Method for estimating power system health

Abstract

A method of monitoring the health of a semiconductor power electronic switch such as an insulated gate bipolar transistor (IGBT) is provided. The method having the steps of: measuring one or more parameters selected from the group consisting of: a rate of change of voltage $\left(\frac{\mathrm{dV}}{\mathrm{dt}}\right)$
across the switch; a rate of change of current $\left(\frac{\mathrm{di}}{\mathrm{dt}}\right)$
through the switch, a charge present on a gate of the switch (Q.sub.G), a peak overshoot voltage (V.sub.PO) across the switch, and a peak overshoot or reverse recovery current (I.sub.RR) through the switch; and estimating the health of the switch based on the measured parameter(s).

Claims

1. A method of monitoring the health of a semiconductor power electronic switch, the method comprising steps of: estimating a junction temperature of the switch; counting a number of temperature cycles based on the junction temperature; when the number of temperature cycles reaches a predetermined number of cycles, measuring one or more parameters selected from the group consisting of: a rate of change of voltage across the switch; a rate of change of current through the switch; a charge present on a gate of the switch; a peak overshoot voltage across the switch; and a peak overshoot or reverse recovery current through the switch; and estimating the health of the switch based on the measured one or more parameters, wherein estimating the health of the switch includes identifying changes in the measured one or more parameters as a function of temperature cycles, and an amount of change exceeding a threshold at higher temperature cycles indicates that the switch is unhealthy.

2. The method of claim 1, wherein estimating the health of the switch includes comparing the measured one or more parameters with a respective pre-defined threshold value.

3. The method of claim 1, wherein the step of measuring the one or more parameters and the step of estimating the health of the switch are repeatedly performed at a sampling frequency, and the sampling frequency is a function of the estimated health of the switch.

4. The method of claim 1, wherein each of the one or more parameters is measured by a respective embedded measuring circuit of a gate driver of the switch.

5. The method of claim 1, wherein estimating the health of the switch includes plotting at least one of the measured one or more parameters with respect to the estimated junction temperature.

6. The method of claim 5, wherein at least two of the parameters of the one or more parameters are measured in the step of measuring the one or more parameters, one of the at least two measured parameters is used to estimate the junction temperature; and the other of the at least two measured parameters is plotted with respect to temperature cycles and results of plotting the parameters are used to estimate the health of the switch.

7. A power conversion system including: at least one semiconductor power electronic switch, the switch having a gate driver; and a processor connected to the switch, wherein the processor is configured to: estimate a junction temperature of the switch; and count a number of temperature cycles based on the junction temperature, and wherein the gate driver includes one or more parameter measuring modules, and when the number of temperature cycles reaches a predetermined number of cycles, the one or more parameter measuring modules measure one or more respective parameters selected from the group consisting of: a rate of change of voltage across the switch; a rate of change of current through the switch; a charge present on a gate of the switch; a peak overshoot voltage across the switch; and a peak overshoot or reverse recovery current through the switch, wherein the processor is further configured to estimate the health of the switch based on the measured one or more parameters, estimating the health of the switch includes identifying changes in the measured one or more parameters as a function of temperature cycles, and an amount of change exceeding a threshold at higher temperature cycles indicates that the switch is unhealthy.

8. The power conversion system of claim 7, wherein the processor is embedded into the gate driver.

9. The power conversion system of claim 7, wherein the processor is configured to estimate the health of the switch by comparing each of the measured one or more parameters with a respective pre-defined threshold value.

10. The power conversion system of claim 7, wherein the one or more parameter measuring modules repeatedly measure the one or more parameters at a sampling frequency which is a function of the estimated health of the switch, and the processor is further configured to estimate the health of the switch at the sampling frequency.

11. The power conversion system of claim 7, wherein the one or more parameter measuring modules are respective embedded measuring circuits of the gate driver.

12. A propulsion system comprising: an electric drive, the electric drive being electrically connected to the power conversion system of claim 7.

13. The power conversion system of claim 7, wherein estimating the health of the switch includes plotting at least one of the measured parameters with respect to the estimated junction temperature.

