Advances in Motor Protection Relay Features

Ricardo Abboud, Paulo Lima, John Needs, and Alejandro Rodriguez

Schweitzer Engineering Laboratories, Inc.

Dibyendu Bhattacharya
BP Exploration Operating Co., Ltd.

Presented at the
PCIC Europe Conference
Berlin, Germany
June 14–16, 2016
 ADVANCES IN MOTOR PROTECTION RELAY FEATURES
Copyright Material PCIC Europe

Ricardo Abboud Dibyendu Bhattacharya Paulo Lima

Schweitzer Engineering BP Exploration Operating Co., Ltd. Schweitzer Engineering
Laboratories, Inc. Chertsey Road, Sunbury Laboratories, Inc.
2350 NE Hopkins Court United Kingdom Avenida Pierre Simon de Laplace, 633
Pullman, WA 99163 Condomínio Techno Park
USA Campinas, São Paulo 13069-320
Brazil

John Needs Alejandro Rodriguez

Schweitzer Engineering Schweitzer Engineering
Laboratories, Inc. Laboratories, Inc.
Unit 19, Hollins Business Centre Hurksestraat 43, 3rd Floor
Rowley Street, Stafford, ST16 2RH Eindhoven 5652 AH
United Kingdom Netherlands

Abstract—The algorithms used by numerical relays in Thermal protection is required to detect and protect
motor thermal protection accurately simulate the electric motors against abnormal conditions like
characteristics of the motor. These algorithms use the overload, locked rotor, frequent starts, unbalance, low-
motor speed to calculate rotor heat. This results in voltage operation, and others.
proper starting for high-inertia loads connected to Installations using electromechanical relays have
motors and minimizes the cooling time, providing limited or no capabilities to accurately track motor
quicker restarts. These algorithms are performed in a heating conditions. In the case of large industrial motors,
numerical relay that also performs logging and plotting only numerical relays or intelligent electronic devices
of starting characteristics. An accurate record of motor (IEDs) with special algorithms are able to adequately
performance can therefore be obtained, providing an simulate actual rotor and stator thermal conditions.
indication of possible motor failure. Broken rotor bars Modern numerical relays are the natural choice for
cause reduced accelerating torque, increased motor retrofit applications, and they offer many improvements
heating, and increased vibrations, which can inflict over electromechanical or static relays. These
severe damage on a motor. Modern numerical motor enhancements include improved thermal modeling of
relays monitor the stator current spectrum for frequency motor heating, event reporting, sequential event
components associated with this phenomenon and use reporting, motor start reports, motor operating statistics,
motor current signature analysis to detect broken rotor additional protection features (such as the detection of
bars. For added safety, these devices can also include broken rotor bars in induction motors), and additional
arc-flash protection, allowing faults in the switchgear to control functions (such as synchronous motor starting).
be quickly detected and cleared. A comprehensive thermal model that precisely
represents motor heating is discussed later in this paper.
Index Terms—Motor protection, thermal model, According to surveys performed by the Electric Power
arc-flash protection, broken rotor bar. Research Institute (EPRI) and IEEE, 5 percent of motor
failures happen because of problems in the rotor cage
I. INTRODUCTION [2]. Early detection of a broken rotor bar is very
important to minimize motor damage and reduce the
Electric motor systems account for about 60 percent time out of operation, which consequently reduces repair
of global industrial electricity consumption [1]. They are and operation costs. The broken-bar condition can be
fundamental to industrial processes. An undesired initiated by a fracture at the junction between the rotor
operation of the system protecting the motor can lead to bar and the end ring as the result of thermal and
substantial economic losses and may even compromise mechanical stressors. Motors with high-inertia loads are
the safe operation of the plant. more susceptible to a broken rotor bar condition when
The main function of industrial electrical power starting [2]. Motor current signature analysis (MCSA) is
systems is to provide energy for these electric motors. the most popular method to detect rotor cage faults and
Motors are subject to faults and abnormal conditions is discussed later in the paper.
that can cause extensive damage. Motor damage can There are ten Occupational Safety and Health
cause delays in industrial processes, with corresponding Administration-reportable (OSHA-reportable) arc-flash
economic losses. For this reason, a reliable motor incidents every day in the United States [3]. In addition,
protection system is fundamental for increasing the up to 80 percent of all electrical worker injuries are due
reliability of industrial processes. This paper discusses to external burns created by the intense radiant heat
key elements of creating a reliable motor protection energy of an electrical arc flash [3].
system, including thermal modeling, the detection of Arc-flash detection sensors provide a cost-effective
broken rotor bars, and arc-flash detection. method to reduce arc-flash energy by minimizing

