Fundamentals of a Motor Thermal Model and its

Applications in Motor Protection

B.Venkataraman, B.Godsey W. Premerlani E.Shulman, M.Thakur, R.Midence
Black & Veatch Corporation GE Global Research GE Multilin
Niskayuna, New York Markham, Ontario

1. Abstract
This paper discusses the fundamentals of a motor thermal 12%, as in the Pulp and Paper industry. Downtime in a factory
model and its mathematical interpretation and physics for the can be very expensive and, in some instances, may exceed
different stages of motor operation. (overload, locked rotor, too the cost of motor replacement. Proper machine protection is
frequent or prolonged acceleration, duty cycling applications). required to minimize the motor failure rate, prevent damage
It explains Thermal Model Time Constants and other technical to associated equipment and to ensure both personnel safety
parameters that cause the biasing of the thermal model and production targets.
algorithm. Other topics covered in this paper show that (a) The document “Report of Large Motor Reliability Survey of
detailed motor data sheet information, and (b) coordination Industrial and Commercial Installations”, published by the IEEE
between the protection engineer and the motor supplier, can Motor Reliability Working Group [3] contains the results of IEEE
lead to proper selection of motor thermal protection parameters. and EPRI surveys on motor reliability and major causes of motor
This paper presents a closer look at motor stall, acceleration failure. The summary of these results is shown in Table I.
and running thermal limit curves. It also explains the concept
of thermal capacity and elaborates on how thermal capacity is In spite of different approaches and criteria (IEEE failure groups
evaluated in motor protection devices. The following points are are formed according to “cause of failure” and EPRI according to
also covered in this paper: “failed component”) both studies indicate a very similar failure
percentage associated with mechanical- and electrical-related
• Discusses some additional methods, such as voltage- machine problems.
dependant and slip-dependant motor overload curves, Analyzing the data from this table we can conclude that
employed to evaluate thermal capacity in many failures are directly or indirectly related to, or caused
nonstandard motor applications,
IEEE Study EPRI Study Average
• Presents the concept of matching thermal Failure Contributor % Failed Component % %
time constants for motor cyclic loads cases.
Persistent Overload 4.2% Stator Ground Insulation 23.00 Electrical
In addition, the response of a thermal Related
model algorithm in practical applications is Normal Deterioration 26.40% Turn Insulation 4.00
Failures
demonstrated. Bracing 3.00
Core 1.00
• Describes a real case example showing how Cage 5.00 33%
to apply and fine-tune the thermal model in Electrical Related Total 30.60% Electrical Related Total 36.00%
high-inertia load application. High Vibration 15.50% Sleeve Bearings 16.00 Mechanical
Related
Poor Lubrication 15.20% Antifriction Bearings 8.00
• Explores in this context, some of the key Failures
Trust Bearings 5.00
topics that will ensure safe operation of the
Rotar Shaft 2.00
motor while promoting satisfactory motor
Rotor Core 1.00 31%
design characteristics.
Mechanical Related Total 30.70% Mechanical Related Total 32.00%
High Ambient Temp. 3 Bearing Seals 6.00 Environmental
2. Introduction Abnormal Moisture 5.8 Oil Leakege 3.00
Maintenance
& Other
Abnormal Voltage 1.5 Frame 1.00 Reasons
Induction motors are the workhorses of any Related
industrial plant. Typical motor applications Abnormal Frequency 0.6 Wedges 1.00
Failures
include pumps, fans, compressors, mills, Abrasive Chemicals 4.2
shredders, extruders, de-barkers, refiners, Poor Ventilation Cooling 3.9
cranes, conveyors, chillers, crushers, and Other Reasons 19.7 Other Components 21.00
blowers. Statistics have shown that despite their Environmental Reasons & 38.70% Maintenance Related & 32.00% 35%
reliability and simplicity of construction, annual Other Reasons Total Other Parts Total
motor failure rate is conservatively estimated Table 1.
at 3-5% per year, and in extreme cases, up to Summary of IEEE and EPRI Motor Reliability Surveys.

Motor Thermal Model Protection Applications 41

 by, extensive heating of the different motor parts involved in In high-inertia load applications voltage monitoring can be
machine operation. That is why we find accurate tracking of used in the thermal model algorithm to dynamically match
motor thermal status and adequate response of the motor the thermal limit to different starting conditions. In some
control system to abnormal situations to be very important. applications speed sensors are employed to detect slow rotor
rotation or motor stall.
Modern trends in motor design and construction are moving
in the direction of making motors more compact and efficient. Another important part of thermal model implementation
The use of inorganic insulation materials such as fiberglass is “Expected values stored in MPD”. This term implies that
and silicon resins provides improved dielectric motor insulation information is available from the motor designer and motor
properties compared to legacy materials such as cotton manufacturer, that is related to the thermal reserve, allowed
and varnish. But at the same time some new materials are performance and thermodynamics of the motor in question.
more vulnerable to excessive heating. Another important
The motor is not a homogeneous body and even one component
consideration that should be consdered in tracking the thermal
can be presented as a combination of nodes connected
state of the motor, is heating overestimation, which can also
via thermal resistance to each other and external ambient
cause undesirable motor stoppage and hence potentially costly
conditions. For example, the stator has slot copper, end-head
interruption of processes. The statements above clearly explain
copper, teeth and a core. Each node is characterized by its own
the importance of an accurate thermal estimate of a motor in
rate of temperature change. [6]
service.
That is why in order to do the full analysis and detect a boundary
Currently this task (precise motor thermal protection) is strongly
for normal operation, motor designers always target the
supported by modern technology. The developed algorithms can
development of the most detailed model including electrical,
be implemented in microprocessor devices, which are capable
mechanical, thermal, and chemical components. But once
of providing a desirable level of accuracy and flexibility.
a motor is properly designed and constructed to its intended
The thermal algorithm operates as per the following sequence: specifications, a less detailed model is adequate to provide
thermal protection by evaluating thermal risk with reference to
• Real-time motor data is supplied to microprocessor
motor data sheets and thermal damage curves.
device.
Common sense dictates reliance on a complete motor analyses
• This data is processed according to the firmware thermal
to determine the correspondence of the MPD algorithm variables
algorithm program and compared with expected values,
to the data typically available from the motor manufacturer.
stored in memory.
MPD also incorporates simplified algorithms modeling physical
• The protection device computes the analog value, which is motor states and processes. This approach allows us to attain
compared with the programmed threshold. an adequate level of thermal protection in modern MPD, for
any application, by handling the available motor information. In
• The protection device triggers the digital outputs if the
trying to keep the algorithm simple we face another challenge.
compared analog value exceeds this threshold.
It is rather difficult to relate the thermodynamic behavior of the
The ideal analog method for modeling the thermal image in the motor under steady-state conditions, with the rapid stator and
Motor Protection Device (MPD) would be to embed non-inertial rotor heating that occurs during thermal motor transients such
temperature sensors into the stationary (stator) and rotating as acceleration, stall and cyclic load change. The algorithm must
(rotor) parts of the motor structure. However, it is not feasible also account for heat transfer from the motor’s winding to the
to install temperature sensors in the rotors for technical housing and from the housing to the free (ambient) air. To resolve
reasons, reliability and cost. An additional reason to reject such this issue the “time before trip” parameter was selected as the
temperature sensors as the main basis for thermal protection, common criterion for thermal condition evaluation. Actually,
is the fact that the traditional Resistance Temperature Detector for motor acceleration and stall conditions, the safe stall time
(RTD) has a relatively slow reaction time and can’t respond specified by motor designers, is the only objective estimate of
adequately to the high speed of the heating process during the maximum allowable motor temperature, because of the
motor acceleration. real difficulty of directly measuring the rotor temperature. [6]
Stator RTDs actually provide realistic results in monitoring the Based on the discussion in this section of the paper, the main
temperature under balanced motor conditions, but again they motor thermal algorithm requirements can be summarized as
are not suitable for monitoring the fast thermal transients. follows:
Alternatively, a main real-time input thermal model could use • Accuracy. A precise estimate of the thermal motor image.
3-phase motor current. The electrical energy applied to the Consideration of different motor applications, such as
motor is partially transformed into heat which is stored in the variable frequency, voltage unbalance, long acceleration,
motor. Thus this heat is a function of current and time. This fact, cyclic loads. Reference to data specified by motor
plus some other factors and assumptions that will be covered designers.
further in this paper, are employed to develop and implement
• Simplicity. The algorithm is easy to understand. A simple
the current-based thermal model. 3-phase current values
way to calculate the thermal estimate of the motor for the
measured in real-time are also used in special algorithms
operational sequence in question.
applied to detect different stages of motor operation: stopped,
start, run, overload. • Dependability. The capability of monitoring the thermal

