Wind turbine generator probabilistic fatigue life prediction method, device, equipment and medium

Abstract

The invention relates to the technical field of wind turbine generator operation and maintenance, in particular to a wind turbine generator probability fatigue life prediction method and device, equipment and a medium. The method comprises the following steps: constructing a digital twinborn model of a wind turbine generator; based on operation load data, collected in real time, of the wind turbine generator, a digital twinborn model is adopted to determine corresponding stress data; and inputting the stress data into a pre-trained probabilistic fatigue life prediction model to obtain a current probabilistic fatigue life result of the wind turbine generator. According to the method, the current stress data of the wind turbine generator are calculated by constructing the digital twinborn model of the wind turbine generator, and the current probability fatigue life result of the wind turbine generator is obtained based on the pre-trained probability fatigue life prediction model. Therefore, according to the method, through real-time interaction of the physical unit and the virtual model, accurate prediction of the fatigue damage of the wind turbine generator is achieved, that is, the method can reflect the operation state and the fatigue damage condition of the wind turbine generator in time, and the real-time performance is high.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine operation and maintenance, and in particular to a method, device, equipment and medium for predicting the probabilistic fatigue life of a wind turbine. Background Art

[0002] During long-term operation, wind turbines are subject to complex environmental loads (such as wind speed, direction, and temperature) and operating conditions. Key components (such as blades, main shafts, gearboxes, and bearings) can suffer fatigue damage, severely impacting their service life and operating efficiency. Existing probabilistic fatigue life prediction methods typically rely on offline data analysis, which struggles to accurately reflect the current health of the equipment. This leads to delayed or overly conservative maintenance strategies and increases operational costs. Summary of the Invention

[0003] In view of this, the present invention provides a method, device, equipment and medium for probabilistic fatigue life prediction of a wind turbine generator set to solve the problem that the probabilistic fatigue life prediction method in the prior art is difficult to reflect the current health status of the equipment.

[0004] In the first aspect, the present invention provides a method for predicting the probabilistic fatigue life of a wind turbine, the method comprising: constructing a digital twin model of the wind turbine, the digital twin model being used to simulate and obtain stress data of the wind turbine; based on the operating load data of the wind turbine collected in real time, using the digital twin model to determine the corresponding stress data; inputting the stress data into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0005] In this method, a digital twin model of the wind turbine is constructed to calculate the wind turbine's current stress data. Based on a pre-trained probabilistic fatigue life prediction model, the wind turbine's current probabilistic fatigue life is obtained. This method, through real-time interaction between the physical turbine and the virtual model, enables accurate prediction of wind turbine fatigue damage. This method can promptly reflect the wind turbine's operating status and fatigue damage, achieving high real-time performance.

[0006] In an optional embodiment, a digital twin model of a wind turbine is constructed, including: obtaining the geometric structure, material parameters and historical operating data of the wind turbine to construct a finite element model of the wind turbine; performing multiple simulations of stress data based on the finite element model to obtain multiple simulation results; and using the Kriging method to construct a proxy model for the multiple simulation results, and the constructed proxy model serves as the digital twin model of the wind turbine.

[0007] In the present invention, a digital twin model is obtained by first constructing a finite element model and then using the Kriging method to construct a proxy model based on multiple simulation results of the finite element model. The finite element model can simulate the complex structure and physical behavior of the wind turbine with high precision, and carefully analyze its mechanical responses such as stress and strain under various operating conditions. Based on the finite element model, the proxy model efficiently fits and approximates the simulation results in a data-driven manner. The combination of the two can accurately reproduce the actual operating status of the wind turbine in digital space, thereby enabling the constructed digital twin model to accurately and quickly determine stress data under different operating loads.

[0008] In an optional embodiment, before inputting the stress data into the pre-trained probabilistic fatigue life prediction model, the method also includes: performing high-frequency fatigue life tests on the materials of the wind turbine generator set under different stress amplitudes to obtain fatigue lives corresponding to different stress amplitudes; using the mean rank estimation method to sort the logarithmic values of the fatigue life and calculate the failure probability of the logarithmic fatigue life; based on the correspondence between the normal distribution quantiles of the failure probability and the logarithmic fatigue life, determining the logarithmic fatigue life corresponding to different stress amplitudes under different failure probabilities or survival rates to obtain an SN curve; based on the SN curve and combined with linear cumulative damage, establishing a probabilistic fatigue life calculation model, the probabilistic fatigue life calculation model is used to provide training data for training the probabilistic fatigue life prediction model.

[0009] In the present invention, a probabilistic fatigue life calculation model is established through high-frequency fatigue life testing, failure probability calculation, normal distribution quantile calculation, and SN curve determination. As a result, the training data obtained through the model can provide life prediction results that are more in line with actual conditions, providing an accurate data basis for the training of the neural network model.

[0010] In an optional embodiment, the training data includes probabilistic fatigue life corresponding to different stress data. Before inputting the stress data into the pre-trained probabilistic fatigue life prediction model, the method also includes: using the training data to train the back propagation neural network model to obtain an initial probabilistic fatigue life prediction model; and correcting the initial probabilistic fatigue life prediction model based on the historical operating data of the wind turbine to obtain a pre-trained probabilistic fatigue life prediction model.

[0011] In an optional embodiment, the stress data is input into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine, including: using the rain flow counting method to extract stress characteristic data of the stress data, the stress characteristic data including the amplitude, average value and number of cycles of the cyclic stress; inputting the stress characteristic data into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0012] In the present invention, stress characteristic data is extracted through the rain flow counting method, which simplifies complex data into key parameters such as the number of cycles, amplitude, and average value. This can more clearly present the fatigue load conditions experienced by the material or structure, facilitate the model to use these parameters for fatigue life prediction and analysis, and improve the prediction accuracy and efficiency.

