High-Impedance Faults: Comparing Algorithms

Abstract— It remains a great challenge to identify an leaning on distribution circuits cause lines to sag and to
arcing, downed conductor. Detecting and removing power conduct current intermittently to ground. If a utility does not
from fallen wires is very important for safety of property and do an adequate job of clearing vegetation around distribution
human life. Recent events in California have spurred the effort lines, then tree limbs touch the line irregularly, causing
to protect against the destruction caused by this situation. The arcing on intact conductors.
challenge is that a downed conductor exhibits an intermittent,
high-impedance, low-current signal that is difficult to sense for Contaminated and failing equipment, such as
conventional relay protection elements. Over the years, disconnects, fuses, and dirty insulators can cause high-
protection engineers have developed algorithms for impedance faults in the distribution system.
recognizing these high-impedance (Hi-Z), or high-impedance-
faults (HIF) events.

This paper describes algorithms used for detecting an

arcing, downed conductor. Comparisons and benefits of each
of these algorithms, and settings methods, is presented. Also
described is testing of these algorithms, including the test
methods and results. Lastly, there are recommendations for
employing HIF protection schemes, including tripping and
alarming for high-impedance faults.

Keywords—arcing, downed conductor, HIF, high-impedance

fault
Fig. 1. Downed conductors
I. INTRODUCTION
The following topics comprise the discussion of high-
impedance faults (HIFs):
 Definition and causes of high-impedance faults
 Methods for HIF detection
 Method comparisons
 Implementation strategies
This paper covers these topics, to provide an
understanding of the challenge and successful detection of
HIFs.
Fig. 2. Broken conductor
II. DEFINITION AND CAUSES OF HIGH-IMPEDANCE FAULTS
A high-impedance fault (HIF) occurs when a primary An HIF event that occurs often is when a conductor
conductor makes unwanted electrical contact with a tree, breaks and falls to the ground. When broken and laying on
pole, structure or with the ground, and there is a large the ground, we call this a “broken conductor,” as shown in
impedance restricting the flow of electrical current. The Error! Reference source not found.. Note that the ends are
fault current can be at a few milliamps to 100 amps primary, separated and slack. The break drops load on the affected
much smaller than the current that standard overcurrent feeder, and sometimes, a momentary overcurrent condition
elements can detect. Even in cases where the instantaneous occurs as the falling conductor comes in contact briefly with
fault current exceeds standard overcurrent thresholds, the an adjacent line or a solidly grounded object. There is little
duration of this transient event is so short that standard fuses effect on the system voltage.
and overcurrent elements will not clear or pick up. There is Fault arcing and currents differ on various surfaces.
little threat of damage to power-system equipment from
Tests indicate that 6085 percent of all downed conductors
these transient events, but these events are a safety and fire
can be detected. This means that the remainder cannot be
hazard. Line crews responding to a downed or broken
detected, because not all surfaces produce arcing. Without
conductor event seldom document these as such on trouble
arcing, HIF (or Hi-Z) detection is unable to declare a downed
reports. It is quite challenging to detect HIFs; it requires
conductor. For example, downed conductors on dry asphalt
special methods, combining multiple techniques.
and dry sand cannot be detected because these surfaces do
Downed and broken conductors are the causes of high- not produce arcing. However, it is difficult to find purely dry
impedance faults. The conductor touching the ground might asphalt and dry sand with no moisture and impurities—the
be intact, or, it could be broken. If the conductor touches the materials that would provide a conduction path. In contrast,
ground or other surface, and remains intact feeding a load, reinforced concrete (lots of metal rebar) provides the most
then we call this a “downed conductor.” Note the tension on arcing. Many utilities have performed staged-fault tests on
the wires in Error! Reference source not found.. This can their systems to test the effectiveness of HIF detection. In
be caused by support failure, heavy-loading sag, or an object most cases, the utility includes a “challenge” test case.
(tree) on the distribution line. Weather, nature, and faulty Typically, a conductor is dropped on asphalt or sand. In one
equipment can cause this problem. Icing and tree limbs test, the conductor was dropped on asphalt with the
expectation of no detection. What occurred, however, was A. Overall fault coverage
that the arc found paths through cracks in the asphalt that Traditional overcurrent elements cover 90 percent of
permitted arcing and subsequent detection by the HIF device. distribution faults, including some downed-wire events, as
Once on the ground, the resulting electrical signature is a shown in Error! Reference source not found.. However,
function of the contacted surface. Surfaces such as concrete, some downed-wire events have a large impedance to ground
grass, dirt, and wet surfaces in general result in an “arcing and go undetected by traditional overcurrent protection
fault” with RMS fault currents in the range of 1–100 amps, (which is the portion in light green). For these high-
whereas surfaces such as dry sand and asphalt result in a zero impedance faults, modern relays employ high-impedance
to constant, low-level current flow. Ground-path resistance fault (HIF) detection to protect people and property.
varies during the arcing fault. The changing current creates
non-linear waveforms. The non-linearities produce large
harmonic and non-harmonic energy in the frequency
spectrum of the current. This is an important factor in
detecting high-impedance faults. Arcing generates
significant noise on the affected feeder; spectral analysis
reveals large harmonics and non-harmonic noise to
approximately 700 Hz. Arcing faults result in definable and
detectable patterns. However, surface types change the
arcing-burst signatures, which presents a challenge to secure
and dependable detection.
Arcing time for HIF currents is erratic; usually these
currents decrease over time because the wire-contacted
material burns. Arcing can stop for minutes, and then start
again. High-impedance faults can persist from seconds to
minutes, and sometimes for days!
Fig. 3. Cleared and uncleared distribution-system faults