42 Motor Thermal Model Protection Applications

 capacity at any moment of motor operation. The thermal • A Start Supervision Element preventing an excessive
estimate is maintained and responds adequately to MPD number of motor starting sequences.
power failure events.
• A mechanical Jam Detector.
• Compliance to industry standards. The algorithm
• An acceleration limit timer.
must meet the requirements, and should follow the
recommendations listed in, IEEE Guide for AC Motor • Phase Short Circuit and Ground Fault Protection Functions.
Protection (Std C37.96-2000) [9] and IEEE Guide For the
• Voltage and Frequency elements ensuring motor operation
Presentation of Thermal Limit Curves for Squirrel Cage
within specified limits. Phase Reversal Detection.
Induction Machines (Std 620-1996) [10].
• Power Elements to monitor and respond to abnormal motor
• Easy Setup. The parameters required to set up the thermal
loading conditions.
model are obtained from the standard set of motor data
readily available from motor manufacturers. • MPD failure detection.
• Reliability. The model is supported by alternative motor • Communication capability to host computers to allow easy
temperature evaluation methods, based on RTD’s integration into existing DCS and SCADA systems.
monitoring. This backup method is extremely useful in
• Cost justifiable.
cases where the thermal process significantly deviates
from what was expected because of abnormal ambient • Can be adapted (retrofitted) to multi-vendor MCC’s and
temperatures or motor cooling impairment. motor starters.
• Flexibility. The possibility of apply the model even in very • Industrially hardened by means of a conforming coating, to
unusual cases. work in mill environment.
In addition to the accurate thermal model the state of the • Highly accurate predictions of mechanical and insulation
art Motor Protection Device should be equipped with the failure, as well as the broken rotor bar condition, without
enhancements and additional functionality listed below. removing the motor from service and without the need for
resident experts.
• RTD Inputs for absolute temperature monitoring, alarming
and tripping of the motor at high temperatures. • The capability of reading/capturing motor currents and
voltages during electrical system faults.
• Temperature-based stator thermal estimate, capable to
correct main thermal model in the abnormal operational • The capability of recording and storing in the device’s
conditions nonvolatile memory, time-stamped events related to
abnormal motor situations.
• A temperature-based stator thermal estimate, capable
of correcting the main thermal model under abnormal Additional protection functions can be provided using expensive
operating conditions equipment such as vibration sensors and/or instruments to
display the current spectrum of the motor, to predict incipient
• Wide selection of thermal overload curves; standard for
failures. These are not covered in this document.
typical applications, user defined for unusual applications
and voltage dependant for special applications, featured 3. Thermal Protection Theory
long starts of high inertia loads.
• A Motor Start Lockout feature inhibiting the start of the There are two main types of thermal risks for an overheated
machine in the case of non-availability of sufficient thermal motor: stator insulation may degrade and/or rotor
reserve to complete the acceleration. The lockout time is conductors may decrease their capability to resist bending
calculated based on the available thermal capacity, the (deformation) forces or even melt. Deterioration of stator
maximum learned value of Thermal Capacity Used (TCU) insulation presents the chemical process that is governed by
during one of the last 5 successful starts and the rate of an Arrhenius equation [6[, [7]. NEMA Motor Insulation Class
temperature change for the motor at standstill. defines the maximum allowable temperature rise above the
• A wide selection of thermal overload curves; standard for
typical applications, user-defined for unusual applications
and voltage-dependant for special applications, featuring
long starts of high inertia loads.
• Thermal model biasing in response to the current unbalance
that causes an extensive heating effect.
• The option to select separate cooling constants for the
motor in the stopped and running states.
• A current unbalance element capable of issuing a warning
about a potentially dangerous level of unbalance and of Fig 1.
tripping the motor off line on single–phasing. Aging Factor of Motor Insulation.

