METHOD AND DEVICE FOR ASCERTAINING THE ROTOR TEMPERATURE OF A PERMANENT-MAGNET SYNCHRONOUS MACHINE

Abstract
A method for ascertaining the rotor temperature of a permanent-magnet synchronous machine (10), in which a first estimate (TR1) for the rotor temperature is ascertained as a function of a remanent flux density of permanent magnets contained in a rotor of the synchronous machine (10). A second estimate (TR2) for the rotor temperature is ascertained via a Kalman filter containing a thermal model of the synchronous machine (10), the first estimate (TR1) for the rotor temperature being supplied at least intermittently to the Kalman filter.
Description
BACKGROUND OF THE INVENTION

1. Field of the Invention


The present invention relates to a method and a device for ascertaining the rotor temperature of a permanent-magnet synchronous machine, in which a first estimate of the rotor temperature is ascertained as a function of a remanent flux density of permanent magnets contained in a rotor of the synchronous machine.


2. Description of the Related Art


In permanent-magnet synchronous machines, the permanent magnets of the rotor are usually the components most sensitive to temperature because if the temperature even briefly exceeds the Curie temperature of the magnetic material used, it will cause damage to the magnets, in particular loss of their magnetization. Therefore, temperature monitoring of the rotor of the synchronous machine is advisable to protect the permanent magnets, so that the synchronous machine may be deactivated, e.g., shortly before reaching the Curie temperature.


Monitoring of the rotor temperature is technically complex due to the rotor movement. Known methods therefore propose ascertaining the rotor temperature from the more easily detectable electrical operating variables of the synchronous machine instead of a direct temperature measurement in the rotor.


Such a method is known from published German patent application document DE 102 54 295 A1 and utilizes the effect whereby the remanent flux density of the permanent magnets of the rotor depends on their temperature or on the rotor temperature. Consequently, the amplitude of the synchronous internal voltage of the synchronous machine is a function of the rotor temperature. The known method has the disadvantage that the rotor temperature is ascertained on the basis of a linear model, which is valid only in a restricted operating range of the synchronous machine due to magnetic saturation effects, among other things, and offers inadequate precision for many applications.


SUMMARY OF THE INVENTION

Accordingly, the object of the present invention is to improve upon a method and a device of the type defined in the introduction, to the effect that increased precision is achieved in ascertaining the rotor temperature.


This object is achieved in a method of the type defined above according to the present invention by ascertaining a second estimate for the rotor temperature via a Kalman filter containing a thermal model of the synchronous machine, the first estimate for the rotor temperature being supplied at least intermittently to this Kalman filter. In other words, according to the present invention, in addition to the first estimate, which is known from the traditional methods, a second estimate is also ascertained. The use of the thermal model of the synchronous machine allows a particularly advantageous estimate of the rotor temperature here based on input variables, which are detectable by measurement with little complexity, e.g., the stator temperature. An increased precision in ascertaining the rotor temperature is advantageously achieved through the combination of the conventional first estimate with the Kalman filter containing the thermal model according to the present invention.


The first estimate for the rotor temperature as well as a measured value for the stator temperature may be advantageously supplied according to the present invention to the Kalman filter, in particular for the correction step, the first estimate for the rotor temperature in particular being also used in the sense of a measured variable to obtain a new second estimate via the newly calculated Kalman gain.


In an advantageous specific embodiment of the method according to the present invention, a particularly simple evaluation of the Kalman filter from the mathematical standpoint is obtained when either only the first estimate for the rotor temperature or the measured value for the stator temperature is supplied to the Kalman filter, in particular for a correction step. It is therefore advantageously possible to omit a matrix inversion of the 2×2 output matrix or measurement matrix in ascertaining the Kalman gain as part of the correction step of the Kalman filter.


The exclusive evaluation of just one of the two variables within one iteration of the Kalman filter proposed according to the present invention advantageously makes it possible to use a simpler output matrix, so that instead of a matrix inversion, a division by a scalar quantity is possible.


