Determination of temperature dependence of on-resistance

Abstract

The invention relates to the determination of a temperature dependency of an on-resistance, in particular to a device (22) for determining a temperature dependency of an on-resistance of a switch (20) in a power semiconductor module (18), comprising: a pulse unit (26) for applying a current pulse to the switch; a measuring unit (28) for measuring the resistance of the switch during and / or after the current pulse is applied to the switch for a measuring time period; and an evaluation unit (30) for determining a resistance-temperature curve of the switch. The invention also relates to a traction converter (12) for a vehicle (10), a training device (34) for providing a predictive model, a method for determining a temperature dependency of an on-resistance of a switch (20), and a computer program product.

Description

Field of the Invention

[0001] The present invention relates to a device for determining the temperature dependence of the on-resistance of a power semiconductor module. The present invention also relates to a training device for providing a prediction model, a traction inverter for a vehicle, a method for determining the temperature dependence of the on-resistance of a power semiconductor module, and a computer program product. Background Art

[0002] The power electronics of electric and hybrid vehicles transmit traction energy from the battery to the electric motor and convert direct current into alternating current in the process. For this purpose, an alternating current converter, an inverter or a traction inverter is provided. Usually, a plurality of transistors or other power semiconductors are used, which are combined into a power semiconductor module and switched at short and regular time intervals. In particular, in this context, MOSFETs (metal-oxide-semiconductor field-effect transistors) and IGBTs (insulated-gate bipolar transistors) are used as switches. In the on-state, the battery current is transmitted to the motor (conduction phase). The voltage profile of the alternating voltage is achieved by these high-frequency switching processes and can then be converted into traction energy in the electric motor. To increase the current-carrying capacity, usually a plurality of power semiconductors are connected in parallel in the (topological) switches.

[0003] In this context or also in other fields, relatively high currents are switched with the aid of such power semiconductor modules or power semiconductors, and the power semiconductors being switched generate high temperatures here. Usually, an active cooling system is used to dissipate the switching and conduction losses. In the event of a failure due to material damage or due to control problems, the temperature may rise excessively here, which causes the relevant components to fail or even causes the module to burn out. The power semiconductors have a maximum operating temperature that is permitted during operation (for example, 150 °C for Si IGBTs, 175 °C for SiC MOSFETs). This maximum operating temperature is specified by the semiconductor manufacturer. When designing the drive device of the power electronics, it must be ensured that this temperature is not exceeded in any operating state.

[0004] For monitoring the status and for recognizing damaged conditions and preventing dangerous situations, there are currently power semiconductor modules with an integrated temperature measurement function in this regard. Here, the prior art is to use NTC (negative temperature coefficient) resistors or PTC (positive temperature coefficient) resistors to monitor the chip temperature. Due to the temperature dependence of the resistance value, temperature measurement can be carried out and the temperature within the power semiconductor module can be inferred. Alternatively, there are also methods of directly measuring the temperature of the power semiconductor by using temperature-sensitive electrical parameters (TSEP).

[0005] The challenge in using additional components for temperature measurement in a power semiconductor or a power semiconductor module lies in the additional space requirement. In addition, it also results in the additional components having to be arranged spatially separately from the actual power semiconductor. In this regard, the temperature measurement is always carried out at a location different from the actual relevant location. Therefore, the actual temperature of the power semiconductor can only be calculated or estimated based on such a measurement.

[0006] In this regard, DE 10 2020 208 167 A1 discloses a power module for operating an electric vehicle drive device, which has an improved function for determining the temperature of the power semiconductor. The power module includes a plurality of power switches, each of which has a power semiconductor. In addition, the power module further includes control electronics for controlling the plurality of power switches in order to generate an output current based on an input current. The control electronics also includes a temperature unit, which is configured to obtain the operating voltage and operating current of the power semiconductor and to determine the temperature of the power semiconductor based on the operating voltage and operating current.

