34 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 19, NO.

1, MARCH 2004

Determination of the Absolute Rotor Temperature

of Squirrel Cage Induction Machines
Using Measurable Variables
Martin Maximini and Hans-Jürgen Koglin

Abstract—This paper presents a new method for observing the

rotor temperature of high-power squirrel cage induction machines
using measurable variables. The method is based on the fact that
the rotor resistance depends on the actual rotor temperature. The
main problem is to separate the changes in the rotor resistance due
to temperature and skin effect. By comparing the input impedance
with a known circle diagram measured during commissioning, it
is possible to calculate changes in the rotor temperature. Further
analyses also make it possible to obtain the absolute rotor temper-
ature at any time. Results of procedure testing are demonstrated
through computer simulations and evaluations of data recorded on
a 75-kW test machine.
Fig. 1. Equivalent circuit diagram of an induction machine.
Index Terms—Differential measurement, induction machine,
rotor temperature, thermal protection.

I. INTRODUCTION corresponding algorithm was developed on the basis of a very

simple first-order thermal model. This kind of modeling is quite

D UE to their high reliability and their robust construction,

induction machines with squirrel cage rotors are installed
in many areas of industrial practice. In the past decade, research
conservative, so that the machine and the process are not used
optimally.
At the Institute of Power Engineering at the Saarland
in protection techniques and instrumentation technology have University, a new method for observing changes in the rotor
brought some new methods for the measurement of protection- temperature using measured variables has been developed.
relevant variables, their link to instrumentation and control tech- This new method, called compensation method, was published
nology and, finally, their analysis. Effective supervisory systems first in [13]. In Section II, the principle of this method and
have two positive effects: the lifetime of the machine can be some new results are presented using a simulation model and
prolonged and the operating range of the machine and of the data recorded on a 75-kW test machine. Subsequently, a new
following process can be expanded. Still, a critical variable for algorithm using the results of the compensation method for
induction machines, in particular for those of high power, is the calculating the absolute rotor temperature is explained and the
rotor temperature. Direct measurement of the rotor temperature results obtained are presented.
requires considerable effort and it is not very reliable. During
the last few years, a number of methods for calculating the
rotor temperature have been published. Prediction procedures
II. COMPENSATION METHOD
calculate the actual rotor temperature by using thermal models
[1]–[5] or neural networks [6], [7], whereas other procedures A. Principle
use the effects of the rotor temperature on directly measurable
variables [8]–[12]. Although some of these ideas are very inno- The idea is based on the fact that during the starting operation
vative, they still do not satisfy the practical requirements. The each induction machine has its own typical impedance circle
reason for this is on one hand the lack of general applicability diagram. This diagram is the fingerprint of the machine. The
of some methods or the complexity of measurement, and, on the target of this method is to determine the actual heating status of
other hand, the sensitivity to disturbances or the non-observance the rotor by evaluating changes in this circle diagram.
of nonlinear effects. The main problem in observing the rotor Fig. 1 shows the well-known equivalent circuit diagram of the
temperature in measurable variables is to separate the effects induction machine.
due to skin effect and temperature. Thus, even today, a primitive It is obvious that the input impedance depends on the
interlocking rule is implemented in most protection devices. The values of the model parameters, on the temperature of stator
and rotor , and on the slip . An increase in the rotor
temperature has the equivalent effect on the input impedance
Manuscript received January 13, 2003.
The authors are with the Saarland University, Saarbrücken 66123, Germany. as a reduction of the slip . This fact gives the presented method
Digital Object Identifier 10.1109/TEC.2003.821833 the name compensation method and it is used in the following
0885-8969/04$20.00 © 2004 IEEE
MAXIMINI AND KOGLIN: DETERMINATION OF THE ABSOLUTE ROTOR TEMPERATURE 35

Fig. 2. Schematic structure of the test machine.

TABLE I
INDUCTION MACHINE NAMEPLATE DATA
Fig. 4. Block diagram of the compensation method.

In more detailed form, (1) can be rewritten as follows:

(2)

In (2), is a reference temperature (generally ).

The absolute difference in the electrical rotor frequency between
the points and can be neglected, so the following holds:
(3)
Thus, the parameter is omitted from (2). This implies
that the skin effect has no influence on the method. Using the
rotor differential temperature as defined in (4), (2) can be
converted into (5)
(4)

Fig. 3. Principle of the compensation method.

