阅读说明：本技术 用于估计半导体器件的温度的方法和设备以及计算机程序 (Method and apparatus for estimating temperature of semiconductor device, and computer program ) 是由 J-P·克雷默 U·哈尔曼 于 2019-06-25 设计创作，主要内容包括：为了操控驱动马达、如电动车辆的电动机,使用半导体模块(200)。这种半导体模块(200)的形式是IGBT模块。在运行时,在功率半导体模块的功率晶体管(210)和二极管中形成热损耗,这些热损耗导致它们的温度升高。因而,IGBT模块的制造商建议：给操控IGBT模块的控制设备的软件配备保护功能,该保护功能持续监控IGBT模块的温度并且必要时当达到IGBT模块的组件的不容许的温度时进行干预。使用温度模型来进行计算。制造商提供更精确的更高阶的温度模型,但是该温度模型造成计算花费增加。按照本提议,使用简化温度模型,该简化温度模型根据“平衡截断”法被算出并且针对某些工作范围被优化。(For controlling a drive motor, such as an electric motor of an electric vehicle, a semiconductor module (200) is used. This semiconductor module (200) is in the form of an IGBT module. In operation, thermal losses are formed in the power transistors (210) and diodes of the power semiconductor modules, which thermal losses lead to an increase in their temperature. Thus, manufacturers of IGBT modules recommend: the software that controls the control device of the IGBT module is equipped with a protective function that continuously monitors the temperature of the IGBT module and, if necessary, intervenes when an inadmissible temperature of the components of the IGBT module is reached. The calculation is performed using a temperature model. Manufacturers provide more accurate higher order temperature models, but the temperature models result in increased computational expense. According to the proposal, a simplified temperature model is used, which is calculated according to the "equilibrium cut-off" method and is optimized for certain operating ranges.)

1. A method for estimating a temperature of a semiconductor element (210), wherein the semiconductor element (210) is handled by a calculation unit (172, 186), wherein temperature estimation is performed in the same calculation unit or in another calculation unit, characterized in that a reduced version of a temperature model explained for the semiconductor element (210) is used for the temperature estimation.

2. The method of claim 1, wherein the higher order temperature model corresponds to a parameterized 3, 4, or 5 order Foster model.

3. The method of claim 1 or 2, wherein the simplified version of the temperature model is computed by a "equilibrium truncation" calculation method.

4. The method according to any of the preceding claims, wherein two or more separate frequency ranges are predefined for the "equilibrium cut-off" calculation method for further optimizing the model simplification of the temperature model.

5. The method of any of claims 2 to 4, wherein the parameterized Foster model is calculated according to the following formula:

wherein Z_thIs a thermal resistance to be applied to the heat exchanger,

wherein t is the time at which the user is expected to,

where N is the order of the Foster model,

wherein r is₁To r_NIs the N thermal resistances of the semiconductor element (210) illustrated in the data page,

and τ₁To tau_NAre the corresponding N thermal time constants,

wherein the transfer function of the Foster model is calculated according to the following formula:

where s is the laplace variable of the transfer function,

wherein the transfer function for the time discrete calculation is calculated according to the following formula:

according to the parameter r₁, ..., r_NAnd τ₁, ..., τ_NAnd calculating the parameter a according to the selected calculation cycle₀, ..., a_NAnd b₀, ..., b_N-1。

6. The method of claim 5, wherein for the time-discrete calculation of the transfer function according to a first-order simplified model, the following formula is used:

according to the parameter r₁To r_NAnd τ₁, ..., τ_NAnd calculating alpha according to the selected calculation period₀To alpha₂And beta₀。

7. The method according to claim 5, wherein for the time-discrete calculation of the transfer function according to the second-order simplified model, the following formula is used:

according to the parameter r₁To r_NAnd τ₁, ..., τ_NAnd calculating alpha according to the selected calculation period₀To alpha₂And beta₀To beta₁。

8. The method according to claim 5, wherein for the time-discrete calculation of the transfer function according to a third-order simplified model, the following formula is used:

according to the parameter r₁To r_NAnd τ₁, ..., τ_NAnd calculating alpha according to the selected calculation period₀To alpha₃And beta₀To beta₂。

9. The method of any of claims 2 to 8, wherein if the deviation in the selected frequency range relative to the more accurate order N Foster model when calculating temperature effects from a lower order simplified model is greater than 5K in value, the temperature estimation is performed using a higher order simplified model relative to a lower order simplified model.

