Methods for estimating operational state of semiconductor chip

Abstract

To provide power converters.SOLUTION: The present disclosure provides examples of methods and systems for estimating an operational state of a semiconductor chip of a power semiconductor device which are based on determining a parameter indicative of an operational state of the semiconductor chip including an indication of uncertainty of the operational state based on the received operational data. The present disclosure further provides examples of methods and systems for predicting a remaining useful life of a semiconductor chip.SELECTED DRAWING: Figure 3A

Description

[Technical Field]

[0001] The present disclosure relates to power converters, and more particularly to methods and systems for determining the operating state of power semiconductor devices in a power converter. The present disclosure further relates to methods and systems for estimating the cumulative life usage or life usage and peak junction temperature of semiconductor chips in power devices, particularly power electronic converters coupled to generators of wind turbines. The present disclosure further relates to methods and systems for predicting the remaining useful life of semiconductor chips, particularly semiconductor chips used in power converters connected to wind turbines. [Background technology]

[0002] Currently used wind turbines are used to provide electrical power to the grid. A wind turbine generally includes a tower having a nacelle supported on top of the tower. A wind turbine rotor, including a hub and a number of wind turbine blades, may be rotatably mounted to the nacelle.

[0003] The wind turbine blades can be moved by the wind. The hub of the wind turbine can be operatively coupled to the rotor of a generator. As the hub and blades rotate, the kinetic energy of the wind is converted into kinetic energy of the wind turbine rotor, and ultimately into electrical energy or power in the generator. The generator can typically be located within a nacelle.

[0004] The wind turbine rotor may be directly coupled to the rotor of a generator in a so-called direct-drive wind turbine. Alternatively, the wind turbine rotor may include a main rotor shaft (a so-called "low-speed shaft") that leads to a gearbox. The high-speed shaft of the gearbox may then drive the generator. Regardless of the form of the wind turbine, the power output of the generator may be fed into a power grid. The connection of the generator to the power grid may include, for example, converters, transformers, medium-voltage lines, and so on.

[0005] Wind turbines also typically include a power electronic converter for converting the electrical power generated by the generator into converted AC ("alternating current") power of a predetermined frequency and voltage. The converted AC power is then fed to a main wind turbine transformer having a low voltage side and a high voltage side, which converts the converted AC power to a higher voltage and supplies the power to the power grid.

[0006] Different topologies for wind turbine generators and power electronic converters are known. One such topology is the DFIG ("Doubly Fed Induction Generator"). In a DFIG configuration, the generator's stator is directly connected to the power grid. The generator's rotor contains multiple coils. These coils are electrically connected to the power grid via power electronic converters, which include a rotor-side converter, a DC link, and a power grid-side converter.

[0007] In another known topology, the rotor of the generator carries multiple permanent magnets. The stator of the generator is connected to the grid via a so-called "full converter," which includes a machine-side converter, a DC link, and a grid-side converter.

[0008] Power converters can include a variety of different switches and rectifiers, including diodes and semiconductor devices such as Insulated Gate Bipolar Transistors (IGBTs), Metal-Oxide-Semiconductor Field-Effect Transistors (MOSFETs), and thyristors.

[0009] Power semiconductor devices may be considered herein as semiconductor devices used as switches or rectifiers in power electronics. Such devices are also referred to as "power devices."

[0010] A wind turbine also typically includes a wind turbine controller, which may be configured to determine appropriate actuator setpoints for the wind turbine based on the prevailing conditions. Current actuator setpoints for variable speed wind turbines include, for example, generator torque and blade pitch angle. Through control of blade pitch angle and generator torque, rotor speed can be controlled, as well as power output, aerodynamic thrust, and additional mechanical loads. The objective of the control system is generally to maximize power output while maintaining wind turbine loads at acceptable levels.

[0011] Normal or standard operation of a wind turbine generally follows a predetermined power curve that defines the operation of the wind turbine as a function of the prevailing wind speed. Normal operation includes different operating ranges. At low wind speeds, the objective is generally to maximize power output. At higher wind speeds, especially above the nominal wind speed, wind turbine operation focuses on controlling the load while maintaining power output at a predetermined level.

[0012] As previously mentioned, the torque and pitch actuator setpoints (but also other actuators such as yaw) can be modified in response to conditions. Such conditions can include, for example, mean wind speed, turbulence, wind shear, air density, and other weather conditions, as well as internal conditions such as vibration, mechanical loads, or component temperatures. They can also include specific external requirements to reduce noise, interruptions in operation for maintenance, grid-based conditions such as a request for active power reduction, or grid events such as low voltage events, zero voltage events, or increases in grid frequency.

[0013] Depending on the operating conditions of the wind turbine, the set points from the wind turbine controller, and the power output of the generator, the currents in the power converter and its components may vary.

[0014] When current flows through a power supply, the chip heats up due to internal losses. Because semiconductors are made of materials with different thermal expansion coefficients (silicon chip, copper heat sink, plastic case, silicone gel, etc.), repeated thermal expansion and contraction can cause thermal stress in these materials, potentially resulting in breakage of aluminum bonding wires or cracks in solder, which can lead to the destruction of the power semiconductor. This phenomenon is called power cycling. It can lead to chip degradation, lift-off of bonding wires, or other degradation or wear of the power device.

[0015] Chip temperature and its changes are important for chip performance and reliability. Temperature and temperature changes also have a significant impact on chip degradation and the end of life of chips housed in power devices. Therefore, it is beneficial to monitor temperature by directly measuring it. However, direct temperature observation can be difficult or even impossible, especially when electric drive systems are used in the field.

[0016] The lifetime of power devices in power converters varies depending on environmental conditions such as wind speed and outside temperature, which are inherently variable, making it difficult to accurately predict cumulative lifetime usage. If the lifetime usage of a power device cannot be accurately predicted, operators may miss the timing to replace the power device, which could lead to reduced operating efficiency or serious accidents. It could also lead to unnecessary power device replacement.

[0017] Understanding the ageing of power converters and implementing predictive maintenance can ensure the highest possible turbine availability. Especially for offshore turbines, which typically require specialized vessels, maintenance should be planned well in advance. Predicting converter problems allows for better and earlier planning of vessel operations, which can have a significant impact on cost savings.

[0018] It is known in the art to use high frequency sampled data, particularly phase current and voltage drop, to estimate junction temperatures. While junction temperatures have been calculated or estimated based on this, it has been found that such estimations or calculations do not necessarily lead to a well-informed decision-making process. In the prior art, this high frequency data can be used for life prediction based on rain cycle counting. However, for wind turbines from other vendors or other wind turbine fleets that do not have these life predictions, service and maintenance teams face significant challenges for performing inventory (surveys of components) and scheduling service.

[0019] Therefore, a need exists for providing estimates of power semiconductor junction temperatures and lifetimes using low frequency sampled data. There is a need to further improve the decision-making process regarding the maintenance or repair of power converters and the replacement of their components.

[0020] There is also a need to predict remaining life using low frequency sampled data.

[0021] Most existing reliability models are deterministic. However, a point estimate indicating that a converter will fail within 30 days may not be sufficient to schedule replacement / repair. Repair timing can be more easily scheduled using time intervals.