[0013] In an optional embodiment, the method also includes: comparing the current probabilistic fatigue life result of the wind turbine with a preset safety threshold to obtain a comparison result; determining whether to generate a warning signal and an optimized operation method of the wind turbine based on the comparison result; controlling the operation of the wind turbine based on the optimized operation method, and the optimized operation method includes adjusting the operating parameters of the wind turbine.

[0014] In the present invention, the operation strategy of the wind turbine is dynamically adjusted according to the fatigue life prediction results, so as to extend the equipment life, reduce unplanned shutdowns caused by component fatigue failure, and improve the economic benefits of the wind farm.

[0015] In an optional embodiment, the method also includes: updating the digital twin model based on the monitoring data and probabilistic fatigue life results of the wind turbine generator set; visually displaying the probabilistic fatigue life results, the probabilistic fatigue life results including a remaining life distribution diagram, a cumulative damage evolution curve and a real-time load distribution diagram.

[0016] In the present invention, by updating the digital twin model, the adaptability of the model to the operating conditions of the wind turbine generator set is improved.

[0017] In the second aspect, the present invention provides a probabilistic fatigue life prediction device for a wind turbine, the device comprising: a model construction module for constructing a digital twin model of the wind turbine, the digital twin model being used to simulate and obtain stress data of the wind turbine; a stress determination module for determining corresponding stress data based on the operating load data of the wind turbine collected in real time using the digital twin model; and a prediction module for inputting stress data into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0018] In a third aspect, the present invention provides a computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the probabilistic fatigue life prediction method for wind turbines according to the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0019] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the wind turbine probabilistic fatigue life prediction method of the first aspect or any corresponding embodiment thereof.

[0020] In a fifth aspect, the present invention provides a computer program product comprising computer instructions for causing a computer to execute the method for predicting the probabilistic fatigue life of a wind turbine generator system according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 is a flow chart of a method for predicting the probabilistic fatigue life of a wind turbine generator system according to an embodiment of the present invention;

[0023] Figure 2 is a flow chart of another method for predicting the probabilistic fatigue life of a wind turbine generator system according to an embodiment of the present invention;

[0024] Figure 3 2 is a schematic diagram of a probabilistic fatigue life back propagation neural network model according to an embodiment of the present invention;

[0025] Figure 4 is a graph showing different survival rates SN curves of a metal material according to an embodiment of the present invention;

[0026] Figure 5 is a complete probabilistic fatigue life curve of a metal material according to an embodiment of the present invention;

[0027] Figure 6 is a schematic diagram of a digital twin architecture according to an embodiment of the present invention;

[0028] Figure 7 is a structural block diagram of a device for predicting probabilistic fatigue life of a wind turbine generator system according to an embodiment of the present invention;

[0029] Figure 8 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of the present invention.

[0031] According to an embodiment of the present invention, an embodiment of a method for probabilistic fatigue life prediction of a wind turbine is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0032] In this embodiment, a method for predicting the probabilistic fatigue life of a wind turbine is provided, which can be used in electronic devices such as computers, mobile phones, tablet computers, etc. Figure 1 FIG. 1 is a flow chart of a method for predicting the probabilistic fatigue life of a wind turbine according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0033] Step S101, construct a digital twin model of the wind turbine, and the digital twin model is used to simulate and obtain stress data of the wind turbine. Specifically, the digital twin model constructed in this embodiment integrates multi-dimensional information such as geometric information, physical characteristics, material properties, operating status, historical data, etc. of the physical entity, i.e., the wind turbine. Specifically, the digital twin model not only contains the precise three-dimensional model of each component of the wind turbine, but also includes the mechanical performance parameters of the material, environmental data such as wind speed, wind direction, temperature during operation, and state data such as stress, strain, vibration, etc. of the components. By integrating this information into a unified model, the various characteristics and behaviors of the physical entity can be described comprehensively and accurately.

[0034] In practical applications, a high-fidelity digital twin model of key wind turbine components, such as blades, main shafts, and gearboxes, can be constructed based on the actual operation and maintenance requirements of wind turbines. Furthermore, the constructed digital twin model can calculate wind turbine stress data based on real-time load data and finite element analysis.

[0035] In step S102, based on the real-time collected operational load data of the wind turbine, the corresponding stress data is determined using a digital twin model. Specifically, corresponding monitoring data can be acquired through a sensor network installed in the wind turbine, such as strain sensors, acceleration sensors, and temperature sensors. The operational load of the wind turbine can then be determined through relevant data analysis. For example, during the operation of a wind turbine, components will experience acceleration due to the effects of wind, turbine vibration, and other dynamic forces. Acceleration sensors are installed on the rotor blades. When the blades vibrate under aerodynamic loads, the acceleration sensors can measure the vibration acceleration of the blades and thereby analyze the dynamic loads borne by the blades. For another example, in a gearbox, a temperature sensor can monitor the temperature of the lubricating oil. When the gearbox is subjected to a large load, the oil temperature will rise. By analyzing the changes in oil temperature and other relevant parameters, the load borne by the gearbox can be determined.

[0036] After determining the wind turbine's operating load data, real-time performance calculations are performed using the constructed digital twin model to obtain stress data. If the acquired load data is load time series data, the corresponding stress data is stress time series data. Furthermore, the acquired load data can be collected through a high-speed data acquisition system connected to each sensor. After acquisition, filtering algorithms can be used to remove noise signals to ensure data authenticity and accuracy.