III. METHODS FOR HIF DETECTION The goal is 100 percent detection of all HIFs, but it is
With wide variations in types of material that an arcing impossible to detect all high-impedance faults. However,
line touches, the high-impedance fault signature changes, experience shows that 60–85 percent of HIFs are detectable.
making HIF detection difficult. While no single parameter It is possible to achieve a large degree of dependability and
classifies an HIF, there are sufficient similarities among HIF security with a reasonably priced detection system.
incidents to have good success at detection. Over time, we If we assume 90 percent of faults are “low impedance”
have developed different measures and strategies to detect faults or overcurrent faults cleared by fuses and conventional
HIF conditions. Using multiple detection algorithms, it is overcurrent protection, and if the remaining 10 percent of the
possible to detect 60–85 percent of high-impedance faults. faults are high-impedance, downed conductors, then a
In addition, it helps to supervise HIF detection with modern microprocessor relay with instantaneous- and time-
power-system load events such as a sudden load loss, or a overcurrent protection, as well as HIF detection, can operate
sudden increase in second-harmonic current (from for 98% of the faults on a utility distribution system.
transformer inrush).
IV. EFFECTIVE HIF DETECTION TECHNIQUES
Ideally, HIF detection should detect all HIFs and should
be secure, while ignoring normal conditions. It should be Modern HIF-detection techniques analyze arcing and
immune to false-positive indications and misoperations. HIF load. Secure and dependable detection is required for high-
detection must differentiate between HIF current signatures impedance fault detectors. Early solutions, like an
and the waveforms from intermittent, noisy loads. Examples electromechanical relay that detected changes in 3I0 current,
of loads that generate power-system noise are arc furnaces, proved insecure, often causing nuisance trips.
welders, capacitor and line switching, load-tap changing, DC Today, microprocessor relays with HIF detectors provide
rectification, motor commutation and starting, etc. Effective secure HIF detector performance under normal system
HIF detection techniques evaluate and subtract load from the conditions, such as noisy feeders, arc furnaces, arc welders,
arcing algorithm. capacitor switching, line switching and load-tap changing.
Detecting HIF is vitally important. Undetected, live, Modern HIF detection analyzes the arcing time-domain and
downed conductors can be fatal to the public and to line frequency-domain components, measures increments in the
crews. Arcing from high-impedance faults cause fires. current inputs (not absolute/peak values) and incorporates
load analysis and pattern recognition.
A line that is arcing results in line failure, which leads to
power outages and loss of production, affecting the economy There are two methods for detecting HIF using existing
and human wellbeing. CT and PT inputs to protective relays. These methods work
for solidly grounded, resistance-grounded, and Petersen-coil-
HIFs cost utilities liabilities and service problems. In grounded systems. The two methods discussed in this paper
2020, PG&E paid 23 billion dollars in fines for the 2018 are Method P and Method F.
California wildfire.
Method P employs a zero-sequence core-balance CT
(CBCT) as shown in Error! Reference source not found..
It detects arcing faults by dividing the analog current input
signal from the CBCT into an overall arcing-detection  Component harmonic analysis (CHA)
portion called Fundamental Analysis (FA), and into a third-
/fifth-harmonics portion called Component Harmonic  Directional analysis (DIR)
Analysis (CHA). HIF detection relies on evaluating fault-current-
waveform signatures. These waveforms differ from fault to
fault (and within the same fault occurrence). However, there
are common attributes in HIFs:
 Transient bursting (intermittent change of amplitude)
on which the relay uses fundamental analysis (FA),
with counters and timers
 Harmonic content (this method uses 3rd and 5th
harmonics) on which the relay uses component
400 A, 20 ohms harmonic analysis (CHA), detecting the periodic
distortions at the zero crossings
To establish the fault direction, Method P uses an
13.8 kV
instantaneous power measurement in the directional-analysis
(DIR) algorithm.
Feeder relay
CBCT, 50:1 A
with HIF The overall block diagram for Method P (Error!
Reference source not found.) shows the voltage input VN,
and the sensitive-earth-fault current input ISEF from the core-
balance CT. Settings and reset are additional inputs to the
HIF function. The two paths to the output are via
fundamental analysis, FA, and component harmonic
analysis, CHA. For each of the FA and CHA paths, the HIF
Fig. 4. Typical Method P installation, with CBCT outputs are transient (arcing detected) and HIF. In addition,
the FA path outputs a Steady Fault operand. The main
In addition, Method P uses voltage for a sensitive power output is the ANDed combination of the FA HIF and the
calculation, to determine direction (DIR). CHA HIF outputs, called HIF Alarm.
Method F employs a special analog-input module that VN
amplifies each of the four inputs for three-phase currents and ISEF
Direction