1
detection times. High-speed light detection combined the motor torque. When the rotor is locked, the stator
with high-speed overcurrent element supervision and mimics a transformer with a resistance-loaded
high-speed output contacts can provide a dependable, secondary and experiences current that is typically
secure, and fast method for tripping. This, in turn, can 6 times the rated current. Because of the rotor
contribute to reducing damage to equipment and resistance during a locked rotor being 3 times greater
significantly increasing personnel safety. Numerical than during running conditions, the effective heating due
motor relays can use multiple sensors for arc-flash to rotor ohmic losses is 108 times that of normal
detection. The most common sensors are lens-point operation [5].
sensors and bare fiber-optic sensors.
B. First-Order Thermal Model
II. THERMAL MODELS
Motor thermal protection is responsible for removing
A. Motor Thermal Limits power before a motor’s temperature reaches values
above of the maximum level permitted by the thermal
The thermal limitations of induction motors are limit curves. The actual motor heating can be calculated
specified by thermal limit curves that are plots of the with a thermal model that represents the motor thermal
limiting temperatures of the rotor and stator in units of system, as shown in Fig. 2.
I2t, where I is the positive-sequence, balanced stator
current for a three-phase motor and t is time. The curves θA
for a 7,000 hp, 6.6 kV, 900 rpm motor are shown in θM
Fig. 1.
1,000 Rth

r
Running Overload Heat
Dissipation
100 Cth
Locked Rotor (Cold)

Starting Current (80% Vn)

Fig. 2 Motor thermal system
Time (s)

10 Locked Rotor (Hot)

Starting Current (100% Vn)
The electric power applied to a motor is partially
converted into heat that is stored in the motor, causing
the temperature to rise. Thus, the temperature is a
function of current and time. These variables are the
1
basis of the thermal model that represents the motor
temperature. A first-order thermal model is used to
calculate the motor heating and is applied to the motor
thermal protection [6].
0.1 Consider the motor heating caused by the current
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5
Current (per unit [pu]) flowing through a resistor (r) that represents the
resistance of the motor windings, as shown in Fig. 2.
Fig. 1 Thermal limit curves for a 7,000 hp motor The environmental temperature is θA and the motor
The starting curves are an indication of the amount of temperature is θM.
time and associated current for the motor to accelerate This simple first-order thermal system is modeled by a
from a stop condition to a full running condition. In Fig. 1 thermal resistance (Rth) to the environment and a
there are two starting curves: the solid curve represents thermal capacitance (Cth), with the motor considered a
the motor starting at rated voltage and the dashed curve homogeneous body.
represents the motor starting at 80 percent of the rated Fig. 3 illustrates the first-order thermal model used to
voltage. represent the motor heating [5]. The major components
Thermal protection is required to detect and protect of the model are as follows:
electrical motors against abnormal conditions. 1. Heat source. Heat flow from the source is
Unbalances produce negative-sequence currents that I2 • r watts (J/s).
can cause rotor overheating [4]. A low-voltage condition, 2. Thermal capacitance (Cth). This represents the
if it occurs during normal motor operation, can cause the capacity of the motor to absorb heat from the
motor to jam. If a low-voltage condition occurs during heat source. The unit of thermal capacitance is
starting, the motor may not start normally because the J/°C.
motor torque might be less than the load torque. In both 3. Thermal resistance (Rth). This represents the
cases, the resulting overcurrent can damage the motor. heat dissipated by a motor to its surroundings.
Motor stall occurs during the start operation when the The unit of thermal resistance is °C/W.
motor torque cannot overpower the load torque and the 4. The comparator. This creates a trip condition
motor cannot start moving. when the calculated motor pu temperature
The cause of a locked rotor may be a failure of the exceeds a preset value that is based on the
load bearings, a failure of the motor bearings, a low motor manufacturer’s data, as explained in more
supply voltage, single phasing, or a load that exceeds detail later in the paper.