[0007] The challenge in temperature measurement of power semiconductors via the temperature-dependent on-resistance of power semiconductor switches lies in the necessary calibration. For the measurement, a resistance-temperature curve is required because each individual power semiconductor or each individual power semiconductor module may have different characteristics in this regard due to material differences or also due to different contact methods, etc. In other words, the production-induced fluctuations of this resistance-temperature curve make it necessary to calibrate each power semiconductor or each switch in order to know this curve and to enable reliable temperature measurement during operation. However, this calibration is time-consuming and causes relatively high costs because, for example, a relatively time-consuming heating process must be carried out when producing power semiconductor modules, and the resistance and temperature are measured simultaneously. In addition, even during the operation of power semiconductor modules, for example, in a traction converter in a vehicle, this curve may change during the service life of the power semiconductor module, for example, due to aging effects. In order to continue to enable reliable temperature measurement, a new calibration is then necessary. Summary of the Invention

[0008] Based on this, the object of the present invention is to provide a method for determining the temperature dependence of the on-resistance of a switch in a power semiconductor module. In particular, compared with the hitherto calibration methods, cost savings should be emphasized while reliably obtaining the corresponding data. A method that can be implemented efficiently should be provided. Finally, reliable temperature measurement should also be possible even throughout the service life of the power semiconductor module or the traction converter in a vehicle.

[0009] To solve this problem, in one aspect, the present invention relates to a device for determining the temperature dependence of the on-resistance of a switch in a power semiconductor module, the device having: a pulse unit for applying a current pulse to the switch; a measurement unit for measuring the resistance of the switch during and / or after the application of the current pulse within a measurement period; and an evaluation unit for obtaining the resistance-temperature curve of the switch based on the measured resistance and a predefined prediction model.

[0010] In a further aspect, the present invention relates to a traction converter in a vehicle, the traction converter having a power semiconductor module and the above device, wherein: the measurement unit is configured to measure the resistance of the switch; and the evaluation unit is configured to obtain the temperature of the switch based on the obtained resistance-temperature curve.

[0011] In a further aspect, the present invention relates to a training device for providing a prediction model, the training device having: An input interface for receiving a plurality of resistance-temperature curves of various switches; and A modeling unit for obtaining a prediction model based on the received resistance-temperature curves, wherein the modeling unit is in particular configured to train an artificial neural network.

[0012] A further aspect of the invention relates to a method configured according to a device and a training device, and a computer program product having program code which, when executed on a computer, is used to perform the steps of the method. Furthermore, an aspect of the invention also relates to a storage medium on which a computer program is stored which, when executed on a computer, causes the implementation of the method described herein.

[0013] Preferred embodiments of the invention are described in the dependent claims. It is to be understood that the features described above and to be explained below can be used not only in the respectively described combinations, but also in other combinations or alone, without departing from the scope of the invention. In particular, the device, the traction converter, the training device, the method and the computer program product can be configured according to the embodiments described for the device in the dependent claims.

[0014] According to the invention, a current pulse is applied to the switches in the power semiconductor module. Thus, the current pulse conducts through the switch, which causes heating. Here, the temperature reached by the heating is (far) below the possible or permissible maximum temperature. In particular, the switch includes one or more power semiconductors. During and / or after the conduction of the current pulse, the resistance of the switch is measured. Based on this resistance measurement, the resistance-temperature curve of the switch is obtained by using a predefined prediction model. The resistance-temperature curve assigns different resistance values to the temperature. According to the invention, the temperature is not measured directly, but is obtained based on a predefined prediction model. The prediction model here includes data on the temperature characteristics of mutually similar components. Thus, by means of the predefined prediction model, an extrapolation method can be achieved, which is based on the assumption that the temperature characteristics over the entire temperature range can also be recognized by means of a current pulse that does not cause the possible maximum temperature to be reached. Starting from the fact that there is a significant scatter in the resistance-temperature characteristics of the individual switches or individual power semiconductors, by using the method according to the invention of applying a current pulse and measuring the resistance during a measurement period, it is possible to arrive at the conclusion of the resistance-temperature curve of the switch. This curve can then be used in further operation for temperature measurement (during operating time).