the residual-current inputs as shown in Error! Reference From CBCT

Average
source not found.. It is wired in series with the existing amplitude
FA decision

relay CT inputs. In addition, the Method F HIF processing Incremental

amplitude
FA analysis
Transient fault FA Transient

Steady fault FA Steady Fault

uses voltage for phase identification and restraint. Method F Average HIF FA HIF
learns the line noise and subtracts it from sensed arcing sample array
AND HIF Alarm
CHA decision
bursts. Incremental
CHA analysis HIF CHA HIF
amplitude
HIF tPREPARE Transient fault CHA Transient
HIF SEF AnyStart

FA settings

CHA settings
HIF function

HIF forced reset Reset buffers

Fig. 6. Method P block diagram

1) Fundamental Analysis (FA)

Fundamental Analysis (FA) captures the intermittent
characteristics of high-impedance fault current. Generally,
the system current is stable and tracks the load conditions.
Method P averages the sensitive-earth-fault current,
summing the latest samples, and storing this value in a
buffer. The relay compares this value continually with the
latest current value. If there is a sudden increase in current,
this value significantly exceeds the average value. It is this
Fig. 5. Method F uses a Hi-Z processing module (M1a—M4c) wired in increment that starts the fault-evaluation process.
series with the regular relay current inputs
Once FA is triggered by any sudden increase of the
amplitude, the algorithm counts bursts in a time window.
A. Method P details: FA, CHA, and DIR Only these sustained series of changes within the time
Method P uses three components to provide a reliable window are evaluated as HIF. If the current-burst increment
HIF detection function: is greater than a start threshold, determined by the setting
FA> Start Thresh, the FA algorithm evaluates the fault.
 Fundamental analysis (FA)
The algorithm applies the Burst Valid (BV) threshold,
determined by the setting FA> Burst Thresh, to judge CHA detects a fault by timing the duration of the
whether the increment indicates fault-current conduction. By Satisfied State (SS). If this time is longer than the HIF
counting the changes of the BV states within a time window, Setting time, an HIF event is reported. If the duration is
the relay establishes that it has detected an intermittent fault. shorter, but still longer than a Transient Setting time, the
algorithm reports a CHA Transient event.
An arcing fault triggers FA detection with any sudden,
current-amplitude increase. However, the relay classifies Similar to FA, a transient event needs further
only sustained series of changes within a specified time as a confirmation. The algorithm activates three separate timers
high-impedance fault (HIF). The following shows the once the CHA starts:
classifications:
 A reset timer
 HIF—bursts exceed set count within time limit
 An HIF-duration timer measures the duration of this
 Transient event—Bursts do not exceed count, but at Satisfied State, to issue HIF
least two bursts counted
 A Transient timer detects any transient event
 Steady event—Burst-count threshold exceeded
If the Satisfied State lasts for the entire time set by the
 Noise—other causes of detected current bursts (e.g., HIF timer, an HIF is reported, and all procedures are reset. If
arc welding) the Satisfied State lasts for less than the HIF duration but is
still more than the transient time, a Transient Suspicion event
A declaration of HIF indicates certain confidence, and a is reported and the detection process evaluates another
steady event has very small arcing confidence. section. If any HIF requirement is satisfied within the reset
These are the settings for fundamental analysis (FA) to time, a HIF is reported, and the detection is reset. If there are
adapt to specific applications: more than three Transient Suspicion events reported within
the reset time, a HIF is reported.
 Burst-current threshold
Flattening across the zero crossings indicates arcing. As
 Burst count the voltage wave passes through the zero crossing, the
 Monitoring time window voltage available to drive the fault current reduces
drastically, and therefore, the current waveform has
It is important to avoid nuisance alarms. Method-P FA ‘shoulders’ around the zero crossings.