The method according to the present invention nevertheless offers an increased precision in comparison with the traditional approach because the first estimate for the rotor temperature or the measured value for the stator temperature may be used for the correction step of the Kalman filter, for example, alternating in successive iterations, and thus both values contribute to the refinement of the second estimate obtained via the Kalman filter. These two values may also be used in an irregular order instead of an alternating use. For example, whenever there is a new first estimate for the rotor temperature, the first estimate for the rotor temperature is used by the Kalman filter as the “measured variable” and otherwise the Kalman filter is evaluated using the stator temperature detected by measurement as the measured variable.


In an advantageous specific embodiment of a variant of the present invention, it is advantageously provided that for ascertaining the second estimate precisely, the thermal model of the synchronous machine has at least the stator, rotor and cooling water as components, heat transfers among all components preferably being taken into account.


According to the present invention, it is also possible to use thermal conductance values among the components of the thermal model as a function of a rotational speed of the synchronous machine to increase precision.


The functionality of the device according to the present invention may advantageously be integrated into a traditional control unit for a synchronous machine, wherein the method steps according to the present invention are executable, e.g., by an arithmetic unit of the control unit.


Additional features, possible applications and advantages of the present invention are derived from the following description of exemplary embodiments of the present invention, which are illustrated in the figures of the drawings. All the features that are described or illustrated here, either alone or in any combination, constitute the subject matter of the present invention, regardless of how they are combined in the patent claims or their references back to preceding claims and regardless of how they are described in the description and/or illustrated in the drawings.





BRIEF DESCRIPTION OF THE DRAWINGS


FIG. 1 schematically shows a function diagram of a specific embodiment of the method according to the present invention.



FIG. 2 shows the chronological sequence of individual steps of the method according to the present invention.





DETAILED DESCRIPTION OF THE INVENTION


FIG. 1 schematically shows a permanent-magnet synchronous machine 10, which is controlled by a control unit 20 assigned to it. Electrical operating variables Ud, Uq, Id, Iq and rotational speed N of synchronous machine 10 are sent to an estimator unit 100, which uses them to calculate a first estimate TR1 for the rotor temperature of synchronous machine 10.


Operating variables Ud, Uq correspond to the voltage of synchronous machine 10 in the d direction and q direction, respectively, and operating variables Id, Iq correspond to the current of synchronous machine 10 in the d direction and q direction, respectively.


The operating principle of estimator unit 100 is based on the temperature dependence of the remanent flux density of the permanent magnets of the rotor of synchronous machine 10. Since the amplitude of the synchronous internal voltage of synchronous machine 10 is also a function of the rotor temperature, estimator unit 100 may form a first estimate TR1 for the rotor temperature according to the equation:







T

R





1


=


T

R





0


+




K
m



(

T

R





1


)


-

K

m





0




α
·

K

m





0









where α is the temperature coefficient of the remanent flux density, Km0 is a reference value of the remanent flux density at reference temperature TR0 and:








K
m



(

T

R





1


)


=




U
q

-

R
·

I
q



ω

-


L
d

·


I
d

.







Temperature coefficient α and reference values Km0, TR0 may be ascertained, for example, by measurement under suitably defined operating conditions and stored for normal operation of synchronous machine 10 and execution of the method according to the present invention.


Temperature-dependent remanent flux density








K
m



(

T

R





1


)


=




U
q

-

R
·

I
q



ω

-



L
d



(

I
d

)


·

I
d







is obtained from the steady-state model of synchronous machine 10 using the simplifications described below:






U
d
=R·I
d
−L
q
·I
q·ω  (equation 1a)






U
q
=R·I
q
+L
d
·I
d
·ω+K
m·ω  (equation 1b)


where Ud, Uq denote the voltage of synchronous machine 10 in the d direction and q direction, respectively, Id, Iq denote the current of synchronous machine 10 in the d direction and q direction, respectively, and Ld, Lq denote the corresponding inductances, where R is a strand resistance, ω is the electrical angular velocity, and Km is the temperature-dependent fundamental-wave amplitude of the field-linked direct-axis flux or the remanent flux density.


The following approximations are preferably made to simplify the evaluation of the steady-state model of synchronous machine 10 as presented above:


a) Experience has shown that it is sufficient to take into account only a dependence on current Iq for temperature-dependent remanent flux density Km(TR1).


b) Estimate TR1 is preferably obtained from equation (1b) so that inductance Lq need not be evaluated.


c). The validity range of the model may be restricted to the case when Iq=0.