[0015] Compared with the corresponding methods for calibrating the temperature sensorless temperature determination of power semiconductors to date (in which heating is carried out to a pre-defined high temperature), the method according to the invention enables time and cost savings. The external temperature sensor for calibration is eliminated. It is possible to determine the resistance-temperature curve of the power semiconductor without using direct temperature measurement or without fully knowing it through measurement during calibration. By using a pre-defined prediction model, statements can be made regarding the temperature characteristics, which enable sufficiently precise operation. This results in efficient and precise calibration possibilities. In addition, the method according to the invention also enables re-calibration at a later time point, for example when the power semiconductor module has already been installed in the traction converter of a vehicle. Therefore, the resistance-temperature curve can be re-determined during working hours or operating time in order to reflect possible changes in the resistance-temperature curve due to aging phenomena. In this regard, improved and more precise temperature measurement can be achieved.

[0016] In particular, a corresponding training device can be provided to determine the pre-defined prediction model. The training device receives the resistance-temperature curves of various switches, which have been obtained, for example, during a conventional measurement process by using a corresponding temperature sensor, and creates a corresponding prediction model based on these training data. In particular, an artificial neural network can be used and trained accordingly here. Therefore, the training data can be used, in which the resistance measurements during the measurement period are assigned to the resistance-temperature curves of the various switches obtained thereby. Then, the pre-trained or pre-defined prediction model can be used in the device according to the invention.

[0017] The resistance-temperature curve obtained by means of the device according to the invention can be used in the traction converter of a vehicle. Then, temperature determination can be carried out in the traction converter, and the temperature of the power semiconductor or the power semiconductor module can be obtained. In this regard, it is particularly advantageous that the power semiconductor module itself can contain all the components of the device according to the invention. Therefore, no external components are required for calibration, and in this way, in-operation calibration or subsequent calibration can be achieved.

[0018] In a preferred design, the pulse unit is configured to apply current pulses to the switch with a duration between 0.5 s and 20 s, preferably between 1 s and 10 s, particularly preferably between 1 s and 2 s, and a current intensity between 50 A and 500 A, preferably between 100 A and 250 A, particularly preferably between 150 A and 200 A. By using relatively short current pulses, sufficient heating and sufficient change in the on-resistance are obtained so that when a corresponding predefined prediction model exists, reliable statements about the resistance-temperature curve can be derived. This results in efficient calibration possibilities.

[0019] In a preferred design, the measuring unit is configured to measure the resistance based on a voltage measurement taken during the application of the current pulse to the switch. Additionally or alternatively, the measuring unit is configured to measure the resistance based on a voltage measurement taken on the measuring current after the application of the current pulse to the switch. Thus, on the one hand, the current of the current pulse can be directly utilized during the voltage measurement, and the voltage can be intercepted during the application of the current pulse. Since the resistance changes during heating due to the current pulse, the voltage also changes even when the current intensity remains constant. The resistance is temperature-dependent. Since the current and the measurement time period are known, the change in the resistance value can be recorded and then used, by means of a predefined prediction model, to obtain the resistance-temperature curve. Alternatively or in addition thereto, it is also possible that during the cooling phase, i.e., after the switch has been applied with a current pulse, a (smaller) measuring current is applied and the change in the resistance is obtained based on this measuring current. During the cooling phase, since the temperature drops again, the resistance also changes after the current pulse is no longer applied. This temperature change and the corresponding change in the resistance can also serve as a basis for obtaining the resistance-temperature curve. On the one hand, the measurements during heating and cooling can be combined. However, on the other hand, it is also possible to perform the measurement only during heating or during cooling. This results in an efficiently achievable determination of the resistance-temperature curve of the switch.