[0022] Examples of the present disclosure provide systems and methods that at least partially solve one or more of the problems discussed above.

[0023] Although this disclosure focuses on power semiconductor devices used in wind turbine power converters, similar problems are found in power semiconductor devices in other applications, and therefore this disclosure is not limited to applications in wind turbines. Summary of the Invention

[0024] In a first aspect, a method for estimating an operational state of a semiconductor chip of a power semiconductor device is provided. The method includes receiving operational data (operational data) related to the semiconductor chip over successive time intervals and determining a parameter indicative of the operational state of the semiconductor chip, the parameter including an indication of uncertainty of the operational state, based on the received operational data. The method further includes generating an alert if alert criteria (warning criteria) are met, the alert criteria being based on the parameter indicative of the operational state of the semiconductor chip and the indication of uncertainty.

[0025] According to this aspect, the operating state is determined and an alert is generated taking into account uncertainty regarding the operating state. Thus, for example, not only a specific value of peak junction temperature or lifetime usage, but rather its possible distribution is considered. The alert can be generated based on, for example, a confidence interval or probability distribution rather than when a specific threshold is reached. This method can be used with data obtained from a low sampling rate.

[0026] The operational data here may be data indirectly related to the semiconductor chip, where indirect operational data may be considered to be operational data of a machine or device in which the semiconductor chip is used or to which the semiconductor chip is connected, but not operational data (e.g., current or temperature) of the semiconductor chip itself.

[0027] An uncertainty measure may be considered herein as a measure, variable or parameter that indicates that the actual, precise state of the semiconductor chip is not known, but rather different states are possible, and these different states have different probabilities of being the actual, precise state.

[0028] In a second aspect, a method for predicting the remaining useful life of a semiconductor chip housed in a power semiconductor device is provided. The method includes estimating a current state of the semiconductor chip and determining potential degradation of the semiconductor chip over time as a function of operating conditions of the semiconductor chip. The method further includes estimating the remaining useful life of the semiconductor chip based on the estimated current state and potential degradation of the semiconductor chip. In the method, determining the potential degradation includes selecting an early-life degradation model or an end-of-life degradation model, wherein the early-life degradation model is based on a slower sampling rate than the end-of-life degradation model.

[0029] This aspect allows for improved predictability of the lifespan of power semiconductor devices, thereby providing an improved method for enabling planned retirement, replacement, or maintenance. The method recognizes that the rate and manner in which a power semiconductor device degrades may differ between early and end-of-life, and is configured to take this different degradation behavior into account.

[0030] In any of these embodiments, operation of the wind turbine and / or power converter and / or semiconductor chip may be adjusted based on a prediction of remaining useful life or based on an estimation of the operating state of the semiconductor chip.

[0031] In further aspects, the present disclosure also relates to computer programs and computer systems configured to carry out such methods.

[0032] Non-limiting examples of the present disclosure are described below with reference to the accompanying drawings. [Brief explanation of the drawings]

[0033] [Figure 1] 1 illustrates a perspective view of a wind turbine according to an embodiment. [Figure 2] 1 illustrates a detailed internal view of a nacelle of a wind turbine according to one embodiment. [Figure 3A] 1 shows a schematic example of the degradation of a power device over its lifetime; [Figure 3B] 1 shows a first example of a method for estimating an operating state of a semiconductor chip of a power semiconductor device. [Figure 3C] 1 shows a schematic example of a peak junction temperature of a semiconductor chip with an indication of the uncertainty. [Figure 3D] 3C shows a schematic example of how the method according to FIG. 3B can be implemented. [Figure 3E] 3C shows a schematic example of how the method according to FIG. 3B can be implemented. [Figure 3F] 1 illustrates schematically the use of a collection of models to determine the operating state of a semiconductor chip of a power semiconductor device. [Figure 4A] 1 illustrates schematically an example of a method for estimating an operational state of a semiconductor chip of a power semiconductor device in a power converter coupled to a wind turbine generator. [Figure 4B] 1 illustrates schematically an example of a method for estimating an operational state of a semiconductor chip of a power semiconductor device in a power converter coupled to a wind turbine generator. [Figure 4C] 10A and 10B illustrate schematically an example of a user interface that may be used in an example method for determining an operational state of a semiconductor chip. [Figure 5A] FIG. 1 is a diagram illustrating an example of estimating the remaining useful life of a semiconductor chip. [Figure 5B] 5A illustrates diagrammatically how the example method of FIG. 5A can be combined with the example methods of FIGS. DETAILED DESCRIPTION OF THE INVENTION

[0034] Reference will now be made in detail to the embodiments of the present disclosure, one or more examples of which are illustrated in the drawings. Each example is provided by way of illustration only, and not by way of limitation. Indeed, it will be apparent to those skilled in the art that various modifications and variations can be made in the present disclosure without departing from the scope or spirit thereof. For example, features illustrated or described as part of one embodiment can be used with another embodiment to yield a still further embodiment. Accordingly, it is intended that the present disclosure cover such modifications and variations as come within the scope of the appended claims and their equivalents.

[0035] FIG. 1 is a perspective view of an example wind turbine 10. In this example, wind turbine 10 is a horizontal axis wind turbine. Alternatively, wind turbine 10 may be a vertical axis wind turbine. In the example, wind turbine 10 includes a tower 100 extending from a support system 14 on ground 12, a nacelle 16 mounted on tower 100, and a rotor 18 coupled to nacelle 16. Rotor 18 includes a rotatable hub 20 and at least one rotor blade 22 coupled to the hub and extending outward from hub 20. In this example, rotor 18 has three rotor blades 22. In other embodiments, rotor 18 includes more or less than three rotor blades 22. Tower 100 may be fabricated from tubular steel to define a cavity (not shown in FIG. 1 ) between support system 14 and nacelle 16. In alternative embodiments, tower 100 may be any suitable type of tower having any suitable height. Alternatively, the tower may be a hybrid tower including a concrete section and a tubular steel section. The tower can also be a partial or full lattice tower.

[0036] Rotor blades 22 are spaced about hub 20 and facilitate rotation of rotor 18, allowing for the transfer of kinetic energy from wind into usable mechanical energy and subsequently electrical energy. Rotor blades 22 are mated to hub 20 by coupling blade root portions 24 to hub 20 at multiple load transfer areas 26. The load transfer areas 26 may include hub load transfer areas and blade load transfer areas (both not shown in FIG. 1 ). Loads induced on rotor blades 22 are transferred to hub 20 through the load transfer areas 26.

[0037] In embodiments, rotor blades 22 may have lengths ranging from approximately 15 meters (m) to approximately 90 meters or more. Rotor blades 22 may have any suitable length that enables wind turbine 10 to function as described herein. For example, non-limiting examples of blade lengths include lengths of 20 m or less, 37 m, 48.7 m, 50.2 m, 52.2 m, or greater than 91 m. When wind impinges on rotor blades 22 from wind direction 28, rotor 18 rotates about rotor axis 30. As rotor blades 22 rotate and experience centrifugal forces, rotor blades 22 also experience various forces and moments. In this manner, rotor blades 22 may deflect and / or rotate from a neutral or unbiased position to a biased position.