[0037] Step S103: Input the stress data into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine. Specifically, the pre-trained probabilistic fatigue life prediction model characterizes the relationship between stress and probabilistic fatigue life. For example, a network model, such as a neural network model, can be trained using the acquired historical stress data and probabilistic fatigue life as training data to obtain the pre-trained probabilistic fatigue life prediction model. After obtaining the real-time stress data of the wind turbine through the above steps, it is input into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0038] The probabilistic fatigue life prediction method for wind turbines provided by the present invention constructs a digital twin model of the wind turbine to calculate the wind turbine's current stress data. Based on a pre-trained probabilistic fatigue life prediction model, the method obtains the wind turbine's current probabilistic fatigue life results. This method achieves accurate prediction of wind turbine fatigue damage through real-time interaction between the physical turbine and the virtual model. Specifically, the method can promptly reflect the wind turbine's operating status and fatigue damage, achieving high real-time performance.

[0039] In this embodiment, a method for predicting the probabilistic fatigue life of a wind turbine is provided, which includes the following steps:

[0040] Step S201: construct a digital twin model of the wind turbine generator set. The digital twin model is used to simulate and obtain stress data of the wind turbine generator set.

[0041] Specifically, the above step S201 includes:

[0042] Step S2011: Obtain the wind turbine's geometric structure, material parameters, and historical operating data to construct a finite element model of the wind turbine. The blades, main shaft, and gearbox are key components of the wind turbine, and this embodiment primarily performs finite element modeling on these structures. In other embodiments, the entire wind turbine can also be modeled, and this embodiment is not limited to this.

[0043] Specifically, when establishing the finite element model of key components, the following steps are mainly included:

[0044] 1. Geometric modeling. Using 3D modeling software, perform geometric modeling according to the component's structural design file (i.e., the acquired geometric structure data). Accurately depict the dimensions, shape, and connection relationships of the blades, main shaft, and gearbox, and obtain the geometric models of each component.

[0045] 2. Meshing. Import the geometric model into the finite element software and perform finite element meshing on each component, discretizing the continuous geometric model into a finite number of elements and nodes. Simultaneously, the mesh is optimized and processed to ensure that the mesh quality meets the calculation requirements, such as avoiding deformed elements and ensuring consistent element size. Finite element calculation models of each component are then generated.

[0046] 3. Set loads and constraints. Based on the actual operating conditions of the wind turbine (i.e., historical operating data), determine the loads acting on the blades, main shaft, and gearbox, such as aerodynamic loads, gravity, and inertia. Then, set the corresponding loads for the finite element calculation model in the finite element software. At the same time, set reasonable constraints based on the installation and connection methods of the components to simulate the boundary conditions of the components in actual operation.

[0047] 4. Material property definition. Based on the obtained material parameters, determine the materials used for the blades, main shaft, and gearbox. Define the material's mechanical properties, such as elastic modulus, Poisson's ratio, density, and yield strength, in the finite element model to accurately calculate the component's response under load.

[0048] 5. Model Verification and Calibration. Perform preliminary calculations on the established finite element model and compare and verify it with actual test data or known theoretical results. If any deviations are found, calibrate and correct the model by adjusting model parameters or meshing until the calculated results are consistent with the actual situation or within an acceptable error range.

[0049] Step S2012 involves performing multiple simulations of stress data based on the finite element model to obtain multiple simulation results. Specifically, after establishing the finite element model, boundary conditions and operating loads can be set in the finite element software, and the range of simulation parameters such as the rotor speed and blade pitch angle can be determined. The stress data is then calculated using the solver and solution algorithm in the finite element software. During the simulation, different parameters can be set based on the different operating conditions of the wind turbine, and multiple simulations can be performed to obtain multiple stress data sets.

[0050] In step S2013, a proxy model is constructed based on multiple simulation results using the Kriging method, and the constructed proxy model is used as the digital twin model of the wind turbine. Kriging is a statistical interpolation method that can construct an approximate model based on limited sample data (here, the finite element simulation results) to replace the complex finite element model for calculation. This is done to improve computing efficiency and ensure the real-time performance of the digital twin model. Because finite element calculations are usually time-consuming, in practical applications, it is difficult to meet real-time requirements if finite element calculations are performed every time, while proxy models can quickly give approximate results.

[0051] Specifically, when using the Kriging method to construct a proxy model, the acquired sample data (including operating loads and corresponding stress data) can be input into the selected Kriging model and trained using the estimated hyperparameters. During the training process, the model calculates the weight of each data point based on the spatial correlation between the data points, thereby constructing a proxy model that can describe the relationship between the input parameters and the output stress data. The constructed proxy model is a high-fidelity digital twin model. In this embodiment, high-fidelity digital twin models of the blades, main shaft, and gearbox were established respectively.

[0052] Step S202: Based on the real-time collected operating load data of the wind turbine, the corresponding stress data is determined using the digital twin model. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0053] In step S203, the stress data is input into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine. In order for the pre-trained probabilistic fatigue life prediction model to be able to predict probabilistic fatigue life, it is necessary to use stress data and corresponding probabilistic fatigue life and other training data for model training. Before model training, the training data must be obtained.

[0054] In an optional embodiment, the following steps are used to obtain training data:

[0055] Step S231, conduct high-frequency fatigue life tests on the materials of the wind turbine under different stress amplitudes to obtain fatigue lives corresponding to different stress amplitudes; specifically, in a wind turbine, different components may use different materials, and the fatigue lives of different materials under the same stress amplitude may also be different. Therefore, fatigue life tests are required for different materials used in the manufacturing process of different components of the wind turbine. For example, if the tower of a wind turbine is made of Q345E steel, fatigue tests are performed on the material Q345E. Among them, high-frequency fatigue life testing refers to the number of cycles before fatigue failure occurs under high-frequency cyclic loads. High-frequency fatigue testing can evaluate the fatigue performance of materials under high-frequency vibrations or alternating loads.

