System and method for controlling a wind power converter

Abstract

A method of operating a wind converter is provided. The method includes receiving a plurality of forecast data sets. The forecast data sets include event signals of the wind converter during fast transient operating conditions (OCs) and operating data of the wind converter having a low sampling rate. The method further includes estimating a converter life consumption during normal OCs and a converter life consumption during fast transient OCs. In addition, the method includes calculating a total converter life consumption of the wind converter. Furthermore, the method includes predicting a remaining useful life (RUL) of the wind converter based on the total converter life consumption using an RUL prediction module. The method further includes adjusting an operation of the wind converter by adjusting an operating variable of the wind converter.

Description

Technical Field

[0001] This disclosure generally relates to systems and methods for operating wind turbine converters, and more particularly to systems and methods for operating wind turbine converters to meet target lifetimes while improving performance. Background Technology

[0002] In power electronic systems, semiconductors and capacitors are susceptible to failure over the system's lifespan. For offshore converters, semiconductor failure can be more common than capacitor failure. For example, wind turbine converter phase modules may be relatively prone to failure and replacement costs may be high.

[0003] Fatigue-related failures of wind turbine converters are relatively rare in recently commissioned wind turbine units. However, due to severe wind conditions, frequent converter tripping events, and other anomalies in offshore wind fields, wind turbine life fatigue issues may become significant in the future. Accordingly, accurate online life monitoring of wind turbine converters is desirable, but this is either unavailable or prohibitively expensive using at least some known systems. Summary of the Invention

[0004] In one aspect, a method for operating a wind turbine converter is provided. The method includes receiving multiple forecasted datasets, wherein the multiple forecasted datasets include event signals of the wind turbine converter during fast transient operating conditions (OC) and operating data of the wind turbine converter with a lower sampling rate. The method further includes using a normal OC lifetime estimation module to estimate the converter lifetime consumption during normal OC based on the operating data. The method also includes using a fast transient OC lifetime estimation module to estimate the converter lifetime consumption during fast transient OC based on the multiple forecasted datasets. Furthermore, the method includes using a hybrid lifetime estimation module to calculate the total converter lifetime consumption of the wind turbine converter by combining the estimated converter lifetime consumption during normal OC and the estimated converter lifetime consumption during fast transient OC. Additionally, the method includes using a remaining useful lifetime (RUL) prediction module to predict the RUL of the wind turbine converter based on the total converter lifetime consumption. The method also includes comparing the predicted RUL with a target RUL of the wind turbine converter. The method further includes using an active lifetime performance control module to adjust the operation of the wind turbine converter based on the comparison by adjusting the operating variables of the wind turbine converter.

[0005] In another aspect, a method for operating a wind turbine is provided. The method includes receiving multiple datasets, wherein the multiple datasets include event signals of the wind turbine during a fast transient open-circuit (OC) event and operational data of the wind turbine with a lower sampling rate. The method further includes using a normal OC lifetime estimation module to estimate the converter lifetime consumption of the wind turbine during a normal OC event based on the operational data. The method also includes using a fast transient OC lifetime estimation module to estimate the converter lifetime consumption of the wind turbine during a fast transient OC event based on the multiple datasets. Furthermore, the method includes using a hybrid lifetime estimation module to calculate the total converter lifetime consumption of the wind turbine by combining the estimated converter lifetime consumption during a normal OC event and the estimated converter lifetime consumption during a fast transient OC event. Additionally, the method includes remotely and in real-time monitoring of the wind turbine's operation based on the calculated total converter lifetime consumption.

[0006] In another aspect, a real-time remote operation monitoring and control system for a wind turbine is provided. The real-time remote operation monitoring and control system for the wind turbine includes the wind turbine, a Remaining Useful Life (RUL) prediction module, and a Remaining Useful Life performance control module. The RUL prediction module includes a hybrid lifetime estimation module. The hybrid lifetime estimation module is configured to receive multiple datasets, which include event signals of the wind turbine during a fast transient open-circuit (OC) event and operational data of the wind turbine with a lower sampling rate. The hybrid lifetime estimation module includes a normal OC lifetime estimation module and a fast transient OC lifetime estimation module. The normal OC lifetime estimation module is configured to estimate the converter lifetime consumption during a normal OC event based on the operational data. The fast transient OC lifetime estimation module is configured to estimate the converter lifetime consumption during a fast transient OC event based on multiple datasets. The hybrid lifetime estimation module is further configured to calculate the total converter lifetime consumption of the wind turbine by combining the estimated converter lifetime consumption during a normal OC event and the estimated converter lifetime consumption during a fast transient OC event. The RUL prediction module is configured to predict the RUL of the wind turbine based on the total converter lifetime consumption calculated using the hybrid lifetime estimation module. The effective life performance control module is configured to adjust the operation of the wind turbine converter by adjusting the operating variables of the wind turbine converter based on a comparison between the predicted RUL and the target RUL of the wind turbine converter.

[0007] This application provides a set of technical solutions as follows.

[0008] Technical Solution 1: A method for operating a wind power converter, the method comprising:

[0009] Receive multiple forecast datasets, wherein the multiple forecast datasets include event signals of the wind turbine during fast transient operating conditions (OC) and operating data of the wind turbine with a lower sampling rate;

[0010] The normal OC lifetime estimation module is used to estimate the converter lifetime consumption during normal OC based on the operating data;

[0011] The fast transient OC lifetime estimation module is used to estimate the converter lifetime consumption during the fast transient OC based on the multiple forecast datasets;

[0012] The total converter lifetime consumption of the wind turbine is calculated by using a hybrid lifetime estimation module to combine the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

[0013] The Remaining Useful Life (RUL) prediction module is used to predict the RUL of the wind turbine based on the total converter lifespan consumption;

[0014] Compare the predicted RUL with the target RUL of the wind turbine converter; and

[0015] The effective lifespan performance control module adjusts the operation of the wind turbine converter by adjusting its operating variables and based on the comparison.

[0016] Technical Solution 2: The method as described in Technical Solution 1 further includes using the normal OC lifetime estimation module to estimate the maximum junction temperature of the wind turbine converter during the normal OC period, wherein:

[0017] Estimating converter lifetime consumption during a fast transient OC further includes using a fast transient OC lifetime estimation module to estimate the converter lifetime consumption during the fast transient OC based on the event signal of the wind turbine converter during the normal OC, the operating data, and the estimated maximum junction temperature.

[0018] Technical Solution 3: The method as described in Technical Solution 1, wherein adjusting the operation of the wind power converter further includes:

[0019] When the predicted RUL is less than the target RUL, the effective lifespan extension control module is used to adjust the operating variables of the wind turbine converter to reduce future lifespan consumption, ensuring that the wind turbine converter meets the target converter lifespan; and

[0020] When the predicted RUL is greater than the target RUL, the effective performance enhancement control module is used to adjust the operating variables to improve performance and increase the future lifetime consumption, while still meeting the target converter lifetime.

[0021] Technical Solution 4: The method described in Technical Solution 3, wherein adjusting the operation of the wind power converter further includes switching between the effective lifespan extension control module and the effective performance enhancement control module.

[0022] Technical Solution 5: The method as described in Technical Solution 1, wherein the target RUL is based on the target converter lifetime of the wind power converter group including the wind power converter.

[0023] Technical Solution 6: The method as described in Technical Solution 1, wherein the operating data includes statistics associated with the operating parameters of the wind power converter.

[0024] Technical Solution 7: The method as described in Technical Solution 6, wherein the statistics include at least one of the following for each of the operating parameters: mean, standard deviation, kurtosis, skewness, pattern, median, quartile, minimum, maximum, range, and interquartile range.