[0038] Furthermore, the pitch angle of the rotor blades 22, i.e., the angle determining the orientation of the rotor blades 22 relative to the wind direction, may be varied by the pitch system 32 to control the load and power generated by the wind turbine 10 by adjusting the angular position of at least one rotor blade 22 relative to the wind vector. A pitch axis 34 of the rotor blades 22 is shown. During operation of the wind turbine 10, the pitch system 32 may vary the pitch angle of the rotor blades 22, among other things, to reduce the angle of attack of (some of) the rotor blades, facilitate a reduction in rotational speed, and / or facilitate a stall of the rotor 18.

[0039] In this example, the blade pitch of each rotor blade 22 is individually controlled by the wind turbine controller 36 or pitch control system 80. Alternatively, the blade pitch of all rotor blades 22 may be simultaneously controlled by the control system.

[0040] Additionally, in this embodiment, as wind direction 28 changes, the yaw orientation of nacelle 16 may rotate about yaw axis 38 to position rotor blades 22 relative to wind direction 28 .

[0041] In the example embodiment, wind turbine controller 36 is shown as being centralized within nacelle 16, but wind turbine controller 36 may be a distributed system throughout wind turbine 10, on support system 14, within a wind farm, and / or at a remote control center. Wind turbine controller 36 includes a processor 40 configured to perform the methods and / or steps described herein. Additionally, many of the other components described herein include a processor.

[0042] As used herein, the term "processor" is not limited to integrated circuits referred to in the art as computers, but refers broadly to controllers, microcontrollers, microcomputers, programmable logic controllers (PLCs), application specific integrated circuits, and other programmable circuits, and these terms are used interchangeably herein. It should be understood that a processor and / or control system may also include memory, input channels, and / or output channels.

[0043] 2 is an enlarged cross-sectional view of a portion of wind turbine 10. In one embodiment, wind turbine 10 includes a nacelle 16 and a rotor 18 rotatably coupled to nacelle 16. Specifically, a hub 20 of rotor 18 is rotatably coupled to a generator 42 located within nacelle 16 by a main shaft 44, a gearbox 46, a high-speed shaft 48, and a coupling 50. In this example, main shaft 44 is disposed at least partially coaxially with a longitudinal axis (not shown) of nacelle 16. Rotation of main shaft 44 drives a gearbox 46, which converts the relatively slow rotational motion of rotor 18 and main shaft 44 into relatively faster rotational motion of high-speed shaft 48, thereby driving high-speed shaft 48. The latter is connected to generator 42 for producing electrical energy with the aid of coupling 50. Additionally, a transformer 90 and / or appropriate electronics, switches and / or inverters may be located within the nacelle 16 to convert the electrical energy produced by the generator 42, having a voltage between 400V and 1000V, into electrical energy having a medium voltage (10-35 KV) or higher voltage, e.g., 66 kV. This electrical energy is conducted from the nacelle 16 to the tower 100 via power cables 160.

[0044] The gearbox 46 within the transformer 90, the generator 42, may be supported by the main support structure frame of the nacelle 16, embodied as a main frame 52, if desired. The gearbox 46 may include a gearbox housing connected to the main frame 52 by one or more torque arms 103. In an embodiment, the nacelle 16 also includes a main forward support bearing 60 and a main aft support bearing 62. Furthermore, the generator 42 may be attached to the main frame 52 by a decoupling support means 54, particularly to prevent vibrations of the generator 42 from being introduced into the main frame 52 and creating a source of noise emissions.

[0045] Optionally, the main frame 52 is configured to carry the weight of the rotor 18 and components of the nacelle 16, as well as all loads caused by wind and rotational loads, and to introduce these loads into the tower 100 of the wind turbine 10. The rotor shaft 44, the generator 42, the gearbox 46, the high-speed shaft 48, the coupling 50, and associated fastening, supporting, and / or securing devices, including but not limited to the supports 52, the forward support bearing 60, and the aft support bearing 62, may be referred to as a drive train 64.

[0046] The nacelle 16 may also include a yaw drive mechanism 56 that can be used to rotate the nacelle 16 , and thus the rotor 18 , about the yaw axis 38 to control the proximity of the rotor blades 22 to the wind direction 28 .

[0047] To properly position the nacelle 16 relative to the wind direction 28, the nacelle 16 may also include at least one meteorological measurement system, which may include a wind vane and an anemometer. The meteorological measurement system 58 may provide information including the wind direction 28 and / or wind speed to the wind turbine controller 36. In this example, the pitch system 32 is at least partially disposed within the hub 20 as a pitch assembly 66. The pitch assembly 66 includes one or more pitch drive systems 68 and at least one sensor 70. Each pitch drive system 68 is coupled to a respective rotor blade 22 (shown in FIG. 1 ) to modulate the pitch angle of the rotor blade 22 along the pitch axis 34. Only one of the three pitch drive systems 68 is shown in FIG. 2 .

[0048] In this example, pitch assembly 66 includes at least one pitch bearing 72 coupled to hub 20 and each rotor blade 22 (shown in FIG. 1 ) for rotating each rotor blade 22 about pitch axis 34. Pitch drive system 68 includes a pitch drive motor 74, a pitch drive gearbox 76, and a pitch drive pinion 78. Pitch drive motor 74 is coupled to pitch drive gearbox 76 such that pitch drive motor 74 applies mechanical power to pitch drive gearbox 76. Pitch drive gearbox 76 is coupled to pitch drive pinion 78 such that pitch drive pinion 78 is rotated by pitch drive gearbox 76. Pitch bearing 72 is coupled to pitch drive pinion 78 such that rotation of pitch drive pinion 78 rotates pitch bearing 72.

[0049] The pitch drive system 68 is coupled to the wind turbine limiting device 36 to adjust the pitch angle of the rotor blades 22 upon receiving one or more signals from the wind turbine limiting device 36. In examples, the pitch drive motor 74 is any suitable motor driven by an electric and / or hydraulic system that enables the pitch assembly 66 to function as described herein. Alternatively, the pitch assembly 66 may include any suitable structure, configuration, arrangement, and / or components, such as, but not limited to, hydraulic cylinders, springs, and / or servo mechanisms. In certain embodiments, the pitch drive motor 74 is driven by the rotational inertia of the hub 20 and / or energy extracted from a stored energy source (not shown) that provides energy to components of the wind turbine 10.

[0050] The pitch assembly 66 may also include one or more pitch control systems 80 for controlling the pitch drive systems 68 in accordance with control signals from the wind turbine limiting devices 36 for certain prioritized conditions and / or during overspeed of the rotor 18. In this example, the pitch assembly 66 includes at least one pitch control system 80 communicatively coupled to each pitch drive system 68 for controlling the pitch drive systems 68 independently from the wind turbine limiting devices 36. In this example, the pitch control systems 80 are coupled to the pitch drive systems 68 and to sensors 70. During normal operation of the wind turbine 10, the wind turbine limiting devices 36 may control the pitch drive systems 68 to adjust the pitch angle of the rotor 22.