10. An apparatus for performing the method according to any one of the preceding claims, the apparatus having a semiconductor module (200) and a calculation unit (172, 186) in which a temperature estimate is calculated based on a reduced-order version of the temperature model explained for the semiconductor module (200).

11. The apparatus of claim 10, wherein the semiconductor module (200) has a plurality of power semiconductor elements (210).

12. The device according to claim 10 or 11, wherein the semiconductor module (200) is an IGBT module, corresponding to an insulated gate bipolar transistor.

13. A computer program having a program code for performing the method steps according to any one of claims 1 to 9 when the program code is run in a computing device.

Technical Field

The invention relates to a method for estimating the temperature of a semiconductor device, wherein a current temperature is estimated by means of a temperature response model. The invention also relates to a correspondingly adapted device and a correspondingly adapted computer program.

Background

Power semiconductor modules are used in a number of technical fields. Mention may be made, as examples: frequency converters, switching power supplies, inverters, induction cookers, and also in terms of drive technology. As in the drive technology, power semiconductor modules are increasingly used in vehicle technology and will be put into more use in the future as the field of electric vehicles continues to evolve.

Of particular emphasis are IGBT modules, corresponding to Insulated-Gate Bipolar transistors, which usually comprise a plurality of IGBT-type power transistors and anti-parallel power diodes, acting as freewheeling diodes. In the automotive field, IGBT modules are used, for example, for controlling electric drive motors or in charging devices for electric vehicles. Electrical components, in particular electric motors, are frequently present in electric vehicles or also in vehicles with combustion engines for certain comfort functions, such as windshield wipers, for power steering, seat adjustments, window lifters, door adjusters, etc. These components typically comprise power semiconductors, such as MOSFET type transistors. These power semiconductors are controlled by a control device, which generates corresponding control signals. The control device usually has one or more microcontrollers, which are correspondingly programmed and generate these control signals.

In operation, thermal losses are formed in the power transistors and diodes of the IGBT modules, which lead to their temperature increasing. At higher temperatures, there is a risk that the transistor or the diode will be destroyed. Temperature cycling also results in a reduction in the useful life of these components. Thus, manufacturers of IGBT modules recommend: the software of the control device which controls the IGBT modules is equipped with a protection function (component protection) which continuously monitors the temperature of the IGBT modules and if necessary intervenes when an inadmissible temperature of the components of the IGBT modules is reached or has already intervened in advance in order to prevent damage. A common measure is to reduce the power during the actuation until the switch-off. Critical for component protection and service life is the junction temperature of the respective semiconductor component.

In order to achieve component protection, the junction temperature of the individual IGBTs and diodes must either be measured or estimated from other available measurement parameters. For cost reasons, mounting the sensors on all components is disadvantageous and also prone to errors. Thus, a computational model approach is sought. For example, it is possible to: by using a temperature model, the junction temperature is estimated from the measured current and voltage. For this purpose, the temperature model is integrated into the software of the control unit, for example into the drive control unit of the vehicle.

The methods and parameters for calculating the junction temperatures of the IGBT and diode are specified by the semiconductor manufacturer, for example in the data sheet of the IGBT module. For this purpose, the so-called Foster model (also referred to as partial fraction network) is often used.

The Foster model represents the thermal impedance between two points, for example, the barrier layer of a semiconductor and a cooling fluid used to cool the semiconductor module. From this thermal impedance and the power loss in the semiconductor, a corresponding temperature difference can be calculated, for example between the barrier layer and the cooling liquid.

A reference temperature, for example the temperature of the cooling liquid, is measured with a temperature sensor. The sum of the reference temperature and the calculated temperature difference yields the absolute temperature of each semiconductor.

For the defined IGBT module, the Foster model can be constructed according to the specifications in the data page of the IGBT module. The description in the data page refers to the higher order Foster model, typically between 3 and 5 orders. Then, the implementation of the model results in high computational expense.

