A method and system for predicting the vibration response of wind turbine blades

Abstract

This invention relates to the field of data processing technology, specifically to a method and system for predicting the vibration response of wind turbine blades. The method includes: determining uncertainty parameters of the wind turbine blade; determining the structural properties of the wind turbine blade based on the uncertainty parameters; generating model samples for vibration response analysis based on the structural properties; determining the vibration analysis results of the wind turbine blade based on the model samples and a physical information neural network model trained on the model samples; and predicting the response of the wind turbine blade based on the vibration analysis results to obtain a target prediction result for the wind turbine blade. This method combines physical information with a neural network model, ensuring that the prediction results not only consider the statistical characteristics of the data but also incorporate the influence of physical laws and system uncertainties, further improving the accuracy of vibration response prediction.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for predicting the vibration response of wind turbine blades. Background Technology

[0002] Wind turbine blades are crucial components of wind turbine generators, and their vibration directly impacts the turbine's operating efficiency and safety. With the continuous development of wind power technology, wind turbine blade designs are increasingly trending towards high efficiency, lightweight construction, and larger dimensions, making their vibration characteristics more complex. Accurate prediction of vibration response during the design and operation of wind turbine blades is essential for ensuring turbine reliability, extending service life, and optimizing maintenance strategies. However, the prediction of wind turbine blade vibration response is influenced by multiple uncertainties, including material properties, manufacturing tolerances, and the operating environment. These factors often make it difficult for traditional vibration analysis methods to provide high-precision prediction results.

[0003] Accurate prediction of the vibration response of wind turbine blades faces numerous challenges. First, the dynamic behavior of wind turbine blades is complex, involving multiple factors such as wind speed variations, blade deformation, and aerodynamic and structural interactions. Second, many key parameters (such as the elastic modulus, damping characteristics, and friction coefficient of the blade material) have significant uncertainties, and traditional vibration response prediction methods often assume that these parameters are known and constant, failing to effectively handle the impact of uncertainties on the prediction results. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for predicting the vibration response of wind turbine blades. The specific technical solution adopted is as follows:

[0005] In a first aspect, embodiments of the present invention provide a method for predicting the vibration response of wind turbine blades, the method comprising:

[0006] Determine the uncertain parameters of the wind turbine blades;

[0007] Based on the aforementioned uncertainty parameters, the structural properties of the wind turbine blades are determined;

[0008] Based on the structural properties, model samples for vibration response analysis are generated;

[0009] Based on the model samples and the physical information neural network model trained by the model samples, the vibration analysis results of the wind turbine blades are determined.

[0010] Based on the vibration analysis results, the response of the wind turbine blades is predicted to obtain the target prediction results of the wind turbine blades.

[0011] Secondly, a wind turbine blade vibration response prediction system is provided, the system comprising:

[0012] The first determining module is used to determine the uncertain parameters of the wind turbine blades;

[0013] The second determining module is used to determine the structural properties of the wind turbine blades based on the uncertainty parameters.

[0014] A generation module is used to generate model samples for vibration response analysis based on the structural properties.

[0015] The third determining module is used to determine the vibration analysis results of the wind turbine blades based on the model samples and the physical information neural network model trained by the model samples.

[0016] The prediction module is used to predict the response of the wind turbine blades based on the vibration analysis results, and obtain the target prediction result of the wind turbine blades.

[0017] Thirdly, a computer program product is provided, comprising: computer program code, which, when run on a computer, causes the computer to perform the method described in the first aspect or any possible implementation thereof.

[0018] Fourthly, a computer-readable storage medium is provided that stores computer program code, which, when executed on a computer, causes the computer to perform the methods described in the first aspect or any possible implementation thereof.

[0019] This invention offers the following advantages: By analyzing the uncertainty parameters of wind turbine blades and determining their structural properties based on these parameters, a model sample for vibration response analysis is generated. This approach, by introducing uncertainty parameters for uncertainty propagation analysis, more accurately reflects the uncertainties present in actual engineering projects, resulting in a more precise model sample. Finally, using this model sample and a physical information neural network model, the vibration analysis results of the wind turbine blades are determined. Based on these results, the response of the wind turbine blades is predicted, yielding the target prediction result. Thus, combining physical information with a neural network model ensures that the prediction results not only consider the statistical characteristics of the data but also incorporate the influence of physical laws and system uncertainties, further improving the accuracy of vibration response prediction. Attached Figure Description

[0020] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram illustrating the implementation process of a wind turbine blade vibration response prediction method provided in an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of a wind turbine blade provided in an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the implementation framework of a wind turbine blade vibration response prediction method provided in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the vibration response prediction results of the GPR method provided in related technologies;

[0025] Figure 5 This is a schematic diagram of the absolute error distribution predicted by the GPR method provided in related technologies.