[0025] Technical Solution 8: The method as described in Technical Solution 1, wherein the normal OC lifetime estimation module includes an alternative model, which is trained using multiple training datasets, having first operational data with a low sampling rate as input and lifetime consumption estimated using the normal OC physics-based lifetime estimation model as output.

[0026] Technical Solution 9: The method as described in Technical Solution 1, wherein the event signal includes data indicating a sudden change in wind speed and a converter trip signal.

[0027] Technical Solution 10: The method as described in Technical Solution 1, wherein calculating the total converter lifetime consumption further includes calculating the total converter lifetime consumption by calculating a weighted sum of the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

[0028] Technical Solution 11: A method for operating a wind power converter, the method comprising:

[0029] Receive multiple datasets, wherein the multiple datasets include event signals of the wind turbine during fast transient operating conditions (OC) and operating data of the wind turbine with a lower sampling rate;

[0030] The normal OC life estimation module is used to estimate the converter life consumption of the wind turbine during normal OC based on the operating data;

[0031] The fast transient OC lifetime estimation module is used to estimate the converter lifetime consumption of the wind turbine during the fast transient OC based on the multiple datasets;

[0032] The total converter lifetime consumption of the wind turbine is calculated using a hybrid lifetime estimation module by combining the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period; and

[0033] The operation of the wind turbine is monitored remotely and in real time based on the calculated total converter lifespan consumption.

[0034] Technical Solution 12: The method as described in Technical Solution 11 further includes using the normal OC lifetime estimation module to estimate the maximum junction temperature of the wind turbine converter during the normal OC period, wherein:

[0035] Estimating converter lifetime consumption during a fast transient OC further includes using the fast transient OC lifetime estimation module to estimate the converter lifetime consumption during the fast transient OC based on the event signal of the wind turbine converter during the normal OC, the operating data, and the estimated maximum junction temperature.

[0036] Technical solution 13: The method as described in technical solution 11, wherein the operating data includes statistics associated with the operating parameters of the wind power converter.

[0037] Technical Solution 14: The method as described in Technical Solution 11, wherein the normal OC lifetime estimation module includes an alternative model, which is trained using multiple training datasets, having first operating data with a low sampling rate as input and converter lifetime consumption during the normal OC estimated using a normal OC physics-based lifetime estimation model as output.

[0038] Technical Solution 15: The method as described in Technical Solution 11, wherein calculating the total converter lifetime consumption further includes calculating the total converter lifetime consumption by calculating a weighted sum of the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

[0039] Technical Solution 16: A real-time remote operation monitoring and control system for a wind power converter, comprising:

[0040] Wind power converter;

[0041] The Remaining Useful Life (RUL) prediction module includes:

[0042] A hybrid lifetime estimation module configured to receive multiple datasets, wherein the multiple datasets include event signals for the wind turbine during fast transient operating conditions (OC) and operating data for the wind turbine with a lower sampling rate, the hybrid lifetime estimation module comprising:

[0043] A normal OC lifetime estimation module, configured to estimate converter lifetime consumption during normal OC based on the operating data; and

[0044] A fast transient OC lifetime estimation module is configured to estimate converter lifetime consumption during the fast transient OC based on the plurality of datasets;

[0045] The hybrid lifetime estimation module is further configured to calculate the total converter lifetime consumption of the wind power converter by combining the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

[0046] The RUL prediction module is configured to predict the RUL for the wind converter based on the total converter lifetime consumption calculated using the hybrid lifetime estimation module; and

[0047] Effective Lifetime Performance Control Module, configured to adjust the operation of the wind power converter by adjusting the operating variables of the wind power converter based on a comparison of the predicted RUL with the target RUL for the wind power converter.

[0048] Technical Solution 17: The system as described in Technical Solution 16, wherein the normal OC lifetime estimation module further includes an alternative model, wherein the alternative model is trained using multiple training datasets, having first operating data with a low sampling rate as input and converter lifetime consumption estimated using a normal OC physics-based lifetime estimation model as output.

[0049] Technical solution 18: The system as described in technical solution 17, wherein

[0050] The normal OC lifetime estimation module is further configured to estimate the maximum junction temperature of the wind turbine during the normal OC period; and

[0051] The fast transient OC lifetime estimation module is configured to estimate the converter lifetime consumption during the fast transient OC based on the event signal, the operating data, and the maximum junction temperature of the wind turbine during the normal OC period, as estimated by the normal OC lifetime estimation module.

[0052] Technical Solution 19: The system as described in Technical Solution 16, wherein the effective lifespan performance control module further includes:

[0053] An effective lifespan extension control module is configured to adjust the operating variables of the wind turbine converter when the predicted RUL is less than the target RUL, thereby reducing future lifespan consumption and ensuring that the wind turbine converter meets the target converter lifespan; and

[0054] An effective performance enhancement control module is configured to adjust the operating variables when the predicted RUL is greater than the target RUL, in order to improve performance and increase the future lifetime consumption, while still meeting the target converter lifetime.

[0055] Technical solution 20: The system as described in technical solution 19, wherein the effective life performance control module is configured to switch between the effective life extension control module and the effective performance improvement control module. Attached Figure Description

[0056] Figure 1A This is a schematic diagram of a demonstration wind turbine converter monitoring and control system.

[0057] Figure 1B This diagram illustrates how to monitor the performance of a wind turbine converter and adjust its operation. Figure 1A The diagram shows the operation of the monitoring and control system.

[0058] Figure 2A It is a diagram. Figure 1A The block diagram shown is an example of a hybrid lifetime estimation module for the system.

[0059] Figure 2B yes Figure 2A The diagram shows a hybrid lifetime estimation module.

[0060] Figure 3A This is a diagram illustrating an example of a wind mission profile at a wind speed of 10 meters per second (m / s).

[0061] Figure 3B This is a diagram illustrating an example of a wind profile at a wind speed of 24 m / s.

[0062] Figure 4 This is a schematic diagram illustrating the demonstration training process of a normal OC lifetime estimation module using an alternative model.

[0063] Figure 5A This is a graph illustrating the maximum converter junction temperature relative to the mean and standard deviation calculated from the wind mission profile.

[0064] Figure 5B The diagram illustrates the lifespan consumption relative to the time spent generating the product. Figure 5A A graph showing the mean and standard deviation calculated from the wind mission profile.

[0065] Figure 6A This is a simplified diagram illustrating a demonstration process for calculating the lifetime consumption of converter fast transient conditions used to calculate converter tripping events.

[0066] Figure 6B This is a simplified diagram illustrating a demonstration process for calculating the lifetime consumption of a converter under rapid transient conditions that cause sudden changes in wind speed.

[0067] Figure 7 This is a schematic diagram illustrating an exemplary method for monitoring the lifespan of a wind turbine converter using the hybrid converter lifespan estimation module shown in Figure 2.

[0068] Figure 8A This is a demonstration flowchart of an exemplary method for predicting the remaining useful life of a wind turbine converter.

[0069] Figure 8B This is a demonstration flowchart of another exemplary method for controlling the operation of a wind turbine converter.

[0070] Figure 9 This is a block diagram of a demonstration computing device. Detailed Implementation

[0071] This disclosure includes systems and methods for monitoring the lifespan of wind turbine converters, predicting their remaining useful life (RUL), and controlling lifespan performance in real time based on hybrid lifespan estimates. The systems and methods described herein are applicable to off-the-shelf wind turbine operating data with low sampling rates (e.g., Global Repository (GR) data collected at 10-minute intervals). The systems and methods described herein do not require any input from the converter supplier. loss / Thermal parameters or additional hardware for implementation. Furthermore, the control algorithm is faster and reduces computational burden. The methods and systems disclosed herein also seamlessly integrate a "life extension control mode" under high stress or use of the wind turbine converter and a "performance enhancement control mode" under low stress or use into a single life performance control module, and can automatically switch between these two modes in real time based on consumer needs. The systems and methods described herein can also be applied to wind turbines in a group to balance power among turbines, enabling the wind turbine converter group to have homogeneous aging based on the target lifespan of the entire group.