In addition, the IGBT module contains a plurality of IGBTs and diodes, which correspondingly further increases the computational expense. If the factorization of the Foster model is multiplied by the number of semiconductor elements, an estimate of the computational expense is derived. For example, a three-phase bridge circuit includes 6 IGBTs and 6 diodes, i.e., 12 semiconductor elements. For high-availability motor control with 6, 9 or 12 phases, an even higher number of semiconductor elements and therefore an even higher computational effort are obtained.

The use of computational models for estimating the junction temperature for protecting components is known from documents DE 10260106 a1 and US 9705487B 2.

Known from DE 102010014070B 4: model simplification is performed on a complex battery model in order to reduce computational expense. Here, a "Balanced Truncation" method is used.

There are also particular features for driving applications in which the motor is operated. In this particular case, the calculation effort for the temperature estimation also increases with the rotational speed of the electric motor, since the phase currents pass at a higher frequency. The heat losses in the semiconductor vary correspondingly with the frequency increase. This requires a short calculation period for the calculation according to the temperature model. In the case of motors for electric vehicles, in particular for electric vehicles, high rotational speeds are required in order to be able to achieve maximum speeds. In the case of these motors, they have to experience a high rotational speed range, so that the gearbox can be saved. Asynchronous motors can reach rated speeds of up to 25000 revolutions per minute, while synchronous motors can achieve rated speeds of about 11000 revolutions per minute.

Therefore, there is a need to simplify the calculation of the junction temperature of the semiconductor module in order to achieve a real-time implementation (algorithm run-time). Here, optimization for certain operating ranges is also important. This is the task of the present invention. Sufficient accuracy in the temperature estimation should likewise be achieved in order to achieve effective component protection.

Disclosure of Invention

This object is achieved by a method for estimating a temperature of a semiconductor module according to claim 1, a corresponding device according to claim 10 and a corresponding computer program according to claim 13.

The dependent claims contain advantageous developments and refinements of the invention in accordance with the following description of these measures.

The solution is as follows: in a method for estimating the temperature of a semiconductor module controlled by a computing unit, wherein a temperature estimation is carried out in the same or in another computing unit, a reduced-order version of the temperature model explained for the semiconductor module is used for the temperature estimation.

For power semiconductor modules, in particular IGBT modules, the Foster model is often used. The Foster model is parameterized in the data page of the IGBT module. Usually, the parameterized Foster model is implemented in software, which is based on the description in the data sheet, but which subsequently results in increased computational expense. According to this proposal, model simplification is performed prior to software implementation. In particular, the Foster model is typically parameterized in order between 3 and 5 in the data page. Lower order models can be parameterized by model simplification. With model simplification, a simplification of the calculation of the junction temperature of the semiconductor module is achieved. The computational expense is correspondingly reduced.

There are different calculation methods for model simplification, which are described in the literature. Some of these calculation methods are already integrated in the development environment, for example, in the Control System Toolbox (Control System Toolbox) of the simulation software Matlab @. A frequently used calculation method is the so-called "equilibrium truncation" method. It is advantageous that: a simplified version of the Foster model is computed by a "balanced truncation" computation. The calculation method enables optimization of model simplification within a defined operating range (frequency range). In this case, the error between the simplified model and the more accurate higher order model is minimized over the defined frequency range. This possibility is advantageous for the application in question, since heat losses in IGBT modules often occur within a defined frequency range.

In the case of this calculation method, too, values of the parameters of the simplified model are obtained. After model simplification, software implementation can be performed using the calculated formulas and parameters. Model simplification can be performed when developing the control device. This can be implemented in a computer of superior performance and need not be performed in the control device itself. Thus, the computational expense for calculating the simplified model is not so important for the subsequent use of the control device, but the computational expense for estimating the temperature from the simplified model plays an important role.

The parameters of the Foster model are illustrated in the data page, where the order of the model is typically between 3 and 5 for IGBT modules. For the Nth order Foster model, N thermal resistances r are illustrated₁To r_NAnd N thermal time constants τ₁To tau_NAs a parameter. The thermal impedance is calculated using the following equation:

for example, for the fourth order model, the thermal impedance is calculated according to the following equation:

the thermal impedance may be presented in an equivalent form as a slave loss power P_LossA transfer function to the temperature difference Δ T. The corresponding formula is:

where s is a Laplace variable.