[0056] During testing, the material specimen to be tested is mounted on a fatigue testing machine. Cyclic loads are then applied to the specimen at each set stress amplitude, while the surface condition of the specimen is observed. When cracks appear on the surface, the corresponding number of cycles is recorded as the fatigue life of the material at that stress amplitude. Multiple specimens of the same material can be tested, and multiple tests can be performed for each stress amplitude. This yields fatigue life test results for different materials at different stress amplitudes.

[0057] Step S232: sort the logarithmic values of fatigue life using the mean rank estimation method and calculate the failure probability of the logarithmic fatigue life. Specifically, based on the logarithmic normal distribution assumption of fatigue life under the same stress amplitude, the logarithmic fatigue life obeys the normal distribution. Therefore, the fatigue life in the fatigue life test results is logarithmically calculated to obtain the logarithmic fatigue life. Then, sort the multiple logarithmic fatigue life data lg(N) under the same stress amplitude using the mean rank estimation method and calculate the failure probability of this batch of logarithmic fatigue life. Since fatigue failure occurred in each pair of stress and life data in the test results, the material samples all broke and failed, and the probability of failure was equal. The calculation formula is as follows:

[0058]

[0059] In the formula, i represents the sequence number after sorting, and n represents the total number of data.

[0060] Step S233, based on the correspondence between the normal distribution quantile of the failure probability and the logarithmic fatigue life, determine the logarithmic fatigue life corresponding to different stress amplitudes under different failure probabilities or survival rates, and obtain the SN curve; wherein, the normal distribution quantile, also called quantile, is a value that divides the probability distribution of the normal distribution into different parts. For the standard normal distribution (a normal distribution with a mean of 0 and a standard deviation of 1), if a probability value p is given, then the quantile u satisfies: from negative infinity to u, the cumulative distribution function value of the standard normal distribution is equal to p. In this embodiment, the normal distribution quantile corresponding to the failure probability is calculated using the following formula:

[0061]

[0062] Where, Is the failure probability, which can be understood as the cumulative distribution probability corresponding to the quantile, F -1 (·) is the inverse of the cumulative distribution function.

[0063] Specifically, the calculated normal distribution quantile represents the specific numerical point corresponding to this failure probability under the normal distribution. This quantile is used to characterize the discreteness, central tendency, and other characteristics of the logarithmic fatigue life data under the normal distribution. Therefore, by fitting different logarithmic fatigue life data under the same stress amplitude with the quantile using the least squares method, the slope and intercept obtained by fitting are the mean and standard deviation of different logarithmic fatigue life data under the same stress amplitude. The fitting result is expressed by the following formula:

[0064]

[0065] In the formula, s and Represent the standard deviation and mean of the logarithmic fatigue life data under the same stress amplitude.

[0066] The logarithmic fatigue life data corresponding to each stress amplitude are processed in the above manner to obtain the mean and standard deviation of the logarithmic fatigue life under different stress amplitudes. Then, a specific failure probability or survival rate (the difference between the survival rate of 1 and the failure probability) is selected and substituted into the fitting results of different stress amplitudes to obtain the logarithmic fatigue life corresponding to different stress amplitudes under a specific failure probability or survival rate. The least squares linear fitting is then performed on this. The slope and intercept obtained by the fitting are the characterization parameters of the SN curve. Thus, a material SN curve under a specific failure probability or survival rate is obtained. The specific process is expressed by the following formula:

[0067] σ m ×N=c

[0068] mlg(σ)+lg(N)=lg(c)

[0069]

[0070] Where m and c are material constants, and σ is the stress amplitude.

[0071] The selected failure probability or survival rate can be adjusted according to the actual situation, thereby obtaining the SN curve under any failure probability or survival rate. In addition, it should be noted that the above process can be used to process each material to obtain the corresponding material SN curve.

[0072] Step S234, based on the SN curve and combined with linear cumulative damage, a probabilistic fatigue life calculation model is established. The probabilistic fatigue life calculation model is used to provide training data for the probabilistic fatigue life prediction model training. Specifically, a probabilistic fatigue life calculation model is constructed by the SN curve and linear cumulative damage (such as Miner linear cumulative damage criterion), wherein the SN curve describes the correspondence between stress amplitude and logarithmic fatigue life, and in the actual operation process of the wind turbine, the components will be subjected to cyclic loads of various stress amplitudes. The Miner linear cumulative damage criterion can accumulate fatigue damage under different stress amplitudes, comprehensively consider various working conditions, and be more in line with actual operating conditions, so that the calculated probabilistic fatigue life is more accurate. Therefore, the probabilistic fatigue life calculation model constructed by the SN curve and linear cumulative damage can accumulate damage under different stress levels, that is, the model calculates the cumulative damage degree of the material or component after a series of load cycles. Based on this probabilistic fatigue life calculation model, training data containing stress amplitude, corresponding cumulative damage value and remaining life can be obtained.

[0073] Step S235 , using the training data to train the back propagation neural network model to obtain an initial probabilistic fatigue life prediction model.

[0074] Step S236 , calibrating the initial probabilistic fatigue life prediction model based on the historical operating data of the wind turbine generator system to obtain a pre-trained probabilistic fatigue life prediction model.

[0075] The back-propagation neural network model is trained using training data, enabling it to learn the relationship between stress characteristics and damage evolution. Furthermore, the initial probabilistic fatigue life prediction model obtained through training is calibrated using historical wind turbine operating data to ensure that the error in the final model prediction does not exceed 5%, thus confirming the accuracy of the final model prediction.

