Patent Application
20230014410
The present invention relates to an inverter and to an estimation of an internal temperature of a semiconductor switch of a power module of an inverter. It is especially intended be used in an automotive vehicle.
It is difficult to directly measure the internal temperature of the semiconductor switch (e.g. its junction temperature) once the inverter is installed in an automotive vehicle. That is why the internal temperature is instead estimated from other measures.
Known techniques to estimate the temperature include using a thermal model of the inverter. The thermal model is based for example on the physical modeling of the inverter losses and the measured thermal impedance of the inverter. The thermal impedance is given to the thermal model as a function of the coolant flow rate. The tolerances of the thermal model function inputs cause big deviation and high inaccuracy, especially when the coolant inlet temperature is being estimated and not measured. This problem causes the system to behave unexpectedly (e.g. an early derating action for the system because of higher estimated power module temperature than the actual temperature or delayed derating, which may cause thermal stress damage).
An object of the invention is to improve the accuracy of the temperature estimation.
The object of the invention may be solved by an inverter comprising:
wherein the temperature model is a polynomial of order three or more having, as arguments, operating parameters including:
Surprisingly, it has been found that a polynomial temperature model may be more accurate than a dedicated thermal model. Advantageously, a polynomial temperature model provides higher accuracy as the terms of the polynomial increase.
Some further optional features of the invention which can be used together or separately are developed below.
The control device may use the estimated temperature for carrying out a derating action, such as a torque derating action or switching frequency derating action.
The control device may use the estimated temperature for estimating losses of the inverter.
The operating parameters may include at least one of, preferably all of:
The operating parameters may include at least one of:
a rotor torque of the electric motor,
The estimated internal temperature may be a junction temperature of the at least one semiconductor switch. Estimating the junction temperature is important because it is often the most affecting parameter for the lifetime of the power module and the inverter. The power module operation is limited to a certain junction temperature value (Tjunction_max). If the junction temperature exceeded this value, a derating action may need to be activated to protect the inverter.
The power module may comprise at least one switch leg comprising two semiconductor switches each having first and second terminals, the two second terminals being connected to each other at a middle point and the DC voltage being intended to be applied between the two first terminals.
The at least one semiconductor switch may be an IGBT or a MOSFET.
The control device may be configured to estimate the internal temperature of each of several semiconductor switches by means of respective temperature models.
The polynomial temperature model may be of degree 3.
Advantageously, it has been found that a polynomial temperature model of degree 3 offers a good compromise with respect to realizing dynamic behavior and polynomial complexity.
The AC voltage may be a multiphase voltage, such as a three phase voltage.
The inverter may further comprise at least one sensor for respectively measuring at least one of the operating parameters, for example a temperature sensor for measuring the temperature of the power module, an AC current sensor for measuring the AC current outputted by the power module, or a DC voltage sensor for measuring the DC voltage.
The temperature model may be trained in advance by machine learning.
The training may comprise:
There may be at least ten times more combinations than operating parameters.
The invention also relates to an electric drive comprising an inverter according to the invention, an electric motor driven by the inverter.
The invention also relates to a vehicle comprising drive wheels and an electric drive according to the invention for driving, at least indirectly, at least one of the wheels.
The invention also relates to a method for estimating an internal temperature of at least one semiconductor switch of a power module of an inverter comprising a power module having at least one semiconductor switch, and a control device configured to control the power module such that, during normal operation of the inverter, the power module converts a DC voltage into an AC voltage by a switching operation of the at least one semiconductor switch, the method comprising:
The present invention will be described more specifically with reference to the following drawings, in which:
Referring to
The vehicle 100 comprises drive wheels 102 for causing the vehicle to move, and an electric drive 104 configured to drive at least one of the drive wheels 102 at least indirectly. The vehicle 100 further comprises a DC voltage source 106, such as a battery, for electrically powering the electric drive 104. The DC voltage source 106 is configured for providing a DC voltage Udc.
The electric drive 104 comprises an electric motor 108 and an inverter 110 configured to drive the motor 108, for instance by supplying electric power from the DC voltage source 106. The motor 108 is a rotary electric motor comprising a stator 112 and a rotor 114 configured to rotate around a rotation axis with respect to the stator 112, at a rotational speed Sp, and to provide a torque Tq with respect to the stator 112. In the described example, the electric motor 108 is a three-phase electric motor comprising three stator phases.