The transfer function of the parameterized order-N Foster model is calculated according to the following formula:

wherein r is₁To r_NAre the N thermal resistances of the corresponding semiconductor elements, and τ₁To tau_NAre the corresponding N thermal time constants and where s is the laplace variable of the transfer function.

The formula is scaled for time discrete calculations. Next, a transfer function for a time-discrete signal is calculated from a Z-transform based on a transfer function for a continuous signal calculated by a laplace transform. Advantageously, the following formula is used for the time discrete calculation:

wherein, according to the parameter r₁, ..., r_NAnd τ₁, ..., τ_NAnd calculating the parameter a according to a desired calculation cycle₀, ..., a_NAnd b₀, ..., b_N-1. There are a number of computational methods for converting from a continuous representation to a discrete representation. Mention is made here of the bilinear transformation method, the impulse response invariant method and the matching z-transformation method. For example, the time-discrete transfer function of a 4 th order Foster model is:

wherein, according to the parameter r₁, ..., r₄And τ₁, ..., τ₄And calculating the parameter a according to a desired calculation cycle₀, ..., a₄And b₀, ..., b₃. This formula also makes the calculation expensive.

The computational expense is further increased if the description in the data page corresponds to a 5 th order Foster model.

It is therefore advantageous: for the time-discrete calculation of the transfer function according to a first-order simplified model, the following formula is used:

wherein alpha is₀、α₁And beta₀Are parameters calculated according to the "balanced truncation" method.

It is also advantageous: for the time-discrete calculation of the transfer function according to the second-order simplified model, the following formula is used:

wherein alpha is₀、α₁、α₂、β₀And beta₁Are parameters calculated according to the "balanced truncation" method.

It is also advantageous: for the time-discrete calculation of the transfer function according to a third-order simplified model, the following formula is used:

wherein alpha is₀、α₁、α₂、α₃、β₀、β₁And beta₂Are parameters calculated according to the "balanced truncation" method. According to the parameter r₁To r_NAnd τ₁, ..., τ_NAnd calculating the corresponding parameter set alpha according to the selected calculation period₀、α₁And beta₀；α₀、α₁、α₂、β₀And beta₁And alpha₀、α₁、α₂、α₃、β₀、β₁And beta₂。

The "balanced truncation" method enables optimization of model simplification over a defined frequency range. Within the defined frequency range, the error between the simplified model and the more accurate higher order model is minimized. This possibility is advantageous for model simplification of the Foster model, since heat losses in the IGBT module occur within a defined frequency range. The first frequency range is defined by the fact that the heat losses in the semiconductor are always positive. The loss thus contains a direct current component, that is to say a component with a frequency of 0 Hz. The second frequency range is defined by the frequency spectrum of the current and voltage in the semiconductor. The frequency spectrum can be derived from characteristic data of the system, for example for use in a drive system, the frequency spectrum is dependent on the motor speed. Other components of the current and voltage at higher frequencies are caused by the form of the signal used to manipulate the semiconductor element. PWM modulation (pulse width modulation) which is often used for the steering is mentioned as an example. However, the thermal impedance of the semiconductor decreases with frequency, so that the higher frequency components of the heat loss have only a smaller effect on the junction temperature. Thus, the higher frequency components can be ignored for model simplification. This can be tolerated at least when the thermal impedance estimated using the simplified model contains a slight deviation at high frequencies. The deviation is only not allowed to become so large that a temperature difference of more than 5K is obtained from this with respect to the more accurate nth order Foster model.

It has proven advantageous for the method to: if the deviation in the selected frequency range relative to the more accurate order-N Foster model when calculating temperature effects from the lower order model is greater than 5K in value, a higher order simplified model is used for temperature estimation relative to the lower order simplified model. This is a good compromise between the required accuracy of the temperature estimation and avoiding high computational costs.

Corresponding advantages apply to a correspondingly adapted device and a corresponding computer program.

Drawings

An embodiment of the invention is shown in the drawings and will be further explained below with reference to the drawings.

