Method for online estimation of the lifetime of a semiconductor switch

Abstract

A method is proposed for online estimation of a lifetime state of a semiconductor switch by determining an error between a measured resistance value and an estimated resistance value by means of an observer, wherein the measured resistance value is determined from a determined current and a determined voltage at the semiconductor switch, and the estimated resistance value is determined from a power loss of the semiconductor switch, wherein the error is determined as the difference between the measured resistance value and the estimated resistance value.

Description

[0001] The present invention relates to the field of power electronics, more specifically to the lifetime monitoring of semiconductor switches.

[0002] Semiconductor switches are used in various power converters such as inverters (DC / AC converters, AC / DC converters) and rectifiers. These are used in a variety of industries, including photovoltaics, wind power generation, the automotive sector, and many other applications.

[0003] In particular, the use of electronic modules, such as power electronic modules, in motor vehicles has increased significantly in recent decades. This is due, on the one hand, to the need to improve fuel economy and vehicle performance, and, on the other hand, to advances in semiconductor technology. The main component of such an electronic module is a DC / AC inverter, which is used to supply electrical machines such as electric motors or generators with a multi-phase alternating current (AC). In this process, a direct current generated by a DC energy source, such as a battery, is converted into a multi-phase alternating current. For this purpose, the inverters comprise a variety of electronic components with which bridge circuits (such as half-bridges) are implemented, for example semiconductor power switches, which are also known as power semiconductors.

[0004] Since all components, including semiconductor switches, age and fail with increasing operating life, it is important to monitor their service life (aging state). This is particularly true for semiconductor switches, where monitoring is currently carried out by modeling the expected service life of the power semiconductors of semiconductor switches using so-called power cycling tests, or by estimating an aging process or calculating it using models. In addition, the semiconductor area is often over-dimensioned to ensure a required service life. However, this requires installation space and is expensive. Since the semiconductor switch must provide as much of the same power as possible over its specified service life, e.g. >8,000 operating hours, measures must be taken to detect premature aging, if possible during operation of the semiconductor switch.However, it is currently not possible to perform an online estimate of the lifetime of a semiconductor switch, i.e. an estimate during operation of the semiconductor switch.

[0005] The invention is therefore based on the object of providing a method for online estimation of the lifetime state of a semiconductor switch.

[0006] This object is achieved by the features of the independent claims. Advantageous embodiments are the subject of the dependent claims.

[0007] A method is proposed for online estimation of a lifetime state of a semiconductor switch by determining an error between a measured resistance value and an estimated resistance value by means of an observer, wherein the measured resistance value is determined from a determined current and a determined voltage at the semiconductor switch, and the estimated resistance value is determined from a power loss of the semiconductor switch, wherein the error is determined as the difference between the measured resistance value and the estimated resistance value.

[0008] In one embodiment, the power loss of the estimated resistance value is determined from the sum of a power loss of conduction losses of the semiconductor switch and switching losses of the semiconductor switch determined from predetermined input variables.

[0009] In one embodiment, the estimated resistance value is determined based on the power loss and a coolant temperature.

[0010] In one embodiment, the estimated resistance value is determined from a look-up table based on the power dissipation and the coolant temperature, or from a temperature swing determined from the coolant temperature.

[0011] In one embodiment, an adjustment factor is provided which supplies the error in determining the estimated resistance value as a further input variable for recalculating the estimated resistance value.

[0012] In one embodiment, the semiconductor switch is a field-effect transistor, and the measured resistance is determined from the source-drain current and the source-drain voltage of the semiconductor switch. In one embodiment, the semiconductor switch is an IGBT, and the measured resistance is determined from the collector-emitter current and the collector-emitter voltage of the semiconductor switch.

[0013] Furthermore, a computer program is provided that is configured to execute individual or all steps of the method using program code on a computing unit. Furthermore, a computing unit is provided on which the computer program is implemented.

[0014] Furthermore, an electronic module for operating an electric drive of a vehicle is provided, comprising a plurality of semiconductor switches and control electronics, wherein the semiconductor switches can be switched via the control electronics in such a way that they pass or interrupt a load current, wherein the electronic module comprises a computing unit on which the computer program is implemented.

[0015] Furthermore, an electric drive of a vehicle is provided, comprising the electronic module.