On the whole, under these assumptions this yields:


for first estimate TR1:







T

R





1


=


T

R





0


+




K
m



(

T

R





1


)


-

K

m





0




α
·

K

m





0










Where







K
m



(

T

R





1


)


=




U
q

-

R
·

I
q



ω

-


L
d

·


I
d

.







In the case of simplification Iq=0, the following equation ultimately applies:








K
m



(

T

R





1


)


=



U
q

ω

-


L
d

·


I
d

.







The advantage of the simplification described above is that first estimate TR1 may be formed only under the corresponding operating conditions, e.g., Iq=0, and therefore cannot always readily be formed anew for a different operating situation.


As an alternative to the simplified consideration described above, equations (1a) and (1b) of this steady-state model for Ud, Uq may also be evaluated without simplifying assumptions, in which case the precision of first estimate TR1 increases accordingly; in particular an instantaneous value may also be obtained in operating situations other than that on which the simplification is based.


In order for the model of equations (1a) and (1b) to also be valid for the case of magnetic saturation, it may be provided according to the present invention that inductances Ld, Lq and variable Km(TR1) are determined as a function of currents Id, Iq, i.e., parameters Id, Iq, Km (TR1) are then functions of two variables, so that a real-time-capable determination with the help of corresponding characteristics maps is available.


In general, to ascertain first estimate TR1 for the rotor temperature, any other method which ascertains the rotor temperature as a function of a remanent flux density may be used additionally or as an alternative.


To allow a more precise determination of the rotor temperature, the method according to the present invention provides, in addition to the consideration of first estimate TR1, that a second estimate TR2 for the rotor temperature is ascertained via a Kalman filter containing a thermal model of synchronous machine 10, first estimate TR1 for the rotor temperature being supplied to this Kalman filter at least intermittently.


In other words, the thermal model according to the present invention for the rotor temperature, which is evaluated by the Kalman filter, is expanded further according to the present invention by first estimate TR1 so that it is possible to ascertain the rotor temperature more precisely than with the traditional methods. In particular, the thermal model according to the present invention advantageously also allows the stator temperature of synchronous machine 10 to be taken into account.


The thermal behavior of synchronous machine 10 may be described approximately by a thermal network, which includes at least the nodes “stator,” “rotor” and “cooling water” but may also be modeled to be more complex, if necessary.


Heat transfers are preferably assumed among all components, so that the following equations are obtained for thermal capacitances cp,R, cp,S:






C
p,R
{dot over (T)}
R2
=G
RS(TS−TR2)+GRK(TK−TR2)+PVR  (equation 2a)






c
p,S
{dot over (T)}
S
=G
RS(TR2−TS)+GSK(TK−TS)+PVS  (equation 2b)


Variable {dot over (T)}R2 here represents the change in second estimate TR2 for the rotor temperature, which is considered according to the present invention, as a function of time, and variable {dot over (T)}S here represents the change in stator temperature TS over time, which can be determined by measurement relatively easily in contrast with the rotor temperature of synchronous machine 10 due to the steady-state stator.


Variables GRS(N), GRK(N), GSK(N) represent the thermal conductance values as a function of rotational speed N of synchronous machine 10 among the rotor (index “R”), stator (index “S”) and cooling water (index “K”) components. Variable GRS(N) denotes the thermal conductance between the rotor and stator, for example.


Temperature TK of the cooling water is assumed to be known. Like stator temperature TS, this may also be detected easily by measurement and therefore together with the latter forms an input variable of the thermal model according to the present invention. Additional input variables of the thermal model according to the present invention include power loss values PVs, PVR, which act as a heat source on the stator and rotor, which are functions of currents Id, Ig, rotational speed N, and the temperatures of the stator and rotor.


In a particularly preferred variant of the method according to the present invention, it is assumed for the purpose of simplification that the dependence of power loss PVS, PVR on the rotor temperature is weak and therefore is negligible in first approximation.