(5)
manner: by having two impedance circle diagrams with addi-
tional information about the slip, the rotor temperature differ- It is obvious that the calculated rotor differential temperature
ence between points of identical slip can be determined. The depends on the measured slip , the temperature coeffi-
first impedance circle diagram is called the reference diagram cient of the rotor material , and the reference rotor temperature
(ref) and the other the measurement diagram (meas). Fig. 2 . The reference rotor temperature has no essential influ-
shows the schematic structure of the test machine. Machine ence on the differential temperature. This will be shown in the
nameplate data are summarized in Table I. next paper section. Another disturbance, the stator temperature
The principle of the compensation method is given in Fig. 3: , can be eliminated by using an impedance on which the stator
There are the measured impedance circle diagrams of two starts has no influence [see (6) and Fig. 1]. The stator resistance
of the test machine presented. The impedance phasor moves and the stator temperature can easily be measured
clockwise during the starting operation. The points of the mea-
(6)
surement diagram and of the reference diagram have iden-
tical slip but not identical phasors. Because In Fig. 4, the realization of the compensation method is sum-
the rotor temperature during the measurement start was about marized. During commissioning—for a reference start—the
hotter than during the reference start, the point significant variables (slip, input impedance, and stator temper-
corresponds to point of the reference. How to use this quali- ature) are recorded. These data can be called the fingerprint of
tative effect to quantify the temperature difference is shown the machine. Later, during operation, the correspondent mea-
in Section III. surement data are obtained. With the compensation method,
these measurement data are compared with the fingerprint
B. Derivation as already explained. The output result of the compensation
The rotor temperature difference between the points with method is the rotor differential temperature.
identical slip (points and ) can be evaluated as follows: the
C. Identification of a Simulation Model
input impedance and the rotor resistance are equal at the points
and To compare practical and theoretical effects of the compensa-
tion method, it is necessary to have a simulation model of the test
(1) machine. For estimating the parameters of the equivalent circuit
36 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 19, NO. 1, MARCH 2004

Fig. 5. Measured and simulated impedance circle diagram.

Fig. 7. Influence of different reference rotor temperatures (left: theory, right:
practice).
TABLE II
RESULTS OF THE PARAMETER ESTIMATION

Fig. 8. Influence of the stator temperature (left: theory, right: practice).

Fig. 6. Used reference rotor temperatures.

diagram of the test machine, a least square algorithm is used.

The objective function is the minimum square deviation be- Fig. 9. Influence of errors in the slip (left: theory, right: practice).
tween the adjusted input impedance during a start measured and
the model impedance [14]. The resistance of the stator can be slip data are the same for the theoretical and the practical exami-
measured directly. The results of the estimation with their 95%- nations; in both cases, the measured data were used. The indexes
confidence-intervals are shown in Table II. Fig. 5 shows the specify the used reference rotor temperature.
measured impedance circle diagram and the impedance circle It can be concluded that theory and practice correspond very
diagram of the identified model. The impedance circle diagrams well. The influence of the used reference rotor temperature on
are corresponding very well. the result is not essential. The influence depends on the me-
This simulation model is used in Section II-D to compare chanical speed (i.e., slip) of the machine and on the size of the
theoretical and practical effects of the compensation method. rotor differential temperature for theory and practice in the same
manner.
D. Examinations of Different Influences In Fig. 8, the influence of the stator temperature on the result
First, the influence of the reference rotor temperature of the compensation method is presented, in Fig. 8(a) the the-
on the result of the compensation method is examined. Thus, oretical effect is seen, in Fig. 8(b) the practical effect, respec-
the compensation method is tested with different reference rotor tively. With being designated, the differential temperature
temperatures shown in Fig. 6. calculated with the adjusted impedances (6), is calculated
is the effectively measured reference rotor tempera- with the original impedances, including the stator influences.
ture, and are the reference rotor temperatures Theory and practice show the same effect: the influence of
simulated for this examination. In Fig. 7, the results obtained by the stator temperature diminishes with the mechanical speed.
the compensation method are presented. The last effect examined is the influence of errors in the slip
Fig. 7(a) and (c) are the theoretical results obtained with the on the compensation method. These results are shown in Fig. 9.
identified model, and Fig. 7(b) and (d) show the corresponding The slip of the measurement is changed for one evaluation: the
results using the measured input impedances. Temperatures and electrical frequency is changed from 50.018 to 50 Hz. It is
MAXIMINI AND KOGLIN: DETERMINATION OF THE ABSOLUTE ROTOR TEMPERATURE 37

Fig. 11. Thermal transfer function.