[0072] Wind turbine fatigue can become significant due to harsh wind conditions, frequent converter tripping events, and other anomalies at offshore wind fields. Therefore, it is important to provide accurate online lifespan monitoring of wind turbines throughout their entire service life. For wind turbines, online junction temperature monitoring and lifespan monitoring are typically unavailable or prohibitively expensive. Therefore, there is a need for online estimation of converter junction temperature and lifespan monitoring using readily available, low-sampling-rate operating data.

[0073] Wind turbine converter phase modules significantly contribute to wind turbine downtime, and converter phase module lifespan failures are typically associated with semiconductor components. Two common failure modes are semiconductor wire bond cracking / peeling and solder cracking / delamination. Both failure modes are fatigue-related and are caused by thermoelectric stresses generated during power / thermal cycling.

[0074] Changes in junction temperature represent thermal cycling and are indicators of thermal stress and lifespan depletion. Junction temperature can be measured using optical measurement devices (such as infrared (IR) cameras) or physical measurement devices (such as resistance temperature detectors (RTDs) or thermocouples). However, these methods require additional sensors pre-installed with semiconductor components, which may not be practical for commercial converter products.

[0075] It can also indirectly estimate the junction temperature. Estimation techniques can be classified into two categories. One category includes estimations using temperature-sensitive electrical parameters (TSEP) (e.g., V). CEon V ge,th V ge,off and I sat This can be used to estimate junction temperature. However, TSEP measurements require additional measurement circuitry and may therefore be difficult to implement in commercial converter products. Another category of techniques includes using dynamic resistance-to-capacitor (RC) network estimation to estimate junction temperature. Accurate thermal RC network estimation requires detailed loss data for each semiconductor (e.g., an insulated-gate bipolar transistor (IGBT) or diode) as well as the thermal RC parameters of the entire power module, which can be challenging to obtain. Furthermore, this category of techniques requires high sampling rates (at least on the order of several seconds) of operating data throughout the converter's lifespan due to high turbulence caused by wind conditions, and also requires significant computational power, as semiconductor loss calculations are performed, for example, on the order of microseconds. These requirements are difficult to meet because original equipment manufacturers (OEMs) can only provide continuously recorded converter data at predetermined sampling intervals (e.g., every 10 minutes), and there are generally limited computational capabilities for cloud-based analytics.

[0076] Therefore, there is a need for a converter lifetime estimation solution that can be directly applied to readily available low-sampling-rate wind turbine operating data (e.g., GR data collected at 10-minute intervals) and is parameter-free, fast, and has a low computational burden.

[0077] As used herein, low-sampling-rate operational data refers to data on operating parameters (e.g., power, voltage, temperature, and flow rate) that reflect the operating status of the wind turbine converter and are acquired at intervals of approximately several minutes or longer. Low-sampling-rate operational data may be collected by the OEM. Compared to mission profile data (which represents the power output of the wind turbine converter and changes every few seconds or sub-seconds, and is typically sampled at a higher sampling rate, such as 20 Hz), low-sampling-rate operational data has a lower sampling rate or frequency, for example, every 10 minutes. Low-sampling-rate operational data as used herein may include statistics on the operational data, such as the mean or standard deviation. It may also include other statistics for a given operational data point, such as kurtosis, skewness, pattern, median, quartiles, minimum, maximum, range, and interquartile range. For example, low-sampling-rate operating data for power can be a time series of power values ​​with data points every 10 minutes, where each data point is based on power data with a high sampling rate (e.g., 20 Hz) and a duration of 10 minutes, and includes the mean, standard deviation, maximum and / or minimum power values ​​within this 10-minute duration.

[0078] In addition to low-sampling-rate operating data, event signals from wind turbine converters are also incorporated into the systems and methods described herein. Event signals can also have low sampling rates, similar to low-sampling-rate operating data. Semiconductor tripping and sudden changes in wind speed constitute events and rapid transient conditions lasting approximately several minutes. Signals of such events are referred to herein as event signals.

[0079] The system and method described in this paper use low-sampling-rate operating data and event signals to estimate the lifetime consumption of a wind turbine converter. Lifetime consumption is the percentage of a wind turbine converter's remaining service life. The minimum lifetime consumption is 0%, corresponding to a wind turbine converter that has never been used. In contrast, the maximum lifetime consumption is 100%, corresponding to a wind turbine converter that has reached the end of its life. In estimating lifetime consumption, a hybrid estimation model can be used, combining a normal operating condition (OC) lifetime estimation module of an alternative model with a fast transient OC lifetime estimation module that uses either a physics-based or alternative model to estimate lifetime consumption when the wind turbine converter is operating in fast transient OC. The physics-based model is also known as an analytical model. The alternative model can be a previously trained neural network-based model. On the other hand, the physics-based model is based on computation and the physical relationship between operating data and event signals, as well as maximum junction temperature and lifetime consumption. The input data for the physics-based model can have a high sampling rate (e.g., 20 Hz). Although not directly related to lifetime consumption, maximum junction temperature is also an indicator of wind turbine converter performance. If the maximum junction temperature is too high, the wind turbine converter may need to be shut down.

[0080] The hybrid lifetime estimation module can be used to monitor the lifetime of wind turbines. In monitoring the lifetime of wind turbines, the lifetime estimation process can be repeated for each time point of low-sampling-rate data. That is, lifetime consumption starts at 0%, and the data point at the first time point is processed to estimate the change in lifetime consumption. Once these data points are processed, the estimated change in lifetime consumption is added to the total lifetime consumption up to the previous time point, and then the data point at the next time point is processed. This process is repeated iteratively until data at all time points has been processed.

[0081] It can also be used to calculate the Remaining Lifetime (RUL) and, based on the calculated RUL, to control the performance of wind turbines. The operating parameters of the wind turbines are adjusted to extend their lifespan or improve their performance while still meeting the target lifespan. The target lifespan can be specified by the OEM. RUL is the remaining lifespan of the wind turbine expressed in time units (e.g., months or years). Low-sampling-rate operating data and event signals are also used to estimate RUL, but for the purpose of estimating RUL, the time series of operating data and event signals may need to be extended to a longer duration. For example, to extend the time series data, low-sampling-rate operating data and event signals can be input into a forecasting function that generates future data points based on current data points and assumptions about trends within those current data points. Similar to converter lifespan monitoring applications, RUL estimation also uses a hybrid lifespan estimation model and iterative process, except that RUL is the output rather than the current lifespan consumption. RUL is calculated based on lifespan consumption using assumptions (e.g., a linear assumption that the wind turbine ages in a synchronized manner throughout its lifespan). Once the Remaining Ultra-Live (RUL) is estimated, the operating parameters of the wind turbine are optimized to minimize the difference between the RUL and the target remaining lifetime. These optimized operating parameters are then used to adjust the operation of the wind turbine.

[0082] Figure 1AA schematic diagram of an exemplary real-time remote operation monitoring and control system 100 for a wind turbine is shown. System 100 includes a wind turbine life monitoring module 122, a converter RUL prediction module 104, and an effective life performance control module 106. The wind turbine life monitoring module 122 monitors the lifespan of the wind turbine. The converter RUL prediction module 104 predicts the RUL of the wind turbine. Both the wind turbine life monitoring module 122 and the converter RUL prediction module 104 may further include a hybrid life estimation module 102, which estimates the total lifespan of the wind turbine. Module 102 uses a hybrid life estimation algorithm to estimate the total lifespan of the wind turbine. The hybrid life estimation module 102 includes a normal OC life estimation module 108 and may further include a fast transient OC life estimation module 110. The effective life performance control module 106 adjusts the operation of the wind turbine based on the estimated RUL, which reflects the stress conditions of the wind turbine, the target converter life, and / or consumer needs. The effective lifespan performance control module 106 may further include: an effective lifespan extension control module 112 for extending the lifespan of the wind turbine converter; and an effective performance enhancement control module 114 for enhancing the performance of the wind turbine converter while still meeting the target lifespan. System 100 or its individual modules may be implemented online, wherein the system or module is web-based and accepts online stored data as input. System 100 or its individual modules may also be implemented directly on a computing device, wherein the computing device receives and transmits data wirelessly or via wired communication.