2
 θtrip + When a motor is de-energized, it does not require
–
Trip thermal protection per se; however, it does need to be
θM locked out and not allowed to re-energize until it cools
down sufficiently to offer further service. When current
Heat Source
ceases to flow in the thermal circuit shown in Fig. 3, the
Cth Rth circuit reconfigures, as illustrated in Fig. 5, and the
I2 • r
capacitor discharges according to the value of Rth.
Motor Heat
θtrip +
Dissipation Trip
–
θM
Fig. 3 First-order thermal model
Heat produced by the heat source is transferred to the
motor, which in turn dissipates the heat to the Cth Rth
surrounding environment. Motor thermal protection is
implemented in modern numerical relays based on this
Motor Heat
thermal model. The relay input current is the phase Dissipation
motor current. The purpose of motor thermal protection
is to allow the motor to start and run within the Fig. 5 Thermal model for motor stopped state
manufacturer’s published guidelines and to trip if the
By applying the presented approach to thermal
motor heat energy exceeds those ratings due to
modeling, it is possible to emulate the dynamic thermal
overloads, negative-sequence current, or locked-rotor
behavior of a motor, as shown in Fig. 6, to prevent
starting.
damaging temperatures for any operating condition.
Because the positive- and negative-sequence rotor
resistances (Rr1 and Rr2) are functions of the motor’s θ
speed, the model becomes nonlinear. An approach used
by some relay designers employs two linear models for
two different stages of the motor, as shown in the Fig. 4.
The limit current (ILIM), which determines when each
Heating Cooling Time
model applies, is defined by the designer. Certain relays
use a limit of 2.5 times the full load current of the motor. Fig. 6 Example of thermal model response for different
motor states
Start/Stall State
C. Stator and Rotor Thermal Models
3 • (I12 + I22 ) C'th R'th θ

For I1 > ILIM To accommodate the differences in stator and rotor

thermal properties, the first-order thermal model can be
Relay
divided into two separate thermal models, as follows:
Running State 1. The rotor model consists of the following
(I 2
1 + 5 • I22 ) Cth Rth θ elements:
− A starting element that protects the rotor
For I1 < ILIM
during the starting sequence.
− A running element that protects the rotor
Fig. 4 Thermal models for start/stall and running states when the motor is up to speed.
Because locked-rotor heating occurs over just a few 2. The stator model protects the stator during
seconds, the start/stall state thermal model assumes starting and when the motor is up to speed.
that no heat is lost to the surroundings and the resistor 1) Rotor Model
is removed from the thermal circuit. The motor’s rated
In the rotor model, the transition from one element to
locked-rotor current defines the thermal trip value.
the other is set at 2.5 times the rated full load current of
When the motor is running, it returns heat energy to
the motor. The high-inertia starting solution using the
its surroundings through radiation, conduction,
slip-dependent thermal model described in the following
convection, and (in some cases) forced cooling. The
subsection only affects the rotor element design.
running state thermal model provides a path for that
It is valid during a starting or stalled-motor condition to
energy return through the thermal resistance (Rth)
neglect ambient heat losses. This results in a
resistor, as shown in Fig. 4.
conservative estimate of the temperature to ensure good
The motor thermal characteristics (Rth and Cth)
operation. This is equivalent to eliminating (making
depend on many design factors. Among others, they
infinite) the thermal resistance from the model.
depend on the motor size (mass). This explains why it is
The rotor starting thermal limit is expressed in terms
so difficult or impossible to emulate large motor thermal
of the maximum time (motor safe stall time) that the
behavior with small bimetal devices. This is a clear
corresponding locked-rotor current (ILRA) can be applied
advantage for numerical relays, in which it is possible to
to a motor, as calculated in (1).
set different values for the motor parameters.
A slip-dependent thermal model of the rotor is 2
ILRA
θtrip = • TSTALL (1)
discussed later in the paper.