Wherein:

FIG. 1 illustrates a typical cockpit of a vehicle;

FIG. 2 shows a block diagram of an in-vehicle electronic device of a vehicle;

fig. 3 shows an inset of a commercially available IGBT module with 6 IGBTs;

FIG. 4 shows a Bode plot in which the results of a 4-order Foster-temperature model are shown in comparison to a first-order simplified model;

FIG. 5 shows a comparative plot of the thermal impedance of the model 4-order Foster model and the 1-order simplified model;

FIG. 6 shows a comparative plot of the impulse responses of the model 4 order Foster model and the simplified model 1 order;

FIG. 7 shows a Bode plot in which the results of a 4 th order Foster-temperature model are shown in comparison to a 2 nd order simplified model;

FIG. 8 shows a comparative plot of the thermal impedance of the model 4-order Foster model and the 2-order simplified model; and

FIG. 9 shows a comparative plot of the impulse responses of the model 4-order Foster model and the 2-order simplified model.

Detailed Description

This specification sets forth the principles of the disclosure in accordance with the invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements which, although not explicitly described or shown herein, embody the principles of the disclosure and are protected thereby.

Fig. 1 shows a typical cockpit of a vehicle 10. A passenger vehicle Pkw is shown. The vehicle 10 is equipped with an electric motor as a drive motor as an electric vehicle. However, any other vehicle is also contemplated as vehicle 10. Examples of other vehicles are: buses, commercial vehicles, in particular trucks Lkw, agricultural machinery, construction machinery, rail vehicles, and the like. The invention may be applied to land vehicles, rail vehicles, ships and aircraft in general.

As described at the outset, different electric motors are used for different comfort functions in the vehicle. The solution described here is less interesting for these motors, since they do not run permanently during driving and therefore hardly exhibit high temperatures in the semiconductor modules that operate them. However, the situation differs in the case of electric motors that drive the vehicle and in the case of electric motors that are used for steering assistance, since these electric motors are permanently activated during operation. In fig. 1, the display unit of the infotainment system is further highlighted with reference numerals. The display unit is a touch sensitive screen 20 mounted in the center console.

Here, the touch-sensitive screen 20 is used in particular for operating functions of the vehicle 10. For example, a radio, a navigation system, a playback of stored musical compositions and/or air conditioning equipment, other electronic devices, or other comfort functions or applications of the vehicle 10 may be controlled in this regard. In general terms, it is often referred to as an "infotainment system". In motor vehicles, in particular passenger vehicles (Pkw), infotainment systems represent an aggregation of vehicle radios, navigation systems, hands-free talking devices, driver assistance systems and other functions in a central operating unit. The term "infotainment" is a hybrid word consisting of the words information and Entertainment (Entertainment). For operating the infotainment system, a touch-sensitive screen 20 ("touch screen") is primarily used, wherein this screen 20 can be viewed and operated particularly well by the driver of the vehicle 10, but also by the co-driver of the vehicle 10. Below the screen 20, mechanical operating elements, such as keys, adjusting knobs, or combinations thereof, such as push-type adjusting knobs, may also be arranged in the input unit 50. Steering wheel operation of portions of the infotainment system is also generally possible. For this reason, the vehicle is equipped with a so-called multi-function steering wheel operation. The unit is not presented separately but is considered part of the input unit 50.

Fig. 2 shows schematically a block diagram of an electronic device of a motor vehicle and illustrates some subsystems or applications of an infotainment system. The infotainment system comprises: a touch-sensitive display unit 20, a computing device 40, an input unit 50, and a memory 60. The display unit 20 includes not only a display area for displaying changeable graphic information; and includes an operation interface (touch-sensitive layer) arranged above the display area for inputting instructions by a user.

The display unit 20 is connected to the computing means 40 by a data line 70. The data lines may be designed according to the LVDS standard, which corresponds to Low Voltage Differential signaling. The display unit 20 receives control data for manipulating the display area of the touch screen 20 from the computing device 40 through the data line 70. Control data of the entered instructions is also transmitted from the touch screen 20 to the computing device 40 via the data line 70. The input unit is denoted by reference numeral 50. The input unit includes the already mentioned operating elements, such as buttons, adjusting knobs, sliders or push-type adjusting knobs, by means of which the operator can input via menu guidance. Input is generally understood to call out a selected menu option, as well as change parameters, turn on and off functions, and the like.