[0083] Figure 1B This diagram illustrates the operation of a system 100 used to monitor and control the performance of a wind turbine converter. To monitor wind turbine converter performance, a hybrid lifetime estimation module 102 within the wind turbine converter lifetime monitoring module 122 estimates the lifetime consumption of the wind turbine converter for each input dataset. The hybrid lifetime estimation module 102 includes a normal OC lifetime estimation module 108 and a fast transient OC lifetime estimation module 110, and fuses the results from the two modules 108 and 110 to obtain a combined lifetime estimation result. The hybrid lifetime estimation module 102 takes low-sampling-rate operating data and event signals as input to each dataset and generates an output including maximum junction temperature and lifetime consumption.

[0084] In an exemplary embodiment, low-sampling-rate operating data and event signals are also used to predict the RUL and / or control the operation of the wind turbine converter. For RUL estimation, the time series of the operating data and event signals may need to be extended. For example, before being input into the hybrid lifetime estimation module 102, the operating data and event signals may be fed into the forecasting module 124 of the converter RUL prediction module 104 to predict and extend the time series of the operating data and event data by applying a forecasting function to generate a forecast dataset. Using the forecast dataset, the hybrid lifetime estimation module 102 predicts the RUL of the wind turbine converter. The predicted RUL, along with operating constraints and operating targets (e.g., converter lifespan targets and power enhancement targets), can be input into the effective lifetime performance control module 106 to control the operation of the wind turbine converter. The effective lifetime performance control module 106 adjusts the operation of the wind turbine converter based on those parameters, thereby switching between an effective lifetime extension control module 112, used to extend the lifespan of the wind turbine converter when converter stress or usage is high, and an effective performance enhancement control module 114, used to improve the performance of the wind turbine converter when converter stress or usage is low. As part of the feedback loop, the effective life performance control module 106 also provides expected operating data as input to the wind turbine converter.

[0085] Figure 2A and Figure 2B This illustrates a sample hybrid lifetime estimation module 102 for a wind power converter given a dataset. Figure 2A This is a block diagram of module 102, and Figure 2B This is a schematic diagram of module 102, including input data and data flow. The hybrid lifetime estimation module 102 includes a normal OC lifetime estimation module 108, in which an alternative model is used. The alternative model can be trained using data generated from the normal OC physics-based lifetime estimation model 120. The hybrid lifetime estimation module 102 may further include a fast transient OC lifetime estimation module 110, which includes a converter trip lifetime estimation module 110a and a sudden wind speed change lifetime estimation module 110b. The fast transient OC lifetime estimation module 110 can be used to estimate lifetime degradation caused by fast transient events different from tripping and sudden changes in wind speed.

[0086] In operation, the normal OC lifetime estimation module 108 using an alternative model takes a dataset of low-sampling-rate operating data as input and outputs the estimated maximum junction temperature and lifetime consumption during normal OC. The fast transient OC lifetime estimation module 110 takes a dataset of low-sampling-rate operating data and event signals, along with the maximum junction temperature estimated by the normal OC lifetime estimation module 108, as input and outputs the lifetime consumption during fast transient OC. The lifetime consumption estimated from the normal OC lifetime estimation module 108 using an alternative model and the fast transient OC lifetime estimation module 110 using a physics-based model or an alternative model is fed into the fused converter lifetime estimation module. Therefore, the hybrid lifetime estimation module 102 outputs the maximum junction temperature and lifetime consumption throughout the entire OC period.

[0087] The hybrid lifetime estimation module 102 can be implemented online by remotely receiving wind turbine converter data or directly on a computing device. The system 100, including the hybrid lifetime estimation module 102, can be implemented using low-sampling-rate data. Low-sampling-rate data, as used herein, has a lower sampling rate compared to the variation or disorder in the wind turbine converter's mission profile data, and may include statistics. For example, mission profile data may be 10 minutes long, changing every few seconds, and sampled at 20 Hz, while the data used herein may only include statistics from the complete 10 minutes of data. In the intended embodiment, the hybrid lifetime estimation module 102 is suitable for datasets of readily available low-sampling-rate wind turbine operating data (e.g., GR data collected at 10-minute intervals) and event signals. Furthermore, the hybrid lifetime estimation module 102 does not require loss / thermal parameters or additional components from the converter supplier to estimate the wind turbine converter's lifetime consumption and is extremely fast and computationally lightweight.

[0088] In an exemplary embodiment, the normal OC lifetime estimation module 108 includes an alternative model. The alternative model is trained and tested before being used to estimate lifetime depletion and / or maximum junction temperature. In one example, a normal OC physics-based lifetime estimation model 120 is used to generate training and testing datasets for the alternative model. Module 120 provides results of a physics-based converter lifetime estimation over a given period (e.g., 10 minutes). The results include converter junction temperature estimates (e.g., in °C) and lifetime estimates (e.g., in %) during normal OC. The results are based on wind mission profiles under different wind speed conditions (e.g., from 4 m / s to 26 m / s) and other operating / design parameters over a given period (e.g., 10 minutes). Operating / design parameters may include active / reactive power of the wind turbine-side / line-side converter, pulse-width modulation (PWM) switching frequency, AC base frequency, AC RMS voltage, DC bus voltage, coolant temperature, coolant flow rate, and coolant ratio. The graphs illustrate two examples of wind profiles at wind speeds of 10 m / s and 24 m / s over a 10-minute period. Figure 3A and Figure 3B The data is provided in the middle, and the wind profile is sampled at 20 Hz. Figure 3B Compared to the wind profile shown, Figure 3A The wind profile shown has greater variation over time, indicating that... Figure 3B Compared to the wind profile shown, Figure 3A The wind profile shown has more thermal cycling and therefore has a faster lifespan.

[0089] The normal OC physics-based lifetime estimation model 120 may include a semiconductor loss calculation module, a thermal RC network module, and a lifetime counting module. The lifetime counting module can use a counting algorithm to calculate lifetime consumption. Module 120 provides power loss, junction temperature, and lifetime consumption estimates for all semiconductor devices of the power converter.

[0090] Figure 4 This diagram illustrates a demonstration training process for an alternative model used in the normal OC lifetime estimation module 108. The alternative model is trained using low-sampling-rate wind converter operating data as input and the maximum junction temperature and lifetime consumption obtained through the normal OC physics-based lifetime estimation model 120 as the target output. The predicted output from 108 is compared with the target output from module 120, and the residual or error between the target and predicted outputs is calculated using a loss function and used to adjust the parameters within the alternative model, reduce the residual or error, and improve the accuracy of the alternative model.

[0091] In some embodiments, design parameters (DP) (e.g., f) PWM, LSC and f PWM, MSCThe parameters are fixed at certain frequencies (e.g., 4000 Hz and 2000 Hz) and cannot be considered as input variables. Different alternative models can be constructed and trained independently if different design parameters (e.g., PWM frequency) are used as inputs. In one embodiment, f MSC The (equivalent wind turbine speed) is correlated with the active power P at each instantaneous time point according to the defined power curve at a given wind station, and therefore cannot be regarded as a separate input variable.