A representation of the state of the thermal model of synchronous machine 10 according to the present invention, which has been discretized over time, according to equations (2a) and (2b) using time index k=0, 1, . . . yields:






X
k+1=AdXk+BdUk  (equation 3)


having the state vector:








x
k

=

[




T


R





2

,
k







T

S
,
k





]


,






x

k
+
1


=

[




T


R





2

,

k
+
1








T

S
,

k
+
1






]


,




of the system matrix defined by the thermal model:








A
d

=

[




1
-


Δ






t


(


G
RS

+

G
RK


)




c

p
,
R








Δ






t
·

G
RS




c

p
,
R









Δ






t
·

G
RS




c

p
,
S






1
-


Δ






t


(


G
RS

+

G
SK


)




c

p
,
S







]


,




of the input matrix:








B
d

=

[





Δ






t
·

G
RK




c

p
,
R






1

c

p
,
R





0






Δ






t
·

G
SK




c

p
,
S





0



1

c

p
,
S






]


,




and having the input vector:







u
k

=


[




T

K
,
k







PV

R
,
k







PV

S
,
k





]

.





On the basis of the thermal model according to equation (3), a Kalman filter is designed according to the present invention and used to ascertain second estimate TR2 for the rotor temperature.


The Kalman filter formed according to the present invention provides for the execution of a prediction step at a sampling time tk according to:






{circumflex over (x)}
k+1

=A
d
x
k
+
+B
d
u
k






P
k+1

=A
d
P
k
+
A
k
T
+Q


and then the correction step for evaluation of the measurement according to:






K
k+1
=P
k+1

H
R(HRPk+1HRT+RM)−1






{circumflex over (x)}
k+1
+
={circumflex over (x)}
k+1

+K
k(yk+1HR{circumflex over (x)}k+1)






P
k+1
+=(I−Kk+1HR)Pk+1,





where:





yk=[01]xk


forms a measured variable in the sense of the Kalman filter,


where first:





HR=[01]  (equation 4a)


forms the output matrix or measurement matrix.


The calculations associated with the prediction step are represented by function block 110 in FIG. 1, and the calculations associated with the correction step are represented by function block 120 in FIG. 1. Function block 130 represents in general all other calculations within the scope of the evaluation of the thermal model of synchronous machine 10 according to the present invention.


Matrix Q, measurement error covariance matrix RM and a starting value for matrix P are to be selected in a suitable manner with which those skilled in the art are familiar. Matrix I is the unity matrix.


Vector of state {circumflex over (x)}k+ at the output of function block 120 contains second estimate TR2 for the rotor temperature at point in time tk.


Through the choice of the output matrix according to equation (4a), this achieves the result that in evaluation of the Kalman filter according to the prediction and correction steps described above as a measured variable for the Kalman filter, only stator temperature TS,k from the following vector is used:








x
k

=

[




T


R





2

,
k







T

S
,
k





]


,




This advantageously yields a particularly simple mathematical evaluation of the Kalman filter because a matrix inversion of a 2×2 output matrix may advantageously be omitted in the correction step of the Kalman filter, and only the simple output matrix HR=[0 1] enters into the correction step of the Kalman filter. Therefore, no complex matrix inversion need be performed but instead only division by a scalar quantity is necessary to ascertain the Kalman gain according to






K
k+1
=P
k+1

H
R(HRPk+1HRT+RM)−1


The special advantage of the method according to the present invention lies in the combination of the thermal model described above with first estimate TR1 ascertained otherwise, i.e., from the model according to equations (1a), (1b).


This combination is advantageously achieved according to the present invention by using, at least intermittently, instead of the output matrix:





HS=[01]


the output matrix:





HR=[10]


which has the effect that now only first estimate TR1 for the rotor temperature is used in the correction step of the Kalman filter in the form of second estimate TR2,k from the vector:







x
k

=

[




T


R





2

,
k







T

S
,
k





]





In other words, for the evaluation of the Kalman filter according to the present invention, TR2,k=TR1 is selected for the transfer of first estimate TR1 of estimator unit 100 into the Kalman filter.


Thus according to the present invention, variable TR2,k is advantageously the time-discrete representation of first estimate TR1 as ascertained by estimator unit 100. There is therefore advantageously the option of forming a second estimate TR2 for the rotor temperature on the one hand via the Kalman filter as a function of the thermal model according to equations (2a) and (2b) and stator temperature TS. On the other hand, second estimate TR2 may be refined advantageously using the output matrix:





HR=[10]


by including first estimate TR1 in the iteration process of the Kalman filter.


