady, A. Taalab, M. Elgeziry, Experimental investigation of high
Energy Conference at Illinois (PECI), 2015 IEEE, 2015, pp. 1–7. impedance faults in low voltage distribution networks, in: Power Energy
[132] Y.-H. Lin, C.-W. Liu, C.-S. Chen, A new PMU-based fault detection/location Society General Meeting, 2009, PES ’09. IEEE, 2009, pp. 1–6.
technique for transmission lines with consideration of arcing fault [160] W.H. Kwon, G.W. Lee, Y.M. Park, M.C. Yoon, M.H. Yoo, High impedance fault
discrimination—Part I: Theory and algorithms, IEEE Trans. Power Deliv. 19 detection utilizing incremental variance of normalized even order harmonic
(4) (2004) 1587–1593. power, IEEE Trans. Power Deliv. 6 (2) (1991) 557–564.
[133] D. Patynowski, J. Cardenas, D. Menendez, Z. Zhang, J. Roca, J. Gardy Germain, [161] A. Sharaf, L. Snider, K. Debnath, A neural network based relaying scheme for
A. Yunta Huete, M. Canales, A. Martinez, J. Rosendo, J. Roldan, J. Montanez, distribution system high impedance fault detection, in: First New Zealand
Fault locator approach for high-impedance grounded or ungrounded International Two-Stream Conference on Artificial Neural Networks and
distribution systems using synchrophasors, in: 2015 68th Annual Expert Systems, 1993. Proceedings, 1993, pp. 321–324.
Conference for Protective Relay Engineers, 2015, pp. 302–310. [162] T. Lai, L. Snider, E. Lo, C. Cheung, K. Chan, High impedance faults detection
[134] T. Gammon, J. Matthews, Instantaneous arcing-fault models developed for using artificial neural network, in: Advances in Power System Control,
building system analysis, IEEE Trans. Ind. Appl. 37 (1) (2001) 197–203. Operation and Management, 2003, ASDCOM 2003, vol. 2, 2003,
[135] M. Jannati, L. Eslami, Precise modeling of high impedance faults in power pp. 821–826.
distribution system in emtpworks software, J. Electr. Eng. (2013) 1–8. [163] M. Eissa, G. Sowilam, A. Sharaf, A new protection detection technique for
[136] W.C. Santos, N.S.D. Brito, B.A. Souza, F.A. Pereira, High impedance fault high impedance fault using neural network, in: 2006 Large Engineering
modeling based on statistical data, in: International Conference on Power Systems Conference on Power Engineering, 2006, pp. 146–151.
Systems Transients (IPST2015), 2015, pp. 1–8. [164] M. Al-Dabbagh, L. Al-Dabbagh, Neural networks based algorithm for
[137] J. Carr, Detection of high impedance faults on multi-grounded primary detecting high impedance faults on power distribution lines, in:
distribution systems, IEEE Trans. Power Appar. Syst. PAS-100 (4) (1981) International Joint Conference on Neural Networks, 1999. IJCNN ’99, vol. 5,
2008–2016. 1999, pp. 3386–3390.
388 A. Ghaderi et al. / Electric Power Systems Research 143 (2017) 376–388

[165] P. Silva, J. Santos, A.F. Jota, An intelligent system for automatic detection of [188] C. Kim, H. Kim, R. Aggarwal, A. Johns, Wavelet transform in the accurate
high impedance faults in electrical distribution systems, in: Proceedings of detection of high impedance arcing faults in high voltage transmission lines,
the 38th Midwest Symposium on Circuits and Systems, 1995. Proceedings, in: Seventh International Conference on Developments in Power System
vol. 1, 1995, pp. 453–456. Protection, 2001. IEE, 2001, pp. 422–425.
[166] D. Jeerings, J. Linders, A practical protective relay for down-conductor faults, [189] C.-H. Kim, H. Kim, Y.-H. Ko, S.-H. Byun, R. Aggarwal, A. Johns, A novel
IEEE Trans. Power Deliv. 6 (2) (1991) 565–574. fault-detection technique of high-impedance arcing faults in transmission
[167] P. Silva, J. Santos, A.W. Boaventura, G. Miranda, J. Scott, Impulse response lines using the wavelet transform, IEEE Trans. Power Deliv. 17 (4) (2002)
analysis of a real feeder for high impedance fault detection, in: Transmission 921–929.
and Distribution Conference, 1994. Proceedings of the 1994 IEEE Power [190] M. Akoreder, J. Katende, Wavelet transform based algorithm for
Engineering Society, 1994, pp. 276–283. high-impedance faults detection in distribution feeders, Eur. J. Sci. Res. 41
[168] H. Khorashadi-Zadeh, A novel approach to detection high impedance faults (2) (2010) 237–247.
using artificial neural network, in: Universities Power Engineering [191] M.-T. Yang, J.-C. Gu, C.-Y. Jeng, W.-S. Kao, Detection of high impedance fault
Conference, 2004. UPEC 2004. 39th International, vol. 1, 2004, pp. 373–376. in distribution feeder using wavelet transform and artificial neural
[169] B. Russell, R. Chinchali, A digital signal processing algorithm for detecting networks, in: 2004 International Conference on Power System Technology,
arcing faults on power distribution feeders, IEEE Trans. Power Deliv. 4 (1) 2004. PowerCon 2004, vol. 1, 2004, pp. 652–657.