[0092] The normal OC lifetime estimation module 108 can use different types of alternative models, such as polynomial response surfaces, radial basis functions, support vector machines, and artificial neural networks. Artificial neural network models, such as shallow feedforward neural networks, deep feedforward neural networks, recurrent neural networks, and long short-term memory, can be used. As another example, multiple neural networks forming an ensemble of neural networks can also be used.

[0093] In the exemplary embodiment, the alternative model for the normal OC lifetime estimation module 108 uses operational data and results from physics-based converter lifetime estimation to train an approximate model to estimate lifetime consumption and maximum temperature (see [link to example model]). Figure 2A and Figure 2B If only the physical lifetime estimation model 120 for normal OC is used to estimate the lifetime consumption and maximum temperature of normal OC, module 120 cannot be applied with low-sampling-rate data, and it can lead to a large computational burden (e.g., performing semiconductor loss calculations in microseconds), and can affect the real-time execution of the method. In contrast, the computational burden of the alternative model described herein is mitigated by using data with a low sampling rate and constructing an approximate model (e.g., an alternative model). The calculations in the alternative model used for converter normal OC lifetime estimation module 108 are therefore faster. For example, the calculations can be completed within milliseconds or seconds, depending on the amount of data. In addition, using the alternative model also eliminates the need for any loss / thermal parameters or additional hardware from the converter vendor.

[0094] In addition to reducing computational burden and increasing computational speed, the low sampling rate operating data leveraged using the systems and methods described in this paper is correlated with lifetime calculations and maximum junction temperature. Figure 5A and Figure 5B The figure shows the maximum converter junction temperature. Figure 5A ) and lifespan consumption ( Figure 5B A graph showing the mean and standard deviation relative to the values ​​calculated from the wind mission profile. This is derived from physics-based converter lifetime simulation results provided by the converter supplier. Figure 5A and Figure 5B The maximum converter junction temperature and lifetime consumption used. Figure 5A and Figure 5BThe average value and standard deviation calculated from the wind power profile (sampled at 20 Hz) are the maximum converter IGBT junction temperature over a 10-minute time period. Figure 5A ) and lifespan consumption ( Figure 5B A good indicator of ). As average power increases, maximum junction temperature increases ( Figure 5A However, lifespan consumption does not necessarily increase. Figure 5B In contrast, as the standard deviation of power increases, lifetime consumption increases ( ). Figure 5B However, the maximum junction temperature ( Figure 5A This does not necessarily increase. These results demonstrate that lifetime degradation depends heavily on the variations experienced by the semiconductor device (i.e., thermal cycling), while the maximum junction temperature depends heavily on the average value (see also...). Figure 3A and Figure 3B ).

[0095] The hybrid lifetime estimation module 102 may further include a fast transient OC lifetime estimation module 110, which considers the impact of event signals on the lifetime consumption of the wind turbine converter. The fast transient OC lifetime estimation module can be constructed using a physics-based model or an alternative model, similar to that in the normal OC lifetime estimation module. In an exemplary embodiment, the maximum junction temperature result output from the normal OC lifetime estimation module 108, along with low-sampling-rate operating data and event signals, is fed into the physics-based model in the fast transient OC lifetime estimation module 110 to estimate the converter lifetime consumption during the fast transient OC for each or selected converter component.

[0096] Figure 6A This is a simplified diagram illustrating a demonstration process for estimating lifetime consumption using a physics-based model of a fast transient OC (overclocking) event involving a converter trip. This process is performed in the converter trip lifetime estimation module 110a. The maximum junction temperature estimated by the normal OC lifetime estimation module 108 is used to approximate the junction temperature before tripping or after rebooting (at 602 and 608, respectively). The converter ambient or coolant temperature can be used to approximate the junction temperature after tripping or before rebooting (at 604 and 606, respectively) because the time between tripping and rebooting is typically much larger than the time constant of the converter phase module. In estimating lifetime consumption, the converter trip lifetime estimation module 110a calculates the output lifetime consumption per trip 610 and the lifetime consumption per reboot 612. This module may also include calculations for the increase in maximum junction temperature caused by trip current overshoot. If no converter trip signal is present, the lifetime consumption due to the trip / rebooting event is assumed to be zero.

[0097] Figure 6BThis is a simplified diagram illustrating a physics-based model of a rapid transient condition involving sudden changes in wind speed to estimate lifetime consumption. This process is performed in the sudden wind speed change lifetime estimation module 110b. An event signal indicating a sudden change in wind speed, or an estimate of such a sudden change using operational data, is input to the sudden wind speed change lifetime estimation module 110b for lifetime estimation. The junction temperature before and after the sudden wind speed change (at 614 and 616, respectively) is approximated using the maximum junction temperature estimated by the normal OC lifetime estimation module 108. The sudden wind speed change lifetime estimation module 110b then outputs the lifetime consumption 618 for each sudden change in wind speed. If no sudden change in wind speed is detected, the lifetime consumption estimate due to the wind speed change is assumed to be zero.

[0098] In some embodiments, alternative models for fast transient operating conditions can be constructed directly, rather than using a physics-based model of fast transient operating conditions as described above. Either a physics-based or alternative model can be applied, provided it can estimate lifetime loss during fast transient operating conditions (OC) with sufficient accuracy and a lighter computational burden using low-sampling-rate operating data and event signals.

[0099] In an exemplary embodiment, once converter lifetime estimates during normal OC (e.g., using an alternative model) and fast transient OC (e.g., using a physics-based model) are completed, a hybrid wind converter lifetime estimate or fused converter lifetime consumption for a dataset is obtained by fusing the lifetime estimates. The fused lifetime consumption is calculated by combining the lifetime consumption during normal OC and the lifetime consumption during fast transient OC. For a given snapshot of converter data collected at time t (where the lifetime consumption during normal OC is denoted as ΔLC)... N (t), the lifetime consumption during the fast transient over-current (OC) period of the converter trip is expressed as ΔLC. FT,1 (t), and the lifespan consumption during sudden changes in wind speed is expressed as ΔLC. FT,2 (t)), then the fusion lifetime consumption ΔLC(t) of the snapshot at time t is calculated as follows:

[0100]

[0101] in , and These are the weights associated with events such as sudden changes in wind speed, converter tripping, and estimated lifetime consumption during normal over-the-air (OC) conditions, respectively. If there are no converter tripping events or sudden changes in wind speed during this time period, then ,as well as The weight w can be determined by the actual duration of each OC during the entire time period (e.g., 10 minutes) of the collection time point t of a given dataset. i The value is calculated for each of the semiconductor devices in the hybrid lifetime estimation module 102 if more semiconductor devices are included.

[0102] Figure 7 This is a schematic diagram illustrating an exemplary method 700 for monitoring a wind turbine converter using a hybrid lifetime estimation module 102. Method 700 can be applied online, wherein it is web-based and accepts online stored data as input. Method 700 can also be implemented directly on a computing device, wherein the computing device receives and transmits data wirelessly or via wired communication.

[0103] In method 700, lifetime consumption ( Figure 7 (abbreviated as LC) starts at 0%. At time points t = t0, t1, t2, ..., t k 、 . . .、t present To collect the converter's operational data set. The operational data includes the operating parameters of the wind turbine converter. At time point t... k Each snapshot of the collected operating data may include low-sampling-rate converter operating data. The collected data may include at least operating parameters such as converter active power output (P), machine-side / line-side converter reactive power (Q). LSC and Q MSC ), Line-side AC (RMS) root mean square voltage (U) LSC ), coolant temperature (T) coolant ), coolant flow rate (V coolant ) and the coolant ratio of the mixed coolant (R coolant See also Figure 2B The data includes the mean and standard deviation (STD) of each time variable. The data may also include event signals, such as signals indicating sudden changes in wind speed and / or converter tripping signals.