For example, the inverter 110 comprises switch legs 1161-3 respectively associated to the stator phases of the electric motor 108. Each switch leg 1161-3 comprises a high side (HS) switch Q′ having a first terminal connected to a positive terminal of the DC voltage source 106 and a low side (LS) switch Q having a first terminal connected to a negative terminal of the DC voltage source 106. In this manner, the DC voltage Udc is applied between the two first terminals of the HS and LS switches Q, Q′. The HS switch Q′ and the LS switch Q have respective second terminals connected to each other at a middle point connected to a respective associated stator phase of the electric motor 108.
The switches Q, Q′ are semiconductor switches comprising for example transistors. Each switch Q, Q′ comprises for example one amongst: a Metal Oxide Semiconductor Field Effect Transistor (MOSFET), an Insulated Gate Bipolar Transistor (IGBT) and a Silicon Carbide MOSFET (SiC MOSFET).
Each switch leg 1161-3 is intended to be controlled to commute between two configurations. In the first one, called high side (HS) configuration, the HS switch Q′ is closed (on) and the LS switch Q is open (off) so that the DC voltage Udc is essentially applied to the associated stator phase. In the second one, called low side (LS) configuration, the HS switch Q′ is open (off) and the LS switch Q is closed (on) so that a zero voltage is essentially applied to the associated stator phase.
The switch legs 1161-3 are included in one or several power modules. Each power module therefore comprises one or several switch legs 1161-3. For example, all the switch legs 1161-3 may be included in a single power module, such as a three phase power module. In the described example, each switch leg 1161-3 is included in a respective power module, so that each power module comprise exactly one switch leg 1161-3. For this reason, in the rest of the description, the references “1161-3” will be used interchangeably for the switch legs and the power modules.
Referring to
Back to
The inverter 110 further comprises a control device 120 configured to control the switches Q, Q′ such that the switches Q, Q′ convert the DC voltage Udc into an AC voltage, which may be a multiphase AC voltage, for example, a three phase voltage. In the described example, the voltage phases of the AC voltage are respectively provided to the stator phases of the electric motor 108. In response to the AC voltage, the electric motor 108 causes the power modules 1161-3 to output a total output AC current Ito the electrical motor 108. In the described example, the total output AC current I is a multiphase AC current (for instant three phase AC current) having phase currents being respectively provided to the stator phases of the electric motor 108.
In the described example, the control device 120 is configured to commute each switch leg 1161-3 between the two configurations mentioned above, at a switching frequency FSW. The AC voltage provided to the stator phase associated with the switch leg 1161-3 is then obtained by varying a duty cycle between the two configurations, according to a Pulse Width Modulation (PWM) scheme.
For example, the Sinusoidal PWM (SPWM) scheme may be used for each switch leg 1161-3. In the SPWM scheme, a sine wave (modulated wave) is compared with a triangle wave (carrier wave). When the instantaneous value of the triangle wave is less than that of the sine wave, the switch leg 1161-3 is switched to a first of the two configurations. Otherwise the switch leg 1161-3 is switched to the second of the two configurations. The switching is produced at every moment the sine wave intersects the triangle wave. Thus the different crossing positions change the duty cycle. To describe the modulation state, a modulation index M is defined by the ratio of the amplitude of the modulated wave to that of the carrier wave.
The control device 120 is further configured for estimating an internal temperature TJ of at least one of the semiconductor switches Q, Q′, by means of a temperature model 122 of the at least one semiconductor switches Q, Q′. The estimated temperature TJ is preferably a junction temperature of the at least one semiconductor switches Q, Q′. When the internal temperature of several semiconductor switches Q, Q′ is estimated, the control device 120 uses preferably a respective temperature model 122 for each semiconductor switch Q, Q′.
The temperature model 122 maybe stored either in the control device 120 or outside of the control device. The temperature model 122 is a polynomial of order three or more having, as arguments, operating parameters of the power module 1161-3 and/or of the electric motor 108.
For example, the control device 120 comprises a computer system comprising a data processing unit (such as a microprocessor) and a main memory (such as a RAM memory, standing for “Random Access Memory”) accessible by the processing unit. The computer system further comprises for example a network interface and/or a computer readable medium, such as for example a local medium (such as a local hard disk). A computer program containing instructions for the processing unit is stored on the medium and/or downloadable via the network interface. This computer program is for example intended to be loaded into the main memory, so that the processing unit executes its instructions so as to carry out the temperature estimation method of
For obtaining the operating parameters to be fed to the temperature model 122, the electric drive 104 further comprises a measure system 124 including sensors 124A-J of at least some of the operating parameters and/or estimators of at least some of the operating parameters from measures made by sensors.
The temperature model 122 is preferably trained in advance by machine learning to provide the estimated temperature TJ from the measured or estimated operating parameters. In particular, the training comprises determining coefficients of each term (also called monomial) of the polynomial.