[0020] In a preferred design, the evaluation unit is configured to obtain the resistance-temperature curve based on machine learning methods. In particular, a pre-trained model can be used. Additionally or alternatively, the evaluation unit can be configured to obtain the resistance-temperature curve based on a pre-trained artificial neural network. Using machine learning methods and in particular the method of artificial neural networks enables a reliable and accurate prediction of the resistance-temperature curve to be achieved when only one resistance is measured during the measurement time period. Thus, by means of the corresponding change in the resistance when the switch is heated or cooled, a complete resistance-temperature curve can be predicted. Here, the use of machine learning methods or the method of artificial neural networks enables efficient implementation.

[0021] In a preferred design, the pulse unit is configured to load a current pulse onto a switch having a plurality of power semiconductors. In particular, a plurality of parallel-connected power semiconductors can be combined to form the switch of the topology and are loaded with the current pulse jointly. The resistances of all the parallel-connected power semiconductors are measured. In this regard, an estimation of the maximum temperature is achieved. The resistance-temperature curve is carried out for all the parallel-connected power semiconductors of the switch. In this way, an efficient measurement or an efficient determination of the temperature dependence of the on-resistance of the switch is obtained.

[0022] In a preferred design, the device comprises a receiving interface for receiving a calibration request. The pulse unit is configured to load a current pulse onto the switch after receiving the calibration request. The measuring unit is configured to measure the resistance after receiving the calibration request. The evaluation unit is configured to obtain the resistance-temperature curve after receiving the calibration request. The calibration request can, to a certain extent, be regarded as a signal for triggering the corresponding calibration process or recalibration. For example, the change in the resistance-temperature curve of the switch due to aging can be taken into account. If the resistance-temperature curve changes, a calibration request can be sent or received, and the re-obtaining of the resistance-temperature curve can be initiated based on the current state of the switch.

[0023] In a preferred design, the receiving interface is configured to receive the calibration request from the vehicle controller via a mobile wireless connection and / or after a predefined operating condition occurs. For example, the calibration request can be stored in the vehicle controller and sent periodically or after a specific condition occurs in order to trigger a new calibration. It is also possible that the calibration request originates from a signal received via the mobile wireless connection. In this regard, a new calibration can also be initiated by remote access. In this way, accurate temperature measurement can be carried out in the traction converter throughout the entire service life of the vehicle.

[0024] In this document, a power semiconductor module particularly refers to a structural component used in an inverter structure for a vehicle or in a traction converter. In this document, a switch includes one or more power semiconductors. In this regard, a power semiconductor module includes multiple power semiconductors or multiple switches. A power semiconductor particularly corresponds to an electronic chip having one or more integrated circuit components. For example, MOSFETs and / or IGBTs can be used. In this document, a current pulse is preferably a current flowing through a switch that is applied as constant as possible within a predetermined (short) time period. Therefore, the current flows through one or more power semiconductors. The on-resistance, or resistance, is measured. A predefined prediction model can particularly refer to an assignment rule that assigns input parameters (i.e., the measured resistance values) to output parameters (i.e., the resistance-temperature curve). In particular, a predefined prediction model can be a pre-trained artificial neural network. In particular, the resistance-temperature curve assigns the (measured) resistance to the temperature. In particular, the resistance measurement is performed within a predetermined period of time. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be described and explained in more detail below with reference to some selected embodiments in conjunction with the drawings. Among them:

[0026] Figure 1 Schematic diagram showing a vehicle having a traction converter according to the present invention;

[0027] Figure 2 Schematic diagram showing a device for determining the temperature dependence of the on-resistance according to the present invention;

[0028] Figure 3 Schematic diagram showing a training device according to the present invention;

[0029] Figure 4 Exemplary qualitative illustration showing various resistance-temperature curves for different switches;