[0104] In an exemplary embodiment, method 700 further includes using a hybrid lifetime estimation module 102 to estimate lifetime consumption at a specific time point 702. Converter low-sampling-rate operating data and event signals are fed into the hybrid lifetime estimation module 102 to provide a converter lifetime estimate. The hybrid lifetime estimation module 102 includes a normal OC lifetime estimation module 108. An alternative model is used to construct the normal OC lifetime estimation module 108. In one example, training and testing datasets are used to train and test the normal OC lifetime estimation module 108 using the alternative model. The training and testing datasets may include low-sampling-rate operating data as input and lifetime consumption during normal OC as output. The training dataset may also include the maximum junction temperature as output. In some embodiments, a normal OC physics-based lifetime estimation model 120 is used with a wind mission profile and other operating / design parameters during normal OC. Figure 2A and Figure 2B (As shown in the diagram) to obtain lifetime consumption and maximum junction temperature. The hybrid lifetime estimation module 102 may further include a fast transient OC lifetime estimation module 110, which is used to consider the impact of event signals on the lifetime consumption of the wind turbine converter. The fast transient OC lifetime estimation module 110 can be constructed using a physics-based model or an alternative model. Additional event signals (such as converter trip signals) or operational data procedures may be required to trigger the calculation of each converter lifetime estimate for the fast transient OC.

[0105] Method 700 further includes calculating 704 up to time point t = t k The total lifetime consumption. The hybrid lifetime estimation module 102 may further include a fusion converter lifetime estimation module, which combines the lifetime estimation results from the normal OC lifetime estimation module 108 with the fast transient OC lifetime estimation module 110. In a contemplated embodiment, t = t k Snapshot lifespan consumption Then it was added up to the previous point in time. Total lifespan consumption The calculation is based on fatigue accumulation assumptions (such as the linear fatigue accumulation assumption) up to t = t. k Total lifespan consumption In other words, total lifespan calculation. Calculated as:

[0106]

[0107] Different fatigue accumulation assumptions (such as the nonlinear fatigue accumulation assumption) can be used to estimate total lifespan loss.

[0108] Method 700 also includes taking the snapshot time point t k Compared to the current time point t present(Or a point in time of interest) compares 706 to determine whether to proceed to the next snapshot or stop calculation, and outputs the total lifetime consumed. In one example, if t k < t present If k = k + 1, then estimate 702, calculate 704, and compare 706 repeatedly to calculate at t = t k+1 The lifetime of the snapshot. If t k t present This indicates that all snapshots have been processed, and LC (t) present ) = LC(t k This is provided as total lifetime consumption for monitoring the operation of the wind turbine converter. In an exemplary embodiment, in monitoring the lifetime performance of the wind turbine converter, if more than one semiconductor device is included, a lifetime estimate (T) for each device is calculated. jmax and LC).

[0109] System 100 may include a converter RUL prediction module 104 for estimating RUL, and may further include an effective lifetime performance control module 106 based on the estimated RUL. Instead of providing lifetime consumption as a percentage of the total lifetime of the wind turbine converter, converter RUL prediction module 104 predicts RUL as a duration (e.g., in months or years).

[0110] Figure 8A This is a demonstrative flowchart of an exemplary method 800 for operating a wind turbine converter. Method 800 can be applied online, wherein it is web-based and accepts online stored data as input. Method 800 can also be implemented directly on a computing device, wherein the computing device receives and transmits data wirelessly or via wired communication. Method 800 includes predicting the RUL of the wind turbine converter 801 and controlling the operation of the wind turbine converter 850 by adjusting the operating variables of the wind turbine converter based on a comparison of the predicted RUL with a target RUL.

[0111] In an exemplary embodiment, method 801 includes providing 802 a forecast wind dataset. In forecast module 124, a forecast function is used to generate the forecast dataset to extend the time series of operational data and event signals, resulting in a forecast dataset with a longer total time period T. win T win This could be within a monthly or yearly range. The generated forecast wind dataset is in T... win Each snapshot includes low-sampling-rate operational data (for...) (the time interval) and the event signal within that time interval.

[0112] In an exemplary embodiment, method 801 further includes using a hybrid lifetime estimation module 102 to calculate the lifetime consumption of snapshot j of 804. An alternative module in the hybrid lifetime estimation module 102 is used to feed each snapshot of the predictor's low-sampling-rate operation data into the normal OC lifetime estimation module 108 to provide a converter lifetime estimate during normal OC at each snapshot j. The maximum junction temperature result obtained from the normal OC lifetime estimation module 108 is then used. Possible converter event signals and low-sampling-rate operating data are then fed into the fast transient OC lifetime estimation module 110 within the hybrid lifetime estimation module 102 for converter lifetime estimation during fast transient OC. A physics-based model or an alternative model can be built in module 110, as long as it can estimate lifetime consumption during fast transient OC with sufficient accuracy and a relatively light computational burden using the low-sampling-rate operating dataset and event signals. The estimated lifetime consumption during normal OC and the estimated lifetime consumption during fast transient OC can be combined in the hybrid lifetime estimation module 102.

[0113] In an exemplary embodiment, method 801 further includes connecting index j with... Compare 806. If If the value of the snapshot is j, then j = j + 1, and the next snapshot j+1 is processed, where the calculation 804 of the hybrid lifetime estimation module 102 is repeated for snapshot j+1. Otherwise, all snapshots in the dataset have been processed, and method 801 proceeds to calculate the predicted RUL of 808.

[0114] In the exemplary embodiment, the predicted RUL of 808 is calculated using the following formula based on the linear fatigue accumulation assumption.

[0115] .

[0116] Here, It is the total duration T win The total lifetime consumption within the period. Different fatigue accumulation assumptions (such as the nonlinear fatigue accumulation assumption) can also be used to calculate the predicted RUL.

[0117] In an exemplary embodiment, method 800 may further include method 850 for controlling the effective lifespan performance of the wind turbine converter. Figure 8B A sample flowchart of method 850 is shown. Method 850 is used to control the performance of a wind turbine converter by adjusting its operating variables based on the predicted RUL predicted by method 801. Method 850 can be implemented using the Effective Lifetime Performance Control Module 106.

[0118] In an exemplary embodiment, method 850 includes calculating target RUL 852 using the following formula.

[0119] ,

[0120] in Indicates the target lifespan, and can be provided by the OEM. This represents the current time (or time of interest) obtained from the converter lifetime monitoring module 122. The calculated target... Then compared with the RUL predicted by method 801 Compare. If If the converter's lifespan is predicted to be unsatisfactory (e.g., due to high converter stress or usage), then method 850 initiates the use of the effective lifespan extension control module 112 to reduce the future lifespan consumption of 854. If If the converter's lifespan is predicted to meet the target (e.g., lower converter stress or usage), and there is a possibility of increasing converter power and increasing future lifespan consumption, while maintaining all other converter stage and turbine stage operating constraints. Accordingly, method 850 includes using an effective performance enhancement control module 114 in this case to enhance the performance of converter 856.

[0121] In the exemplary embodiment, to reduce the future lifespan consumption of 854, the effective lifespan extension control module 112 is used to meet the target lifespan. First, and current lifetime consumption LC(t) k The time period T is used to express the time period T using the equation shown below. win The total lifespan consumption within the target is calculated as follows: This is represented by the target lifetime consumed in each time interval Δt. The lifetime consumption vector of the number of elements Vector sum:

[0122] .

[0123] Then calculate the time period T. win The target lifetime consumption for each element (i.e., each snapshot of the dataset).