3
 The rotor resistance at a speed of zero is typically θtrip –
3 times that of the rotor resistance when the motor is at +
Trip
its rated speed. For this reason, the effect of the Running
State
positive- and negative-sequence currents is multiplied
by a heat source factor of 3 in the starting motor rotor (I
2
1 + 5 • I22 ) Cth Rth θ

thermal model.
Incorporating all of these changes results in the I2t
starting element of the first-order thermal model Fig. 9 Running motor rotor thermal element
illustrated in Fig. 7.
2) Stator Model
θtrip – The running motor overload curves show the stator
Trip
+ thermal limit. These curves fit the time-current equation
Start/Stall
State (2), where τ is the stator thermal time constant, I is the
3 • (I12 + I22 ) Cth Rth θ stator current in pu of rated current, I0 is the initial
current in pu of rated current, and SF is the motor
service factor. This equation has the form of a first-order
thermal model.
Fig. 7 I2t starting element
 I2 − I02 
To match the heat source factor of 3, the selected t = τ • ln  2 2 
(2)
thermal capacitance is also 3. When the motor positive-  I − SF 
sequence current is equal to the locked-rotor current,
the estimated heat reaches the trip value within the The motor stator thermal time constant is a setting
locked-rotor time limit. Therefore, for starting protection, parameter for the running motor thermal model, and it
only the motor nameplate data are needed for the can be calculated from the stator thermal limit curves by
starting motor rotor thermal model. applying (2).
If the temperature response of this model is plotted Fig. 10 depicts the equivalent circuit that corresponds
against the line current of the motor, the response curve to the stator first-order thermal model. In this case, the
is a straight line, as illustrated in Fig. 8. thermal capacitance (Cth) equals the stator thermal time
constant (τ). Assigning a value of 1 to the thermal
1.0 10
resistance (Rth) resistor provides a value equal to τ for
Temperature (U/UL), Rotor R • 0.01

0.9 9 θ the time constant (Rth • Cth) of the equivalent circuit.

0.8 8
Because the positive- and negative-sequence currents
0.7 7 Rotor Temperature
have the same heating effect on the stator, the heat
0.6 6
source equals I12 + I22 . When Rth = 1, the trip threshold
Current

0.5 5 Motor Current

0.4 4 should equal SF2.
0.3 3
R1 = RM Constant SF2 –
0.2 2
Trip
0.1 1 Motor Torque +
0 0
0 5 10 15 20 25 30
Time (s) (I 2
1 + I22 ) Cth Rth θ

Fig. 8 I2t starting element response curve

Note that this model keeps the rotor resistance
Fig. 10 Stator running thermal model
constant at RM, which occurs at a standstill, where the
slip (S) is equal to 1.0 pu. D. Slip-Dependent Thermal Model
Fig. 9 depicts the electric analog of the first-order rotor
thermal model for the motor running condition. When the Reference [7] derives an expression for slip-
motor is running, it returns heat energy to its dependent rotor resistance [Rr(S)] in terms of the
surroundings. The running motor rotor thermal element maximum rotor resistance (RM), which occurs at a
provides a path for that energy return through the standstill (S = 1), and the normal rotor resistance (RN),
thermal resistance (Rth) resistor. In this state, the trip which occurs at the rated motor speed (S = rated slip).
threshold “cools” exponentially from a locked-rotor This expression is shown in (3):
threshold to the appropriate threshold for the running
condition using the motor thermal time constant. This Rr (S) =
(RM − RN ) • S + RN (3)
emulates a motor temperature that cools to the steady-
state running condition. In the running condition, the
model considers the rotor resistance to have the rated
speed value (Rr = RN).

4
 Fig. 11 shows the rotor resistance during starting. θtrip –
Trip
0.024 Start/Stall State +

0.022  2 Rr1 ( S ) 2 Rr 2 ( S ) 
 I1 • R + I2 •
RM 
 Cth Rth θ
Rotor Resistance Rr (pu)

 M
0.02

0.018
Fig. 12 Slip-dependent thermal model
0.016
E. High-Inertia Starting
0.014
In high-inertia starting, the time to accelerate a motor
0.012 up to its rated speed is equal to or longer than its
0.01
locked-rotor time limit. High-inertia loads, such as
0 5 10 15 20 25 30 induced-draft fans, require long accelerating times and
Time (s) can exceed the allowable locked-rotor thermal limit.
Fig. 11 Rotor resistance during starting Prolonged starts are safely permitted in some situations
because the rotor resistance Rr(S) is a function of slip
The values in pu of the maximum rotor resistance and decreases as the motor accelerates.
(RM) and the normal rotor resistance (RN) can be The starting current of an induction motor at the
calculated by using (4) and (5) [8]. beginning of the start nearly equals the locked-rotor
current magnitude but has a lesser heating effect during
SLR
RM = TLR • 2
(4) the start because rotor resistance decreases as the
ILRA motor accelerates to rated speed.
A comparison of the conventional I2t starting element
where:
response curve to the slip-dependent starting element
TLR is the locked-rotor torque in pu. response curve is shown in Fig. 13. The comparison
SLR is the slip when the locked-rotor clearly shows that, because of the decreasing rotor
condition = 1. resistance as the motor accelerates, rotor temperature is
ILRA is the locked-rotor current in pu. not a linear relationship. This provides the ability to
facilitate high-inertia starts without premature motor
SN
RN = TN • 2
(5) trips.
IFLA
1.0 10
Temperature (U/UL), Rotor R • 0.01