14. The power conversion system of claim 13, wherein the parameter measuring modules measure at least two parameters of the one or more parameters, and wherein the processor is further configured to: estimate the junction temperature using one of the at least two measured parameters; and estimate the health of the switch using results of plotting the other of the at least two measured parameters with respect to the estimated junction temperature.

15. A gate driver for a semiconductor electronic power switch, the gate driver being suitable to be installed in a power conversion system, the gate driver comprising: an embedded processor configured to: estimate a junction temperature of the switch; count a number of temperature cycles based on the junction temperature; when the number of temperature cycles reaches a predetermined number of cycles, execute one or more parameter measuring modules for measuring one or more respective parameters selected from the group consisting of: a rate of change of voltage across the switch; a rate of change of current through the switch; a charge present on a gate of the switch; a peak overshoot voltage across the switch; and a peak overshoot or reverse recovery current through the switch; and estimate the health of the switch based upon the measured one or more parameters, wherein estimating the health of the switch includes identifying changes in the measured one or more parameters as a function of temperature cycles, and an amount of change exceeding a threshold at higher temperature cycles indicates that the switch is unhealthy.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Embodiments of the invention will now be described by way of example with reference to the accompanying drawings in which:

(2) FIG. 1 shows a block diagram of a converter including a health prognosis system;

(3) FIG. 2 shows a schematic of a gate driver with embedded prognostic capability;

(4) FIGS. 3A-3D are plots showing the variation of certain parameters with respect to junction temperature for both aged and new devices;

(5) FIG. 4 shows a flow chart of the inverter prognosis system;

(6) FIG. 5 shows a control schematic of a junction temperature estimator;

(7) FIG. 6 is a plot showing the dependence of three parameters with temperature cycle;

(8) FIG. 7 is a plot showing the dependence of two other parameters with temperature cycle;

(9) FIG. 8 is a diagram showing a gate charge measurement circuit as connected to an IGBT;

(10) FIG. 9 contains two plots of gate-emitter voltage and gate-current against time;

(11) FIG. 10 is a diagram showing the process for extracting the gate charge from the measured gate-emitter voltage under constant gate current;

(12) FIG. 11 is a diagram showing a rate of change of current measure en circuit as connected to an IGBT;

(13) FIG. 12 is a plot of collector-current from an oscilloscope and collector-current from a current mirror against time;

(14) FIG. 13 is a diagram showing the process for extracting the rate of change of current from the current mirror input;

(15) FIG. 14 is a diagram showing a rate of change of voltage measurement circuit as connected to an IGBT;

(16) FIG. 15 is a plot of collector-emitter voltage from an oscilloscope and collector-emitter voltage from a miller capacitor against time;

(17) FIG. 16 is a diagram showing the process for extracting the rate of change of voltage from an external miller capacitor circuit;

(18) FIG. 17 is a plot of voltage against time for an IGBT before and after power cycle testing, illustrating the change in gate charge characteristics;

(19) FIG. 18 is a plot of voltage against time for an IGBT before and after power cycle testing, illustrating the change in rate of change of current characteristics;

(20) FIG. 19 is a plot of voltage against time for an IGBT before and after power cycle testing, illustrating the change in rate of change of voltage characteristics; and

(21) FIG. 20 is a schematic of a propulsion system.

DETAILED DESCRIPTION AND FURTHER OPTIONAL FEATURES

(22) In the description below, an IGBT (insulated gate bi-polar transistor) is considered. The skilled person will of course appreciate that the methods, circuits, and results discussed below are equally applicable (with any appropriate modification) to other semiconductor power electronic switches such as MOSFETs and SiC MOSFETs.

(23) FIG. 1 is a block diagram of an example power converter (in this case an inverter) including a health prognosis system i.e. a system for monitoring the health of the power converter. A DC source 101 is connected via a capacitor 102 to a series of insulated gate bipolar transistors (IGBTs) 103, which are connected to a load 104. The IGBTs can be referred to as power converting devices. The IGBTs are switched such that the output current to the load is an AC current.