Motor Thermal Model Protection Applications 43

 ambient or thermal limit, if temperature exceeds this limit it stalled times. Stalled thermal calculations are usually performed
doesn’t cause immediate insulation failure but decreases the assuming adiabatic conditions. The designer often concedes
insulation’s expected lifetime. A fairly accurate approximation the fact that the peak temperature of the stator winding may
of Arrhenius equation states that an operating temperature temporarily exceed the steady state capability of the insulation
increase of 10°C in excess of the thermal limit cuts the life of system, taking into account the expected application of the
stator insulation by half. The percent of life vs temperature motor and how many times it is expected to be stalled cold
characteristics for different classes of insulation are shown at or hot in a lifetime, in making a design compromise. After a
Figure 1. design is complete, a summary of the thermal model becomes
available. Basic information includes the steady state thermal
The thermal risk for a squirrel cage rotor is that the rotor
rating of the motor, hot and cold stall times, and the cooling
conductors may deform or melt. Since there is no insulation,
time constants of the motor. For medium and large motor
the rotor conductors can be operated at a much higher
designs complete thermal damage curves of allowable time
temperature than the stator conductors. It is difficult and
versus current are provided as a standard.
impractical to provide a numerical temperature value defining
the rotor thermal limit. Motor designers address the maximum
allowable rotor temperature under stall, acceleration or any
Once the motor has been designed, and the basic operational
other fast transient conditions by stating the stall time thermal
parameters have been established for steady state load and
limits for a hot or cold motor. These values must correspond to
cold and hot stall times, the responsibility shifts to thermal
the system voltage level during the stall event. For the majority
protection for the motor. For majority of service conditions
of applications, the safe stall time defines the rotor thermal
the operating profile of the motor matches the assumptions
limit, but in some special cases motor capability during stall
made by the motor designer, so that the main job of thermal
and acceleration is dictated by the stator thermal limit. A rule
protection is to stay out of the way and let the motor run.
of thumb to define a stator-limited motor says: “When the
However, if motor is abused by mechanical breakage or human
voltage rating of the motor is equal to or greater than 10 times
error then protection steps in to assure there is no risk of thermal
the horsepower rating, the motor is stator limited.” For example:
damage.
500 HP, 6900 V. [8]
The question is, what model should be used to protect the motor
Steady-state operations such as running overloads are usually
when it is running? What is a reasonable compromise between
not an issue for the rotor. Under running conditions the stator is
accuracy and complexity? What physics should be included?
subject to extensive heating. Thus a stator overload protection
What should be used as an estimate of operation limit?
element ensures an overall sufficient level of thermal protection
for a rotor rotating at near synchronous speed. As we mentioned before the ideal method would be to have
the direct accurate temperature measurement and use aging
factor to estimate the consumed motor thermal capacity. But
The steady state and transient thermal behavior of the stator temperature sensors (RTD) have a delayed response to thermal
and rotor conductors of a motor depends on the details of the transients such as stall and acceleration and can’t serve as a
motor thermal circuit. The motor designer typically uses a rather basic criterion for a thermal model.
detailed thermal circuit, including separate representations
How detailed should the model be?
of stator iron, rotor iron, stator conductors, rotor conductors,
internal air, external air, motor shell and end shields. Details of We should certainly provide a model with enough flexibility
the thermal circuit depend on the ventilation construction of the to protect motors that have a dynamic thermal response
motor, including “drip proof”, “totally enclosed fan cooled”, and represented by several time-constants. A single time-constant
“totally enclosed non-ventilated”. For example, heat storage in is not always adequate [6]. Physics shows that there are at least
each circuit element as well as convective or conductive heat 4 distinct thermal time-constants: 2 for the stator conductors,
transfer between various pairs of circuit elements is included and 2 for the rotor conductors. For example, when heat is
in the model. A typical motor thermal circuit used by a motor generated in the stator conductors, the first effect is to raise
designer may have on the order of 20 nodes and 20 branches, the temperature of the conductors. The stator winding in the
resulting in a dynamic response characterized by several time stator slots are surrounded by a steel magnetic core. Therefore,
constants. as the windings get hot, heat begins to flow from the windings
into the steel core. The combination of the thermal capacity of
the winding and the thermal conductivity/impedance between
Motor designers are typically interested in a few standard the winding and the steel core establishes a short time-
thermal scenarios including steady state loading, cold, hot and constant. Heat that continues to flow from the winding into the
successive starting. The designer checks the computed steady- surrounding core is stored in the core, causing its temperature
state temperature of the stator winding to make sure it is within to rise, but more gradually than the initial rate of rise of the
the capability of the selected insulation system, designers also windings, because of the greater thermal capacity of the core.
define the time limits to withstand overloads. It is also very Eventually, the temperature of the core (and the motor frame,
important to determine running and stopped motor cooling etc.) also rises, causing heat transfer by convection to the
rates especially for “totally enclosed non-ventilated” motor surrounding air. The combination of the thermal capacity of the
designs and in some applications with intermittent use ratings. core and the frame and the thermal impedance between them
The motor designer is also interested in allowable cold and hot and the cooling air establishes a time-constant that is much

44 Motor Thermal Model Protection Applications

longer than that of the winding-core interaction.
dT t  2
˜R
I 2 t  ˜  T t 
I rated
W˜ (3)
So, the next question is, what is the best way to go beyond a dt H ˜ Tmax
single time-constant model?
C
W
The most reasonable way to model the thermal state of the H
motor is to measure motor current and to correlate it in real
The maximum temperature is related to the rated current such
time to motor thermal damage curves. The manufacturer’s
thermal damage curves represent the results of simulations
2
I rated ⋅R
that = 1 . In that case, equation (3) can be rewritten
of a complete motor model, including a multi-node thermal as: H ⋅ T max
model. The curves capture the multi-time-constant parameters
and thermal damage times for running, stall and sometimes dT t 
acceleration conditions of the motor. Typical curves are shown W˜ I 2 t   T t  (4)
dt
at Figure 3. Any point on the motor thermal damage curve
represents a thermal time limit at a specific level of current, Equation (4) can be used to analyze the thermal response of
or in other words: “The thermal limit defines how long a motor a single time-constant model to a steady overload. It can be
can withstand the corresponding level of stator current without shown that the solution of equation (4) for a steady overload,
exceeding the thermal boundary specified by the motor starting from a cold initial condition, is given by:
manufacturer.” Details of the thermal model implementation,
based on overload curves are given in the next section. T t  
I 2 ˜ 1  et / W 
I per unit motor current (a constant) (5)
In this section we answer two important theoretical questions T t  per unit motor temperature rise
concerning a thermal model based on motor thermal damage
curves (overload curves): Equation (5) can be solved for the amount of time needed for
the temperature rise to reach the thermal limit of the motor, i.e.
1. What is the relationship between standard overload curves T(t)=1:
and a single time-constant thermal model? § I2 ·
tmax I  W ˜ ln¨ 2 ¸
¨ I 1¸
2. Does an overload curve based thermal model behave © ¹
correctly when it is used in applications in which the load is tmax I  time estimated by a simple thermal (6)
not constant? model for the motor temperature to
We turn to mathematical analyses of the physics to answer reach thermal limit
theses two questions, starting with an analysis of a single time-
constant model. The thermodynamic behavior of homogeneous To develop a comparison between a single time constant
body at rest (motor) heated by electrical current can be described thermal model and overload curves, we now turn our attention
by a single time-constant thermal equation: to standard overload curves, which are given by:
dT ct 
C˜ I c2 t  ˜ R  H ˜ T ct  tmax I 
87.4 ˜ CM
dt
I 2 1 (7)
T ct  motor temperature rise above ambient (1) tmax I  trip time, seconds
I ct  motor current
CM curve multiplier
C specific heat capacity of the motor
H running heat dissipation factor To compare standard overload curves with the behavior of a
R electrical resistance single time constant model, it is useful to start by recognizing
that the numerator of the right hand side of equation (7)
It is convenient to rewrite equation (1) in terms of per unit corresponds to a time constant:
temperature rise and per unit current:
W CM
t max I 
T t  T ct  / Tmax per unit temperature rise I 2 1 (8)
I t  I ct  / I rated per unit current W CM 87.4 ˜ CM
(2)
I rated rated current
Tmax motor temperature at thermal limit Equation (6) and equation (8) are plotted in Figure 2. In order to
make the curves align for large values of current, it is necessary
trip condition to satisfy the following constraint:
In that case, equation (1) can be rewritten as:
W
C
W CM 87.4 ˜ CM (9)
H