[0030] Figure 5 Schematic diagram showing a curve of the variation of current over time for an exemplary current pulse;

[0031] Figure 6 Schematic diagram showing an exemplary curve of the resistance of a switch during a measurement period; and

[0032] Figure 7 Schematic diagram showing a method according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] In Figure 1Vehicle 10 with a traction converter 12 according to the present invention is schematically shown. The traction converter 12 is arranged between the battery 14 and the electric motor 16 of the vehicle 10 in order to convert the direct current of the battery 14 into the alternating current required by the electric motor 16. The traction converter 12 includes a power semiconductor module 18 having switches 20 and a device 22 for determining the temperature dependence of the on-resistance of the switches 20. Figure 1 The illustration in should be understood as a schematic side sectional view. In the illustrated embodiment, the device 22 is integrated into the power semiconductor module 18 or is configured as part of the power semiconductor module 18. In particular, it is possible here that the device 22 according to the present invention is integrated into the microprocessor of the power semiconductor module 18 or is integrated into the control electronics of the power semiconductor module 18. It is also possible in other embodiments of the present invention that the device 22 according to the present invention is configured as a separate component. In addition, in the illustrated embodiment, the switch 20 is shown having two parallel power semiconductors 24. It is also conceivable in other embodiments that the switch 20 includes only a single power semiconductor 24 or multiple power semiconductors 24.

[0034] It is provided according to the present invention that the temperature dependence of the on-resistance of the switch 20 is determined by means of the device 22. If this temperature dependence is known, the temperature of the switch 20 or the power semiconductor module 18 can be determined by means of a resistance measurement. This temperature measurement or temperature determination carried out in the power semiconductor 24 is safety-related for the operation of the inverter or the traction converter. In the case of overheating of the power semiconductor, the inverter has to throttle the output power if necessary in order to avoid failure. By means of the temperature measurement that can be achieved according to the present invention, the operation of the inverter can be carried out closer to its limit range. Thereby, the output power can be increased, or it can be achieved that smaller power semiconductors are installed to save costs. The method according to the present invention for determining the temperature of the power semiconductor 24 via the temperature-dependent on-resistance of the power semiconductor is equivalent to an indirect temperature measurement. A temperature sensor is not necessary. During operation, the current-dependent voltage of the power semiconductor 24 in the on-state is measured. Then, the resistance is obtained based on this measurement. In the case of a known resistance-temperature curve, this resistance can be assigned to the temperature. The device 22 according to the present invention is used to obtain this resistance-temperature curve. The fluctuations caused by production in the resistance-temperature curve result in that in the prior art methods, this curve has to be obtained by means of calibration for each power semiconductor. According to the present invention, it is now provided that this relatively time-consuming calibration can be dispensed with. Here, full use is made of the fact that the resistance of the switch 20 or the power semiconductor 24 changes in a characteristic manner when a current pulse is applied. It can be said that this resistance-time curve contains the characterization of the resistance-temperature curve of the tested or determined component.

[0035] Figure 2 Device 22 according to the invention is schematically shown. Device 22 includes a pulse unit 26, a measurement unit 28, and an evaluation unit 30. In addition, device 22 may optionally include a receiving interface 32. The various interfaces and units can be implemented partially or fully in hardware and / or software here. In particular, it is possible that device 22 is partially or fully equivalent to a microcontroller or software for a microcontroller. Here, in particular, the microcontroller or the software for the microcontroller can be equivalent to the processor of the power semiconductor module 18 or the software of the processor. Here, the functions according to the invention can be implemented regardless of whether the power semiconductor module 18 has already been installed in the traction converter 12 of the vehicle or not. The method according to the invention can be carried out directly after the manufacture of the power semiconductor module 18 or during ongoing operation in the installed state.