uses these techniques:
Staged, downed-conductor testing shows that the CHA
 RMS measurements, not instantaneous values algorithm picks up for a conductor arcing in sand.
 Measuring magnitude increments, not absolute values 3) Directional Analysis (DIR)
Directional analysis helps with resistance-/impedance-
 Adaptive monitoring time, following start grounded systems. The FA algorithm has no capability of
2) Component harmonic analysis (CHA) detecting direction. FA-only is best used in a system with
Component harmonic analysis (CHA) measures the limited capacitance, or a system with a directly grounded
sensitive earth fault (SEF) current from the CBCT, compares neutral point. In these cases, the fault current on healthy
this with the average current value, and uses the increment of lines is limited.
the sampled value to extract the 3rd- and 5th-harmonic However, when a system is resistance-grounded, the
component. CHA evaluates the phase and amplitude fault-generated transient might be dispersed along both
differences between the fundamental and the third/fifth healthy and faulted lines because of large, distributed
harmonics to determine the presence of HIFs. capacitances. Therefore, a directional element enhances FA
The Satisfied State (SS) is the value that indicates HIF performance for these systems.
non-linearity; CHA measures the duration of the Satisfied The relay uses fault instantaneous power direction to
State to evaluate and classify faults. The fault-evaluation obtain the transient direction, calculated directly from the
process can be triggered internally or externally. fault-component (zero-sequence) samples. In transient
The criteria determining HIF non-linearities are the situations, this is a more accurate method than using phasor-
following: based power calculations. The fault component circuit is
used for analysis. The source is the fault itself. The
 The fundamental amplitude is greater than a set capacitive branch produces the reactive power while the
threshold (setting CHA> Fund Thresh) inductance branch absorbs the reactive power. The resistance
branch absorbs the active power, from the source. The
 The amplitude ratio between the 3rd/5th harmonics
reactive power from the source balances the total
and the fundamental is greater than a set threshold
consumption of the reactive power by the other part of the
(setting CHA>3rdHarmThrsh), and is greater than 90
circuit.
percent of the fundamental
Generally, the reactive power is more distinctive, because
 The phase difference between the 3rd/5th harmonics the distributed capacitance is greater than the distributed
and the fundamental is within a range of conductance. Therefore, in resistance-grounded and in
approximately 180° (settings CHA Del Ang180-x and isolated systems, the reactive power direction is used for
CHA Del Ang180+x) transient-direction detection. In Petersen-coil-grounded
 These effects last for a significant time systems, the active-power direction detects the direction
because the Petersen coil cancels reactive power flow.
 The output of the direction detection function (DIR) are Zoom in on the rocky-sand waveform, Error! Reference
flags indicating the fault direction: FA DIR Forward and FA source not found., to view the FA increment increasing to
DIR Reverse. The relay sets these flags if the algorithm is in yield “FA HIF” outputs.
the Start stage and the criteria have been met. The FA
function uses the flag status to determine whether it is a
forward fault or a reverse fault. When counting a spike into
the FA function's counter, the FA first refers to the direction
flag. Only spikes with forward direction (Forward transient)
are counted for fault evaluation. An alarm can also be set to
indicate the faulted line.
4) Selecting the Method P solution
Select the HIF solution according to the system
grounding, as shown in Table I. FA detects intermittent
faults where the fault current is changing between
conducting and non-conducting. This can be used in any Fig. 8. . FA HIF outputs
system grounding conditions.
The component harmonic analysis (CHA) algorithm
produces an HIF output for arcing in the rocky sand, as
TABLE I. METHOD P SOLUTIONS shown in Error! Reference source not found..
Solid Resistor Petersen Isolated
coil