(24) Connected to each IGBT is a gate driver 105, which includes an embedded health prognosis system. Each gate driver 105 includes a module for measuring: the gate charge Q.sub.G, rate of change of voltage across the IGBT

(25) $\frac{\mathrm{dV}}{\mathrm{dt}},$
and rate of change of current

(26) $\frac{\mathrm{di}}{\mathrm{dt}}.$
The gate driver 105 may also include an analogue to digital converter and/or digital to analogue converter, as well as a module for directly measuring the voltage across the collector and emitter of the IGBT when it is switched on: V.sub.CE.sub.ON. At least some of the outputs of the IGBTs are connected to current sensors 106, and both the current sensors and gate drivers can be connected to a digital processor 107. The output of this digital processor may be subjected to signal conditioning and data management. The health status of the IGBTs may therefore be determined.

(27) Here each power device (i.e. IGBT) can be monitored periodically, and any changes in the measured characteristics can be logged by the gate driver or by a central management computer. A gate driver with the capabilities discussed above can be referred to as a gate driver with embedded prognosis capability (GDEPC). The GDEPC can transmit the observed variation to a processing system such as the digital processor, which can allow the device health status to be determined and analysed. Alternatively the device health status can be determined by a processor embedded in the gate driver. The digital processor, either external or embedded, can intelligently manage the data acquisition from the GDEPCs and communicate with a converter system controller to enable the derivation of the system level health status and also a remaining useful lifetime (RUL) estimation for the system. The system level health indication and RUL estimation of the system can either be done in the digital processor connected to each converter or at a separate dedicated processor responsible for monitoring the health of the converter. This separate dedicated processor can derive the RUL and system health of the entire converter by considering the individual component health, RUL status, and expected usage profiles.

(28) FIG. 2 shows in more detail a gate driver with embedded prognostic capability (GDEPC). The GDEPC manages the switching time, isolation, and other basic functions needed to operate an IGBT 201 as well as measuring the electrical parameters needed to detect the ageing (i.e. deterioration in health) of the power module, The term ageing can be considered to define a degradation process that leaves the device operating in an un-safe mode i.e. it is unhealthy. The GDEPC primarily detects the ageing of power modules due to temperature and power cycling, and does so by measuring one or more of the following parameters: the gate charge Q.sub.G, rate of change of voltage across the IGBT

(29) $\frac{\mathrm{dV}}{\mathrm{dt}},$
and rate of change of current

(30) $\frac{\mathrm{di}}{\mathrm{dt}}.$

(31) It may do so by provision of: a Miller capacitor 203 connected to a

(32) $\frac{\mathrm{dV}}{\mathrm{dt}}$
module 207, a gate resistor 202 connected to a gate driver with gate charge Q.sub.G measurement module 206, and a

(33) $\frac{\mathrm{di}}{\mathrm{dt}}$
measurement module 205 connected to a Kelvin emitter 204. The Kelvin emitter 204 is connected in series with the emitter terminal of the IGBT 201.

(34) Generally the rate of change of current, rate of change of voltage, and gate charge measurements require high bandwidth sensors. This would be expensive and difficult to implement as an integral part of the converter. Therefore measurement circuits may be implemented in the gate drive of each IGBT to measure each of these parameters. In addition to the three parameters discussed above, it is also possible to determine the peak overshoot voltage and peak overshoot or reverse recovery current from the circuits used to measure

(35) $\frac{\mathrm{dV}}{\mathrm{dt}}\mathrm{and}\frac{\mathrm{di}}{\mathrm{dt}}$
respectively. Alternatively, as will be appreciated by one skilled in the art, separate circuits can be provided to directly measure the peak overshoot voltage and/or the peak overshoot or reverse recovery current.

(36) In general, the derivation of the health status of the devices may have two components, The first monitors the abovementioned parameters to detect any significant change in the most recently measured parameters as compared to historically measured parameters. The second, based upon the most recently measured parameters, specifies the current operating condition of the system.

(37) The health status of the device can be determined by observing the magnitude of the abovementioned parameters. For example the value of these observed parameters can be compared with pre-defined threshold values that indicate the health condition of the device. A look-up table can be provided in the GDEPC which enables the GDEPC to ascertain the health condition of the device. The data in the look-up table can be obtained by performing accelerated aging tests on representative examples of the device (e.g. IGBTs of the same or similar specification).