where: 0.9 9
I2t
TN is the nominal torque in pu. 0.8 8
0.7 7 Slip
SN is the slip at the rated speed. Dependent
0.6 6
IFLA is the motor’s rated current in pu.
Current

0.5 5 Motor Current

To establish the slip-dependent thermal model, it is 0.4 4
necessary to incorporate the slip-dependent rotor 0.3 3
resistance into the heat source of the thermal model 0.2 2
shown in Fig. 7. 0.1 1 Motor Torque
Expressing the slip-dependent resistance value Rr(S) 0 0
in terms of its maximum value (RM) and substituting it 0 5 10 15 20 25 30
Time (s)
into the heat source equation provides (6):
Fig. 13 Comparison of starting element response
Rr (S)
W I2=
= • r I2 • (6) curves
RM
III. DETECTING BROKEN ROTOR BARS
Breaking (6) down into positive-sequence (Rr1) and
negative-sequence (Rr2) components accommodates The detection of a broken rotor bar as soon as it
motor heating caused by balanced current (positive occurs is essential to minimize damage and reduce the
sequence) and any current unbalance (negative time and cost to repair the motor. According to [2], the
sequence) that is present. broken-bar condition results from mechanical and
Replacing the heat source of Fig. 7 with (7) provides thermal stresses that lead to a fracture at the junction
the slip-dependent thermal model shown in Fig. 12. between the rotor bar and the end ring.
In order to detect a broken-bar condition, MCSA can
Rr1(S) 2 Rr 2 (S) be applied [9] [10]. With this method, the frequency
WTOTAL I12 •
= + I2 • (7)
RM RM spectrum of the stator current is calculated and analyzed
to check whether lower and upper sidebands (i.e.,
[1 ± 2S]f0, where f0 is the nominal frequency) are present
in the stator current, indicating that the rotor has broken
bars. The magnitude of the sidebands is proportional to
the number of bars that are broken.