In other words, in order for an overload curve to match a single

time-constant thermal model during a simple step overload, the
Motor Thermal Model Protection Applications 45
 We then take the first two terms in a Taylor’s expansion of
ln (1 − x ) with respect to x around the point x = 0:

ln(1  x ) | ln(1)  x |  x x  1
W (12)
? t max I  | I 2 !! 1
I2

Equation (8) describes how long it will take a standard overload

curve to reach thermal limit for a constant overload. We now
turn our attention to how a standard overload curve behaves
during cycling loads. We start with the differential equation that
is used to implement standard overload curves:

dT t  I 2 t   1 (13)
dt W CM
Fig 2.
Single Constant Thermal Model vs Relay Overload Curve Comparison To gain insights into what the response is to a cycling load,
we will consider a simple cycling load in which the current
time-constant implied by the curve multiplier of the overload alternates between no load and an overload value:
curve must be set equal to the time-constant of the single time- Ilow 0 | current during the low cycle
constant model. In Figure 2, the ratio of the time divided by the
time-constant is plotted against per unit current. It can be seen I high current during the high cycle
that although equation (6) is not exactly the same as equation (14)
tlow time interval for the low cycle
(8), the approximation is very close, particularly for large values
of current. For lower values of current, the standard overload thigh time interval for the high cycle
curves are a better approximation to typical motor overload
curves than a single time-constant model. That is because Motor heating is proportional to the square of the current, so
there are at least two time-constants in the thermal response the effective current for heating over the cycle is:
of a motor. Over short time intervals, the thermal response
of a motor is dominated by heat transfer from the stator and
rotor conductors to iron. Over longer time intervals, the thermal
response is dominated by heat transfer from the iron to cooling (15)
air. A single time-constant model cannot be accurate over the
full range of operation and tends to overprotect a motor when
it is operated near its rated load. A standard overload curve
provides protection that is a closer match to a motor’s thermal
limit. The close proximity of the two curves for large values of Equation (15) can also be expressed in terms of a duty cycle
current is not a coincidence because both models are equivalent ratio:
to an adiabatic model for large values of current. This can be H effective D ˜ I high 2  1  D  ˜ I low 2
shown mathematically by finding the asymptotic behavior of
the two curves. First, equation (8) is given approximately by: t high (16)
D duty cycle ratio
W CM t low  t high
t max I 
I 1
2
If the current and heating are expressed in per unit and the low
(10)
W CM cycle current is approximately equal to zero, the steady state
t max I  | I !! 1
2
I2 boundary condition for tripping the motor becomes:

A similar approximation can be shown to hold for equation 1 = D ⋅ I high 2 (17)

(6) by rewriting and taking a Taylor’s expansion in terms of
the reciprocal of the square of the current. First, we rewrite to Equation (17) defines the appropriate response to a duty cycle.
explicitly show the dependence on the reciprocal of the square It can be shown that a single time-constant model provides
of the per unit current: approximately this response. The next question is what is the
response of a standard overload curve to a duty cycle? Analysis
§ I2 ·
t max I  W ˜ ln¨ 2 ¸ of a standard curve under load cycling conditions will show
© I  1¹ that the response is correct, and will reveal how to properly
set an overload curve model to match the behavior specified
§
t max I  W ˜ ln¨
1 ·
¸
© 1  1/ I 2 ¹

W ˜ ln 1  1 / I 2  (11) by equation (17). We must consider values of current below
pickup, during which our motor thermal model is defined by the
t max I  W ˜ ln1  x  following differential equation:
x 1/ I 2

46 Motor Thermal Model Protection Applications

 equal to one resulting in the following consistency constraint:
dT t  § § hot · ·
¸  T t ¸¸
1
˜ ¨¨ I ˜ ¨1  87.4 ˜ CM
dt W cool © © cold ¹ ¹ (18) W cool (min) (24)
60
W cool cooling time constant
Equation (24) represents a consistency constraint relating the
hot hot stall time cooling time-constant and the curve multiplier of a standard
cold cold stall time overload curve. Figure 9 shows what can happen if it is not
satisfied. There are three cases shown for a cycling load with an
approximate per unit heating value of one. In the first case, the
⎛ hot ⎞
1− cooling time-constant is set too long resulting in over-protection
The factor ⎜⎝ ⎟
cold ⎠ is included to match the hot and cold and early motor tripping. In the second case, the cooling time-
stall times specified by the motor manufacturer. By including
constant is set according to equation (24) to match the implied
the factor in the cooling computation, the hot overload curve
time-constant of the curve multiplier, and the protection is
is effectively shifted down by the correct amount relative to the
correct. In the third case, the cooling time-constant is set
cold overload curve to account for the difference in “time to
too short, resulting in under-protection and possible motor
trip” of hot and cold motor conditions.
overheating.
For the load cycle under consideration, the current during the
unloaded part of the cycle is approximately equal to zero, so the
differential equation given by (18) reduces to: 4. Thermal Model Algorithm
dT t  T t 
 (19)
dt W cool The thermal model algorithm was developed in order to create
Taken together, equations (19) and (13) describe the behavior the thermal image of the motor and closely track the thermal
of our model during the assumed load cycle. The approximate conditions for all states of motor operation. The following states
temperature rise during the overload portion of the load cycle of motor operation are recognized:
estimated by the overload curve is computed by multiplying • Motor Stopped: Current is below zero level threshold and
equation (13) by the overload time: motor switching device indicates the open status.

'Thigh |
1
W CM

˜ I high
2

 1 ˜ t high (20) • Motor Starting: State is declared when previous state was
“Stopped” and current greater than 2% of the motor full
load amps has been detected. The motor current must
The approximate temperature drop estimated by the cooling
increase to the level of overload pickup (service factor
model during the unloaded portion of the duty cycle is
times full load amps) within 1 second otherwise motor will
computed by multiplying equation (19) by the appropriate time,
transfer into the next state: “Running”
with per unit temperature equal to 1, because that is what it will
be approximately equal to during a limit cycle that approaches • Motor Running: State is declared when previous state was
tripping: “Starting” and motor current drops below overload pickup
1 level.
'Tlow |  ˜ tlow (21)
W cool • Motor Overloaded: State is declared when previous state
The overload detection boundary is determined by setting the was “Running” and motor current raises above the overload
net temperature change equal to zero. This implies that the pickup level.
total of the right hand sides of equations (20) and (21) taken
together is equal to zero:

˜ I high  1˜ t high 
1 1
'Thigh  'Tlow 2
˜ t low 0 (22)
W CM W cool
Equation (22) can be rearranged to show that standard overload
curves respond correctly to cycling loads. Equation (22) also
reveals how to properly select parameters for a load cycling
applications:
W cool
1 ˜ D ˜ I high
2
(23)
W CM
Equation (23) expresses the actual overload detection boundary
of an overload curve model in terms of its settings, the duty
cycle, and the amount of overload. Except for the factor
of ττ cool
CM
, equation (23) is exactly the same as the ideal overload
detection boundary, specified by equation (17). Equation (23)
and equation (17) will be identical, provided that ττ coolCM
is set Fig 3.
Motor Thermal Limit Curves
Motor Thermal Model Protection Applications 47
 Thermal Capacity Used (TCU) evaluates the thermal condition Thus HCR is 8 sec / 10 sec = 0.8 and the level of stabilized TCU
of the motor. TCU is expressed as percentage of the thermal featuring the hot motor is equal to 20%, or in other words the
limit used during motor operation. Per IEEE Std 620-1996 (10) allowed motor thermal withstand time at overload conditions
the motor thermal limit is presented in the form of a time- will effectively decrease by 20%. If the motor load is lower then
current curve for 3 possible motor overload conditions: locked 100% the TCU level corresponding to the hot motor condition
rotor, acceleration and running overload. Every point on this is proportionally lower: 75% load – 15% TCU, 50% load – 10%
curve represents the maximum allowable save time at a stator TCU and so on.
current above normal load.
The unbalanced stator phase current will cause additional
rotor heating due to the developed negative sequence current
TCU is incrementally updated every 100 milliseconds and the
and flux rotating in the opposite direction to rotor rotation with
integrated value of TCU is stored in the thermal memory register
approximately double the power system frequency. The skin
of MPD according to the following equation.
effect in the rotor bars at this frequency will cause a substantial
TIME INTERVAL increase in rotor resistance and hence increased heating, which
TCU @ T TCU @ T 1  u 100% (25)
TIME TO  TRIP is not accounted for by the regular thermal model. In order
to account for this additional heating factor the Equivalent
Current concept is introduced. The idea is that the current
The following example can be a good illustration of TCU
input into the thermal model is biased to reflect the additional
accumulation during the cold motor start; initial TCU is equal to
heating caused by the negative sequence component of the
0%. Motor starting pattern (1) and relay overload curve (2) are
load current.
shown at Figure 3.
For simplicity assume that the time interval for TCU update is 1 I EQ 2
IM u (1  K u ( I 2 I 1 ) 2 ) (26)
second. Every point of motor current on this plot corresponds
where:
to the number of seconds that motor can withstand before
I EQ - equivalent motor heating current
tripping on overload. The numerical values showing the progress
of TCU accumulation during 17 seconds of motor acceleration I M - real motor current
are presented in table 2. We can observe that by the end of a I1 - positive sequence component of real motor current
successful starting the thermal memory of the motor protection I 2 - negative sequence component of real motor current
device (MPD) accumulates 46.7% of TCU.
K - unbalance bias factor
The Unbalance Bias K factor reflects the degree of extra heating
caused by the negative sequence component of the load
current and can be defined as the ratio of Positive Sequence
Rotor Resistance to Negative Sequence Rotor Resistance. It
is practical and quite accurate to use the estimate method
to define the K factor. Equations for typical and conservative
estimates are presented below.

(27)

where I LRC is the motor locked rotor current.

Of cause, in order to provide a complete thermal model of
the motor in service, the cooling process must be taken into
account. Cooling is characterized by Cooling Time-constants.
Table 2.
Thermal Capacity Used (TCU) calculation. These constants define the rate of cooling under stopped and
running operating conditions.
Typically the motor manufacturer provides locked rotor thermal
When the motor is running at rated load, TCU accumulated
limit curves or locked rotor safe stall time values for 2 motor
during the motor start will decay exponentially and will stabilize
conditions: cold motor (motor @ ambient temperature) and hot
at the level of TCU matching hot motor thermal conditions. If
motor (motor @ ambient + rated rise temperature). In order to
the motor load is lower, then obviously the thermal balance
distinguish between the 2 aforementioned motor conditions
point is proportionally reduced.
the additional motor parameter, Hot/Cold Stall Time Ratio (HCR)
is included in MPD algorithm. The stopped motor will also be subjected to the exponential
decay of TCU stored in MPD thermal memory during motor
These parameters define the proportional increase of TCU of the
operation. Natural cooling of the rotating motor or forced
motor running fully loaded at a settled temperature compared
cooling by means of the special fans installed on the machine
to the motor resting at ambient temperature. For example let us
shaft cause a much higher cooling rate of the running machine
assume that according to the motor data sheets the Cold Safe
compared to the motor at standstill, typically the ratio is 2:1.
Stall Time is 10 seconds and the Hot Safe Stall Time is 8 seconds.

48 Motor Thermal Model Protection Applications

Thus 2 separate Cooling Time-Constants are used in the Thermal 3. RTD bias maximum
Model Algorithm. The equations to calculate TCU decay of the
cooling motor are as follows: This point is set to the temperature rise equal to the motor
insulation thermal limit.
(28)
Typically for NEMA B class motors insulation class is F with
temperature rise above ambient of 115ºC. The TCU at maximum
Where: temperature point is equal to 100%.
TCU START (%) is the initial value of TC accumulated Rate of change of TCU between the adjacent points is
by the moment the cooling starts; approximated as linear.
t (min) is duration of cooling;
is the Cooling Time Constant;
TCU END (%) is the steady state level of TC

The steady state thermal condition for the motor at stand

still is the ambient temperature, which is corresponding to
TCU END (%) = 0 . (29)
The steady state thermal condition for the running motor is
calculated as:

I eq
TCU END u 1  HCR  u 100% (30)
SF u FLA
Where:

In some unforeseen situations, when the motor cooling is Fig 4.

blocked or ambient temperature deviates significantly from RTD Bias of Thermal Model
the industry standard value (40ºC), it becomes difficult to
accurately replicate the motor’s thermal condition based solely 5. Thermal Model Behavior at Different
on the measured current. That is why it is practical to apply Operational Conditions
an independent algorithm, calculating TCU by means of stator
RTDs (resistance temperature detectors) and correcting the In order to illustrate how TCU varies during motor operation let us
thermal model upwards if needed. review the following motor data and operational sequences.

The RTD-TCU Curve is constructed based on the 3 key points. Let us assume that the following motor information is available
See Figure 4 for details. to us.
• Motor thermal limit curves are as presented at Figure 3.
1. RTD bias minimum
• Motor Cold and Hot Locked Rotor Times at 100% of the
Set to 40°C or another value of ambient temperature, if the system voltage are 34 and 26 seconds respectively. At 80%
appropriate RTD is available. TCU is equal to 0%. of the system voltage Cold and Hot Locked Rotor Times are
50 and 38 seconds respectively.
2. RTD bias mid point • Motor Acceleration at 100% of the system voltage is 17
The mid-point temperature is set according to the motor’s hot seconds. Maximum locked rotor current is 6 times that of
running temperature and is calculated as follows: full-load amperes (FLA). The MPD overload curve that we
employ as a limit to calculate TCU, is shown in Figure 3.
Rated Temperature Rise + Ambient Temperature Please note that the location of this curve is between the
For example: The temperature rise for NEMA Class B motors hot and cold thermal limit curves supplied by the motor
with 1.15 Service Factor, is 90ºC. Thus the temperature value for manufacturer. The time-current relation in this curve is per
this point is 130ºC. The TCU quantity for this point is the value of following equation:
a steady-state running condition @ rated motor load, and can
be found as: 87.4 u 12
Tto trip (sec) (32)
2
I EQ 1
TCU CENTER = (1 − HCR ) × 100% (31)

Motor Thermal Model Protection Applications 49

 • The Running and Stopped Motor Cooling Constants are
respectively 20 and 40 minutes. Motor Service Factor =
1.15.
• Current Unbalance Factor: 6

Sequence 1: Combined operation (Figure 5)

State A. Initially the motor is at ambient temperature.
TCU = 0%. The motor is ready to start.
Section AB. The motor is successfully started at 100% voltage.
Acceleration time = 17.1 seconds, TCU accumulated during
start is 46.7% (details are in Table 2)
Section BC. The motor runs for 45 minutes at a steady load of
80% with 10% current unbalance. TCU by the end of the period,
exponentially decays to level of 19.5 %. TCU is calculated per
equation 25.
Section CD. The Motor runs at 125% balanced overload for 15 Fig 6.
minutes. TCU increments to the level of 67.7%. Stall Trip. 100% Voltage

Section DE. The Motor runs at 125% overload with 10%

current unbalance until the thermal capacity reaches 100%
and the relay trips the motor offline in 8.5 minutes. It is not well
illustrated on the graph, but the addition of current unbalance
at the running overload state decreases the trip time by 1.5
minutes or 15% (the calculated balanced overload trip time for
the section DE is 10 minutes).
Section EF. The motor is at standstill and cools down to ambient
temperature for 150 minutes. TCU decays to approximately 0
level. The rate of cooling is 2 times slower than of the running
motor.