(38) A problem in the use of these parameters to indicate the health of power devices can be their dependency on the age of the device as well as the current operating conditions (such as junction temperature). Therefore in some circumstances unprocessed measurements of the parameters may not be suitable for prognostic indicators, and may require normalizing. In some examples, an electro-thermal model is used in a processor embedded in the gate drive to normalise the parameters. In such examples, all the measured prognostic indicators can be normalised with respect to variation in temperature. It is also possible to normalise the parameters with respect to the device operating current as well.

(39) FIGS. 3A-3D show the dependency of the parameters for both new/healthy and aged/unhealthy modules (i.e. components) on junction temperature. As can be seen in all of the plots, the rate of change of the respective parameter with temperature is approximately the same for both new and aged modules. For example, as is clear from FIG. 3A, the rate of change of voltage decreases with increasing junction temperature for both old and new components. The same applied for FIGS. 3B-3D and their respective parameters.

(40) Another potential problem may arise from the fact that the power devices (i.e. IGBTs) may be switched incredibly fast, for example in the range of megahertz. Therefore the amount of measurement data could be very high and, accordingly, difficult to manage. The use of an embedded processor in the gate driver can enable the triggering of data capture which may facilitate data management. The sampling frequency (i.e. the rate at which data capture is triggered) can be varied based upon the health status of the component, as shown in flow diagram in FIG. 4. Here x represents the number of cycles that should have occurred before measurement is to take place. x can be varied in light of the comparison between the normalised measurements and the pre-defined thresholds.

(41) As is also shown in FIG. 4, the case temperature and phase current of the device can be used to estimate the junction temperature using a compact thermal model. From the estimated junction temperature, a number of temperature cycles can be counted using a rain flow algorithm. When the temperature cycle counts reach a predetermined number of cycles x, the controller sends an enabling signal to trigger the collection of the various parameters. The measured parameters can then be normalised and compared with previously observed parameters. If any changes are detected a signal can be transmitted which is indicative of the health status of the device. The number of cycles that can be used for sampling the data can be linear or it can be changed considering the health status of the device to improve the accuracy of the RUL estimation. For example more samples can be taken when the device is near the end of its RUL.

(42) FIG. 5 is a block schematic of a junction temperature estimation method. Any one of the three measured parameters can be taken as inputs together with or instead of the case temperature and used life. A look-up table can then be used to estimate the junction temperature, and the estimated value stored in memory.

(43) The three parameters are able to be used to estimate the health of a component due to their predicable behaviour with respect to the number of cycles (i.e. the age of the component). This is shown in FIG. 6, which is a plot of the three parameters: rate of change of voltage, rate of change of current, and gate charge as a function of temperature cycle. In the beginning (i.e. at low temperature cycle values) the parameters show small changes. However once the temperature cycle value surpasses 20,000 cycles there are significant changes in comparison to the previously observed values and the observed values indicate that the device is no longer healthy. In this ageing characteristics curve, pre-selected threshold values are stored and used as health indicators and also to calculate the used life as well as the remaining useful life. It is also possible to use a particle filter to predict the remaining life by considering variations in the parameters towards pre-defined threshold values. FIG. 7 is a similar plot but indicating the dependence of peak over voltage V.sub.PO and reverse recovery current I.sub.RR with respect to temperature cycle value.

(44) The three parameters in FIG. 6 can be measured utilizing simple circuits which are embedded with the gate driver. In principle, only one or just two of the parameters can be measured, but a more accurate estimate of the health of the IGBT can be obtained by measuring all three.

(45) FIG. 8 shows a test circuit 803 suitable to measure the gate charge on an IGBT 802. The IGBT is controlled by voltage source 801. In this circuit, an approximately constant current (shown in FIG. 9) is supplied to the gate of the IGBT from two switches s1, s2 and two current diodes. During normal operating conditions, switch s1 is closed and s2 is open. When gate charge is to be measured, switch s2 is closed and switch s1 opened. The net measurement of the charge consumed by the gate is relative to a given current and voltage in the gate-to-emitter path (where i.sub.G is the approximately constant current and T(V.sub.GE) is the time required to achieve the given voltage V.sub.GE from an initial voltage):
Q.sub.G#V.sub.GE=i.sub.GT(V.sub.GE)(1)

(46) The gate charge at a particular gate voltage is the product of the gate charge with respect to the time required to achieve that particular gate voltage from an initial voltage value. The gate-emitter voltage V.sub.GE and the gate current i.sub.G are shown by the plots in FIG. 9.