Thus, through an appropriate choice of output matrix H, either first estimate TR1 or stator temperature TS may be selected for the correction step of the Kalman filter. The choice may be made in alternation, according to some other regular scheme, which is based on the time grid of the estimation by estimator unit 100, for example, or may also take place randomly.


The exclusive evaluation, proposed according to the present invention, of just one of two values TR1, TS within an iteration of the Kalman filter advantageously makes possible the use of a simpler output matrix H, so that instead of a matrix inversion, division by a scalar quantity is made possible.


Nevertheless, the method according to the present invention offers increased precision in comparison with the traditional approach because, for example, both values TR1, TS may be used for the correction step of the Kalman filter in alternation in successive iterations, and thus both values TR1, TS contribute toward the refinement of second estimate TR2, which is obtained via the Kalman filter.


Second estimate TR2 therefore represents the value ascertained according to the present invention for the rotor temperature, which corresponds better to the actual value of the rotor temperature than do values ascertained conventionally according to investigations by the present applicant.


Both values TR1 and TS may also be used in irregular sequence instead of an alternating use, as already described. For example, whenever a new first estimate TR1 is available, first estimate TR1 may be used by the Kalman filter and otherwise the Kalman filter is evaluated using stator temperature TS.



FIG. 2 shows the chronological sequence according to the present invention in the evaluation of the Kalman filter, a first prediction step p0 (time index k=0) being performed first according to the equations:






{circumflex over (x)}
k+1

=A
d
x
k
+
+B
d
u
k






P
k+1

=A
d
P
k
+
A
k
T
+Q


First prediction step p0 is followed by first correction step k0 according to the equations:






K
k+1
=P
k+1

H
R(HRPk+1HRT+RM)−1






{circumflex over (x)}
k+1
+
={circumflex over (x)}
k+1

+K
k(yk+1HR{circumflex over (x)}k+1)






P
k+1
+=(I−Kk+1HR)Pk+1,


This first iteration p0, k0 of the Kalman filter takes place as an example by using an output matrix of type HR=[0 1], just as in the following second iteration p1, k1.


However, a new first estimate TR1 of estimator unit 100 is available at point in time t0 (see the arrow in FIG. 2), so that the following third iteration of the Kalman filter is performed using a different output matrix of the type HS=[1 0] to refine second estimate TR2 available at the output of the Kalman filter by using newly formed first estimate TR1.


Next, in the fourth iteration and following iterations p3, k3, . . . of the Kalman filter, the output matrix of type HR=[0 1] is again used, for example, until a new first estimate TR1 is available.


If error covariance matrix P stabilizes quickly at a steady-state value at a constant rotational speed N of the synchronous machine, then the transient response may be disregarded, and it is always possible to work using steady-state matrix P, so that a new calculation is not constantly necessary, e.g., in control unit 20. In this case, the elements of matrix P may be stored as characteristic curves depending on the rotational speed.


In another specific embodiment of the present invention, which makes it possible to ascertain the rotor temperature in a particularly accurate manner, a third form is selected for the output matrix:







H
RS

=

[



1


0




0


1



]





This output matrix HRS allows both values TR2,k, TS,k or vector








x
k

=

[




T


R





2

,
k







T

S
,
k





]


,




which are used as “measured values” in the sense of the Kalman filter, to be taken into account simultaneously within one iteration of the Kalman filter, but this requires an enhanced computation complexity accordingly because a matrix inversion is to be performed to form the Kalman gain:






K
k+1
=P
k+1

H
R(HRPk+1HRT+RM)−1


The method according to the present invention thus allows a reliable and precise means of ascertaining the rotor temperature of synchronous machine 10 on the whole and offers in particular a greater accuracy than those methods based only on an evaluation of electrical operating variables of synchronous machine 10.


The method according to the present invention is preferably executed by control unit 20.