[0051] According to one embodiment, a power supply 84, including, for example, a battery, an electrical capacitor, or a generator driven by the rotation of the hub 20, is disposed at or within the hub 20 and is coupled to the sensors 70, the pitch control system 80, and the pitch drive system 68 to provide power to these components. In this example, the power supply 84 provides a continuous source of power to the pitch assembly 66 during operation of the wind turbine 10. In another embodiment, the power supply 84 provides power to the pitch assembly 66 only during a power loss event of the wind turbine 10. A power loss event may include a loss or dip in the power grid, a malfunction of the electrical system of the wind turbine 10, and / or a failure of the wind turbine limiting device 36. During a power loss event, the power supply 84 operates to provide power to the pitch assembly 66 so that the pitch assembly 66 can operate during the power loss event.

[0052] In this example, pitch drive system 68, sensor 70, pitch control system 80, cables, and power supply 84 are each disposed within a cavity 86 defined by an inner surface 88 of hub 20. In another embodiment, the components may be disposed relative to and directly or indirectly coupled to the outer surface of hub 20.

[0053] 3A shows a schematic example of degradation that may occur over the life of a power device, which may be a power device in a power converter connected to a wind turbine generator.

[0054] Power supplies have an expected maximum useful life. However, the rate at which a power device degrades may depend on the operation of the power device, particularly the level of current and associated thermal cycling. To avoid failure of the device in which the power device is used (e.g., a power converter), an end-of-life (EOL) or point in time at which the power device should be discarded or replaced may be estimated or calculated. EOL here refers to the moment at which a power device no longer functions according to its specifications. A threshold can be defined below EOL to ensure that failures are avoided.

[0055] In FIG. 3A , at time t=0, the life of the power supply begins. After some time, initial degradation in the performance of the power supply may be detected. "Now" indicates the current time. Based on the detected values ​​of different measured and operating parameters, a remaining useful life (RUL) may be calculated. However, due to uncertainties in both the measurement of the operating parameters (operating parameters that are measured or can be measured, sampling rates, and / or inherent imprecision in the measurements) and the rate and manner of degradation exhibited by the power supply, the actual or real remaining useful life may be shorter or longer than the calculated value.

[0056] Example methods and systems of the present disclosure aim to improve decision-making regarding the disposal, repair or replacement of powered devices or devices incorporating such powered devices by taking such uncertainties into account.

[0057] In a first aspect, a method 200 for estimating an operational state of a semiconductor chip of a power semiconductor device is provided. The method 200 includes receiving operational data associated with the semiconductor chip over successive time intervals at block 210. The operational data may be operational data indirectly associated with the semiconductor chip, such as measured or derived parameters or variables not directly available from the power semiconductor device.

[0058] The method then includes determining, based on the received indirect operational data, parameters indicative of the operational state of the semiconductor chip, including an indication of uncertainty in the operational state, at block 220. The method further includes generating an alert, at block 240, if alert criteria (evaluated at block 230) are met, the alert criteria being based on the parameters indicative of the operational state of the semiconductor chip and the indication of uncertainty.

[0059] In an example, the continuous time interval may be between 1 minute and 20 minutes, specifically between 5 minutes and 15 minutes.

[0060] As can be seen in the example of FIG. 3B , the method illustrated herein may be executed continuously. That is, if the alert criteria of block 230 are not met, the method returns to block 210. Also, if the alert criteria are met in block 230 and an alert is generated in block 240, the method may return to block 210. In addition to being executed continuously, several loops of the method may be executed in parallel. That is, before one loop ends, another loop may be started.

[0061] In some instances, a parameter that indicates the operating state of a semiconductor chip is the peak junction temperature. Here, the peak junction temperature can be considered as the maximum operating temperature of the actual semiconductor in an electronic device. High junction temperatures can damage or degrade the semiconductor. Temperature is important for the performance, reliability, and end-of-life of the chip housed in the power device. Here, both the maximum temperature and the temperature change are important.

[0062] Direct observation or measurement of temperature can be very difficult or impossible. Prior art techniques use high frequency sampled data, such as phase current and voltage drop, to derive or estimate peak junction temperature. Such high frequency data is not always available. Furthermore, determining peak junction temperature by such methods is never 100% accurate; that is, even if such high frequency data is available, a level of uncertainty remains.

[0063] 3B, the alert criteria incorporate not only the temperature value but also an indication of the temperature uncertainty. In some examples, the operating condition uncertainty value and indication may include two or more of the mean, standard deviation, variance, lower and upper bounds with confidence intervals, minimum, maximum, median, quantiles, and interquartile range for the parameter indicative of the operating condition (in this case, peak junction temperature).

[0064] 3C graphically illustrates how to determine, for each time period (e.g., 10 minutes), the mean or average peak junction temperature as well as a measure of the uncertainty in the peak junction temperature (e.g., a confidence interval, a standard deviation, a multiple of the standard deviation, or the like). Lines 290 and 295 represent point estimates of the peak junction temperature from two independent models, and line 297 represents the mean peak junction temperature with a 95% confidence interval. Note that line 297 and its values ​​correspond to the output of block 220 in FIG. 2A. Note that the two independent point estimates of the peak junction temperature are covered within the confidence interval output from line 297 in this example.

[0065] In one example, the alert criterion may be the upper limit of a 95% confidence interval of the peak junction temperature reaching a particular threshold. When the alert criterion is met, an alert may be generated. In some examples, the alert criterion may be a predetermined number of times (e.g., three, five, 10, or any suitable number) that the upper limit of the peak junction temperature exceeds a predetermined temperature threshold.

[0066] The alert can take various suitable forms, such as a warning message or alarm signal in a user interface. Such a user interface can be located, for example, in a remote operation center. From such a remote operation center, multiple wind turbines, or even multiple wind farms, can be monitored. An operator can take note of such an alert and, if necessary, adjust the operation of the wind turbine or the power converter based on the alert.

[0067] The user interface may be a graphical user interface. The alert may be in the form of a graph, diagram, text message, table, or any other suitable format. The alert may additionally or alternatively be audible.

[0068] In some examples, the parameter indicative of the operating state of the semiconductor chip may be the lifetime usage of the chip, and optionally may be the cumulative lifetime usage of the chip, or the lifetime usage percentage of the chip. Instead of or in addition to the peak junction temperature, the lifetime usage of the chip may also be used to generate an alert. In particular, a threshold may be defined relative to the cumulative lifetime usage or the lifetime usage percentage.

[0069] Here, lifetime consumption refers to the amount (e.g., percentage) of the lifetime of a semiconductor chip that has been consumed within a specified period or used up to a specified point in time (cumulative lifetime consumption or cumulative lifetime usage).

[0070] In some examples, a lifetime usage alert criterion is when the difference between the upper limit of accumulated lifetime usage and the lower limit of accumulated lifetime usage for a period of time exceeds a predefined lifetime usage threshold. The difference between the upper limit at the end of the period and the lower limit at the start of the period is an indicator of the rate of deterioration.

[0071] Alerts can be generated at the machine level, e.g., at the power converter level or wind turbine level. Alerts can also be generated at the individual phase module level. For example, if either the peak junction temperature or cumulative life usage of the semiconductor chips of an individual phase module meets the alert criteria, the specific individual phase module can be identified in the user interface. Thus, the operator can be informed of the overall degradation or life usage at the converter level, as well as at the individual phase module level.