[0124] In the exemplary embodiment, for each snapshot of a low-sampling-rate dataset with a time interval Δt, the operational data estimation function is used to calculate the operational data input, given in each snapshot. And the optimal set of operating constraints at the converter and turbine stages. To achieve the target lifetime consumption target, the impact on converter performance should be minimized. The optimal values ​​of relevant operating variables (including, for example, converter active power, reactive power, voltage, and coolant temperature) are determined based on the target lifetime consumption and converter performance constraints. This problem is addressed by using a static constraint optimizer (expressed by the following equation) for each snapshot or a dynamic constraint optimizer (e.g., model predictive control):

[0125] ,

[0126] in ,and c1 is the cost-related penalty coefficient for deviations from the target lifetime, and c2 is the cost-related penalty coefficient for performance violations.

[0127] The above function ensures that the effective lifespan extension control module considers operational lifespan consumption and performance constraints. This is done when estimating each time interval. Optimal lifespan consumption In this process, the hybrid lifetime estimation module 102 can be used again to calculate the lifetime consumption of 804. It is a function of operating parameters and design parameters. In this optimization context, xc represents the controllable operating parameters related to the converter's lifetime, which are also the decision variables of the optimization problem. xnc represents the uncontrollable parameters affecting the converter's lifetime and performance, including design parameters and uncontrollable operating parameters. Uncontrollable operating parameters are caused by, for example, device failure or malfunction and are not adjustable. Similarly, Based on a similar alternative model used to transform operating parameters into converter performance constraints, any standard nonlinear least-squares algorithm (such as Gauss-Newton, Levenberg-Marquardt, conjugate gradient search, and Nelder-Mead search) can be used to derive the optimal operating variables using all defined objective functions, constraints, and decision variables. The optimization problem can also be solved using moving horizon techniques with standard model predictive control packages.

[0128] The optimization framework described above can be used for lifespan extension and performance improvement control, as described below.

[0129] In the exemplary embodiment, in improving the performance of the 856, the effective performance enhancement control module 114 is used to increase the power output while still ensuring that the converter meets the target lifespan. First, and current lifetime consumption LC(t) k Used to calculate the time period T win The total lifespan of the target within the time limit.

[0130] .

[0131] Then calculate the time period T. win The target lifetime consumption for each element (i.e., each snapshot of the dataset).

[0132] In an exemplary embodiment, for each snapshot of a minute-level dataset, the operational data estimation function is used to calculate the operational data input of the device (including generators, cables, and transformers) given in each snapshot. The optimal set of other operational objectives (such as power enhancement objectives) and operational constraints at the converter and turbine stages. This solution can be obtained from static constraint optimizers or dynamic constraint optimizers (such as the model predictive control described in Reducing Future Lifetime Consumption of 854).

[0133] Method 850 may further include online updates 858 to the converter control commands based on optimal operating variables obtained by reducing future lifetime consumption 854 through the effective lifetime extension control module 112 or improving performance 856 through the effective performance enhancement control module 114. A control feedback loop can be implemented by adjusting the wind turbine converter operating variables using the effective lifetime performance control module 106. In one example, if more semiconductor devices are included, the optimal operating variables are determined based on the semiconductor devices that consume the majority of their lifetime among the included devices.

[0134] The Effective Life Performance Control Module 106 seamlessly integrates a "Life Extension Control Mode" (if stress or usage is relatively high) and a "Performance Enhancement Control Mode" (if stress or usage is relatively low). The Effective Life Performance Control Module 106 can also switch between the two modes in real time based on consumer needs and / or predicted RUL results.

[0135] In some embodiments, system 100 integrates with the lifespan models of other wind turbines or converters in the group and facilitates power balancing among the wind converters to achieve homogeneous aging of the group based on the target lifespan of the entire group.

[0136] The modules described herein can be implemented on any suitable computing device and the software implemented therein. Figure 9 This is a block diagram of an exemplary computing device 900. In an exemplary embodiment, the computing device 900 includes a user interface 904 that receives at least one input from a user. The user interface 904 may include a keyboard 906 that enables the user to input coherent information. The user interface 904 may also include, for example, a pointing device, a mouse, a stylus, a touch-sensitive panel (e.g., a touchpad and a touchscreen), a gyroscope, an accelerometer, a position detector, and / or an audio input interface (e.g., including a microphone).

[0137] Furthermore, in an exemplary embodiment, the computing device 900 includes a presentation interface 907 that presents information (e.g., input events and / or verification results) to a user. The presentation interface 907 may also include a display adapter 908 coupled to at least one display device 910. More specifically, in an exemplary embodiment, the display device 910 may be a visual display device, such as a cathode ray tube (CRT), liquid crystal display (LCD), organic LED (OLED) display, and / or an "electronic ink" display. Alternatively, the presentation interface 907 may include an audio output device (e.g., an audio adapter and / or a speaker) and / or a printer.

[0138] The computing device 900 also includes a processor 914 and a memory device 918. The processor 914 is coupled to the user interface 904, the presentation interface 907, and the memory device 918 via a system bus 920. In an exemplary embodiment, the processor 914 communicates with the user, for example, by prompting the user via the presentation interface 907 and / or by receiving user input via the user interface 904. The term "processor" generally refers to any programmable system, including systems and microcontrollers, simplified instruction set circuits (RISC), application-specific integrated circuits (ASICs), programmable logic circuits (PLCs), and any other circuitry or processor capable of performing the functions described herein. The examples above are merely exemplary and are therefore not intended to limit the definition and / or meaning of the term "processor" in any way.

[0139] In an exemplary embodiment, memory device 918 includes one or more means that enable the storage and retrieval of information such as executable instructions and / or other data. Furthermore, memory device 918 includes one or more computer-readable media, such as, but not limited to, dynamic random access memory (DRAM), static random access memory (SRAM), solid-state disks, and / or hard disks. In an exemplary embodiment, memory device 918 stores, but is not limited to, application source code, application object code, configuration data, additional input events, application state, assertion statements, verification results, and / or any other type of data. In an exemplary embodiment, computing device 900 may also include a communication interface 930 coupled to processor 914 via system bus 920. Furthermore, communication interface 930 is communicatively coupled to data acquisition means.

[0140] In an exemplary embodiment, the processor 914 can be programmed by encoding the operation using one or more executable instructions and by providing the executable instructions in a memory device 918. In an exemplary embodiment, the processor 914 is programmed to select a plurality of measurements received from a data acquisition device or a wind turbine.

[0141] In operation, a computer executes computer-executable instructions embodied in one or more computer-executable components stored on one or more computer-readable media to implement aspects of the invention described and / or illustrated herein. The order of operation or execution of the operations in the embodiments of the invention illustrated and described herein is not essential unless otherwise specified. That is, operations may be performed in any order unless otherwise specified, and embodiments of the invention may include additional operations or fewer operations than those disclosed herein. For example, it is contemplated that a particular operation be run or performed before, after, or simultaneously with another operation, which is within the scope of the invention.

[0142] At least one technical effect of the system and method described herein includes: (a) real-time monitoring of the performance of the wind turbine converter; (b) using low-sampling-rate operating data to reduce computational burden and increase computational speed; (c) real-time control of the performance of the wind turbine converter to extend its lifespan or maximize its performance while still meeting the target lifespan; and (d) using existing available operating data and eliminating the need for loss / thermal parameters or additional hardware.

[0143] The foregoing has described in detail exemplary embodiments of systems and methods for operating wind turbine converters. The systems and methods are not limited to the specific embodiments described herein, but can be implemented independently of and using the components and / or methods of the system, regardless of other components and / or operations described herein. Furthermore, the components and / or operations may also be defined in or used in conjunction with other systems, methods, and / or apparatuses, and are not limited to implementation using only the systems described herein.

[0144] While certain features of the various embodiments of the invention may be shown in some figures but not in others, this is merely for convenience. According to the principles of the invention, any feature in the figures may be referenced and / or claimed in conjunction with any feature in any other figure.