FA and Yes Yes Recomm Yes

DIR (active, P) ended

FA and Yes Recomm No Recom

DIR (reactive, ended mended
Q)

FA (no Recomm No No No
DIR) ended

CHA Recomm Recomm No No

ended ended
Fig. 9. CHA output on rocky sand

CHA detects situations where there is a regular earth- V. METHOD F: HARMONICS AND NON-HARMONICS; POWER-
fault harmonic. CHA should only be used for directly SYSTEM LEARNING; LOAD SUBTRACTION
grounded and for low-resistance-grounded systems. As
noted in this table, the directional element helps with Nine algorithms comprise Method F HIF detection.
Petersen coil and isolated systems. High-impedance fault detection is accomplished through a
variety of techniques, all coordinated by an expert system.
This HIF protection system monitors and learns the power-
B. Method P testing—field trials
system nominal behavior. At the heart of the HIF detection
Method P HIF detection field trials were performed on a system is identification of arcing on a feeder. If arcing is
13.8-kV feeder, with a resistance-grounded source. A 50:1 detected, the Expert Arc Detector determines whether the
core-balance CT fed the protection relay. arcing persists for a significant period. If it does, it analyzes
Error! Reference source not found. shows the load current at the beginning of the arcing to determine
waveforms recorded from downed-conductor tests. These whether persistent arcing is from a downed conductor or
waveforms show the erratic signature of arcing in rocky from an intact conductor. A collection of sensitivity and
sand. Note the transition from FA Transient to FA HIF, timing settings tune dependability and security.
marked in this graphic as “FA detects HIF.” If the HIF element determines that a downed conductor
exists, oscillography and fault data are captured. In addition,
target messages and appropriate LEDs are activated on the
relay faceplate.
The detection of a downed conductor or arcing condition
is accomplished through the following algorithms:
 Energy algorithm
 Randomness algorithm
 Expert Arc Detector
 Load Event Detector
 Load Analysis
Fig. 7. Method P downed-conductor test result
  Load Extraction element recognition of such sporadic arcing closes that
contact and appropriate investigation can be taken.
 Arc Burst Pattern Analysis
 Spectral Analysis B. Downed Conductor output
The basis for downed-conductor detection is event
 Arcing-Suspected Identifier
dependent. Differentiation between an arcing, intact
 Even Harmonic Restraint conductor and an arcing, downed conductor is determined by
looking at patterns in the load current at the beginning of the
 Voltage Supervision fault. User settings determine what constitutes a loss of load
The Method F block diagram in Error! Reference or an overcurrent condition. When a conductor breaks,
source not found. shows signal processing through the generally there is a loss of load or an overcurrent condition
energy and randomness algorithms to the Expert Arc when a phase hits another phase. This gives an indication
Detector. The relay processes phase and neutral current that the conductor broke. The detection method looks for
inputs for 60-Hz and arcing to determine whether the conductor has fallen to the
odd-, even-, and non-harmonics. Voltage inputs have a ground or hit a grounded object. A downed conductor is
confirmation check. The relay processes load extraction, only when a loss of load or an overcurrent condition
arc-burst pattern, and spectral analysis. The results of these precedes arcing detection.
analyses proceed to the energy, randomness, and pattern- Avoiding nuisance outputs depends on many strategies.
analysis algorithms. Next, the Expert Arc Detector decision HIF detection monitors RMS currents, not instantaneous
structure provides separate HIF outputs for arcing detected values that could false the HIF algorithms. The variable arc-
alarm, arcing suspected, downed-conductor conditions, and inception voltage causes a changing arcing-current shape,
fault identification. and thus, fluctuating harmonic content. Measured
increments track bursts over time, not absolute values. This
HIF method adapts to changing conditions and loads. Upon
Expert arc
Load-extraction
Arc-burst pattern
Energy
algorithm
detector
algorithm Outputs
power-up, Method F learns the feeder ambient harmonic-
Spectral analysis
algorithms Identify Arcing-
energy level. It determines an ambient average noise level
60-Hz Convert
large energy
changes
suspected
identifier
Arcing suspected
for the odd-, even-, and non-harmonic energy components of
Ia, Ib, Ic, In process odd / even /
and non
Identify
the currents each hour for three days. Then, it takes the
amplifier harmonics
random
energy
Expert arc
detector
Arcing alarm worst average value over this three-day period and uses it as
changes
Randomness
the harmonic noise threshold. The HIF algorithms ignore
algorithm
any harmonic energy patterns in the current below this
Ia, Ib, Ic, In RMS current
detector
ambient level. If the harmonic energy on the feeder changes,
Method F re-determines a new ambient level to adapt to
In Downed changing conditions. Another method for avoiding nuisance
Confirmation Pattern conductor
Voltages checks analysis outputs is to use even-harmonic restraint and voltage
Load-analysis
Phase
supervision as final tests before declaring a downed
algorithm
identification conductor.