[0076] Specifically, the above step S203 includes:

[0077] Step S2031 : extracting stress characteristic data of the stress data using a rain flow counting method. The stress characteristic data includes the amplitude, average value, and number of cycles of the cyclic stress.

[0078] Step S2032: input the stress characteristic data into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0079] Specifically, based on the probabilistic fatigue life prediction model obtained through training in this embodiment, before the stress data obtained through the digital twin model is input into the model prediction, the stress characteristic data of the stress data is first extracted using the rain flow counting method. Among them, because the stress time series data obtained through the digital twin model is continuous and complex, it is difficult to effectively extract key information related to fatigue by directly inputting it into the model. The rain flow counting method simplifies complex data into key parameters such as the number of cycles, amplitude, and average value by identifying and counting load cycles, which can more clearly present the fatigue load conditions experienced by the material or structure, making it easier for the model to use these parameters for fatigue life prediction and analysis, thereby improving prediction accuracy and efficiency. In addition, the specific extraction process of the rain flow counting method can be implemented with reference to relevant technologies and will not be described here.

[0080] In this embodiment, a method for predicting the probabilistic fatigue life of a wind turbine is provided, which includes the following steps:

[0081] Step S301: construct a digital twin model of the wind turbine. The digital twin model is used to simulate and obtain stress data of the wind turbine. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0082] Step S302: Based on the real-time collected operating load data of the wind turbine, the corresponding stress data is determined using the digital twin model. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0083] Step S303: Input the stress data into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine. Figure 1 Step S103 of the illustrated embodiment will not be described in detail here.

[0084] Step S304: compare the current probabilistic fatigue life result of the wind turbine with a preset safety threshold to obtain a comparison result; determine whether to generate a warning signal and an optimized operation method for the wind turbine based on the comparison result; and control the operation of the wind turbine based on the optimized operation method, which includes adjusting the operating parameters of the wind turbine.

[0085] Specifically, after the current probabilistic fatigue life result is predicted, it is compared with a pre-set safety threshold. When the life of different key components in the wind turbine is obtained, it can be compared with the corresponding safety threshold. When the comparison result shows that the life is critical to the safety threshold (for example, the difference between the life and the safety threshold is less than the threshold), an early warning signal is issued and an optimized operation method for the wind turbine is generated. The optimized operation method specifically includes adjusting the operating parameters of the wind turbine according to the corresponding components, such as adjusting the speed and pitch angle, to slow down the fatigue evolution rate of high-damage components.

[0086] Step S305: Update the digital twin model based on the monitoring data and probabilistic fatigue life results of the wind turbine generator system; visualize the probabilistic fatigue life results, which include a remaining life distribution diagram, a cumulative damage evolution curve, and a real-time load distribution diagram.

[0087] Specifically, when updating the digital twin model, the elastic modulus, Poisson's ratio, yield strength, and other mechanical performance parameters of key component materials such as blades, main shafts, and gearboxes can be re-evaluated based on the new stress data and cumulative damage values. If the cumulative damage to a component causes a degradation in material performance, the material parameters in the digital twin model are adjusted accordingly to make them more realistic. Furthermore, when a component experiences geometric structural changes such as wear and deformation due to fatigue, the monitoring data is used to calculate the change in structural dimensions and update the component's geometric parameters in the digital twin model, such as blade thickness, main shaft diameter, and the shape and dimensions of the gearbox's internal components.

[0088] Furthermore, since digital twin models are derived from finite element models and proxy models, they can be further modified. For example, if new data indicates that the original finite element model inaccurately simulates certain areas, the mesh can be re-divided or the element type adjusted to improve the local accuracy of the model. For Kriging proxy models, the model parameters and weights are re-determined based on new simulation data and actual monitoring data, optimizing the proxy model's ability to approximate complex mechanical behaviors and ensuring real-time performance and accuracy.

[0089] In addition, the probabilistic fatigue life prediction results can also be visualized through the human-computer interaction interface, such as the remaining life distribution diagram of key components, the cumulative damage evolution curve and the real-time load distribution diagram, and the prediction results can be stored in the wind turbine operation and maintenance management system database.

[0090] As a specific application example of the embodiment of the present invention, Figure 2 As shown in Figure 2, the wind turbine probabilistic fatigue life prediction method is implemented using the following process:

[0091] Step S1: Based on the actual operation and maintenance management needs of the wind turbine, the geometric structure, material parameters and historical operation data of the key components of the wind turbine (such as blades, main shaft, gearbox, etc.) are collected to build a high-fidelity digital twin model of the wind turbine.

[0092] Specifically, step S1 includes:

[0093] In step S1.1, the stress time series data of the blades, main shaft and gearbox are simulated multiple times based on the finite element modeling technology, and the proxy model is constructed based on the multiple simulation results of the finite element model using the Kriging method to ensure the real-time performance of the digital twin. The constructed proxy model is the high-fidelity digital twin model. This process establishes high-fidelity digital twin models of the blades, main shaft and gearbox respectively, where the inputs of the finite element, proxy model and digital twin model are all load time series data, and the outputs are all stress time series data.

[0094] Step S1.2: Real-time collaborative simulation of the wind turbine overall model and each component model is achieved through dynamic updating of actual operation data.

[0095] Step S2: Real-time monitoring and processing of fatigue loads are carried out. Based on the sensor network installed on the key components of the wind turbine (such as strain sensors, acceleration sensors, temperature sensors, etc.), the operating load of the wind turbine is monitored in real time. In combination with the high-fidelity digital twin model in step S1, real-time performance calculation of the load time series data is performed to obtain stress time series data. The stress time series is processed using the rain flow counting method to extract the amplitude, average value and number of cycles of each cyclic stress, providing input for probabilistic fatigue life prediction.