[0016] Further features and advantages of the invention will become apparent from the following description of exemplary embodiments of the invention, with reference to the figures of the drawing, which illustrate details of the invention, and from the claims. The individual features can be implemented individually or in combination in a variant of the invention.

[0017] Preferred embodiments of the invention are explained in more detail below with reference to the accompanying drawings. Fig. 1 shows a basic structure of an observer model from the method according to an embodiment of the present invention. Fig. 2 shows a basic flow of the error calculation from the method according to an embodiment of the present invention.

[0018] In the following descriptions of the figures, the same elements or functions are provided with the same reference symbols.

[0019] It is known that during operation of components such as a semiconductor switch, the internal resistance changes over the lifetime of the semiconductor switch. The internal resistance of the semiconductor switch deteriorates (increases) over its lifetime. Towards the end of the specified lifetime of a semiconductor switch, deterioration may also occur more rapidly, so that the control electronics should no longer demand the semiconductor switch's full power to prevent the lifetime from being shortened even more rapidly. This is also referred to as derating.

[0020] As already mentioned, it is currently not possible to perform an online estimate of the lifetime of a semiconductor switch, i.e., an estimate during operation of the semiconductor switch. However, this would be desirable in order to be able to assess excessively rapid aging and thus the risk of premature failure of the semiconductor switch.

[0021] Therefore, a method is described below that enables an online estimation of the service life of a semiconductor switch 1. The method is advantageously implemented as a computer program on a computing unit for controlling the semiconductor switch(es), wherein the computing unit may also include the control electronics. The computer program can execute individual or all steps of the method using program code.

[0022] The inventive online estimation of the lifetime state of a semiconductor switch 1 is carried out by determining an error e between a measured resistance value R_Mess and an estimated (predicted) resistance value R_Predict of the semiconductor switch 1 by means of an observer B, as in Fig. 1. An observer B is a control system that can (re)construct, essentially estimate, non-measurable quantities from known input and output variables. The observer B comprises an algorithm A that determines an estimated resistance value R_Predict from given input variables, or more precisely, a resistance value predicted based on the input variables.

[0023] At least a power loss P_In of the semiconductor switch 1 is required as input variables for the algorithm for determining the estimated resistance value R_Predict. This is obtained from a power loss of conduction losses of the semiconductor switch 1 and switching losses of the semiconductor switch 1 determined from specified input variables by summing them.

[0024] The conduction power dissipation of semiconductor switch 1 is obtained from a measured drain-source voltage (in the case of an FET) or collector-emitter voltage (in the case of an IGBT) and the known phase current (usually referred to as I_rms). When the method is applied to other semiconductor switches 1, e.g., freewheeling diodes or thyristors, the voltage between the component's terminal pins is measured.

[0025] The switching losses of semiconductor switch 1 are determined from the specified input variables: current (known) intermediate circuit voltage, known phase current, switching frequency, and current (measured) temperature of semiconductor switch 1, preferably from a (pre-created) look-up table. A calculation would also be possible, but would generally exceed the computing power and available time.

[0026] In addition to the power loss P_In of the semiconductor switch 1 as the input variable of the algorithm, a current (measured) coolant temperature T_C of the coolant cooling the semiconductor switch 1 can serve as a further input variable to establish a suitable system of equations. Thus, a thermal model can be applied to determine the estimated resistance value R_Predict. The estimated resistance value R_Predict can be read directly from a (pre-created) look-up table based on the power loss P_In and the coolant temperature T_C. Alternatively, a temperature swing can be determined in a known manner from the coolant temperature T_C, or more precisely from the thermal impedance of the cooling path (from the semiconductor switch 1 to the cooling medium), in which, for example, RC elements (Foster networks) are provided.Based on this temperature swing and the power dissipation P_In, the estimated resistance value R_Predict can again preferably be read from a (pre-created) look-up table.

[0027] Alternatively, algorithm A can also include a neural network that directly determines the estimated resistance value R_Predict based on the power loss P_In as input variable and, if applicable, other previously described input variables, as well as optionally the operating time already achieved. Learning is performed, for example, using resistance measurement data from power cycling tests or other data from endurance tests. The neural network can then provide the analytical description of the thermal impedance and create lookup tables, as well as derive the model description. However, this implementation requires very high computing power.