Fig. 10. Measured and calculated rotor differential temperature for different
operating conditions.

visible that theory and practice correspond once again, errors

in the electrical frequency (i.e., in the slip) show the same in-
fluence on the calculated differential temperature : an effect
can only be observed in the range of higher mechanical speed.
It can be summarized that all of the presented studies have
shown that theory and practice correspond excellently and that
the examined disturbances have no essential influence on the
result of the compensation method.
Finally, the compensation method is tested for different op-
erating conditions of the machine. The machine must run under
load, because running at no load, the rotor temperature is not
observable in the input impedance. The results of the analyses
are presented in Fig. 10. The evaluated start was a start with the Fig. 12. Algorithm for reference temperature reconstruction (a) Model
motor initially at ambient temperature. Therefore, the rotor dif- parameter for the heating of the machine during the starting operation. (Stator
temperature directly measurable) (b) Model parameter for the cooling of the
ferential temperature during this time is about 0 . After start, machine during the steady-state operation, identified with the differential
a load application of 21 kW was made. Due to the machine run- temperatures. (Rotor differential temperature from the compensation method).
ning with rated speed, the rotor is cooling off by 50 during
16 min. After that, a load change to 69 kW was made and the the square of the stator current . The output variables are the
rotor is warming up by 20 . Subsequently, a load change to stator and the rotor temperatures ( , ).
29 kW was made. The key idea of the algorithm is to estimate the parameters of
It can be concluded that the presented method gives excel- the rotor model by using the result of the compensation
lent results under all tested operating conditions of the machine. method and the stator model . The stator model can be
Only for a short period after load changes, differential temper- easily identified by means of the measured stator temperature.
ature cannot be observed. This is due to the electromechan- The principle of the algorithm for the reference rotor tem-
ical transient reactions which are caused by these load changes. perature reconstruction is shown in Fig. 12. It is assumed that
Under these conditions, the equivalent circuit diagram does not the proportions of the model parameters for the heating of the
describe the machine. The method is applicable during steady machine during the starting operation and for the cooling of
states or quasi-steady states. the machine during the steady-state operation are identical. For
If the absolute reference temperature is known in advance, it identifying the thermal models, a least square algorithm is used.
would be also possible to calculate the actual absolute rotor tem- In the next step, the dependency of the differential rotor tem-
perature by summation of the actual differential temperature and perature on the reference rotor temperature, given by (5) and in
reference temperature at points of identical slip. In Section III, Fig. 7, can be used to estimate the reference rotor temperature.
an algorithm for absolute reference temperature estimation is By combining the algorithm shown in Fig. 12 with the compen-
presented. The data measured during two similar starts repre- sation method described in Section II, it is possible to obtain
sent the data basis of this algorithm. an algorithm for adaptive calculation of the reference rotor tem-
perature. As already mentioned, the measured data of two sim-
ilar starts represent the data basis of the algorithm. In Fig. 13,
III. DETERMINATION OF THE ABSOLUTE ROTOR TEMPERATURE
the principle of this adaptive reference rotor temperature recon-
As shown in Fig. 11, the thermal transfer function in the struction is shown.
Laplace-domain is the starting point of the algorithm for refer- The compensation method is initialized with a random ref-
ence rotor temperature reconstruction. Here, the input variable erence rotor temperature . The result of the compensation
is a current dependent power loss which is proportional to method, the differential rotor temperature , is used as an input
38 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 19, NO. 1, MARCH 2004

Fig. 15. Measured and calculated absolute rotor temperature.

method. As an example, the result of such a calculation—using

Fig. 13. Principle for adaptive reference temperature reconstruction.
the differential temperature from Fig. 10—is given in Fig. 15.
It can be concluded that it is possible to calculate the rotor
temperature very exactly, the difference between the calculated
and the measured rotor temperature averages less than 10 .

IV. CONCLUSION
In this paper, a new method for observing the rotor tempera-
ture using the input impedance, the slip, and the stator temper-
ature of an induction machine is presented. In the first step, the
temperature difference between the actual rotor temperature and
a reference temperature is detected by using the rotor tempera-
ture dependency of the input impedance. This method is called
compensation method. Because this method has the character-
istic of a differential measurement, the main problem, separating
the changes in the rotor resistance due to skin effect and temper-
ature, can be solved.
In the second step, the dependency of the result of the com-
pensation method on the reference rotor temperature and the
measured stator temperature are used in an algorithm for ref-
erence rotor temperature reconstruction. Here, the assumption
Fig. 14. Result of the adaptive reference temperature reconstruction k : was established that the proportions of the thermal stator and
iteration counter. rotor model parameters for the heating of the machine during
the starting operation and for the cooling of the machine during
variable for the algorithm described in Fig. 12. The result of the steady-state operation are equal. By combining these two
this procedure is again the absolute rotor temperature . This procedures, it is possible to adaptively reconstruct the reference
loop is repeated until the calculated rotor temperature does not rotor temperature. Thus, the absolute rotor temperature can be
change anymore. obtained by addition at any time.
For testing the algorithm reliability, it is necessary to check The methods are tested by simulations and evaluations of data
whether the finally calculated rotor temperature is independent recorded on a 75-kW test machine. In both cases, the efficiency
of the algorithm initialization. of the methods is shown.
In Fig. 14, the results of two runs with different initializa-
tions of the reference temperature are presented. The first cal- REFERENCES
culation (A) started with a reference temperature ramp from [1] A. Eltom and N. Moharai, “Motor temperature estimation incorporating
dynamic rotor impedance,” IEEE Trans. Energy Conversion, vol. 6, pp.
20 to 30 , the second calculation (B) started with a ramp 107–113, Mar. 1991.
temperature from 20 to 200 . [2] A. Bousbaine, “Thermal modeling of induction motors based on accu-
The simulations showed that the calculated rotor reference rate loss density distribution,” Elect. Mach. Power Syst., vol. 27, no. 3,
pp. 311–324, 1999.
temperatures converge to the same result after three iterations, [3] E. G. Egorov, V. S. Genin, and N. M. Mikhailov, “Microprocessor
independent of the selected initial values (see Fig. 14). thermal-safety relay for induction motors with short-circuit rotor,”
Thus, the fingerprint of the machine is complete and it is pos- Russian Elect. Eng., vol. 68, no. 1, pp. 68–71, 1997.
[4] M. S. Rajagopal and K. N. Seetharamu, “Transient thermal analysis of
sible to calculate the absolute rotor temperature by adding the induction motors,” IEEE Trans. Energy Conversion, vol. 13, pp. 62–69,
rotor reference temperature and the result of the compensation Mar. 1998.
MAXIMINI AND KOGLIN: DETERMINATION OF THE ABSOLUTE ROTOR TEMPERATURE 39