Fig 7.
Stall Trip. 80% Voltage

Figures 6 and 7 demonstrate how the thermal model provides

an adequate protection where the motor is taken offline before
the thermal limit is reached. This situation has been evaluated
for hot and cold motor conditions at both 100% and 80% of the
system voltages applied to the motor.

Sequence 3: Running overload

Fig 5. Three different scenarios are considered:
Thermal capacity used during motor operation.
• The motor is overloaded immediately after a cold start.
Sequence 2: Motor stall • An overload is applied to the motor that was started, and,
The motor can be seriously damaged if a rotor stall occurs prior to overload, run unloaded for 2 hours.
during the start attempt. Stall can occur due to a mechanical • An overload is applied to the motor that was started and,
breakage or a human mistake. The stalled motor draws current prior to overload, run at full load for 2 hours.
equal to locked rotor amps. Locked rotor time (LRT) values
provided by the motor manufacturer specify the thermal limit The overload that was applied in all three cases was 125%
for the motor at ambient and rated conditions. Typically LRT is of motor full-load amps.The motor thermal limit time values
specified for the motor starts performed at 80% and 100% of allow for applying a 125% overload to the cold and hot motor
the system voltages. for 50 and 29 minutes respectively (data can be found in

50 Motor Thermal Model Protection Applications

Figure 3). The first case is characterized by severe heat protection devices are capable to learn and store, in the non-
generation in the rotor bars during the startup. Immediately volatile memory, TC value utilized by motor during successful
following, the motor startup, the overload heats up the stator start and use this value in the start inhibit algorithm.
windings preventing heat transfer to the environment. This
Sequence 5: Cyclic load
situation presents a serious thermal impact and the motor is
taken offline faster in comparison to the other two cases. Trip According to considerations discussed in a previous section of
time in this case is 16.3 minutes. this paper, the main criterion for a thermal model’s adequate
response to cyclic load is the matching of the implied heating
The second scenario presents an overload of the motor at time-constant to the explicit running cooling time-constant
ambient temperature. Initial TCU is 0%. According to the thermal (see equation 24). Let us review a balanced cyclic load (i.e. the
model algorithm computations, the trip will be implemented 31 effective heating) (equation 16) of 1.
minutes after the overload is applied; which is lower than the
After the cold start, the motor varies the load every 30 seconds
cold motor limit (50 min). In a real application, the temperature
at between 20% and 160% of the full-load current. Per equation
of the unloaded running motor is typically higher than the
24, the running cooling constant is calculated as follows
ambient temperature, because of the associated motor losses.
This fact explains why the significant margin between the cold
87.4 u CM 87.4 u 12 (33)
overload trip time (31 min) and the cold thermal limit (50 min) is W cool 17.5 (min)
required. 60 60

In order to provide a more accurate thermal model response

The third scenario shows the hot overload (i.e. the motor is
to cyclic load conditions, the cooling time-constant should be
assumed to be at the rated temperature). The initial TCU in this
adjusted to the calculated value. At the same time this change
case, the moment before the overload is applied, is 25%, so
(from 20 to 17.5 minutes) would cause no significant impact to
the tripping time is proportionally lower, compared to the cold
the other motor operating sequences.
overload. Tripping time in this case is 23 minutes, which is lower
than the hot thermal limit (29 minutes). Figure 9 demonstrates the importance of cooling constant
value in the thermal model response to cyclic load conditions.
Sequence 4: Consecutive starting Three cases are shown for a cycling load with an approximate
Per NEMA MG1 standard (11) medium and large induction per unit effective heating value of one. In the first case, the
motors are required to withstand thermally: cooling time-constant is set long, resulting in over-protection
and premature thermal model triggering. In the second case, the
• 2 consecutive starts, coasting to rest between starts, with cooling time-constant is set to match the implied time-constant
the motor initially at ambient temperature (cold starts) of the curve multiplier, and the thermal model adequately
• One start with the motor initially at rated load operating responds. In the third case, the cooling time constant is set short,
temperature (hot start) resulting in under-protection and possible motor overheating.

An illustration of the thermal model response to consecutive

starting is shown on Figure 8.
As you can see, the thermal model provides the start sequence
required by NEMA.
An important enhancement to the thermal algorithm is the Start
Inhibit function, which is employed to prevent excessive motor
starting in cases where there is not enough thermal capacity
available to perform a successful start. Modern intelligent

Fig 9.
Thermal model response to cyclic load

Sequence 6: Starting of high inertia loads

The thermal model algorithm has an additional enhancement
that allows the coordination of protection with high-inertia long
starts, while acceleration time is greater than the safe motor
stall time. The voltage-dependant dynamic thermal limit curve
is employed to account for varying thermal limits corresponding
to the acceleration current levels at the different terminal motor
voltages.
Fig 8.
Hot and Cold Consecutive Starts Figure 10 shows an example of a 100% voltage high inertia start
Motor Thermal Model Protection Applications 51
 lasting 17 seconds (curve 1), and a locked rotor time limit of 8 same application reduced to 80% voltage start (Figure 10,
seconds (curve 4). Actually curve 4 implies the line of the same curve 7). LRC at 80% is 4.8 times of FLC. From the 100% voltage
I²T. In many short motor start applications it is reasonable to case we know that the locked rotor condition is referenced to
conservatively approximate that the thermal limit remains the the thermal limit of 288, and the allowed locked rotor safe time
same during motor acceleration. In the short start applications for an 80% voltage start yields 12.5 seconds, but according to
an error introduced by this assumption doesn’t prevent the the acceleration thermal limit curve (Figure 10, section 5) the
motor from successfully starting. The thermal limit curve is thus thermal limit time corresponding to 4.8FLC is 40 seconds which
constructed from sections 2, 3 and 4. If the same approach is is much higher than the allowed value. This means that if the
applied to the case shown on Figure 10, it will result in the TCU
reaching 100% in the middle of the acceleration (Figure 11,
curve 1).
As we mentioned in previous sections of this paper, as the
thermal limit is a function of motor speed during acceleration,
the acceleration thermal limit (curve 5) shows up differently
from the locked rotor limit. Each point on curve 5 corresponds
to current value which, in turn, corresponds to motor rotation
speed during startup. Based on this, we can indirectly find
the reference between motor speed and thermal limit, and
construct an updated motor thermal limit curve which will
include sections 2 and 5 shown on Figure 10.
The new curve helps achieve a successful motor start (Figure 11,
curve 2) despite the fact that locked rotor safe time is shorter
than acceleration period. The protection method described
above is relevant for an ideal situation with a constant terminal Fig 11.
voltage of 100%. Thermal Model Response to High Inertia Load Starts