[0096] Specifically, real-time load time series data can be obtained through a high-speed data acquisition system, and filtering algorithms can be used to remove noise signals to ensure the authenticity and accuracy of the data.

[0097] Step S3, using the back propagation neural network algorithm to train the probabilistic fatigue life model, the input of the model is the stress parameter, and the output is the cumulative damage value and the remaining life.

[0098] Specifically, if Figure 3 As shown, step S3 includes:

[0099] S3.1: Conduct high-frequency fatigue life tests under different stress amplitudes on the materials of different key components of the wind turbine, and establish a material fatigue life test database based on the test results. It is recommended to conduct at least 40 fatigue life tests, i.e. 40 material test specimens, specifically T fatigue life results (T ≥ 5) under no less than 4 different stress amplitudes.

[0100] In this embodiment, a high-frequency fatigue life test is performed on a certain metal material under a stress ratio R=0.1, and 40 sets of logarithmic fatigue life test data are obtained. The data are shown in Table 1 below:

[0101] Table 1

[0102]

[0103] S3.2: Based on the assumption of log-normal distribution of fatigue life N under the same stress amplitude, the logarithmic fatigue life obeys the normal distribution. Therefore, the data in the material fatigue life test database are first logarithmically processed. After logarithmic processing, the fatigue life lg(N) of different materials under the same stress amplitude obeys the normal distribution.

[0104] S3.3: Use the mean rank estimation method to sort the T logarithmic fatigue life data lg(N) under the same stress amplitude and calculate the failure probability of this batch of logarithmic fatigue life Since each pair of stress and life data in the material fatigue life test database has fatigue failure, the material specimens have all broken and failed, and the probability of failure is equal, the failure probability The calculation formula is as follows:

[0105]

[0106] S3.4: The normal distribution quantile u corresponding to the above failure probability i Calculate the failure probability during this process is the quantile u i The corresponding normal distribution cumulative distribution probability value is calculated as follows:

[0107]

[0108] in, The failure probability is the cumulative distribution probability corresponding to the quantile, F -1 (·) is the inverse of the cumulative distribution function.

[0109] For the test data in Table 1 above, the corresponding failure probability, survival probability, and quantiles are shown in Table 2 below:

[0110] Table 2

[0111]

[0112] S3.5: In order to accurately describe the failure probability of the logarithmic fatigue life under the same stress amplitude, it is necessary to calculate the mean and standard deviation of this batch of data. This process is based on the standard normal distribution quantile calculated in step S3.4, combined with the conversion relationship from normal distribution to standard normal distribution. The calculation formula is as follows. The least squares method is used to fit the different logarithmic fatigue life data and quantiles under the same stress amplitude. The slope and intercept obtained by fitting correspond to the mean and standard deviation of this batch of data.

[0113]

[0114] Among them, s and Represent the standard deviation and mean of the logarithmic fatigue life data under the same stress amplitude. The variables are obtained by least squares linear fitting. The fitting accuracy R 2 Greater than 0.98, otherwise the data needs to be preprocessed to filter out data with a high degree of deviation.

[0115] S3.6: Repeat steps S3.3-S3.5 to calculate the mean and standard deviation of the logarithmic fatigue life under 8 different stress amplitudes to accurately describe the probability distribution of this batch of data.

[0116] S3.7: Select a specific failure probability p f Or the survival rate p s , and calculate the logarithmic fatigue life value of different stress amplitudes under a specific survival rate or failure probability based on the known mean and standard deviation data. The conversion formula between failure probability and survival rate is as follows:

[0117] p s =1-p f

[0118] In this embodiment, the survival rate p is selected s The logarithmic fatigue life values of different stress amplitudes corresponding to 99.9% and 50% are shown in Table 3 below:

[0119] Table 3

[0120]

[0121]

[0122] S3.8: Perform logarithmic processing on the SN curve calculation method based on the Basquin formula. The calculation process is shown in the following formula. Combined with the logarithmic fatigue life values of different stress amplitudes under specific failure probabilities and survival rates in step S3.7, a least squares linear fit is performed. The slope and intercept obtained from the fitting are the characterization parameters of the SN curve. At this point, a material SN curve under a specific failure probability or survival rate is obtained.

[0123] σm ×N=c

[0124] mlg(σ)+lg(N)=lg(c)

[0125]

[0126] Where m and c are material constants, and σ is the stress amplitude.

[0127] Specifically, for materials with survival rates of 99.9% and 50%, the SN curves are as follows: Figure 4 shown.

[0128] S3.9: Based on the normal distribution assumption and the mean and standard deviation data of the logarithmic fatigue life data, repeat step S3.8 to obtain the SN curve under any failure probability or survival rate of the material, that is, the probabilistic fatigue life estimate. The probabilistic SN curve is as follows: Figure 5 shown.

[0129] S3.10: Repeat the above steps for the various material data in step S3.1 to obtain a probabilistic fatigue life database of different materials of the wind turbine.

[0130] S3.11: Call the probabilistic SN curve of the probabilistic fatigue life database of different materials in step S3.10, combine it with the Miner linear cumulative damage criterion, and establish a basic probabilistic fatigue life calculation model to provide preliminary probabilistic fatigue life calculation results for different component materials;

[0131] S3.12: Based on step S3.11, the historical operating data and material test data are used to train the back propagation neural network model to learn the relationship between the modified stress characteristics and damage evolution. During the training process, the historical operating data of the wind turbine is used to calibrate the model with an error of no more than 5%, ensuring the accuracy of its prediction.