(1989) 132–140. [192] N. Subramaniam, K. Bagan, Analysis of high impedance transients and
[170] M. Zanjani, H. Karegar, H. Niaki, M. Zanjani, Application of PMUS for high improved data compression using wavelet transform, Serb. J. Electr. Eng. 3
impedance fault detection of distribution network by considering effect of (1) (2006) 19–31.
transformer vector group, in: 2013 18th Conference on Electrical Power [193] H. Yong, C. Minyou, Z. Jinqian, High impedance fault identification method
Distribution Networks (EPDC), 2013, pp. 1–7. of the distribution network based on discrete wavelet transformation, in:
[171] J. Lorenc, J. Handke, A. Kwapisz, B. Staszak, P. Balcerek, New solution of 2011 International Conference on Electrical and Control Engineering
identification of high-impedance earth-fault in compensated MV network, (ICECE), 2011, pp. 2262–2265.
in: 22nd International Conference and Exhibition on Electricity Distribution [194] N. Elkalashy, M. Lehtonen, H. Darwish, M. Izzularab, A.-M. Taalab,
(CIRED 2013), 2013, pp. 1–4. DWT-based investigation of phase currents for detecting high impedance
[172] E. Aboul-Zahab, E.-S. Eldin, D. Ibrahim, S. Saleh, High impedance fault faults due to leaning trees in unearthed MV networks, in: Power
detection in mutually coupled double-ended transmission lines using high Engineering Society General Meeting, 2007, IEEE, 2007, pp. 1–7.
frequency disturbances, in: Power System Conference, 2008. MEPCON 2008. [195] N.I. Elkalashy, M. Lehtonen, H.A. Darwish, A.-M.I. Taalab, M.A. Izzularab,
12th International Middle-East, 2008, pp. 412–419. DWT-based extraction of residual currents throughout unearthed MV
[173] M.-T. Chen, H.-Y. Chu, C.-L. Huang, F.-R. Wu, Performance evaluation of high networks for detecting high-impedance faults due to leaning trees, Eur.
impedance fault detection algorithms based on staged fault tests, Electr. Trans. Electr. Power 17 (2007) 597–614.
Power Syst. Res. 18 (1) (1990) 75–82. [196] F. Lopes, W. Santos, D. Fernandes, W. Neves, N. Brito, B. Souza, A transient
[174] L. Snider, Y.Y. Shan, The artificial neural networks based relay algorithm for based approach to diagnose high impedance faults on smart distribution
distribution system high impedance fault detection, in: Fourth International networks, in: 2013 IEEE PES Conference on Innovative Smart Grid
Conference on Advances in Power System Control, Operation and Technologies Latin America (ISGT LA), 2013, pp. 1–8.
Management, 1997. APSCOM-97 (Conf. Publ. No. 450), vol. 1, 1997, pp. [197] N. Elkalashy, M. Lehtonen, H. Darwish, A.-M. Taalab, M. Izzularab, Feature
100–106. extraction of high impedance arcing faults in compensated MV networks.
[175] H. Mokhtari, R. Aghatehrani, A new wavelet-based method for detection of Part I: DWT-based analysis of phase quantities, in: Power Engineering
high impedance faults, in: 2005 International Conference on Future Power Society Conference and Exposition in Africa, 2007. PowerAfrica ’07. IEEE,
Systems, 2005, p. 6. 2007, pp. 1–6.
[176] I. Baqui, A. Mazon, I. Zamora, R. Vicente, High impedance faults detection in [198] D. Ibrahim, E.-S. Eldin, E. Aboul-Zahab, S. Saleh, High-impedance fault
power distribution system by combination of artificial neural network and detection in EHV transmission lines, in: Power System Conference, 2008.
wavelet transform, in: 18th International Conference and Exhibition on MEPCON 2008. 12th International Middle-East, 2008, pp. 192–199.
Electricity Distribution, 2005. CIRED 2005, 2005, pp. 1–4. [199] E. Eldin, D. Ibrahim, E. Aboul-Zahab, S. Saleh, High impedance fault detection
[177] M.-T. Yang, J.-C. Gu, J.-L. Guan, C.-Y. Cheng, Evaluation of algorithms for high in EHV series compensated lines using the wavelet transform, in: Power
impedance faults identification based on staged fault tests, in: Power Systems Conference and Exposition, 2009. PSCE ’09. IEEE/PES, 2009, pp.
Engineering Society General Meeting 2006, IEEE, 2006, p. 8. 1–10.