5
A. Broken-Bar Detection Element Low-frequency load oscillations can cause current
signatures similar to those of a motor with broken bars
Reference [2] describes a broken-bar detection [2]. One method of differentiating the two is to apply an
element (BBDE) with zero settings. The BBDE algorithm algorithm that detects the presence of a greater-than-
runs periodically to detect a broken-bar condition. It is normal frequency component, which may indicate a
composed of three steps: initialization, data collection, broken-bar condition.
and data processing. During the initialization step, the To detect broken rotor bar conditions in different
algorithm calculates a new current from the phase situations and monitor how they evolve, the event
currents that does not include the portion that flows to history and fast Fourier transform function can be
ground. It then records this current magnitude and the applied in conjunction. This makes it possible to
system frequency. A motor-running condition is detected differentiate situations involving voltage sources with
during the data collection step using these values as a low-frequency components and oscillating loads from
reference. the broken-bar condition.
During the data collection step, the algorithm squares
the current calculated in the initialization step to IV. ARC-FLASH DETECTION
decouple the frequency of interest from the power
system frequency and to move both sidebands into the Applying traditional time coordination for industrial
same frequency. This squared current is also passed systems, like motor control centers (MCCs), can lead to
through a low-pass filter. Finally, the algorithm stores a high fault-clearing times. Fault clearing times are
collection of consecutive samples, referred to as a typically between 0.5 and 1.0 second. However, high
samples window, in digital memory. fault current in combination with long fault clearing times
During the data processing phase, the algorithm causes a very high arc-flash energy, which is a highly
computes the fast Fourier transform of the samples undesirable situation [11]. Therefore, the goal is to
window data and then calculates the magnitude reduce the fault clearing time in order to reduce the arc-
associated with each frequency component. Finally, the flash energy.
average magnitudes of the frequency components are One option to reduce arc-flash energy in radial
compared with a healthy motor threshold. Figure 14 in substations is to apply a simple and economical zone-
[2] shows the thresholds on the frequency spectrum of a interlocked blocking scheme, sometimes called a fast
motor running at 50 percent load with one broken bar. bus-tripping scheme. This scheme provides relatively
high-speed fault clearing for buses that do not have
B. Experimental Results differential protection. Instead of relying on a traditional
coordination interval in the bus main relay, this scheme
Reference [2] also presents some experimental only requires a short delay to allow the feeder relays to
results of the method described in this section using block the bus main relay for a fault external to the bus.
actual broken rotor bars. For the experimental tests, a The scheme can operate for bus faults in approximately
healthy motor and motors with one, two, or three broken 2 to 3 cycles.
bars had their current frequency spectrums compared Fig. 14 shows an event report for a real fault on a
when running at 50 percent of the nominal load. The 480 V bus with an arc flash. The fault started as a
results are shown in Figure 27 in [2]. The values of the single-line-to-ground (SLG) fault on Phase B. After
sideband peaks clearly increase as the number of 1 cycle it evolved into a three-phase fault with a
broken bars increases. considerable increase in the fault current level. Even in
Different load conditions were also experimentally impedance-grounded systems that have low current
tested, and the results are shown in Figure 28 in [2]. The levels for SLG faults, such faults represent a high risk in
frequency of the sideband peaks decreases as the load terms of arc flash because of the fault evolving.
level decreases, and the peaks become undetectable Another interesting observation about Fig. 14 is the
when the motor is unloaded. fact that the fault current is not a pure sine wave at the
Reference [2] describes how broken bars can be fundamental frequency. This is because the arc
erroneously detected during low-frequency source resistance is not constant, and it plays an important role
voltage oscillations. It also recommends some strategies in low-voltage systems. Overcurrent relays that operate
to differentiate a broken-bar condition from a voltage based on fundamental components calculate an
oscillation (e.g., verify if all motors connected to the incorrectly low value for the fault current, which can
same feeder show the same current spectrum, and compromise the tripping of the instantaneous
measure the voltage farther away from the motor and overcurrent element.
closer to the source to confirm the presence of low- The most effective method to reduce fault clearing
frequency components on the power supply). In times is applying arc-flash protection with light sensors
addition, these low-frequency voltage oscillations may combined with fast overcurrent elements. Some modern
not be present in the system all of the time. They numerical motor relays have incorporated arc-flash
typically appear when the system is heavily loaded or detection and support the connection of multiple
very lightly loaded. sensors.

6
 Fault Inception
(Phase B to Ground)

Fault Evolves to
Three-Phase

Fig. 14 An event report for a real fault on a 480 V bus with an arc flash

The purpose of detecting arc flashes is to accelerate makes possible the detection of arc flashes in large
accurate decisions to trip the circuit breaker and areas with only one sensor.
interrupt the fault. Arc-flash detection in a protective
relay minimizes trip time, cost, and complexity. Enabling
arc-flash detection in the relay makes use of the current
monitoring and protection already in the circuit.
Arc-flash detection sensors provide a clear
measurement of an arc flash because the light emitted
during an arc-flash event is significantly brighter than the
normal background substation light. It is also possible to
supervise their operation with a fast overcurrent
element, as discussed later in this section. The light
surge is visible from the initiation of the flash and is
easily detected using proven technology. The most
common sensors are lens-point sensors and bare fiber-
optic sensors.
The light is channeled from the sensor to the detector
located in the protective relay. Monitoring the system
Fig. 16 Bare fiber-optic cable
integrity is accomplished using a fiber-optic loop. In the
case of lens-point sensors (see Fig. 15), each lens has Arc-flash detection systems typically use a
an input and an output connection. The input is combination of lens-point and bare fiber-optic sensors.
connected to a transmitter in the relay, and the output is Proper installation of the sensors and relays provides
connected to a detector in the relay. This loop logical detection and trip points in any system.
connection allows periodic testing of the system by Sensors should be located where arc-flash detection
injecting light from the transmitter through the loop and can trip the corresponding upstream circuit breaker.
back to the detector. This loop connection system works Using multiple sensors and having motor and feeder
with either a lens-point sensor or a bare fiber-optic relays that support connections to light sensors, as
sensor. shown in Fig. 17, provides 100 percent coverage for arc-
flash protection that operates in the order of 2 to 3 ms.