(47) FIG. 10 is a circuit which can be used to extract a value for the gate charge from the measured gate-emitter voltage under a constant gate current. The measured gate-emitter voltage 1003 is compared 1007 with an input voltage reference which is either an on state voltage 1001 or off state voltage 1002. In some examples this is assumed to be 15V for an on state, and 5V or 0V for an off state. The output of the comparator generates a logic pulse which enables an integrator 1007 to estimate the gate charge 1008 of the device.

(48) Similarly, the rate of change of current can be estimated using the circuit shown in FIG. 11. When a power device, such as an IGBT, is forward biased and is turned on by a gate pulse, the collector current will start rising and reach a steady state value in the immediate neighbourhood of the gate to the emitter junction. A minimum time is required for the device to spread the current conduction consistently through the conduction channel. However the current can spread across the whole area of the junction which may surpass the transistor design limits. Due to the rate of increase in collector current, local hot spots can be formed close to the gate connection. These local hot spots can destroy the transistor. Therefore this rate of increase of current is an important factor for analysing the deterioration in health of an IGBT. It could be measured directly by the current sensor 106, however this would require a very high bandwidth current sensor and associated controller. As an alternative, the rate of change of current can also be determined by using the rising time t.sub.r (where I.sub.C is the current through the collector):

(49) $\begin{array}{cc}\frac{{\mathrm{dl}}_c}{\mathrm{dt}}=\frac{0.8I_C}{t_r}& \left(2\right)\end{array}$

(50) The sub-circuit 1103 shown in FIG. 11 is a current mirror connected across a sensing resistor parallel to the IGBT's 1102 Kelvin emitter, where the IGBT 1102 is controlled by voltage source 1101. If the IGBT 1102 is lacking a Kelvin emitter terminal, an emitter can be connected with a small inductance to ground in order to measure the rate of change of current.

(51) FIG. 12 shows two waveforms: the collector current as measured in a simulation of the IGBT, and the collector current as measured in a simulation of the current mirror connected to the simulated IGBT. This plot shows that the error between the measured current using the oscilloscope and the current mirror is very small over the applicable regime (during the switch-on period). Therefore the current measured from the current mirror can be used as a representative value of the actual collector current through the IGBT over the applicable regime.

(52) FIG. 13 is a circuit which can be used to estimate the rate of change of current 1310 from the current mirror output 1302. As with FIG. 10, the input 1302 is compared 1306 to a reference value 1301 which in this case is the reference operating current limit 1301. The output of the comparator 1306 is a pulse with a modified pulse width which allows an integrator 1308 to estimate the value of the rate of change of current 1310. Similarly, a peak detector 1309 can be used to estimate the reverse recovery current I.sub.rr of the device.

(53) As with the two previous parameters, the rate of change of voltage can be measured through the use of a sub-circuit 1404 as shown in FIG. 14. When a forward voltage is applied across the collector and emitter of a transistor, the outer junction of the transistor is forward biased and the inner junction is reverse biased due to the space-charges exiting the depletion region. Therefore it behaves like a capacitance, as illustrated by Miller capacitor 1405, and induces a potential stress across the IGBT module.

(54) The rate of change of voltage can be determined by using a falling time t.sub.f and a turn off time t.sub.doff:

(55) $\begin{array}{cc}\frac{{\mathrm{dv}}_{\mathrm{CE}}}{\mathrm{dt}}=\frac{0.9v_{\mathrm{CE}}}{t_{\mathrm{doff}}}& \left(3\right)\end{array}$

(56) This can be measured by the circuit shown in FIG. 14, where an external Miller capacitor 1405 (which represents the Miller effect) is connected in series with the collector terminal of the IGBT 1402 (the IGBT being controlled by a voltage source 1401). The output of the capacitor 1405 is amplified using an amplifier.