The storage device 60 is connected to the computing device 40 by a data line 80. In the memory 60, a pictogram list and/or symbol list is stored, which has pictograms and/or symbols for possible fading in of additional information.

Other parts of the infotainment system, such as the camera 150, the radio 140, the navigation device 130, the telephone 120 and the combination meter 110 are connected via the data bus 100 to devices for operating the infotainment system. A high-speed variant of the CAN bus according to ISO standard 11898-2 CAN be considered as data bus 100. Alternatively, for example, bus systems based on ethernet technology, such as IEEE 802.03cg, are also conceivable. It is also possible to use a bus system in which the data transmission takes place via optical waveguides. Mention is made, for example, of the MOST (Media Oriented System Transport) Bus or the D2B Bus (home Digital Bus). For wireless communication inward and outward, the vehicle 10 is equipped with a communication module 160. This module is often also referred to as an On-Board Unit (On-Board Unit). The module may be designed for mobile radio communication, for example according to the LTE standard (corresponding to Long Term Evolution). Likewise, the module may be designed for WLAN communication, corresponding to a Wireless LAN, whether for communication with the occupant's equipment in the vehicle or for vehicle-to-vehicle communication, etc.

The infotainment system communication bus 100 is connected to the gateway 30. Other parts of the automotive electronics are also connected to the gateway. One aspect is the powertrain's communication bus 104, which is typically implemented in the form of a CAN bus. As an example, a control device of a powertrain is mentioned and shown: an engine control device 172, an ESP control device 174 and a gearbox control device 176. A further aspect is a communication bus 102 for a driver assistance system, which may be designed in the form of a FlexRay bus. Here, three driver assistance systems are shown: a driver assistance system 182 for automatic distance adjustment ACC, corresponding to Adaptive Cruise Control; a driver assistance system DCC for adaptive Chassis adjustment 184, corresponding to Dynamic Chassis Control; and a steering assist system 186. Further, a communication bus 106 is connected to the gateway 30. The communication bus connects the gateway 30 with the onboard diagnostic interface 190. The tasks of the gateway 30 are: format conversion is performed for the different communication systems 100, 102, 104, 106 so that data can be exchanged between each other.

The semiconductor module mentioned is a power semiconductor module. Such modules are typically implemented as IGBT modules, corresponding to insulated gate bipolar transistors. The IGBT modules are usually mounted directly at the respective electric motors. As described, the engine control device 172 will operate the corresponding IGBT module of the drive motor, and the steering assist control device 186 will operate the IGBT module of the electric motor of the steering assist system. Fig. 3 shows a view of a typical IGBT module with 6 IGBTs. Additionally, each IGBT in an IGBT module contains freewheeling diodes, which are likewise semiconductor elements and which protect the IGBTs against overvoltage.

The temperature response of the respective IGBT module is described in the respective data page of the semiconductor module manufacturer.

An example of a data page is illustrated by the company Infineon (Infineon). This data page is the data page for IGBT module FS820R08A6P 2B. Reference is made in this respect to the Foster model, and the partial fraction coefficients are illustrated in r and τ pairs.

Foot mark	Thermal resistance	Time constant
				[k/W]	[s]
1	r1=0.005	τ1=0.001
			2	r2=0.05	τ2=0.03
3	r3=0.065	τ3=0.25
			4	r4=0.02	τ4=1.5

The parameters of the 4 th order Foster model are illustrated in the data sheet of the module. These parameters are the four thermal resistances r₁, ..., r₄And corresponding four time constants τ₁, ..., τ₄. The transfer function for the laplace transform of the 4 th order Foster model in the frequency range (s is the laplace variable) with partial fraction decomposition is:

for calculations in a digital computer, a continuous transfer function is converted to a time-discrete transfer function. The 4 th order time discrete model is in the z-domain:

according to the data page parameter r₁, ..., r₄And τ₁, ..., τ₄And calculating the parameter a according to the calculation cycle₀, ..., a₄And b₀, ..., b₃。

First, the result of simplifying the model to a1 st order model according to the "balanced truncation" method is presented. For time discretization, the calculation period is assumed to be 0.5 ms.