[0036] The switch 20 is loaded with a current pulse via the pulse unit 26. The current pulse is applied. The current pulse can be generated in the pulse unit 26 or can be said to be switched on. In this regard, the pulse unit 26 can only fulfill the function of switching the switch 20 or the current. For example, a current pulse with a duration between 0.5 s and 20 s and a current intensity between 50 A and 500 A can be switched out. In particular, a current pulse with a constant current can be switched out. The current pulse can be applied to a single power semiconductor 24 here, or can also be applied to the switch 20 with multiple power semiconductors 24. Usually, multiple power semiconductors 24 are connected in parallel and form a (topological) switch. Then, the temperature measurement indicates the temperature of the switch 20.

[0037] The measurement unit 28 is used to measure the resistance of the switch 20 during a measurement period. The resistance is known, for example, based on the voltage drop. The resistance changes due to the applied current pulse. In particular, the applied current pulse causes the resistance to increase. This increase is caused by heating. The measurement of the resistance can be carried out during the measurement period while the current pulse is being loaded. However, it is also possible that the resistance measurement is carried out within a period after the application of the current pulse. In this regard, the measurement can be equivalent to the measurement during heating, or can also be equivalent to the measurement during cooling.

[0038] In the evaluation unit 30, the resistance-temperature curve of the switch 20 is determined based on the measured resistance. For this purpose, a predefined prediction model is used. In particular, machine learning methods can be implemented here. For example, a pre-trained artificial neural network is used as the predefined prediction model, which receives the resistance values measured for the current pulses as input data and, based on this, determines the assignment or the resistance-temperature curve as output data. A pre-trained artificial neural network is used here, which can, for example, already have been generated in a previously performed calibration or can have been generated based on real measurement values in a previously defined training phase.

[0039] A calibration request can be received via the (optional) receiving interface 32. The calibration request can, for example, be received from the vehicle controller. It is also possible for the calibration request to be received via a mobile data connection. The calibration request can trigger the re-implementation of the method according to the invention. Thus, based on an external or internal signal, the switch 20 can be reloaded with a current pulse and the resistance measured and the resistance-temperature curve determined. In this regard, the receiving interface 32 or the calibration request received via the receiving interface enables a way of re-calibrating during operation. During operation of the power semiconductor module 18, the resistance-temperature curve can be determined again in order, for example, to reflect aging effects and thus enable correct and appropriate temperature measurement even after a long operating time.

[0040] Figure 3 FIG. schematically shows a training device 34 for providing a prediction model according to the invention. The training device 34 includes an input interface 36 and a modeling unit 38. The input interface 36 and the modeling unit 38 can likewise be implemented partly or fully in hardware and / or software here. The training device 34 can be implemented as part of the device 22 or integrated into such a device. However, it is also possible for the training device 34 to perform its corresponding functions as a separate unit. A plurality of (in particular, real measured) resistance-temperature curves of various switches 20 are received via the input interface 36. In addition, various measured values for the resistance or on-resistance characteristics of the respective switch 20 when it is loaded with a current pulse can also be received. The data received can be used in the modeling unit 38 to determine the prediction model. In particular, an artificial neural network can be trained here. For example, the corresponding resistance-temperature curves for various switches 20 in the case of the respective current pulses and also the corresponding measured values for the development of the resistance during the loading with the current pulse are thus determined. These training data are then used to generate a general assignment rule.

[0041] Figure 4 FIG. schematically (qualitatively) shows five resistance-temperature curves for different switches 20. FromFigure 4 It can be seen that various individual components will show significant dispersion here, which makes it necessary to characterize the components (e.g., during calibration) for better accuracy in temperature measurement.

[0042] According to the present invention, for example, after the production of the power semiconductor module 18, the switch 20 is operated with a current pulse for a specific time directly. By applying a current pulse to the power semiconductor 24, the power semiconductor is heated and its resistance changes. This can be measured. Thus, a measured value of the change of the resistance characteristic over time (resistance-time curve) is created, and this resistance-time curve indirectly contains the characterization of the resistance-temperature curve.