[0132] Step S4: Input the stress data acquired in real time in step S2 into the trained probabilistic fatigue life prediction model to calculate the cumulative damage value and remaining life of key components, and dynamically update the digital twin model based on the real-time monitoring data and prediction results to improve the model's adaptability to the current operating conditions;

[0133] Step S5: In order to monitor the degradation performance of the current wind turbine in real time, the real-time predicted probabilistic fatigue life results are visualized through the human-computer interaction interface, including the remaining life distribution diagram of key components, the cumulative damage evolution curve and the real-time load distribution diagram, and the prediction results are stored in the wind turbine operation and maintenance management system database. The entire digital twin platform structure, such as Figure 6 shown.

[0134] In step S6, the real-time predicted probabilistic fatigue life result is compared with the set safety threshold. When the component life approaches the critical value, an early warning signal is issued and an optimized operation plan is generated. The operating parameters such as speed and pitch angle are dynamically adjusted through the wind turbine control system to slow down the fatigue evolution rate of high-damage components.

[0135] In the present invention, real-time data-driven analysis based on digital twin technology can timely reflect the operating status and fatigue damage of wind turbines with high real-time performance.

[0136] In the present invention, the physical model and the data-driven model are combined, and the mechanical property description of the physical model is combined with the nonlinear characteristic fitting ability of the data-driven model. The physical model provides the key parameter boundaries, and the data-driven model efficiently processes and predicts the real-time data, finally forming a dynamically interactive hybrid prediction framework, which significantly improves the accuracy and adaptability of probabilistic fatigue life prediction.

[0137] In the present invention, the wind turbine operation strategy is dynamically adjusted according to the fatigue damage and remaining life prediction results, thereby extending the equipment life, reducing unplanned downtime due to component fatigue failure, and improving the economic benefits of the wind farm.

[0138] In the present invention, by constructing a high-fidelity digital twin model, real-time monitoring of the operating loads of key components, training a probabilistic fatigue life model, dynamically updating the prediction model, and feedback optimization of the operation strategy, a high-precision real-time prediction of the cumulative damage and remaining life of the key components of the wind turbine is achieved. First, based on the geometric structure, material parameters and operating data of the key components of the wind turbine, a high-fidelity digital twin model is constructed to provide data support for life prediction. Then, the operating load is monitored through a sensor network, and the cyclic load characteristics are extracted by combining the rain flow counting method to construct a basic probabilistic fatigue life calculation model. The model is trained using a back-propagation neural network algorithm to learn the nonlinear relationship between load characteristics and damage evolution under complex working conditions. On this basis, the digital twin model is dynamically updated based on real-time monitoring data and prediction results, and a feedback optimization strategy is generated through the real-time predicted probabilistic fatigue life results to dynamically adjust the operating parameters to slow down fatigue evolution. Finally, the life prediction results are visualized with the help of a human-computer interaction interface, and the data is stored in the operation and maintenance management system to provide support for the operation and maintenance decision-making of the wind turbine. The present invention can significantly improve the operational reliability and life prediction accuracy of key components of wind turbines, while reducing operation and maintenance costs, providing strong technical support for the health management and life optimization of wind turbines.

[0139] This embodiment also provides a wind turbine probabilistic fatigue life prediction device, which is used to implement the above-mentioned embodiments and preferred implementations. Details already described will not be repeated here. As used below, the term "module" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0140] This embodiment provides a wind turbine probabilistic fatigue life prediction device, such as Figure 7 As shown, including:

[0141] A model building module 71 is used to build a digital twin model of the wind turbine generator set, and the digital twin model is used to simulate and obtain stress data of the wind turbine generator set;

[0142] A stress determination module 72 is configured to determine corresponding stress data using a digital twin model based on the real-time collected operating load data of the wind turbine generator set;

[0143] The prediction module 73 is used to input the stress data into the pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0144] In an optional embodiment, the model construction module is specifically used to: obtain the geometric structure, material parameters and historical operating data of the wind turbine to construct a finite element model of the wind turbine; perform multiple simulations of stress data based on the finite element model to obtain multiple simulation results; use the Kriging method to construct a proxy model for the multiple simulation results, and use the constructed proxy model as the digital twin model of the wind turbine.

[0145] In an optional embodiment, the device also includes: a training data determination module, which is specifically used to perform high-frequency fatigue life tests on the materials of the wind turbine generator under different stress amplitudes to obtain fatigue lives corresponding to different stress amplitudes; use the mean rank estimation method to sort the logarithmic values of the fatigue life and calculate the failure probability of the logarithmic fatigue life; based on the correspondence between the normal distribution quantiles of the failure probability and the logarithmic fatigue life, determine the logarithmic fatigue life corresponding to different stress amplitudes under different failure probabilities or survival rates to obtain an SN curve; based on the SN curve, combined with linear cumulative damage, establish a probabilistic fatigue life calculation model, and the probabilistic fatigue life calculation model is used to provide training data for training the probabilistic fatigue life prediction model.

[0146] In an optional embodiment, the training data includes probabilistic fatigue life corresponding to different stress data, and the device also includes: a model training module, specifically used to use the training data to train the back propagation neural network model to obtain an initial probabilistic fatigue life prediction model; based on the historical operation data of the wind turbine, the initial probabilistic fatigue life prediction model is corrected to obtain a pre-trained probabilistic fatigue life prediction model.