Fig. 10. Method F block diagram C. Arcing Alarm output

If there is only arcing and no loss of load or an
Arcing causes bursts of energy to register throughout the overcurrent condition preceding the arcing detection, an
frequency spectrum of the currents. These arcing bursts have Arcing Alarm is indicated (not a Downed Conductor output).
distinct qualities and signatures; several of the HIF It is assumed that the line is intact, with arcing present. This
algorithms detect these patterns. The odd-, even-, and non- might be a bad insulator or a tree rubbing an intact
harmonic components of the phase and neutral currents are conductor. If the detected arcing is persistent, and an output
analyzed for these distinct patterns. The algorithms look for contact closes.
energy and randomness patterns. Separate outputs of the arc
detection algorithms feed into the Expert Arc Detector.

A. Arcing Suspected output D. Method F testing—field trials

In some cases, arcing is present, but not persistent. For Many utilities have performed their own stage fault tests
example, when it is caused by tree limb contact or insulator on both grounded and ungrounded distribution systems.
degradation, arcing typically is present intermittently, Some utilities have performed “drop” tests, where a line is
interspersed with relatively long periods of inactivity (such dropped intentionally to test the operation of the HIF
as minutes). In these cases, arcing can be affected by such detector.
factors as the motion of a tree limb or the moisture and Recent, in-field testing at PEPCO, where the Method F
contamination on an insulator. Conditions such as these, relay detected 82 percent of downed conductors. The study
characterized by a large number of brief occurrences of was based on 280 installed Method F, HIF relays on a
arcing over an extended period (for example from a fraction distribution system of 620, 13-kV overhead feeders.
of an hour to one or two hours), lead the HIF element to
recognize and flag an “arcing suspected” event. If an output
contact is configured to indicate Arcing Suspected, the HIF
 VI. METHOD COMPARISONS It compares the averaged current with latest current value. If
Table II is a summary of the two methods for detecting there is a sudden increase in current, its value exceeds the
and acting upon high-impedance faults. Shown are Method average value significantly, starting the fault-evaluation
F and Method P. process.
3) Counters (employed for security)
TABLE II. SUMMARY OF METHOD F AND METHOD P The random on-off nature of HIF bursts means that the
HIF protection methods cannot trip instantly. The methods
Feature Method F Method P Comments
employ counters to assure confidence that the burst is an arc
Method or a downed conductor. Unlike traditional protection, HIF
elements take 20 seconds and longer for an alarm / trip
Connection Phase / Zero-
residual sequence output.
CT /
CBCT 4) Arc-burst pattern analysis
Method F inspects the incoming current bursts for
Difference Yes, 64 Yes, 24 Greater sampling rate catches fast transients
(incremental) samples/cycle samples/cycle randomness. HIFs display unsystematic timing and
currents magnitude changes. Method F counts these random events
in a time window, to build a confidence bias, toward
Spectrum Harmonics 3rd and 5th Process arcing frequency spectrum
monitor and non- only declaring arcing and an HIF. Method P counts transients in
harmonics to
780 Hz
time window; periodic events do not add to the count.
5) Directional analysis
Arcing-level 72 hours Average Ignore normal line noise and activity
reference recent samples There is no need to use directionality in a system with
limited capacitance, or a system with directly grounded
Randomness x x Quantify sudden, erratic changes neutral point. In these cases, the fault current on healthy lines
algorithm
is limited. However, when there is arcing or a downed
Counters, x x Build certainty over time
conductor, the healthy lines can carry the burst signal as
trending, and well. This problem is prevalent in isolated, high-resistance-
memory
grounded, and Petersen-coil-grounded systems with a
Arc-burst x x Identify faulted phase relatively large distributed capacitance; the fault-generated
pattern transient can be distributed along both the healthy and
analysis
faulted lines. Therefore, a directional element can enhance
Load-event x Block HIF for power-system events HIF detection performance. Method P has a special
detection directionality determination, to increase security for HIF
outputs. Add directionality with logic to Method F.
Load learning x Remove normal load
and extraction
6) Decision logic
Both methods employ outputs for specific HIF events.
Voltage x Limit false alarms (adds security)
supervision
a) Output for arcing: Transient (suspicion)/Arcing
Directional logic x Isolates faulted line
suspected
analysis The methods have an output for transient events or arcing