In reality the system voltage can deviate from 100% because of

the voltage drop during motor startup. The locked rotor current motor stalls under the reduced voltage conditions it becomes
(LRC) is almost directly proportional to the voltage applied to underprotected and appears to be in real danger of burning.
the motor terminals during acceleration, this fact must be taken
into consideration when the acceleration portion of the thermal
limit is used in the thermal model algorithm. To handle this situation the thermal model is equipped with
a mechanism capable of dynamically responding to voltage
variations during motor startup. Line 6 on Figure 10, shows the
new position of the acceleration curve 4, shifted in response to
the voltage reduction to 80%. The successful start under these
operational conditions is shown on Figure 11, curve 3.
This technique provides adequate thermal protection in cases
of high-inertia load application. In some cases where the
thermal limit difference between locked rotor and acceleration
conditions is not clearly identified, this element should be
supported with a zero speed sensor.

6. Application Description

This case study examines the Induced Draft (ID) fan application
Fig 10. on the A. B. Brown Unit 2 Selective Ccatalytic Reduction (SCR)
Voltage Dependent Thermal Limit Curves Project, located in Evansville, Indiana. Unit 2 is owned by Vectren
Corporation, and the role of Black & Veatch (B&V) on this project
For example, for a 100% voltage start (Figure 10, curve 1) the
was to construct an SCR facility in this plant.
locked rotor thermal limit is calculated based on a LRC of 6 times
full load current (FLC) and 8 seconds of the allowed locked rotor The SCR Project Scope of Work included modifying both ID fans
safe time, and I²T is equal to 288. After 14 seconds the motor for catalyst draft losses. The motors were powered from 13.8
accelerates to approximately 80% of the rated speed and the kV switchgear.
current drops to the level of 4.8 times that of FLC.
The allowed time to withstand 4.8 FLA for this stage of
acceleration is 40 seconds; I²T=922. Now let us consider the

52 Motor Thermal Model Protection Applications

Motor ratings and data • The relay provides thermal protection of the motor during
abnormal starting or running conditions, preventing
Motor Parameter Value thermal damage to the motor (i.e., the MPD curve is placed
Motor Horse Power 5500 HP below the motor thermal damage curves).
Rated Voltage 13200 V
• The relay allows the motor to be started successfully
Phases 3
without nuisance trips, in accordance with the number of
Motor Full Load 893 RPM
starts and thermal/cooling characteristics recommended
Service Factor 1.15
by the manufacturer
Frequency 60 Hz
Rated Full Load Current 226 A • The relay settings allow proper coordination with the
Rated Locked Rotor Current 1205 A respective tie and main circuit breakers on the 13.8 KV bus
Insulation Class F to which the motors are connected.
Ambient Temperature 43° C Typically the B&V specifications require the motor design to
Temperature Rise @ SF=1.0 77° C meet the following criterion:
Temperature Rise @ SF=1.15 87° C
“Motor safe stall time at minimum starting voltage shall not be
Table 3. less than motor acceleration time at minimum starting voltage,
Basic Motor Data plus 2 seconds.” The motor manufacturer could not meet this
Table 3 presents the motor information pertaining to the ID fan requirement for this high-inertia application and indicated that
motors. a speed switch would be provided in lieu of this requirement. The
speed switch option was not used because the MPD provided
Motor starting and thermal characteristics a range of setting options for the overload feature. The MPD
was originally set using the custom overload curve feature to
The motor manufacturer provided the thermal limit curve under
match the motor characteristics, in addition to all the above
locked rotor, acceleration, and running overload conditions,
listed protection criteria.
as well as time-current curves during acceleration at rated
voltage and at minimum specified starting voltage. Some of the
important motor characteristics (from the manufacturer’s data Problems during startup of ID fans
sheet and curves) are summarized in Table 4. The problem that the commissioning team faced during startup
was that the successive motor start-time delays determined
Motor Data
Value by the MPD thermal model were inconsistent with what was
Description Voltage allowed by the motor manufacturer. The motor data sheet
Acceleration time @ Rated Voltage in Seconds 100% 28.0 allowed the following operational characteristics:
Acceleration time @ minimum Voltage in Seconds 80% 53.0
• Two successive cold starts or one hot start.
Cold Locked Rotor Safe Stall Time @ rated voltage and 100% 26.0
ambient temperature in seconds • Following this sequence a new start would be allowed after
Cold Locked Rotor Safe Stall Time @ min voltage and 80% 47.0 any of the following:
ambient temperature in seconds
- A cooling period of 40 minutes if the motor was running
Hot Locked Rotor Safe Stall Time @ rated voltage at 100% 23.0 at service factor load and then stopped.
service factor load operating temperature in seconds
Hot Locked Rotor Safe Stall Time @ min voltage at 80% 42.0 - A cooling period of 10 minutes if the motor was running
service factor load operating temperature in seconds unloaded and then stopped.
Maximum Starts Per Hour N/A 2 - A cooling period of 20 minutes if the motor was de-
Maximum Cold Consecutive Starts @ rated Voltage 100% 2 energized, coasted to rest, and left idle.
Maximum Cold Consecutive Starts @ min Voltage 80% 2
It was observed that the MPD was delaying restart by 40 to 43
Maximum Hot Consecutive Starts @ rated Voltage 100% 1 minutes after every start attempt, regardless of whether it was
Maximum Hot Consecutive Starts @ min Voltage 80% 1 a second cold restart or the first hot restart. This performance
Running Cooling Time Constant in minutes N/A 9 was unacceptable to the client who wanted reliable cold starting
as well as a restart time consistent with that indicated by the
Initial / Modified (during Startup) Stopped Cooling Time N/A 16 / 12
Constants in minutes motor manufacturer. Some of the motor parameters recorded
in the MPD during the startup of one of the fans are as follows:
Table 4. • Hottest RTD Values: 70° C.
Motor Starting and Thermal Limit Characteristics
• Learned Starting Current: 1.085 A.
Protection philosophy
• Average Motor Load: 60 percent of the rated current.
The ID fans on the A. B. Brown Project are fed from a 13.8 kV
auxiliary electric system and are protected by a multifunction It was also noticed that the MPD thermal model was
motor protection device (MPD). The fundamental philosophies accumulating almost all the available thermal capacity
used in setting the MPD are as follows: even during the first cold successful start, thus preventing