[0072] In examples, the parameter indicative of the operating condition of a semiconductor chip may be based on lifetime usage or peak temperature of one or more of a transistor, a diode, a heat sink, or a capacitor.

[0073] 3D and 3E diagrammatically show examples of how the method according to FIG. 3B may be implemented. FIG. 3D schematically shows a method for determining the peak junction temperature. At block 206, user input may be obtained. The user input may include, for example, a search time duration, a phase module of interest, a temperature threshold, and a maximum number of times the temperature threshold may be reached (e.g., within a predetermined period of time).

[0074] A phase module is herein considered to be the converter module of a given electrical phase or group of phases of the wind turbine generator.

[0075] In block 208, it may be determined whether all relevant phase modules have been looped in (covered). In block 210, the average peak junction temperature and standard deviation may be obtained for the phase module. Based on the average peak junction temperature and standard deviation, a confidence interval (e.g., a 90 or 95% confidence interval) for the peak junction temperature (Tjpk) may be determined. The upper limit of the confidence interval may be compared to a temperature threshold in block 230. If the upper limit of the confidence interval exceeds the temperature threshold, the particular phase module may be added to an alert table 245.

[0076] The alert table 245 in this example includes a timestamp (date, time) of the event, the ID or identifier of the wind turbine in question, and a description of the event.

[0077] The method can proceed to the next phase module.

[0078] 3E shows a similar example of a method for determining the life usage of a semiconductor chip. At block 207, user input can be obtained. The user input can include, for example, a search time, phase modules of interest, and a maximum allowable life usage. At block 209, a check is made to see if all relevant phase modules have been looped through.

[0079] In block 211, an estimated average cumulative lifetime usage may be determined for the power devices in the phase module, along with a standard deviation of the cumulative lifetime usage. In blocks 221 and 223, an upper limit for the cumulative lifetime usage at the end of the period and a lower limit for the cumulative lifetime usage at the beginning of the period may be determined. From these, a lifetime usage percentage may be calculated. If the lifetime usage percentage exceeds the lifetime usage percentage threshold, the phase module may be added to the alert table 247.

[0080] As with the previous example, the alert table 247 in this example includes a timestamp, the identification of the wind turbine in question, and a description of the corresponding event. In both example tables 245 and 247, instead of identification at the wind turbine level, the identification may be at the level of an individual phase module.

[0081] The method can proceed to the next phase module.

[0082] For example, determining a parameter indicative of the operational state of the semiconductor chip, including an indication of uncertainty in the operational state, based on the received operational data includes determining the operational state of the semiconductor chip using a collection of models that estimate the operational state of the semiconductor chip.

[0083] Ensemble modeling is the process of creating multiple diverse models to predict outcomes using different modeling algorithms or using different training datasets. The ensemble model aggregates the predictions of each base model to generate a final prediction. Such a final prediction may include a specific value and an uncertainty measure.

[0084] The ensemble of models can be either implicit or explicit. An explicit ensemble may have multiple values ​​of the state distribution, and aggregation of the multiple values ​​may be performed as a post-processing step. An implicit ensemble has multiple ensemble models within a single model, and the aggregation step may be self-contained within the same model.

[0085] In another example, determining a parameter indicative of the operational state of the semiconductor chip, including an indication of uncertainty in the operational state, based on the received operational data includes determining the operational state of the semiconductor chip using a Bayesian inference model.

[0086] In these examples, the Bayesian inference model may be a probabilistic artificial neural network trained using Bayesian inference, and the activation functions and / or weights in the artificial neural network may be probabilistic with associated probability distributions.

[0087] FIG. 3F illustrates schematically the use of a collection of models to determine the operating state of a semiconductor chip of a power semiconductor device.

[0088] In block 210, operating parameters of the wind turbine generator and / or converter are obtained, for example, at a sampling rate of 10 minutes. The operating parameters in this example include maximum power output (Pmax), average power output (Pmean), standard deviation of power output (Pstd), power converter coolant temperature, voltage at the line side converter (ULSC), and reactive power at the line side converter (QLSC). Additionally, the percentage of glycol in the cooling medium (refrigerant) (Rgly) can also be input.

[0089] Based on these parameters, the average peak junction temperature and the standard deviation of the peak junction temperature can be calculated using the first model 250. Similarly, the average and standard deviation of the peak junction temperature can be calculated using the different model 255 (and other models). Based on the results of the different models in the model ensemble, the average and standard deviation of the peak junction temperature can be obtained from the model ensemble module 275 in block 260.

[0090] Similarly, the collection of models 271, 272, . . . 27n can be used to determine the mean and standard deviation of lifetime usage.

[0091] FIG. 4A illustrates an example of a method for estimating the operating state of a semiconductor chip of a power semiconductor device, where the power semiconductor device is a power semiconductor device in a power converter of a wind turbine connected to a power grid.

[0092] 3B, the method may include receiving operational data indirectly related to the semiconductor chip over successive time intervals, at block 310. In this particular case, the operational data may be operational data of a power converter or a wind turbine.

[0093] In some examples, the operational data may include one or more of: a maximum active power of the generator during the time interval, an average active power of the generator during the time interval, a standard deviation of the active power of the generator, a coolant temperature of the converter, an average voltage of the wind turbine, and an average reactive power supplied to the grid. Other or additional operational data may also be used.

[0094] Generally, indirect operational data can be used that can indicate the current levels in the power semiconductor devices. The above data are typically already available from the wind turbine anyway; that is, these operational data are measured or registered for other purposes, including wind turbine control or wind farm control. For example, the continuous time interval for obtaining these operational data (sampling rate) may be between 1 and 20 minutes, specifically between 5 and 15 minutes. Most of the aforementioned operational data may be measured, for example, at 10-minute intervals.

[0095] 4A, based on such operational data, both the peak junction temperature and lifetime usage, and the associated uncertainty, can be determined in blocks 320 and 330. Again, uncertainty can be any measure of the potential distribution, including confidence intervals, variance, standard deviations, etc.

[0096] In the example of Figure 4A, independent alert criteria may be defined for both lifetime use at 340 and peak junction temperature at 350. If one or both of these alert criteria are met, an alert may be generated at 360 and 370, respectively.

[0097] FIG. 4B illustrates an example of how the method of FIG. 4A can be implemented. In FIG. 4B, the same or similar operating parameters as those of FIG. 3F can be obtained from a wind turbine, specifically a wind turbine generator, and a power converter connected between the generator and the power grid at 10-minute intervals. For example, the maximum power output (Pmax), average power output (Pmean), standard deviation of power output (Pstd), temperature of the power converter cooling water (coolant), line-side converter voltage (ULSC), and line-side converter reactive power (QLSC). Additionally, the glycol percentage in the cooling water (Rgly) can be input.

[0098] Based on these operational data, the peak junction temperature with an uncertainty measure (or possible distribution of peak junction temperatures), which is the output of block 330, can be derived as shown in block 370. Meanwhile, the life usage with an uncertainty measure, which is the output of block 325, can also be derived.