(57) FIG. 15 shows the results of a simulation which directly measured the rate of change of collector-emitter voltage as compared to the collector-emitter voltage measured from the external Miller capacitor 1405. This plot shows that over the relevant regime (switch-off) the rate of change of the voltage measured over the external Miller capacitor 1405 is very similar to the directly measured rate of change of collector-emitter voltage. Therefore the rate of change of voltage measured over the external Miller capacitor 1405 can be used as a representative value of the actual rate of change of collector-emitter voltage over the applicable regime.

(58) The rate of change of voltage can then be estimated using the circuit shown in FIG. 16, which takes as its inputs the measurement voltage from the Miller circuit 1602 and an input voltage reference 1601. The values are compared 1606 and the output is a pulse with a modified pulse width. This output can be used by an integrator 1608 to estimate the rate of change of voltage 1610. A peak detector 1609 can also be used to extract the off-state overshoot voltage (V.sub.PO) 1611 of the device.

(59) FIGS. 17 to 19 show ageing test results that confirm the suitability of the three parameters for estimating aging.

(60) FIG. 17 is a plot showing the change in gate charge characteristics before 1702 and after 1701 a power cycling test. As can be seen, whilst the gate voltage takes the same switching time to reach the operating voltage both before and after, the overall gate charge has increased after a power cycling tests.

(61) Similarly, FIG. 18 shows that the rate of change of current before 1802 and after 1801 a power cycling test are appreciably different. The results of the test is that the rate of change of current increases after being subjected to a power cycling test.

(62) Finally, FIG. 19 shows that the rate of change of voltage before 1902 and after 1901 a power cycling test are appreciably different. The result of the test is that the rate of change of voltage decreases after being subjected to a power cycling test. From equation (3) it can be seen that the reduction of rate of change of voltage is the effect of decreasing the turn off delay time t.sub.doff. Asides from the rate of change of voltage, the peak overshoot voltage (highest value before reaching a steady state) also decreases.

(63) From these plots, it can be seen that the rate of change of current

(64) 0$\frac{\mathrm{di}}{\mathrm{dt}},$
reverse recovery current I.sub.rr, rate of change of voltage

(65) $\frac{\mathrm{dv}}{\mathrm{dt}},$
peak overshoot voltage V.sub.PO and gate collector capacitance (and therefore gate charge) vary significantly as the device ages.

(66) FIG. 20 shows a schematic of an example propulsion system. A power supply 2001, in this example an AC supply, connects to a power device 2002 (e.g. of the type shown in FIG. 1) which includes at least one IGBT. The power device 2002 provides an electrical supply to a propulsion component 2003 e.g. a motor, fuel pump etc. This propulsion component 2003 is functionally linked to a propulsion drive 2004 e.g. a propeller, gas-turbine engine etc.

(67) Advantageously, the invention as described above may provide the following advantages: The knowledge of the health of the power system and/or devices, e.g. in the form of the RUL, can facilitate early identification of failure conditions. This can enable condition based control and scheduling for maintenance. The estimation of junction temperature can allow the system to prevent the operation of the power devices beyond their safe operating range. This can stop premature ageing of the devices. Significant deviations in the average values of the parameters measured can be used to identify weaker devices in the system. This can facilitate early identification of failure conditions. The use of relatively simple measuring circuits can avoid a need for high bandwidth sensors.

(68) The invention may be applicable in a number of industrial, automobile, and aero electric converter systems to monitor the health and life of the system. The converter systems can be, for example, DC/DC, DC/AC, or AC/DC converters. The converters can use different types of power modules, for example Si, SiC, GaN, etc.

(69) In addition to monitoring the devices and modules, the health monitoring system can be expanded to include other component monitoring capabilities. For example the monitoring of capacitor and filter performance etc.

(70) While the invention has been described in conjunction with the exemplary embodiments described above, many equivalent modifications and variations will be apparent to those skilled in the art when given this disclosure. Accordingly, the exemplary embodiments of the invention set forth above are considered to be illustrative and not limiting. Various changes to the described embodiments may be made without departing from the spirit and scope of the invention.

(71) All references referred to above are hereby incorporated by reference.