Relay Relay

Bare Fiber- Lens-Point

Optic Sensor Sensor

Fig. 15 Lens-point sensor

Relay Relay Relay Relay
A bare fiber-optic sensor consists of a high-quality
plastic fiber-optic cable without a jacket (see Fig. 16).
The clear fiber-optic cable becomes a lens that captures
Fig. 17 Typical arc-flash detection system with sensors
light from the area. Using a bare fiber-optic sensor and relay-to-relay communication

7
 The installation of sensors varies depending on the designs, processes, and procedures. The addition of
switchgear manufacturer, type of gear, and number of arc-flash detection improves the safety of installations.
sections. Multiple sensor inputs provide coverage and Arc-flash detection systems can be designed into new
sectioning options. One bare fiber-optic sensor can switchgear or retrofitted into existing gear. The security
provide excellent coverage for an entire bus section. of arc-flash detection systems can be increased by
Lens-point sensors provide better detection in small, parallel overcurrent and light-detection systems.
confined spaces.
One obstacle to using light sensors is the need to VI. REFERENCES
measure and adjust for changing ambient light levels.
Relays store analog measurements of light and current [1] T. Fleiter, W. Eichhammer, and J. Schleich,
values. Users can view these measurements and set the “Energy Efficiency in Electric Motor Systems:
normal light levels for the application. Relay event Technical Potentials and Policy Approaches for
reporting also provides a commissioning and Developing Countries,” United Nations Industrial
troubleshooting tool with time-tagged events, including Development Organization, Vienna, Austria,
sensor light levels. November 2011. Available: https://www.unido.org.
In order to add security to an arc-flash detection [2] C. Pezzani, P. Donolo, G. Bossio, M. Donolo,
scheme, a high-speed overcurrent element can be A. Guzmán, and S. E. Zocholl, “Detecting Broken
applied in conjunction with the light sensors, as shown in Rotor Bars With Zero-Setting Protection,”
Fig. 18, without sacrificing trip speeds. The high-speed proceedings of the 48th Industrial & Commercial
overcurrent element is based on raw samples in order to Power Systems Technical Conference, Louisville,
avoid the long delays of filtering. The added advantage KY, May 2012.
of processing the arc-flash detection in the protective [3] T. E. Neal and R. B. Hirschmann, “The Myths and
relay is the ability to use a true overcurrent Realities of Arc Flash Protection,” Electric Energy
measurement as a supervising element to improve Magazine, May/June 2004.
security. Setting the current level below the normally [4] P. Whatley, M. Lanier, L. Underwood, and
expected load enables the arc-flash detector as the trip S. Zocholl, “Enhanced Motor Protection With the
mechanism and removes any time lag; however, it Slip-Dependent Thermal Model: A Case Study,”
sacrifices security and makes the system dependent on proceedings of the 34th Annual Western Protective
light detection alone and must be avoided. Relay Conference, Spokane, WA, October 2007.
[5] S. E. Zocholl, AC Motor Protection. Schweitzer
Light Sensor Engineering Laboratories, Inc., Pullman, WA,
& Arc-Flash
Protection
2003.
[6] S. E. Zocholl, “Optimizing Motor Thermal Models,”
Current Input
proceedings of the 53rd Annual Petroleum and
Chemical Industry Conference, Philadelphia, PA,
Fig. 18 Light detection in combination with high-speed September 2006.
overcurrent element [7] S. E. Zocholl and G. Benmouyal, “Using Thermal
Limit Curves to Define Thermal Models of
V. CONCLUSIONS
Induction Motors,” proceedings of the 28th Annual
Motor protection is greatly enhanced by numerical Western Protective Relay Conference, Spokane,
relays. Induction motors require thermal protection to WA, October 2001.
prevent overheating for cyclic as well as steady-state [8] S. E. Zocholl and E. O. Schweitzer, III, “Protection
overloads. of Large Induction Motors – Practice vs. the Actual
The heat rise in a motor caused by I2 • r watts is a Model,” proceedings of the 39th Annual Georgia
first-order process that can be represented by a first- Tech Protective Relaying Conference, Atlanta, GA,
order thermal model, which a motor relay can use to May 1985.
continuously calculate the temperature in real time. The [9] H. W. Penrose, Electrical Motor Diagnostics,
calculated temperature is monitored to prevent 2nd ed. Success by Design, 2008.
overheating. [10] G. B. Kliman, R. A. Koegl, J. Stein, R. D. Endicott,
The slip-dependent thermal model tracks the motor and M. W. Madden, “Noninvasive Detection of
temperature more accurately than the I2t model, thus Broken Rotor Bars in Operating Induction Motors,”
facilitating high-inertia starts without the use of speed IEEE Transactions on Energy Conversion, Vol. 3,
switches. Issue 4, December 1988, pp. 873–879.
BBDE algorithms that apply MCSA in modern [11] J. Buff and K. Zimmerman, “Application of Existing
numerical motor relays, in conjunction with the event Technologies to Reduce Arc-Flash Hazards,”
history and the fast Fourier transform function, permit proceedings of the 60th Annual Conference for
the detection of broken rotor bars under a wide variety of Protective Relay Engineers, College Station, TX,
motor conditions. The detection element identifies the March 2007.
most common broken-bar cases. The event history
records and makes possible more accurate analysis of
when problems start and how they evolve.
Arc flashes present a clear danger to personnel.
Worker safety should always be at the forefront of