The simplified model of order 1 is presented in the z-domain as:

with a simplified model of order 1, the computational expense (compared to a more accurate model of order 4) is reduced to approximately one-fourth.

The simplification is optimized for a defined frequency range. Consider as an application an inverter for a drive motor in an electric vehicle 10. Due to the design of the drive arrangement, thermal losses are formed in the semiconductor element in the range of approximately 22 to 6280 rad/s during driving, when the vehicle speed changes, in the range of 3.5 to 1000 Hz or as a circular frequency ω. Note here that: the illustrated circular frequency does not correspond to the rotational speed of the motor. More precisely, the circular frequency is the circular frequency of the power loss in the semiconductor, in rad/s. In general, the following conversion equation applies: the circular frequency ω = 2 pi f, where f corresponds to the frequency in Hz.

Thus, in the illustrated example, the simplified model is optimized for two frequency ranges as follows:

-direct current component of the loss power: 0 to 0.01 Hz (circle frequency 0 to 0.0628 rad/s);

-variable frequency of power loss in driving range: 3.5 to 1000 Hz (circular frequency 22 to 6280 rad/s).

Here, for comparison, mention is made of: the parameters 3.5 Hz and 1000 Hz correspond to the driving range of the electric vehicle, which extends from the walking speed up to about 70 km/h.

Fig. 4 shows the results of the 1 st order simplified model in the form of a bode plot. Thereby showing the magnitude and phase of the transfer function. The circular frequency is illustrated on the abscissa in rad/s. In the upper part, the amplitude in decibels (Dezibel) is plotted along the ordinate, while in the lower part the phase in deg is plotted. Curve a corresponds to the results of a 4-step continuous Foster model. Curve B corresponds to the result of a simplified continuous model of order 1. Curve C corresponds to the result of a simplified time discrete model of order 1. It can be clearly seen that: in the important range from 0 to 0.1 rad/s, these curves coincide very well and do not deviate much from the 4 th order continuous Foster model. This applies not only to the amplitude but also to the phase. In the high frequency range starting from about 2000 rad/s, there are deviations in both amplitude and phase. In the case of these two 1 st order simplified models (continuous and time discrete), there is little difference in amplitude. The two curves B and C almost exactly overlap. Only for the phase, significant differences occur in the range of 2000 rad/s and higher. In the frequency range from 22 to 6280 rad/s, especially around 22 rad/s, the accuracy of the 1 st order simplified model is insufficient. In the case of a power loss of 700W corresponding to the specification of the maximum power in the data page of the IGBT module, the error of the calculated temperature is 8.5K.

Fig. 5 shows a comparison of the thermal impedance of the 1 st order simplified model with the 4 th order continuous Foster model. No significant difference can be seen in the two simplified models. In both cases, these two curves almost coincide with the curves of the 4 th order Foster model in such an important range with low steering speed, i.e. at t = 4 s.

Fig. 6 shows a comparison of the impulse response of the simplified model of order 1 with the continuous model of order 4. No significant difference can be seen in the two simplified models. Here, the two curves B, C also coincide with the 4 th order Foster model for large time values after the pulse. In the case of short time values close to zero, the 1 st order simplified time discrete model provides a result of 11.8K/s, while the 4 th order continuous Foster model provides a result of 7K/s.

Next, for the illustrated example, the model is simplified to a 2 nd order model by a "balanced truncation" method is illustrated. The 2 nd order simplified Foster model is presented in the z-domain as:

with a 2 nd order simplified model, the computational expense (compared to a more accurate 4 th order model) is reduced by approximately one-half.

This simplification is optimized for the same frequency range as the simplification to the 1 st order model.

For time discretization, a calculation period of 0.5ms is again used.