suspected. This output indicates that there are many brief
Decision logic Arcing Transient Provide action outputs
suspected, suspicion,
arcing events over an extended period (for example, from a
Arcing alarm,
alarm— fraction of an hour to one or two hours). None of these brief
for FA and
CHA occurrences of arcing indicate a downed conductor. When
Downed
conductor considered cumulatively, however, these arcing events need
attention.
The methods share similar algorithms to detect high-
impedance faults. These similarities center on sampling and b) Output for high-impedance fault: HIF/Downed
tracking arcing bursts in the time and frequency domains. conductor
Both methods use an arcing-level reference, a randomness Once the burst sequence has passed all of the detection
algorithm, and counters, trending, and memory to classify qualifications and the security algorithms, the HIF element
high-impedance faults. The following is a discussion of the declares an HIF, or Downed conductor, event. At this point
similarities and differences. you must determine the action, to alarm, or to trip.
1) Difference incremental currents
B. Differences between the two methods
The methods sample the current-amplitude differences.
Thus, the methods track the fast changes in HIF bursts. Along with similarities, the two methods exhibit
differences in detecting and acting for high-impedance faults.
2) Arcing-level reference
Method F monitors the power-system phase and neutral 1) Connections
currents to set a background level, as a reference for normal A major difference is in the connection for monitoring
activity on the system. This period lasts 72 hours. A testing HIF on the power system. Connections for Method F are to
setting reduces the time to one hour. a high-sensitivity module (easily added to an existing relay)
in series with existing phase inputs and summing through the
Method P calculates a reference current by continually residual input. It does not require special CTs in the
averaging the latest samples and stores this value in a buffer.
substation yard. The Method P relay must be ordered with damage. In determining whether to employ HIF detection
the HIF function, and it requires a core-balance CT, which you should review the circuit to determine whether
can be difficult to retrofit into existing systems. conventional protection is adequate. Apply HIF tripping /
alarming depending upon experience with the soil
Sampling rate conditions, type of circuit construction, experience with
Method F has the greatest sampling rate. This is an energized downed conductors, and the nature of the circuit
advantage for sensing fast, transient arcing. Also, data load.
processing is faster. One popular application is sectionalizing the distribution
2) Spectrum coverage (harmonics/non-harmonics) system when an arc occurs. Upon detecting serious arcing or
Arcing HIFs produce a wide noise spectrum in the power a downed conductor the closest recloser controller senses the
system, both on harmonic multiples and between harmonics. arc and disconnects the line before a fire can start. This is a
The arcing energy between integer harmonics are called one-shot, trip and lockout sequence.
“non-harmonics.” Sensing and processing both harmonics Recent power-line-caused wildfires in windy conditions
and non-harmonics gives Method F more opportunity to shows that settings groups with faster response in windy
detect HIF events. weather are a good idea.
3) Arcing-level reference Consider alarming and tripping options for HIF events. It
Method P averages a period of previous samples, is not practical or advisable to trip the circuit quickly—some
whereas Method F learns a baseline for nominal power- delay is appropriate for power continuity. However, tripping
system noise over a 72-hour period. Method F compensates decreases the danger to human and animal life. It prevents
for load events that mimic line arcing. wildfires and damage to property. In addition, tripping
4) Load-event detection sooner limits legal liability and litigation.
Method F blocks operation (per-phase) for power-system Under non-wildfire conditions, consider alarming for
changes, settings flags for each phase current and for the arcing faults, and sending a crew immediately to investigate.
neutral based on the following events: Sectionalized tripping relieves the impact on supply
continuity.
 An overcurrent condition
As stated previously, the small currents in arcing faults
 A precipitous loss of load
are a small risk of power-system asset damage. Depending
 A high rate-of-change upon experience with soil conditions and the line build out
you can evaluate the likelihood of unwanted trips. Replace
 A significant three-phase event fuses with HIF-detection-equipped relays to manage
 A breaker-open condition distribution arcing protection, especially at the ends of lines.