[178] I. Baqui, I. Zamora, J. Mazón, G. Buigues, High impedance fault detection [200] D. khalil Ibrahim, E.S.T. Eldin, E.M. Aboul-Zahab, S.M. Saleh, Real time
methodology using wavelet transform and artificial neural networks, Electr. evaluation of DWT-based high impedance fault detection in {EHV}
Power Syst. Res. 81 (7) (2011) 1325–1333. transmission, Electr. Power Syst. Res. 80 (8) (2010) 907–914.
[179] M. Sarlak, S.M. Shahrtash, High impedance fault detection in distribution [201] T. Cui, X. Dong, Z. Bo, S. Richards, Integrated scheme for high impedance
networks using support vector machines based on wavelet transform, in: fault detection in MV distribution system, in: Transmission and
2008 IEEE Electrical Power & Energy Conference, 2008, pp. 1–6. Distribution Conference and Exposition: Latin America, 2008 IEEE/PES,
[180] V. Torres Garcia, H. Ruiz Paredes, High impedance fault detection in 2008, pp. 1–6.
distribution system using wavelet transform, in: 2011 8th International [202] S. Ebron, D. Lubkeman, M. White, A neural network approach to the
Conference on Electrical Engineering Computing Science and Automatic detection of incipient faults on power distribution feeders, IEEE Trans.
Control (CCE), 2011, pp. 1–6. Power Deliv. 5 (2) (1990) 905–914.
[181] V.G. Torres, H.P. Ruiz, High impedance fault detection using discrete wavelet [203] D. Hou, Detection of high-impedance faults in power distribution systems,
transform, in: Electronics, Robotics and Automotive Mechanics Conference in: Power Systems Conference: Advanced Metering, Protection, Control,
(CERMA), 2011 IEEE, 2011, pp. 325–329. Communication, and Distributed Resources, 2007. PSC 2007, 2007, pp.
[182] A. Kannan, A. Rathinam, High impedance fault classification using wavelet 85–95.
transform and artificial neural network, in: 2012 Fourth International [204] I. Nikoofekr, M. Sarlak, S. Shahrtash, Detection and classification of high
Conference on Computational Intelligence and Communication Networks impedance faults in power distribution networks using art neural networks,
(CICN), 2012, pp. 831–837. in: 2013 21st Iranian Conference on Electrical Engineering (ICEE), 2013, pp.
[183] A. Hossam Eldin, E. Abdallah, N. Mohamed, Detection of high impedance 1–6.
faults in medium voltage distribution networks using discrete wavelet [205] T.M. Rao, Digital/Numerical Relays, Tata McGraw-Hill Education, 2005.
transform, in: 22nd International Conference and Exhibition on Electricity [206] M. Hassan, A.-M. Bayoumi, Y.-J. Shin, Quadratic-nonlinearity index based on
Distribution (CIRED 2013), 2013, pp. 1–4. bicoherence and its application in condition monitoring of drive-train
[184] J. Chen, B. Phung, D. Zhang, T. Blackburn, E. Ambikairajah, Arcing current components, IEEE Trans. Instrum. Meas. 63 (3) (2014) 719–728.
features extraction using wavelet transform, in: Proceedings of 2014 [207] H. Mesgarnejad, S. Shahrtash, Power system voltage stability assessment
International Symposium on Electrical Insulating Materials (ISEIM), 2014, employing phasor measurement units, in: 2010 The 2nd International
pp. 221–224. Conference on Computer and Automation Engineering (ICCAE), vol. 5, 2010,
[185] P. Escalante Farias, A. Peres de Morais, G. Cardoso, A. Oliveira, Transients pp. 13–19.
detection and classification in distribution networks for high impedance [208] J. De La Ree, V. Centeno, J. Thorp, A. Phadke, Synchronized phasor
faults identification, in: Power Engineering Conference (UPEC), 2014 49th measurement applications in power systems, IEEE Trans. Smart Grid 1 (1)
International Universities, 2014, pp. 1–6. (2010) 20–27.
[186] J. Chen, B. Phung, H. Wu, D. Zhang, T. Blackburn, Detection of high [209] P.V. Subramanian, L. Ajitha, Practical considerations for high impedance
impedance faults using wavelet transform, in: Power Engineering restricted earth fault relay settings, J. Inst. Eng. Ser. B (2015) 107–110.
Conference (AUPEC), 2014 Australasian Universities, 2014, [210] R. Ferraz, L. Iurinic, A. Bretas, A. Bettiol, A. Carniato, L. Passos, I. Khairalla, R.
pp. 1–6. Homma, A. Baader, High impedance fault detection in distribution systems:
[187] D.C.T. Wai, X. Yibin, A novel technique for high impedance fault adaptive approach considering noising environment, in: 23rd International
identification, IEEE Trans. Power Deliv. 13 (3) (1998) 738–744. Conference on Electricity Distribution, 2015, pp. 1–5.