Fig. 7 shows the results of the simplified model of order 2 in the form of a bode diagram. Curve a corresponds to the results of a 4-step continuous Foster model. Curve D corresponds to the result of a simplified continuous model of order 2. Curve E corresponds to the result of a 2 nd order simplified time discrete model. It can be clearly seen that: in the important range from 0 to 0.1 rad/s, these curves coincide very well and do not deviate much from the 4 th order continuous Foster model. This applies not only to the amplitude but also to the phase. In the high frequency range starting from about 2000 rad/s, there are deviations in both amplitude and phase. In the case of these two 2 nd order simplified models (continuous and time discrete), there is little difference in amplitude. The two curves D and E almost exactly overlap. Only for the phase, significant differences occur in the range of 2000 rad/s and higher. The deviation in the frequency range from 22 to 6280 rad/s is smaller than in the case of the simplified model of order 1. In the case of a power loss of 700W corresponding to the specification of the maximum power in the data page of the IGBT module, the error of the calculated temperature is numerically less than 5K.

Fig. 8 shows the thermal impedance of the 2 nd order simplified model compared to the 4 th order continuous model. No significant difference can be seen in the two simplified models. In both cases, these two curves almost coincide with the curves of the 4 th order Foster model in such an important range with low steering speed, i.e. at t = 4 s.

Fig. 9 shows a comparison of the impulse response of a 2-order simplified model with a 4-order continuous Foster model. No significant difference can be seen in the two simplified models. Here, the two curves d), e) also coincide with the 4 th-order Foster model for large time values after the pulse. In the case of short time values close to zero, the 2 nd order simplified time discrete model provides a result of 8.2K/s, while the 4 th order continuous model provides a result of 7K/s. For this model, the deviation is smaller.

Overall, the 2 nd order simplified model exhibits higher accuracy than the 1 st order model.

Using a definite normTo illustrate the accuracy of the simplified model. The 2 nd order model is more accurate than the 1 st order model.

		1 st order model	0.0419 K/W
2 nd order model	0.0229 K/W

The choice of the order of the simplified model, for example 1 or 2, depends on a compromise between the required accuracy on the one hand and the computational expense on the other hand. Here, suggested as guiding principles are: the higher order simplified model is used whenever the deviation in the selected range when calculating the temperature effect from the lower order model is greater than 5K in value relative to the more accurate 4 th order Foster model.

It is also possible to: the quality of the simplified model is checked by simulation. The power electronic circuit under consideration is simulated for the planned operating conditions. The heat loss in the semiconductor element was also calculated in the simulation. The calculated heat losses are used as input variables for the thermal model. This enables a comparison between a more accurate higher order Foster model and a simplified thermal model under near-realistic conditions.

It is to be understood that the proposed method and the pertaining apparatus can be implemented in different forms of hardware, software, firmware, special processors, or a combination thereof. The special processors may include Application Specific Integrated Circuits (ASICs), Reduced Instruction Set computers (RSICs), and/or Field Programmable Gate Arrays (FPGAs). Preferably, the proposed method and apparatus are implemented as a combination of hardware and software. Preferably, the software is installed as an application program on the program storage device. Generally, to a machine based on a computer platform having hardware such as one or more Central Processing Units (CPUs), a Random Access Memory (RAM), and one or more input/output (I/O) interfaces. In addition, an operating system is typically installed on the computer platform. The different processes and functions that have been described herein may be part of an application or may be part of an implementation through a manipulation system.

The present disclosure is not limited to the embodiments described herein. There is room for various matches and modifications that one of ordinary skill in the art would consider based on his or her expertise and that which is part of this disclosure.

List of reference numerals

10 vehicle

20 touch sensitive display unit

30 gateway

40 calculation unit

50 input unit

60 memory cell

70 data line for display unit

80 data line for memory cell

90 data line for input unit

100 first data bus

102 second data bus

104 third data bus

106 fourth data bus

110 combination meter

120 telephone

130 navigation device

140 radio

150 vidicon

160 communication module

172 engine control apparatus

174 ESP control device

176 gearbox control device

182 distance adjustment control equipment

184 chassis control device

186 steering assist control apparatus

190 vehicle-mounted diagnosis plug

200 IGBT module

210 IGBT

Temperature response of A4 th order Foster model (continuous)

Temperature response of B1 order simplified model (continuous)

Temperature response of C1 order simplified model (time discrete)

Temperature response of D2 order simplified model (continuous)

Temperature response of E2 order simplified model (time discrete)