It has been found that the polynomial could be limited to degree 3, while still achieving good results. A polynomial of degree 3 comprises at least one term with one of the operating parameters cubed, and no term with a higher powered operating parameter.
The operating parameters include preferably at least one amongst: the switching frequency FSW of the inverter 110, the rotational speed Sp of the electric motor 108, a temperature TS of the power modules 1161-3, an current outputted by the power modules 1161-3, the flow rate FR of the coolant, the temperature TC of the coolant, the DC voltage Udc of the inverter 110, the torque Tq of the rotor 114, the power factor PF of the electric motor 108, and the modulation index M of the inverter 110.
For example, the temperature TS of the power module 1161-3 may be an ambient temperature of the switches Q, Q′ or a temperature of the substrate 406.
Also for example, the output current outputted by the power modules 1161-3 may be the total output AC current I (as in the described example) or one or several phase current. The output current outputted by the power modules 1161-3 is for example expressed as Root-Mean-Square (“RMS”).
In the described example, the measure system 124 then comprises a temperature sensor 124C for measuring the temperature TS of the power modules 1161-3, an AC current sensor 124 for measuring the total output AC current I, and a DC voltage sensor 124G for measuring the DC voltage Udc.
It has been found that, in general, the most important operating parameters are: the switching frequency FSW, the output current, the coolant entry temperature TC, the coolant flow rate FR, the DC voltage Udc and the rotational speed Sp of the electric motor 108. The remaining operating parameters mentioned above have in general a small impact on the temperature estimation, but may be used for extra precision.
Referring to
At a step 204, the temperature model 122 is trained using the training data. The training comprises for example a polynomial regression. In this case, a coefficient is obtained for each term of the polynomial. Preferably, the initial temperature model 122 (before training) comprises terms for all the operating parameters and for all power (for example until the third power, for a degree 3 polynomial). The terms with very low (near zero) coefficients may be removed.
The terms of the initial temperature model are for example randomly initialized.
The machine learning method 200 has been applied by the invertors to a specific electric drive, using 389 combinations of the 10 operating parameters listed above. In these 389 combinations: the rotational Sp takes 12 different values, the torque takes 49 different values, the DC voltage Udc takes 13 different values, the flowrate FR takes 5 different values, the coolant temperature takes 21 different values, and the switching frequency FSW takes 30 different values.
The result was the following temperature model:
T
J=θ1+θ2·Tq+θ3·Tq2+θ4·TS+θ5·TS2+θ6·TS3+θ7·I+θ8·I2+θ9·I3+θ10·Udc+θ11·Udc2+θ12·M+θ13·PF+θ14·Sp+θ15·Sp2+θ16·Sp3+θ1 7·TC+θ18·TC3+θ19·FR+θ20·FR2+θ21·FR3+θ22·FSW+θ23·FSW2+θ24·FS W3
where θk (k=1 . . . 24) are fixed coefficients. The modulus (i.e. absolute value) of the coefficients represents the weight of the associated operating parameter in the estimation, and is indicated in the following table:
The operating parameters of the previous temperature model are expressed in a scaled version X′ according to the following formula:
X′=(X−μi)/sigma2
where X is the operating parameter in its measured/estimated version, and μi and sigma2 are constants associated with this operating parameter so that all operating parameters are in the same scale. μi and sigma2 are for instance determined during training, so that μi is the mean of the operating parameter values in the training data, and sigma 2 is the standard deviation of the operating parameter values in the training data. In the described example, μi and sigma2 are:
Referring to
At a step 304, the control device 120 receives the measured or estimated operating parameters.
At a step 305, the received operating parameters are scaled according to the constants μi and sigma2.
At a step 306, the control device 120 applies the scaled operating parameters to the temperature model 122 to provide an estimation of the temperature TJ.
The temperature model 122 detailed above has been evaluated by using a quantified value for the difference between measurements and forecast (error). In particular, the mean absolute error (MAE) has been used, defined by:
MAE=(1/m)*Σ|Measurement−Forecast|
The MAE has been found to be 1.8° C., with the maximum absolute difference being 8° C., which shows the accuracy of the temperature model 122.
It will be noted that the invention is not limited to the embodiments described above. It will indeed appear to those skilled in the art that various modifications can be made to the embodiments described above, in the light of the teaching which has just been disclosed.
In the previous detailed description of the invention, the terms used should not be interpreted as limiting the invention to the embodiments presented in the present description, but should be interpreted to include all the equivalents within the reach of those skilled in the art by applying their general knowledge to the implementation of the teaching which has just been disclosed.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10 2021 207 232.1 | Jul 2021 | DE | national |