[0147] In an optional embodiment, the prediction module is specifically used to: extract stress characteristic data of stress data using a rain flow counting method, the stress characteristic data including the amplitude, average value and number of cycles of cyclic stress; input the stress characteristic data into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

[0148] In an optional embodiment, the device also includes: an optimization module, specifically used to compare the current probabilistic fatigue life result of the wind turbine with a preset safety threshold to obtain a comparison result; determine whether to generate a warning signal and an optimized operation method of the wind turbine based on the comparison result; control the operation of the wind turbine based on the optimized operation method, and the optimized operation method includes adjusting the operating parameters of the wind turbine.

[0149] In an optional embodiment, the device also includes: an update module for updating the digital twin model based on the monitoring data of the wind turbine and the probabilistic fatigue life results; a display module for visually displaying the probabilistic fatigue life results, the probabilistic fatigue life results including a remaining life distribution diagram, a cumulative damage evolution curve and a real-time load distribution diagram.

[0150] The further functional description of each of the above modules is the same as that of the above corresponding embodiments and will not be repeated here.

[0151] The embodiment of the present invention also provides a computer device having the above Figure 7 The probabilistic fatigue life prediction device for wind turbines shown.

[0152] See also Figure 8 , Figure 8 is a structural diagram of a computer device provided by an optional embodiment of the present invention, such as Figure 8As shown, the computer device includes: one or more processors 10, memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components utilize different buses to communicate with each other and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in the memory or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Equally, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 8 A processor 10 is taken as an example.

[0153] The processor 10 may be a central processing unit, a network processor, or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic, or any combination thereof.

[0154] The memory 20 stores instructions that can be executed by at least one processor 10, so as to enable at least one processor 10 to execute the method shown in the above embodiment.

[0155] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created based on the use of a computer device for displaying a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0156] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0157] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0158] The embodiment of the present invention also provides a computer-readable storage medium. The above-mentioned method according to the embodiment of the present invention can be implemented in hardware, firmware, or implemented as a computer code that can be recorded in a storage medium, or implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method shown in the above embodiment is implemented.

[0159] A portion of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the form in which the computer program instruction exists in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc. Accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium that can be accessed by the computer.

[0160] Although the embodiments of the present invention have been described with reference to the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention. Such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for predicting the probabilistic fatigue life of a wind turbine generator system, characterized in that: The method comprises: Constructing a digital twin model of the wind turbine generator set, wherein the digital twin model is used to simulate and obtain stress data of the wind turbine generator set; Based on the real-time collected operating load data of the wind turbine, the digital twin model is used to determine the corresponding stress data; The stress data is input into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine generator set.

2. The method according to claim 1, characterized in that Build a digital twin model of the wind turbine, including: Obtain the geometric structure, material parameters and historical operating data of the wind turbine and build a finite element model of the wind turbine; Perform multiple simulations of stress data based on the finite element model to obtain multiple simulation results; The Kriging method is used to construct a proxy model for multiple simulation results, and the constructed proxy model is used as the digital twin model of the wind turbine.

3. The method according to claim 1, characterized in that Before inputting the stress data into a pre-trained probabilistic fatigue life prediction model, the method further includes: Conduct high-frequency fatigue life tests on wind turbine materials under different stress amplitudes to obtain fatigue life corresponding to different stress amplitudes; The logarithmic fatigue life values are ranked using the mean rank estimation method to calculate the failure probability of the logarithmic fatigue life. Based on the correspondence between the normal distribution quantiles of failure probability and logarithmic fatigue life, the logarithmic fatigue life corresponding to different stress amplitudes under different failure probabilities or survival rates is determined to obtain the SN curve; Based on the SN curve and combined with linear cumulative damage, a probabilistic fatigue life calculation model is established, and the probabilistic fatigue life calculation model is used to provide training data for training a probabilistic fatigue life prediction model.

4. The method according to claim 3, characterized in that The training data includes probabilistic fatigue life corresponding to different stress data. Before inputting the stress data into the pre-trained probabilistic fatigue life prediction model, the method further includes: Using the training data to train a back propagation neural network model to obtain an initial probability fatigue life prediction model; The initial probabilistic fatigue life prediction model is corrected based on the historical operating data of the wind turbines to obtain a pre-trained probabilistic fatigue life prediction model.

5. The method according to claim 1, wherein The stress data is input into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine, including: Extracting stress characteristic data of the stress data using a rain flow counting method, wherein the stress characteristic data includes the amplitude, average value and number of cycles of cyclic stress; The stress characteristic data is input into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

6. The method according to claim 1, wherein The method further comprises: Compare the current probabilistic fatigue life result of the wind turbine with the preset safety threshold to obtain a comparison result; determining whether to generate a warning signal and an optimized operation method of the wind turbine generator system according to the comparison result; The operation of the wind turbine generator set is controlled according to the optimized operation method, and the optimized operation method includes adjusting the operation parameters of the wind turbine generator set.

7. The method according to claim 1, characterized in that The method further comprises: Update the digital twin model based on wind turbine monitoring data and probabilistic fatigue life results; The probabilistic fatigue life results are visualized, including the remaining life distribution diagram, the cumulative damage evolution curve and the real-time load distribution diagram.

8. A wind turbine probabilistic fatigue life prediction device, characterized in that: The device comprises: A model building module is used to build a digital twin model of the wind turbine generator set, and the digital twin model is used to simulate and obtain stress data of the wind turbine generator set; A stress determination module, configured to determine corresponding stress data using the digital twin model based on the real-time collected operating load data of the wind turbine generator set; The prediction module is used to input the stress data into a pre-trained probabilistic fatigue life prediction model to obtain the current probabilistic fatigue life result of the wind turbine.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the probabilistic fatigue life prediction method for a wind turbine according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the wind turbine probabilistic fatigue life prediction method according to any one of claims 1 to 7.

Images