[0099] Uncertainty regarding lifespan usage can be propagated to determine cumulative lifespan usage, i.e., individual determinations of lifespan or (lifespan usage rates) that include uncertainty distributions of lifespan can be taken into account in determining cumulative lifespan (which has its own uncertainty distribution).

[0100] The uncertainty in the cumulative life usage may be propagated from the uncertainty in the life usage and / or junction temperature estimation, where the uncertainty propagation may use a first order derivative approach or a Monte-Carlo simulation approach.

[0101] For example, the cumulative life estimation may include a life reset module that resets the life usage value to 0 when a power semiconductor device is replaced. When an individual power semiconductor device is replaced with a new one, its cumulative life usage again becomes 0. Such resetting may occur at the level of an individual power semiconductor device.

[0102] As shown in FIG. 4B, the cumulative life usage in block 328 can distinguish and account for different types of events. While "normal" or standard operation results in a particular usage in block 325, non-standard conditions can significantly impact the life usage. Non-standard conditions can include transients, such as grid events including voltage dips, voltage sags, etc. In block 327, a particular life utilization model (or ensemble of models) can be used to determine the life utilization based on the occurrence of transient events. Input for the transient life utilization model can be derived from the determination of peak junction temperature in block 330.

[0103] Similarly, a particular model (or particular model) can be used to determine the lifetime usage (or lifetime usage rate) of power shift events in block 329. Power shift events can be considered here as transients where the power output changes in response to changes in wind speed under the control of the power converter. Inputs to these power shift event models can also include peak junction temperature in block 330.

[0104] An alert can be generated related to the peak junction temperature of the semiconductor chip at block 350, and an alert can be generated related to lifetime usage at block 340. The alert can include, for example, a lifetime usage related alert table 360 ​​and / or a diagram showing peak junction temperature variation over time for a particular semiconductor chip (or phase module).

[0105] Based on the determination of the cumulative useful life, an estimate of the remaining useful life (RUL) can be made. The RUL can include a mean value and an uncertainty measure. Quantifying the accuracy of the prediction of the reliability measure (failure rate) is just as important as the prediction itself. As mentioned earlier, estimating that a converter will fail within 30 days may not be sufficient to schedule replacement / repair. However, specifying a prediction interval of 45-60 days with a 99% probability makes the timing of repairs more easily determined.

[0106] 4C schematically illustrates an example of a graphical user interface that may be used in connection with the methods disclosed herein. The graphical user interface may be hierarchical, including information at the level of, for example, an entire fleet of wind turbines, or at the level of a group of wind turbines, e.g., all wind turbines in a wind farm. From a fleet perspective, all wind turbines in the fleet may be shown, and for wind turbines, for example, peak junction temperatures (alerts) and cumulative lifetime usage may be shown. A portion of the fleet-level interface may be displayed at the top of FIG. 4C.

[0107] In one example, a user can select a wind turbine that they are interested in. At the individual wind turbine level, information about each converter (e.g., two or three) of the wind turbine may be shown. The user can then select a particular converter and retrieve information about, for example, peak junction temperature over its lifetime or cumulative lifetime usage and its possible sources.

[0108] In a further aspect of the present disclosure, there is provided a method 400 for predicting the remaining useful life of a semiconductor chip housed in a power semiconductor device. Such a method is illustrated in Figure 5A.

[0109] As discussed with respect to the examples of Figures 3 and 4, the power semiconductor device may be a power semiconductor device located in a power converter of a wind turbine, however, the method for predicting remaining useful life may also be used in other devices.

[0110] The method includes estimating a current state of the semiconductor chip at block 410 and determining potential degradation of the semiconductor chip over time as a function of operating conditions of the semiconductor chip at block 430 .

[0111] The method further includes, at block 440, estimating the remaining useful life of the semiconductor chip based on the estimated current state and potential degradation of the semiconductor chip.

[0112] Here, as shown in FIG. 5A, determining potential degradation includes selecting, in block 420, an early-life degradation model or an end-of-life degradation model, where the early-life degradation model is based on a slower sampling rate than the end-of-life degradation model.

[0113] Here, the current state of the semiconductor chip may particularly indicate the accumulated lifetime usage.

[0114] A prediction of the potential degradation of the semiconductor chip over time can be made as a function of the operating conditions of the semiconductor chip. For example, when implemented in a power converter coupled to a wind turbine generator, the operating conditions of the semiconductor chip may depend on the operating conditions of the wind turbine, such as, for example, ambient temperature, wind speed, etc., as these affect the output of the generator, thereby affecting the load on the semiconductor chip of the power converter. Based on the estimated future load and environmental exposure, it can be predicted when the IGBT will no longer operate within its specified specifications.

[0115] Such estimation may, in some instances, be performed at the level of individual phase modules.

[0116] The remaining useful life may include uncertainty. As in the previously illustrated example, an indication of uncertainty may include, for example, the mean value, standard deviation, variance, a lower and higher bound with a confidence interval, a minimum, a maximum, a median, a quantile, and an interquartile range of the remaining useful life.

[0117] For example, if the 5% quantile of remaining useful life (RUL) is less than one week, an alert can be raised for immediate replacement. If the 5% quantity of RUL is more than one week but less than three months, an alert for planned replacement can be generated. These are just examples, different alerts based on different criteria can be configured based on specific needs.

[0118] For example, projections can be made based on different scenarios, such as a "best case" or light load scenario and a "worst case" heavy load scenario, each of which can generate a respective remaining useful life estimate.

[0119] In an embodiment, the method may further include indicating the remaining useful life of the semiconductor chips in a user interface at block 450. As previously mentioned, the user interface may be provided, for example, at a location remote from the wind turbine, particularly at a remote operations center that monitors the wind turbine or multiple wind farms. Based on such indications in the user interface, an operator may take various actions, including, for example, planned maintenance that takes into account the aging of the entire fleet or at least a substantial portion of the fleet.

[0120] By way of example, an end-of-life degradation model may have a sampling rate of 5 minutes to 2 days, more specifically 10 minutes to 1 day, for each sample. In an example, an early-life degradation model may have a sampling rate of, for example, 1 week or more. Thus, different sampling rates may be selected based on the stage of life and the need to monitor degradation more closely towards the end of life.

[0121] In an embodiment, estimating the cumulative lifetime usage of the semiconductor chip may include any of the methods illustrated with reference to Figures 3 and 4. In particular, the cumulative lifetime usage of the semiconductor chip may be based on an estimate based on, for example, uncertainty in the lifetime usage at each moment, and the uncertainty may be propagated to determine an estimate of the cumulative lifetime usage and an indication of the uncertainty.

[0122] Such estimates can be supplemented with laboratory-based data.

[0123] For example, the selection of an early-life or end-of-life degradation model may be based on changes in the life usage trends of the semiconductor chip, and if a sudden change in degradation is observed and found to be sustained, this may be an indication to switch to the end-of-life degradation model.

[0124] Different events can result in different lifetime usage rates. For example, wind turbine power converters are known to consume relatively large amounts of semiconductor chip lifetime due to transients and grid faults. Such large lifetime consumption does not necessarily indicate the aging of the semiconductor chip. However, if a high lifetime usage rate occurs under normal circumstances, it may signal a trend breakdown and be an indicator for switching to an end-of-life degradation model.