8
 VII. VITAE Alejandro Rodriguez graduated from the Universidad
Metropolitana, Venezuela, in 2006 with an M.Sc. in
Ricardo Abboud received his B.S.E.E. degree in electrical engineering. In 2006, he joined Honeywell in
electrical engineering from Universidade Federal de Caracas, Venezuela, as a junior engineer in sector
Uberlândia, Brazil, in 1992. In 1993, he joined CPFL automation, programming and configuration of control
Energia as a protection engineer. In 2000, he joined strategies, and human-machine interface development.
Schweitzer Engineering Laboratories, Inc. (SEL) as a In 2009, he joined Gruppo Loccioni in Angeli di Rosora,
field application engineer in Brazil, assisting customers Italy, as a junior electrical engineer in electrical industrial
in substation protection and automation. In 2005, he system design and power generation systems. He joined
became the field engineering manager, and in 2014, he Schweitzer Engineering Laboratories, Inc. in 2010 as a
became the engineering services manager. In 2016, he junior engineer and is currently working as an
transferred to Pullman, Washington, and is currently an application engineer in protection in Eindhoven,
international technical manager. He is a certified Netherlands. He is a licensed professional electrical
instructor at SEL University, and has authored and engineer.
coauthored several technical papers. alejandro_rodriguez@selinc.com
ricardo_abboud@selinc.com

Dibyendu Bhattacharya graduated from Jadavpur

University, India, in 1991 with a bachelor of electrical
engineering first-class honours degree. He worked for
14 years with the Indian Oil Corporation Ltd. refineries
division, testing, commissioning, troubleshooting, and
maintaining electrical equipment and power systems. He
worked for the Kuwait National Petroleum Company, as
well as Fluor Limited and KBR, as a lead engineer for oil
and gas projects before joining BP. He has presented
papers at PCIC London and coauthored one paper. He
is a Chartered Engineer and a Fellow of the Institution of
Engineering and Technology (IET) in the United
Kingdom.
dibyendu.bhattacharya@uk.bp.com

Paulo Lima received his B.S.E.E. from Federal

University of Itajubá (UNIFEI), Brazil, in 2012. He has
been an application protection engineer for Schweitzer
Engineering Laboratories, Inc. since 2012, providing
technical support for industrial customers in Brazil.
paulo_lima@selinc.com

John Needs graduated from the University of Bath in

1981 with a degree in physics with physical electronics.
He started work with GEC Measurements in 1982 as a
development engineer, initially working in type testing
and then programming distance relays. Next, he was a
relay engineer for National Grid and later joined Alstom,
first as an application engineer and then as an instructor
in the training department. In 1998, Mr. Needs joined
Schweitzer Engineering Laboratories, Inc., where he
currently works as a regional technical manager.
john_needs@selinc.com

Previously presented at the 2016 PCIC Europe Conference,

Berlin, Germany, June 2016.
© 2016 PCIC-Europe – All rights reserved.
20160419 • TP6759