[0043] In this regard, in Figure 5 an exemplary current pulse is schematically shown. Time is plotted on the x-axis and current intensity is plotted on the y-axis. The shown current pulse includes a constant current intensity with respect to time.

[0044] Figure 6 An example of the resistance-time curve is shown in. The shown example shows a semi-logarithmic x-axis for time and a dimensionless y-axis for resistance. Different components show different resistance-temperature curves. For example, a current pulse of 180 A for 10 s can be used. It has also been proven advantageous to apply a current pulse of 200 A with a duration of 1 s. For example, the recording of such a resistance-time curve can correspond to the resistance measurement of the measuring unit.

[0045] Here, according to the present invention, it is possible to detect the measured value for the heating curve (when applying a current pulse) and / or detect the measured value for the cooling curve (when using an additional small sensing current). The values of the heating curve and the cooling curve are characteristic of the thermal path and can be used, for example, to detect faults in the thermal path or to improve accuracy.

[0046] For example, based on various measurements, the resistance-temperature curve can be determined with high accuracy by means of deep learning methods. Various data sets are required for training the deep learning method here. These data can be recorded by using corresponding temperature sensor devices. In the case of a sufficient number of data sets, even when the switch does not reach the desired temperature during the application of the current pulse, the resistance-temperature curve can be predicted by extrapolation with such a predefined prediction model. It should be understood that the training data required here reflects the numerical range to be extrapolated. Compared with the hitherto methods, the method according to the present invention focuses on cost savings because calibration is not required in the case of high-temperature measurement. The internal voltage measurement of the power semiconductor module can be used, so that no additional hardware is required.

[0047] The method according to the invention can also be used, for example, for rapidly measuring temperature-dependent parameters in manufacturing. In addition, the method according to the invention can also be used for rapidly measuring and checking thermal paths in manufacturing. It is possible to perform recalibration during operation. For example, the pulse mode can be called when the vehicle is stationary, whereby the temperature measurement in the power semiconductor module can be recalibrated. For example, this may be meaningful due to the drift of electrical or thermal parameters. In addition, in principle, it is possible to use the cooling curve in the stationary state in order to detect the aging of the thermal path at an early stage. Therefore, it is possible to provide for regular monitoring of the resistance or the change in resistance during switch cooling in order to detect changes at an early stage. In principle, by means of a corresponding recalibration, it is also possible to improve the accuracy of the temperature measurement.

[0048] What can also be achieved by the method according to the invention is to check a capacitor (intermediate circuit capacitor, chip capacitor) or an inductor (motor winding) via a current pulse or a voltage pulse or a triangular signal.

[0049] Figure 7 FIG. schematically shows a method according to the invention for determining the temperature dependence of the on-resistance of a switch 20. The method comprises the step of applying a current pulse S10 to the switch 20. The method further comprises the step of measuring the resistance of the switch 20 S12 during a measurement period. In addition, the method further comprises the step of obtaining the resistance-temperature curve of the switch 20 S14. For example, the method can be implemented in software implemented on a processor of the power semiconductor module 18. It should be understood that it is also conceivable to implement the method in other devices or other units.

[0050] The invention has been fully described and explained with reference to the drawings and the description. The description and the explanation should be understood as examples and without limitation. The invention is not limited to the disclosed embodiments. Other embodiments or variants will be obtained by those skilled in the art in the context of using the invention and in the context of a careful analysis of the drawings, the disclosure and the following claims.

[0051] In the claims, the words "comprising" and "having" do not exclude the presence of further elements or steps. The indefinite article "a" or "an" does not exclude the presence of a plurality. A single element or a single unit can perform the functions of a plurality of units mentioned in the patent claims. Elements, units, interfaces, devices and systems can be implemented partly or wholly in hardware and / or software. The fact that some measures are only mentioned in several different dependent patent claims should not be understood as meaning that a combination of these measures cannot be used advantageously. The reference signs in the patent claims should not be construed as limiting.