[0125] A decay model may be used to fit the dynamic time evolution of, for example, a phase module health indicator during the degradation process. The decay model may be based on linear fitting. Throughout the lifetime, there may be instances of higher loads, such as transients, grid events, etc. In such cases, the lifetime is expected to be higher than under standard operating conditions.

[0126] However, using the decay model may result in a change in the usage trend of the lifespan. Based on this change in trend, it may be determined that the semiconductor chip is no longer operating as if it were at the beginning of its lifespan, and no longer operating as if it were at the end of its lifespan. Based on this determination, the sampling rate and degradation model for determining the remaining useful life estimate may be adapted.

[0127] In an embodiment, an estimate of the cumulative lifetime usage of a semiconductor chip may be based at least in part on data from testing of the power converter. Such testing may include, for example, manual testing. The testing may reveal that the condition of the semiconductor chip is better or worse than assumed based on measured operating parameters and a model used to estimate those conditions.

[0128] For example, an early life degradation model and / or an end of life degradation model determine a measure of uncertainty in degradation and potential degradation.

[0129] In an example, the end-of-life degradation model and / or the early-life degradation model may be updated based on actual usage data that takes into account similar environmental conditions and / or the recurrence of the actual usage data. Based on the actual data, the degradation model used to estimate the remaining useful life may be modified. Furthermore, the most reliable data may be selected as data based on environmental conditions that are most similar to those under which the power converter or semiconductor chip is monitored. Furthermore, recent data may be determined to be more reliable than historical data. Based on the more reliable data, the degradation model can be updated to achieve better results.

[0130] FIG. 5B shows diagrammatically how the method for determining RUL can be combined with the method for determining the operating state of a semiconductor chip of the previous embodiment.

[0131] Various example methods disclosed herein may be implemented using hardware, software, firmware, or any combination thereof.

[0132] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with this disclosure can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability of hardware and software, the various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in a variety of ways for each particular application.

[0133] The various illustrative logic blocks, modules, and circuits described in connection with the disclosure herein may be implemented or performed using one or more general-purpose processors, digital signal processors (DSPs), cloud computing architectures, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but alternatively, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, such as, for example, a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in combination with a DSP core, or any other such configuration.

[0134] The present disclosure also relates to computing systems (e.g., data processing devices, devices, or systems) adapted to perform any of the methods disclosed herein. A computing system configured to perform these methods may include dedicated modules for each of the method steps described herein, such as, for example, a "life usage module" for determining life usage including an uncertainty indicator, a "peak junction temperature module" for determining peak junction temperature including an uncertainty indicator, an "alert criteria" module, an "alert generation" module, etc.

[0135] The present disclosure also relates to a computer program or computer program product comprising instructions (code) which, when executed, perform any of the methods disclosed herein. The present disclosure also relates to a computer-readable data carrier having such a computer program stored thereon.

[0136] A computer program may be in the form of source code, object code, an intermediate code between source code and object code such as partially compiled form, or any other form suitable for use in carrying out a process. The carrier may be any entity or device capable of carrying a computer program.

[0137] If implemented in software / firmware, the functions may be stored on or transferred as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media, including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a general-purpose or special-purpose computer. By way of example, and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD / DVD or other optical disk storage, magnetic disk storage or other magnetic storage devices, or other media that can be used to carry or store desired program code means in the form of instructions or data structures and that can be accessed by a general-purpose or special-purpose computer or a general-purpose or special-purpose processor. Additionally, any connection is properly termed a computer-readable medium. For example, if the software / firmware is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, microwave, etc., then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, microwave, etc. are included in the definition of media. As used herein, disk and disc include compact discs (CDs), laser discs (registered trademark), optical discs, digital versatile discs (DVDs), floppy discs (trademark), and Blu-ray discs, where disks typically reproduce data magnetically and discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0138] For completeness, many aspects of this disclosure are set out in the following numbered clauses. [Embodiment 1] A method for estimating an operating state of a semiconductor chip of a power semiconductor device, comprising: receiving operational data associated with the semiconductor chip over successive time intervals; determining a parameter indicative of an operational state of the semiconductor chip, the parameter including an indication of uncertainty of the operational state, based on the received operational data; Raising an alert when the alert criteria are met; Including, The alert criteria are based on parameters indicative of the operating state of the semiconductor chip and an uncertainty indicator. [Embodiment 2] 2. The method of claim 1, wherein the parameter indicative of the operating state of the semiconductor chip is a peak junction temperature. [Embodiment 3] 3. The method of embodiment 2, wherein the alert criteria include a predetermined number of times that an upper limit of the peak junction temperature and an indicator of its uncertainty exceed a predetermined temperature threshold. [Embodiment 4] The method according to any one of embodiments 1 to 3, wherein the parameter indicating the operating state of the semiconductor chip is the lifetime usage of the chip, and optionally the lifetime usage rate or the cumulative lifetime usage of the chip. [Embodiment 5] 5. The method of embodiment 4, wherein the alert criteria include a difference between an upper limit of cumulative lifetime usage at the end of a period and a lower limit of cumulative lifetime usage at the beginning of a period exceeding a predetermined lifetime usage threshold. [Embodiment 6] 6. The method of claim 4 or 5, wherein the parameter indicative of the operating state of the semiconductor chip is based on lifetime usage or peak temperature of one or more of a transistor, a diode, a heat sink, or a capacitor. [Embodiment 7] 7. The method according to any one of embodiments 1-6, wherein the power semiconductor device is a power semiconductor device in a power converter of a wind turbine connected to a power grid. [Embodiment 8] 8. The method of embodiment 7, wherein the operational data associated with the semiconductor chip includes operating dates from the wind turbine and / or the power converter, and optionally includes one or more of: a maximum active power of the generator during the time interval, an average active power of the generator during the time interval, a standard deviation of the active power of the generator, a coolant temperature of the converter, an average voltage at the wind turbine, and an average reactive power supplied to the grid. [Embodiment 9] 9. The method according to any one of the preceding claims, wherein the indicator of uncertainty of the operating state comprises two or more of the following indicators for the parameter indicative of the operating state: mean value, standard deviation, variance, lower and upper limits with confidence intervals, minimum value, maximum value, median, quantile and interquartile range. [Embodiment 10] A method according to any one of embodiments 1 to 9, wherein determining a parameter indicating the operational state of the semiconductor chip, including an indicator of uncertainty in the operational state, based on the received operational data includes determining the operational state of the semiconductor chip using a collection of models that estimate the operational state of the semiconductor chip. [Embodiment 11] A method according to any one of embodiments 1 to 9, wherein determining a parameter indicative of the operational state of the semiconductor chip, including an indicator of uncertainty in the operational state based on the received operational data, comprises determining the operational state of the semiconductor chip using a Bayesian inference model, and optionally the Bayesian inference model is a probabilistic artificial neural network trained using Bayesian inference, and the activation function and / or weights in the neural network are probabilistic with associated probability distributions. [Embodiment 12] 1. A method for predicting the remaining useful life of a semiconductor chip housed in a power semiconductor device, comprising: estimating a current state of the semiconductor chip, optionally according to any of the methods of embodiments 1 to 11; determining potential degradation of the semiconductor chip over time as a function of operating conditions of the semiconductor chip; estimating a remaining useful life of the semiconductor chip based on the estimated current state and potential degradation of the semiconductor chip; The method of determining the potential degradation includes selecting an early-life degradation model or an end-of-life degradation model, the early-life degradation model being based on a slower sampling rate than the end-of-life degradation model. [Embodiment 13] 13. The method of embodiment 12, wherein the selection of the early life degradation model or the end of life degradation model is based on changes in the life usage trend of the semiconductor chip. [Embodiment 14] 14. The method of embodiment 12 or 13, further comprising indicating the remaining useful life of the semiconductor chip on a user interface. [Embodiment 15] The method of any one of embodiments 12 to 14, wherein estimating the current state of the semiconductor chip includes estimating the cumulative life usage of the semiconductor chip, optionally including the method of embodiment 4. [Embodiment 16] 16. A method according to any one of embodiments 12 to 15, wherein estimating the current state of the semiconductor chip is based at least in part on data from manual inspection of the power converter. [Embodiment 17] 17. The method according to any of embodiments 12 to 16, wherein an early life degradation model and / or an end of life degradation model determines a measure of uncertainty in degradation and potential degradation. [Embodiment 18] 18. The method according to any one of embodiments 12 to 17, wherein the end-of-life degradation model has a sampling rate of 5 minutes to 2 days, more particularly 10 minutes to 1 day, for each sample. [Embodiment 19] A method according to any one of embodiments 12 to 19, wherein the end-of-life degradation model and / or the early-life degradation model are updated based on similar ambient conditions and / or life data taking into account the recency of the life data. [Embodiment 20] 20. The method according to any one of embodiments 12-19, wherein the power semiconductor device is a power semiconductor device disposed in a power converter of a wind turbine.