Claims
  • 1-11. (canceled)
  • 12. A method for ascertaining the rotor temperature of a permanent-magnet synchronous machine comprising: determining a first estimate (TR1) for the rotor temperature as a function of a remanent flux density of permanent magnets contained in a rotor of the synchronous machine, and determining a second estimate (TR2) of the rotor temperature by using a Kalman filter containing a thermal model of the synchronous machine, the first estimate (TR1) for the rotor temperature being supplied to the Kalman filter at least intermittently.
  • 13. The method as recited in claim 12, wherein the first estimate (TR1) for the rotor temperature and a measured value (TS) for stator temperature are both supplied to the Kalman filter for a correction step.
  • 14. The method as recited in claim 12, wherein either only the first estimate (TR1) for the rotor temperature or a measured value (TS) for stator temperature is supplied to the Kalman filter for a correction step.
  • 15. The method as recited in claim 12, wherein the thermal model of the synchronous machine includes at least stator, rotor and cooling water components, and heat transfers among all components is taken into account.
  • 16. The method as recited in claim 13, wherein the thermal model of the synchronous machine includes at least stator, rotor and cooling water components, and heat transfers among all components is taken into account.
  • 17. The method as recited in claim 14, wherein the thermal model of the synchronous machine includes at least stator, rotor and cooling water components, and heat transfers among all components is taken into account.
  • 18. The method as recited in claim 12, wherein the thermal model uses thermal conductance values between its or the components as a function of a rotational speed of the synchronous machine.
  • 19. The method as recited in claim 12, wherein the thermal model takes into account heat losses of a stator and a rotor of the synchronous machine, and a dependence of the heat losses on the rotor temperature are disregarded.
  • 20. The method as recited in claim 13, wherein the thermal model takes into account heat losses of a stator and a rotor of the synchronous machine, and a dependence of the heat losses on the rotor temperature are disregarded.
  • 21. The method as recited in claim 14, wherein the thermal model takes into account heat losses of a stator and a rotor of the synchronous machine, and a dependence of the heat losses on the rotor temperature are disregarded.
  • 22. The method as recited in claim 15, wherein the thermal model takes into account heat losses of a stator and a rotor of the synchronous machine, and a dependence of the heat losses on the rotor temperature are disregarded.
  • 23. The method as recited in claim 12, wherein a steady-state model of the synchronous machine is used for determining the first estimate (TR1) for the rotor temperature according to: Ud=R·Id−Lq·Iq·ωUq=R·Iq+Ld·Id·ω+Km·ωwhere Ud, Uq denote the voltage of the synchronous machine in the d direction and q direction, respectively, Id, Iq denote the current of the synchronous machine in the d direction and q direction, respectively, and Ld, Lq denote the corresponding inductances, where R is a strand resistance, w is the electrical angular velocity and Km is a temperature-dependent fundamental-wave amplitude of the field-linked direct-axis flux.
  • 24. The method as recited in claim 23, wherein for determining the first estimate (TR1) for the rotor temperature, the synchronous machine is operated temporarily at an operating point at which no saturation prevails.
  • 25. The method as recited in claim 23, wherein inductances Ld, Lq and the fundamental-wave amplitude of field-linked direct-axis flux Km are each ascertained as a function of currents Id, Iq, using characteristics maps.
  • 26. The method as recited in claim 24, wherein inductances Ld, Lq and the fundamental-wave amplitude of field-linked direct-axis flux Km are each ascertained as a function of currents Id, Iq, using characteristics maps.
  • 27. The method as recited in claim 23, wherein the first estimate (TR1) for the rotor temperature is obtained according to the equation:
  • 28. The method as recited in claim 24, wherein the first estimate (TR1) for the rotor temperature is obtained according to the equation:
  • 29. A device for determining the rotor temperature of a permanent-magnet synchronous machine, the device being capable of determining a first estimate (TR1) for the rotor temperature as a function of a remanent flux density of permanent magnets contained in a rotor of the synchronous machine, and the device being capable of determining a second estimate (TR2) for the rotor temperature via a Kalman filter containing a thermal model of the synchronous machine, the first estimate (TR1) for the rotor temperature being supplied at least intermittently to the Kalman filter.
Priority Claims (1)
Number Date Country Kind
10 2008 040 725.9 Jul 2008 DE national
PCT Information
Filing Document Filing Date Country Kind 371c Date
PCT/EP2009/058759 7/9/2009 WO 00 4/5/2011