[5] A. Boglietti, M. Lazzari, G. Luparia, and V. Hollo, “A simplified thermal [14] S. Spannhake, “Eine Methode Zur Identifikation Der Parameter Der
model for the power derating prediction of variable speed self-cooled Asynchronmaschine mit Doppelkäfigläufer Zur Verlustleistungsberech-
induction motor,” in Proc. Int. Conf. Elect. Mach., Espoo, Finland, Aug. nung,” Ph.D. dissertation, Berlin, T. Univ., Germany, 1994.
28–30, 2000, pp. 348–352.
[6] R. Busch and S. Hilfert, “On-line bestimmung der temperatur von elek-
trischen maschinen mittels neuronaler netze,” ELEKTRIE, vol. 52, no.
1–2, pp. 56–61, 1998.
[7] E. Aviolio and C. Dias, “Development of a thermal-electric mathemat-
ical model, to analyze and specify induction motors using artificial Martin Maximini received the Dipl.-Ing. and Dr.-Ing. degrees from the Saar-
neural networks,” in Proc. Int. Conf. Elect. Mach., Espoo, Finland, land University, Saarbruecken, Germany, in 1995 and 2001, respectively.
Aug. 28–30, 2000, pp. 363–367. Currently, he is Senior Consulting Engineer in the Electrical System Con-
[8] B. Amin, “Principle of the direct determination of the temperature-de- sulting Group of ABB Utilities GmbH, Mannheim, Germany. He is responsible
pendant rotor and stator resistance of an induction motor,” Eur. Trans. for the planning of power systems. His main research interests include reliability
Elect. Power, vol. 8, no. 1, pp. 69–71, 1998. of power systems, protection, and electrical machines.
[9] M. Schrödel, “Identifikation der rotortemperatur einer käfigläufer-asyn-
chronmaschine,” E&I, vol. 105, no. 7/8, pp. 315–318, 1998.
[10] O. V. Thorsen and M. Dalva, “Methods of condition monitoring and fault
diagnosis for induction motors,” Eur. Trans. Elect. Power, vol. 8, no. 5,
pp. 383–395, 1998.
[11] C. Ghita and H. G. Schuster, “Short-circuit/overload protection of in- Hans-Jürgen Koglin was born in 1937. He received the Dipl.-Ing. and Dr.-Ing.
duction motors using a calorimetric method,” in Proc. 31st Int. Intell. degrees from the Technical University Darmstadt, Darmstadt, Germany, in 1964
Motion Conf., Nürnberg, Germany, 1997, pp. 317–321. and 1972, respectively.
[12] Th. Vetter, “Überwachung Und Prädiktion Des Erwärmungsverhaltens Currently, he is a Full Professor at Saarland University, Saarbruecken, Ger-
Einer ASM Mit Käfigläufer Mittels Parameterschätzung,” Ph.D. disser- many, where he has been since 1983. From 1973 to 1983, he was a Professor at
tation, Technical Univ. Darmstadt, Darmstadt, Germany, 1998. Saarland University. His main research interests include the planing and oper-
[13] M. Maximini, H.-J. Koglin, and M. Igel, “New method for rotor temper- ation of power systems and specially optimal MV and LV networks, visibility
ature estimation,” in Proc. Int. Measure. Confederation, Vienna, Austria, of overhead lines, state estimation, reliability, corrective switching, protection,
Sept. 25–28, 2000, pp. 393–398. and fuel cells.