[0139] This written description uses examples to disclose the teachings, including preferred embodiments, and to enable one of ordinary skill in the art to practice the teachings, including making and using any devices or systems and performing any incorporated methods. The patentable scope is defined by the claims and may include other examples that occur to those skilled in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they contain equivalent structural elements that do not differ substantially from the literal language of the claims. Aspects from the various described embodiments, as well as other known equivalents to each such aspect, can be mixed and matched by one of ordinary skill in the art to construct additional embodiments and techniques in accordance with the principles of the present application. Where reference signs relating to the drawings are placed in parentheses in the claims, they are intended only to enhance comprehension of the claims and should not be construed as limiting the scope of the claims. [Explanation of symbols]

[0140] 10: Wind turbine 12: Ground 14: Support system 16: Nacelle 18: Rotor 20: Hub 22: Rotor blade 24: Blade root 26: Load transfer area 28: Wind direction 30: Rotor shaft 32: Pitch system 34: Pitch axis 36: Wind turbine controller 38: Yaw axis 40: Processor 42: Generator 44: Main shaft 46: Gearbox 48: High speed shaft 50: Coupling 52: Main frame 54: Decoupling support means 56: Yaw drive mechanism 58: Meteorological measurement system 60: Main forward support bearing 62: Main aft support bearing 64: Drive train 66: Pitch assembly 68: Pitch drive system 70: Sensor 72: Pitch bearing 74: Pitch drive motor 76: Pitch drive gearbox 78: Pitch drive pinion 80: Pitch control system 84: Generator 86: Cavity 88: Inner surface 90: Transformer 100: Tower 103: Torque arm 160: Power cable 250: First model 255: Other models 275: Model ensemble module 360: Alert table

Claims

1. A method for estimating an operating state of a semiconductor chip of a power semiconductor device, comprising: receiving (210) operational data associated with a semiconductor chip over successive time intervals; determining (220) a parameter indicative of the operational state of the semiconductor chip, including an indication of uncertainty of the operational state, based on the received operational data; generating an alert (240) if the alert criteria are met; Including, The method (200), wherein the alert criteria are based on a parameter indicative of the operating state of the semiconductor chip and an uncertainty indicator.

2. 2. The method (200) of claim 1, wherein the parameter indicative of the operating condition of the semiconductor chip is a peak junction temperature.

3. 3. The method (200) of claim 2, wherein the alert criteria include a predetermined number of times that an upper limit of the peak junction temperature and a measure of its uncertainty exceed a predetermined temperature threshold.

4. The method (200) of claim 1, wherein the parameter indicative of the operating condition of the semiconductor chip is a lifetime usage of the chip, optionally a lifetime usage percentage or a cumulative lifetime usage of the chip.

5. 5. The method (200) of claim 4, wherein the alert criteria comprises a difference between an upper limit of cumulative lifetime usage at the end of the period and a lower limit of cumulative lifetime usage at the start of the period exceeding a predetermined lifetime usage threshold.

6. The method (200) of claim 4, wherein the parameter indicative of the operating condition of the semiconductor chip is based on lifetime usage or peak temperature of one or more of a transistor, a diode, a heat sink, or a capacitor.

7. The method (200) of claim 1, wherein the power semiconductor device is a power semiconductor device in a power converter of a wind turbine connected to a power grid.

8. 8. The method (200) of claim 7, wherein the operational data associated with the semiconductor chip includes dates of operation from the wind turbine and / or the power converter, and optionally includes one or more of: a maximum active power of the generator during the time interval, an average active power of the generator during the time interval, a standard deviation of the active power of the generator, a coolant temperature of the converter, an average voltage at the wind turbine, and an average reactive power supplied to the grid.

9. 2. The method (200) of claim 1, wherein the measure of uncertainty of the operating state comprises two or more measures of mean, standard deviation, variance, lower and upper limits with confidence intervals, minimum, maximum, median, quantile and interquartile range for the parameter indicative of the operating state.

10. 2. The method (200) of claim 1, wherein determining a parameter indicative of the operational state of the semiconductor chip, including an indication of uncertainty in the operational state, based on the received operational data comprises determining the operational state of the semiconductor chip using a collection of models that estimate the operational state of the semiconductor chip.

11. 2. The method (200) of claim 1, wherein determining a parameter indicative of the operational state of the semiconductor chip, including an indication of uncertainty in the operational state, based on the received operational data comprises determining the operational state of the semiconductor chip using a Bayesian inference model, and optionally, the Bayesian inference model is a probabilistic artificial neural network trained using Bayesian inference, wherein activation functions and / or weights in the neural network are probabilistic with associated probability distributions.

12. 1. A method for predicting the remaining useful life of a semiconductor chip housed in a power semiconductor device, comprising: A method according to any one of claims 1 to 11, comprising the step of estimating (410) a current operating state of a semiconductor chip; determining (430) potential degradation of the semiconductor chip over time as a function of operating conditions of the semiconductor chip; estimating (440) the remaining useful life of the semiconductor chip based on the estimated current state and potential degradation of the semiconductor chip; A method (200) comprising:

13. A computing system including a processor configured to perform the method of any of claims 1 to 11.

14. A computer program comprising instructions which, when executed by a computer, cause the computer to carry out a method according to any one of claims 1 to 11.

15. A computer readable data carrier storing a computer program according to claim 14.

Images