Motor Thermal Model Protection Applications 53

 the motor from performing a successive start. The MPD Did it work?
did not allow a restart for 40 to 43 minutes because of the
start-inhibit feature that prevented further restarts when the The synergistic efforts to problem solving between B&V, the
available thermal capacity was not sufficient for a successful relay manufacturer, and the motor manufacturers paid off.
start. This performance was found to be inconsistent with the The implementation of voltage-dependent overload curves
motor manufacturer’s recommendations. B&V requested the resolved the issue of unreliable motor starting, and the motor
involvement of the motor and relay manufacturers to resolve successive restart time delay was reduced to around 20 minutes
this apparent problem. to the satisfaction of the client. The fans have been successfully
commissioned and are now running without problems at the A.
Solution to the problem B. Brown plant.
B&V closely coordinated with all the parties, and the following Lessons Learned from This Experience
measures were put in place in sequence:
The cited case study highlights the need for flexibility and
• The motor manufacturer provided the following collaboration between all parties and, above all, a customer-
recommendations regarding the thermal model settings of oriented approach to relay coordination studies for complex
the MPD: motor applications. The following are some of the salient
- Use the voltage-dependent overload curve option lessons:
available in the MPD, and set it with reference to the • The relay application engineer should obtain accurate
motor thermal damage curve. information about the motor before designing the relay
- To better model the motor during the 9 minutes of settings. In this respect, it is important that the motor’s
coast-down period, decrease the stopped cooling certified acceleration curves provided by the motor vendor
time-constant to 12 minutes. match actual conditions. In addition, the proper cooling
time-constants recommended by the motor manufacturer,
- Decrease the safety margin in the start-inhibit function must be programmed into the relay so that motor behavior
to shorten the lockout time between starts. (The safety may be simulated accurately.
margin was changed from 25 to 5 percent.)
• The motor manufacturer must be consulted and requested
• Acting on these recommendations, the relay manufacturer, to concur on overload relay selection curves in applications
B&V, and the motor manufacturer collaboratively reviewed where the motor starting curves and the thermal damage
the motor protection coordination and relay set points. The curves are very close to one other. The solution that was
following actions were taken: arrived at in the A. B. Brown case study, where a portion
- The relay overload curve was changed to the voltage of the overload curve was set on the motor thermal
dependent overload curve, with the revised relay curves, would not have been possible without the motor
points on the hot thermal curve for the starting manufacturer’s concurrence.
zone. • All efforts must be made to achieve safe operation of the
- The stopped cooling time-constant was programmed motor and adequate coordination with other devices and,
as 12 minutes, reduced from 16 minutes. at the same time, meet the operational expectations of the
application.
- Acceleration time was changed to 35 seconds because
the motor was observed to start and accelerate 7. Conclusions
satisfactorily within approximately 28 seconds.
The modern industrial marketplace has a strong demand for
- The thermal capacity used margin set point was
a simple, reliable, accurate, multifunctional MPD designed in
changed from 25 to 5 percent.
accordance with industry standards. The major element of
- The jogging block function was left on, and the a MPD is the thermal model, which must create an accurate
maximum number of starts was programmed to be 2; image of the motor thermal conditions at any stage of the
time between starts was programmed to 0. protected machine’s operation.
- The restart block was enabled and set to 10 minutes Theoretical considerations prove that a simplified thermal
(600 seconds) to comply with the motor requirements model based on an equivalent single time-constant model, and
of coasting to rest (9 minutes) after being de-energized, overload curves matching motor manufacturers standards
before the new start would be allowed. for thermal limits, can provide adequate protection at a level
of accuracy desirable for this type of application. It can be
- The RTD bias of the thermal model was disabled
clearly demonstrated that if the implied thermal constant of
because it was inappropriate for this application. It
the overload curve matches the explicit cooling constant of
is important to mention here that B&V, as part of the
the running motor, the relay algorithm computes the correct
control design, always programs an alarm in the relay
thermal image of the motor during a cycling load.
for high temperature based on RTD inputs.
- The thermal capacity alarm level setting was left at 85
percent to be readjusted later if required.

54 Motor Thermal Model Protection Applications

 8. References
[1] Sam F. Farag and M. K. Jhaveri, “Intelligent Microprocessor-
Based Devices Provide Advanced Motor Protection,
Flexible Control, and Communication in Paper Mills” IEEE
Transactions On Industry Applications, Vol. 33, No. 3, May/
June 1997.

[2] Gabriel J. Paoletti and Alex Rose, “Improving Existing Motor

Protection for Medium V oltage Motors”, IEEE Transactions
On Industry Applications, Vol. 25, No. 3, May/June 1989.

[3] Motor Reliability Working Group, “Report Of Large Motor

Reliability Survey Of Industrial And Commercial Installations,
Part I”. IEEE Trans. Ind. Appl., vol. IA-21, no. 4, pp. 853-864,
July/Aug. 1985.

[4] James H. Dymond, “Stall Time, Acceleration Time, Frequency

of Starting: The Myths and The Facts” IEEE Transactions On
Industry Applications, Vol. 29, No. 1, Jan/Feb 1993.

[5] Sam F. Farag, Robert G. Bartheld, and William E. May,

“Electronically Enhanced Low Voltage Motor Protection and
Control,” IEEE Industrial and Commercial Power Systems
Conference, May 1993, pp. 166-174.

[6] David R. Boothman et al, “Thermal Tracking – A Rational

Approach to Motor Protection”, T74 029-5, IEEE PES Winter
Meeting, 1974.
Fig 12.
Coordination of ID Fan Motor Thermal Limit Curves and MPD Thermal
Protection Curves. [7] Emanuel L. Brancato, “Estimation of Lifetime Expectancies
of Motors”, IEEE Electrical Insulation Magazine, Vol. 8 No. 3,
May/June 1992.
A detailed explanation of the thermal algorithm actually
provides the tools required to calculate the thermal capacity for [8] B. H Moisey, “Concepts of Motor Protection. Edmonton,
literally any application. In many cases, it is useful to evaluate Alberta, 1997.
thermal model behavior before motor energizing and compare
the results with operation restrictions dictated by the motor [9] IEEE Guide for AC Motor Protection. Std C37.96-2000.
manufacturer. Analysis of typical motor sequences based on
real motor specification data shows how the thermal model [10] IEEE Guide For the Presentation of Thermal Limit Curves for
algorithm implemented in an MPD can successfully handle Squirrel Cage Induction Machines. Std 620-1996.
excessive motor operation duty and avoid stator and rotor
overheating as well as premature machine tripping because of [11] NEMA Standards Publication No. MG 1-1993. Motors and
thermal overestimation. Generators.
Thousands of motor protection systems employing the proposed
thermal algorithm have been successfully installed in various [12] Grainger, L.G.; McDonald, M.C. “Increasing refiner
motor applications. However, in a few instances, the setting of production by using motor thermal capacity for protection
motor thermal protection is not as straightforward as it is in the and control”, IEEE Transactions On Industry Applications,
majority of cases. Volume: 33 , Issue: 3 , May-June 1997.

The A. B. Brown case study presents a unique situation where the [13] GE Multilin, “Setting The 469 Motor Management Relay for
close coordination between the parties involved (the end-user, A Cycling Load Application”, GE Publication GET-8478.
the application engineer, the relay manufacturer and the motor
manufacturer) allowed the refinement of application-related
information so that proper thermal protection and coordination
during commissioning could be provided.

Motor Thermal Model Protection Applications 55