[0052] List of reference signs

[0053] 10 Vehicle

[0054] 12 Traction Converter

[0055] 14 Battery

[0056] 16 Electric Machine

[0057] 18 Power Semiconductor Module

[0058] 20 Switch

[0059] 22 Device

[0060] 24 Power Semiconductor

[0061] 26 Pulse Unit

[0062] 28 Measuring Unit

[0063] 30 Evaluation Unit

[0064] 32 Receiving Interface

[0065] 34 Training Device

[0066] 36 Input Interface

[0067] 38 Modeling Unit

Claims

1. A device (22) for determining the temperature dependency of the on-resistance of a switch (20) in a power semiconductor module (18), the device comprising: A pulse unit (26), the pulse unit being used to apply a current pulse to the switch; a measuring unit (28) for measuring the resistance of the switch during and / or after the current pulse is applied to the switch within a measuring time period; as well as An evaluation unit (30) is used to obtain a resistance-temperature curve of the switch based on the measured resistance and a predefined prediction model.

2. The device (22) according to claim 1, wherein: The pulse unit (26) is designed to apply a current pulse of a duration between 0.5 s and 20 s, preferably between 1 s and 10 s, particularly preferably between 1 s and 2 s, and a current intensity between 50 A and 500 A, preferably between 100 A and 250 A, particularly preferably between 150 A and 200 A, to the switch (20).

3. The device (22) according to any one of the preceding claims, wherein: The measuring unit (28) is configured to measuring resistance based on a voltage measurement made during the time the switch is loaded with the current pulse; and / or The resistance is measured based on a voltage measurement of a measurement current after the switch is loaded with the current pulse.

4. The device (22) according to any one of the preceding claims, wherein: The evaluation unit (30) is configured to Determining the resistance-temperature curve based on machine learning methods, in particular pre-trained models; and / or The resistance-temperature curve is obtained based on a pre-trained artificial neural network.

5. The device (22) according to any one of the preceding claims, wherein: The pulse unit (26) is designed to apply current pulses to a switch (20) having a plurality of power semiconductors (24), in particular a plurality of power semiconductors connected in parallel.

6. The device (22) according to any of the preceding claims, comprising a receiving interface (32) for receiving a calibration request, wherein: The pulse unit (26) is configured to apply the current pulse to the switch (20) after receiving the calibration request; The measuring unit (28) is configured to measure the resistance after receiving the calibration request; as well as The evaluation unit (30) is designed to determine the resistance-temperature curve after receiving the calibration request.

7. The device (22) according to claim 6, wherein The receiving interface (32) is configured to receiving the calibration request from a vehicle controller; receiving the calibration request via a mobile wireless connection; and / or The calibration request is received after a predefined operating condition occurs.

8. A traction converter (12) for a vehicle (10), comprising a switch (20) in a power semiconductor module (18) and a device (22) according to any one of the preceding claims, wherein: The measuring unit (28) is configured to measure the resistance of the switch; as well as The evaluation unit (30) is designed to determine the temperature of the switch based on the determined resistance-temperature curve.

9. A training device (34) for providing a prediction model, the training device comprising: an input interface (36) for receiving a plurality of resistance-temperature curves of various switches (20); and A modeling unit (38), the modeling unit is used to obtain a prediction model based on the received resistance-temperature curve, wherein: The modeling unit is designed in particular to train an artificial neural network.

10. A method for determining the temperature dependency of an on-resistance of a switch (20), the method comprising the following steps: Applying (S10) a current pulse to the switch; measuring (S12) the resistance of the switch during a measuring time period during and / or after applying the current pulse to the power semiconductor (24); and A resistance-temperature curve of the switch is learned ( S14 ) based on the measured resistance and a predefined prediction model.

11. A computer program product comprising instructions which, when the computer program is executed by a computer, cause the computer program to perform the steps of the method according to claim 10.

Images