[0145] This written description uses examples including the best mode to disclose the invention and also enables those skilled in the art to practice the invention, including making and using any device or system, and performing any combination method. The patentable scope of the invention is defined by the claims and may include other examples that may 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 are exactly the same as the wording of the claims, or if they include equivalent structural elements that have a non-substantially different wording from the claims.

Claims

1. A method for operating a wind turbine converter, the method comprising: Receive multiple forecast datasets, wherein the multiple forecast datasets include event signals of the wind turbine during fast transient operating conditions (OC) and operating data of the wind turbine with a lower sampling rate; The normal OC lifetime estimation module is used to estimate the converter lifetime consumption during normal OC based on the operating data; The fast transient OC lifetime estimation module is used to estimate the converter lifetime consumption during the fast transient OC based on the multiple forecast datasets; The total converter lifetime consumption of the wind turbine is calculated by using a hybrid lifetime estimation module to combine the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period. The Remaining Useful Life (RUL) prediction module is used to predict the RUL of the wind turbine based on the total converter lifespan consumption; Compare the predicted RUL with the target RUL of the wind turbine converter; and The effective lifespan performance control module adjusts the operation of the wind turbine converter by adjusting its operating variables and based on the comparison. The normal OC lifetime estimation module is used to estimate the maximum junction temperature of the wind turbine during the normal OC period, wherein: Estimating converter lifetime consumption during a fast transient OC further includes using a fast transient OC lifetime estimation module to estimate the converter lifetime consumption during the fast transient OC based on the event signal of the wind turbine converter during the normal OC, the operating data, and the estimated maximum junction temperature. as well as The predicted RUL of the wind turbine based on the total converter lifetime consumption is determined based on the allocation of higher converter lifetime consumption during periods of lower average power and higher standard deviation of power, compared to periods of higher average power and lower standard deviation of power.

2. The method as described in claim 1, wherein, Adjusting the operation of the wind power converter further includes: When the predicted RUL is less than the target RUL, the effective lifespan extension control module is used to adjust the operating variables of the wind turbine converter to reduce future lifespan consumption, ensuring that the wind turbine converter meets the target converter lifespan; and When the predicted RUL is greater than the target RUL, the effective performance enhancement control module is used to adjust the operating variables to improve performance and increase the future lifetime consumption, while still meeting the target converter lifetime.

3. The method as described in claim 2, wherein, Adjusting the operation of the wind power converter further includes switching between the effective lifespan extension control module and the effective performance enhancement control module.

4. The method of claim 1, wherein, The target RUL is based on the target converter lifetime of the wind turbine converter group including the wind turbine converter.

5. The method of claim 1, wherein, The operational data includes statistics associated with the operating parameters of the wind turbine converter.

6. The method of claim 5, wherein, The statistics include at least one of the following for each of the operating parameters: mean, standard deviation, kurtosis, skewness, pattern, median, quartile, minimum, maximum, range, and interquartile range.

7. The method of claim 1, wherein, The normal OC lifetime estimation module includes an alternative model, which is trained using multiple training datasets. The alternative model has first operational data with a low sampling rate as input and lifetime consumption estimated using the normal OC physics-based lifetime estimation model as output.

8. The method of claim 1, wherein, The event signals include data indicating sudden changes in wind speed and converter trip signals.

9. The method of claim 1, wherein, Calculating the total converter lifetime consumption further includes calculating the total converter lifetime consumption by calculating a weighted sum of the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

10. A method of operating a wind turbine converter, the method comprising: Receive multiple datasets, wherein the multiple datasets include event signals of the wind turbine during fast transient operating conditions (OC) and operating data of the wind turbine with a lower sampling rate; The normal OC life estimation module is used to estimate the converter life consumption of the wind turbine during normal OC based on the operating data; The fast transient OC lifetime estimation module is used to estimate the converter lifetime consumption of the wind turbine during the fast transient OC based on the multiple datasets; The total converter lifetime consumption of the wind turbine is calculated by using a hybrid lifetime estimation module to combine the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period. as well as The operation of the wind turbine is monitored remotely and in real time based on the calculated total converter lifespan consumption. The normal OC lifetime estimation module is used to estimate the maximum junction temperature of the wind turbine during the normal OC period, wherein: Estimating converter lifetime consumption during a fast transient OC further includes using the fast transient OC lifetime estimation module to estimate the converter lifetime consumption during the fast transient OC based on the event signal of the wind turbine converter during the normal OC period, the operating data, and the estimated maximum junction temperature; and The estimated maximum junction temperature of the wind turbine is determined based on the higher maximum junction temperature during periods of higher converter power, compared to the period of increase in the standard deviation of the wind turbine power.

11. The method of claim 10, wherein, The operational data includes statistics associated with the operating parameters of the wind turbine converter.

12. The method of claim 10, wherein, The normal OC lifetime estimation module includes an alternative model trained using multiple training datasets. The alternative model has first operating data with a low sampling rate as input and converter lifetime consumption during the normal OC period estimated by the normal OC physics-based lifetime estimation model as output.

13. The method of claim 10, wherein, Calculating the total converter lifetime consumption further includes calculating the total converter lifetime consumption by calculating a weighted sum of the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period.

14. A real-time remote operation monitoring and control system for a wind power converter, comprising: Wind power converter; The Remaining Useful Life (RUL) prediction module includes: A hybrid lifetime estimation module configured to receive multiple datasets, wherein the multiple datasets include event signals for the wind turbine during fast transient operating conditions (OC) and operating data for the wind turbine with a lower sampling rate, the hybrid lifetime estimation module comprising: A normal OC lifetime estimation module, configured to estimate converter lifetime consumption during normal OC based on the operating data; and A fast transient OC lifetime estimation module is configured to estimate converter lifetime consumption during the fast transient OC based on the plurality of datasets; The hybrid lifetime estimation module is further configured to calculate the total converter lifetime consumption of the wind power converter by combining the estimated converter lifetime consumption during the normal OC period and the estimated converter lifetime consumption during the fast transient OC period. The RUL prediction module is configured to predict the RUL for the wind converter based on the total converter lifetime consumption calculated using the hybrid lifetime estimation module; and An effective lifetime performance control module is configured to adjust the operation of the wind turbine converter by adjusting its operating variables based on a comparison between a predicted range-limited lifetime (RUL) and a target RUL for the wind turbine converter. The normal OC lifetime estimation module is further configured to estimate the maximum junction temperature of the wind turbine during the normal OC period; and The fast transient OC lifetime estimation module is configured to estimate the converter lifetime consumption during the fast transient OC based on the event signal, the operating data, and the maximum junction temperature of the wind turbine during the normal OC period, as estimated by the normal OC lifetime estimation module. In addition to predicting the RUL of the wind power converter based on the presence of higher converter lifetime consumption during periods of lower average power and higher standard deviation of power, the RUL prediction module also predicts the RUL of the wind power converter based on the presence of higher converter lifetime consumption during periods of lower average power and higher standard deviation of power.

15. The system of claim 14, wherein, The normal OC lifetime estimation module further includes an alternative model, wherein the alternative model is trained using multiple training datasets, having first operating data with a low sampling rate as input and converter lifetime consumption estimated using the normal OC physics-based lifetime estimation model as output.

16. The system of claim 14, wherein the effective lifespan performance control module further comprises: An effective life extension control module is configured to adjust the operating variables of the wind turbine converter when the predicted RUL is less than the target RUL, so as to reduce future life consumption and make the wind turbine converter meet the target converter life. as well as An effective performance enhancement control module is configured to adjust the operating variables when the predicted RUL is greater than the target RUL, in order to improve performance and increase the future lifetime consumption, while still meeting the target converter lifetime.

17. The system of claim 16, wherein, The effective life performance control module is configured to switch between the effective life extension control module and the effective performance enhancement control module.

Images