These flags feed the load learning and extraction VIII. CONCLUSIONS
algorithm, increasing security.
Special techniques are required to detect high-impedance
5) Load learning and extraction faults (HIF). HIF methods identify significant percent of Hi-
Method F uses load learning and extraction differentiates Z faults, modern methods at 80 to 85 percent. HIF protection
between arcing, downed conductors and arcing, intact prevents fires and increases safety. Testing and experience
conductors by looking for a precipitous loss of load and/or an improves application effectiveness.
overcurrent disturbance at the beginning of an arcing
episode. This algorithm divides arcing into “arcing This paper presents two methods, Method P and Method
suspected” intact conductors and downed conductors, as well F.
as reducing false positives. Common to these methods are the following:
6) Voltage supervision / even-harmonic restraint  Arcing sense and level settings to adjust to your
This algorithm was implemented to minimize the power-system conditions
probability of a false HIF indication due to bus voltage dips,
for example, from parallel feeder faults. A fault on a parallel  Counts and other security measures to identify arcing
line can cause voltage dips that produce a decrease in the line faults and broken conductors accurately
load that can be mistaken by the HIF element as Loss of  There are outputs for arcing suspected and downed
Load. conductor, so that you can take appropriate mitigation
Even-harmonic restraint inhibits setting the overcurrent actions
flags. This is to prevent a cold-load-pickup event (lots of Differences in the two methods are the following:
motor starting and transformer inrush) from starting the HIF
logic sequence.  Connections—Method P is via a core-balance CT,
and Method F has a module in series with the phase-
VII. IMPLEMENTATION STRATEGIES current and residual current inputs. (Both can use
voltage.)
When determining how to apply HIF detection for arcing
and downed conductors, consider all the trade-offs involved.  Spectrum/harmonics—Method F looks at more of the
Unlike most forms of relay protection, downed-conductor spectrum
protection is employed primarily for safety reasons; high-
impedance faults rarely cause distribution equipment
  Load subtraction is a function in Method F that makes “High Impedance Fault Detection Technology Report of
it secure against regular feeder noise events PSRC Working Group D15,” March 1, 1996,
https://www.pes-
Applying HIF protection takes some finesse. Testing and psrc.org/kb/published/reports/High_Impedance_Fault_Detect
experience improves application effectiveness. ion_Technology.pdf

IX. REFERENCES CIGRE Working Group B5.94, "High Impedance

Faults," December 2009
B.M. Aucoin, R.H. Jones, “High Impedance Fault
Implementation Issues,” IEEE Transactions on Power General Electric MiCOM, “P40 Agile P14DInstruction
Delivery, January 1996, Volume 11, Number 1, pp 139–148 Manual,” Publication Reference P14D-TM-EN-8
B. D. Russell, B. M. Aucoin, T. J. Talley, "Detection of S. Subramanian, K. Venkataraman, Detection of High
Arcing Faults on Distribution Feeders," Texas A&M Impedance Faults in MV Distribution Systems,
University, EPRI Final Report EL-2767, December 1982 11thInternational Conference on Developments in Power
System Protection, April 2012
W. Tyska, B. D. Russell, B. M. Aucoin, "A
Microprocessor-Based Digital Feeder Monitor with High Wester, C., “High Impedance Fault Detection on
Impedance Fault Detection," 47th Annual Texas A&M Relay Distribution Systems,” GER-3993, General Electric,
Conference, March 21-23, 1994
.