[0028] The measured resistance value R_Mess is determined from a measured voltage and a measured current at semiconductor switch 1. Corresponding measuring devices are already available in current applications, so they will not be discussed further here. To determine the measured resistance value R_Mess, a function (mathematical relationship) between the measured voltage and the measured current is determined in a known manner. The resistance value R_Mess corresponds to the first derivative of the function.

[0029] The error e is defined as the difference between the resistance value R_Predict determined by the observer B and the measured resistance value R_Mess, as in Fig. 2, i.e.: e=R_Measure−R_Predict

[0030] The error e between the resistance value R_Predict determined by observer B and the measured resistance value R_Mess increases over the operating life of semiconductor switch 1, since the model is initially calibrated to the resistance value and will always calculate as initially specified without adjustment. However, due to aging-related deterioration, the actually measured resistance R_Mess becomes increasingly larger and thus also the systematic deviation between R_Mess and R_Predict. Therefore, an adjustment factor K can be provided, which feeds the error e in determining the estimated resistance value R_Predict to algorithm A as a further input variable for the next calculation of R_Predict, as shown in Fig. 2. This prevents the error e from drifting away from the measured resistance value R_Mess.

[0031] The larger the error e calculated by the adjustment factor K, the faster the semiconductor switch 1 ages. Thus, the method for online estimation of the service life of the semiconductor switch 1 can be used to determine that the semiconductor switch 1 is aging faster than expected. Based on this finding, measures can be taken to delay the aging and thus the end of the service life of the semiconductor switch 1, e.g., by no longer demanding full power from the semiconductor switch 1.

[0032] The process can be used in all applications that involve semiconductor switches, particularly in power converters (DC / AC converters, inverters, or AC / DC converters), as well as in rectifiers such as DC / DC or AC / AC converters. Its use is not limited to the automotive sector. Applications in other industries, such as the solar industry (e.g., inverters for photovoltaic systems), can also benefit from the process. List of reference symbols 1 semiconductor switch B Observer K adjustment factor R_Mess measured resistance value R_Predict estimated resistance value e errors P_In power loss at 1 T_C coolant temperature A algorithm

Claims

[1] Method for online estimation of a lifetime state of a semiconductor switch (1) by determining an error (e) between a measured resistance value (R_Mess) and an estimated resistance value (R_Predict) by means of an observer (B), wherein - the measured resistance value (R_Mess) is determined from a determined current and a determined voltage at the semiconductor switch (1), and - the estimated resistance value (R_Predict) is determined from a power loss (P_In) of the semiconductor switch (1), whereby - the error (e) is determined as the difference between the measured resistance value (R_Mess) and the estimated resistance value (R_Predict). [2] Method according to claim 1, wherein the power loss (P_In) of the estimated resistance value (R_Predict) is determined from the sum of a power loss of conduction losses of the semiconductor switch (1) and switching losses of the semiconductor switch (1) determined from predetermined input variables. [3] Method according to claim 1 or 2, wherein the estimated resistance value (R_Predict) is determined based on the power loss (P_In) and a coolant temperature (T_C). [4] Method according to claim 1 or 2, wherein the estimated resistance value (R_Predict) is - a look-up table is determined based on the power loss (P_In) and the coolant temperature (T_C), or - a temperature lift determined from the coolant temperature (T_C). [5] Method according to one of the preceding claims, wherein an adjustment factor (K) is provided which supplies the error (e) of the determination of the estimated resistance value (R_Predict) as a further input variable for recalculating the estimated resistance value (R_Predict). [6] Method according to one of the preceding claims, wherein the semiconductor switch (1) - is a field-effect transistor and the measured resistance value (R_Mess) is determined from the source-drain current and the source-drain voltage of the semiconductor switch (1), or - is an IGBT and the measured resistance value (R_Mess) is determined from the collector-emitter current and the collector-emitter voltage of the semiconductor switch (1). [7] Computer program which is designed to carry out individual or all steps of the method according to one of the preceding claims by means of program code on a computing unit. [8] Computing unit on which the computer program according to claim 7 is implemented. [9] Electronic module for operating an electric drive of a vehicle, comprising a plurality of semiconductor switches (1) and control electronics, wherein the semiconductor switches (1) can be switched via the control electronics in such a way that they pass or interrupt a load current, wherein the electronic module comprises a computing unit on which the computer program according to claim 7 is implemented. [10] Electric drive of a vehicle, comprising the electronic module according to claim 9.

Images