[0026] Figure 6 This is a schematic diagram of the vibration response prediction result of a wind turbine blade vibration response prediction method provided in an embodiment of the present invention;

[0027] Figure 7 This is a schematic diagram of the absolute error distribution results of a wind turbine blade vibration response prediction method provided in an embodiment of the present invention.

[0028] Figure 8 A schematic diagram of the probability distribution of the maximum vibration response under uncertainty conditions provided in an embodiment of the present invention;

[0029] Figure 9 This is a schematic diagram of the composition structure of a wind turbine blade vibration response prediction system provided in an embodiment of the present invention;

[0030] Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0031] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a wind turbine blade vibration response prediction method proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments may be combined from any suitable form.

[0032] In the description of the embodiments of the present invention, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present invention, "multiple" means two or more.

[0033] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] In some embodiments, vibration prediction models based on machine learning methods such as Physical Information Neural Network (PINN) and Gaussian Process Regression (GPR) have attracted widespread attention. PINN combines neural networks and physical constraints, enabling it to fully utilize the physical model of the wind turbine blades during the prediction process and optimize network parameters through a backpropagation algorithm. This approach not only effectively improves the accuracy of vibration response prediction but also handles complex boundary and constraint conditions.

[0036] Gaussian process regression (GPR) is a nonparametric regression method based on a probabilistic model. It utilizes a Gaussian process as a prior distribution to describe the latent form of a function, effectively modeling the relationship between input and output data. Unlike traditional regression methods, GPR does not assume that the data follows a specific functional form. Instead, it provides predictions of function values ​​and their uncertainties by modeling the covariance structure of the training data. In GPR, each training data point is associated with a random variable in a Gaussian process, and the relationship between these random variables is described by a covariance function (also called a kernel function). By optimizing the parameters of this covariance function, a best fit to the data can be obtained. The advantage of GPR lies in its ability to provide not only predicted values ​​but also estimates of the uncertainty of those predictions, which is particularly important for many practical problems requiring uncertainty analysis. Applying GPR based on PINN can effectively improve accuracy.

[0037] Based on this, embodiments of the present invention provide a method for predicting the vibration response of wind turbine blades, combining a surrogate model of Physical Information Neural Network (PINN) and Gaussian Process Regression (GPR) to accurately predict the vibration response of wind turbine blades. This method can be widely applied in the design, optimization, and operation and maintenance of wind turbine generator sets, and has significant application value, especially in the vibration analysis and structural health monitoring of wind turbine blades. By quantifying the uncertainties in the system and integrating physical constraints, embodiments of the present invention provide a more accurate and reliable solution for predicting the vibration response of wind turbine blades, which can improve the operating efficiency and safety of wind turbines, optimize maintenance strategies, and extend the service life of wind turbines. The specific scheme of the wind turbine blade vibration response prediction method provided by the present invention is described in detail below with reference to the accompanying drawings. Please refer to... Figure 1 The diagram illustrates a flowchart of a wind turbine blade vibration response prediction method according to an embodiment of the present invention. The method includes:

[0038] 101. Determine the uncertainty parameters of the wind turbine blades.

[0039] Here, the wind turbine blade model is as follows: Figure 2 As shown. Uncertainty parameters can be parameters that may change or are not completely determined during design and operation. These parameters have a significant impact on vibration response.

[0040] In some possible implementations, step 101 above can be achieved through the following steps 111 to 113 (not shown in the figures):

[0041] 111, Determine the modal frequency data of each order of the wind turbine blade.

[0042] Here, by combining the modal frequency data of the wind turbine blades, we designed the wind turbine blades into a multi-segment structure, with each segment having a different Young's modulus property. The reliability of the method is verified by predicting the response of the multi-segment blade system within a certain frequency range.

[0043] 112. The optimal Latin hypercube is used to set the data range of each modal frequency data.

[0044] Here, the optimal Latin hypercube (OLHD) is used to design the range of parameters. Samples are generated using the optimal Latin hypercube design based on the range of uncertain variables.

[0045] 113. Based on the data range, determine the uncertainty parameters of the wind turbine blades.

[0046] Here, the data range of each modal frequency is determined as the data range of the uncertainty parameters of the wind turbine blade, thus obtaining the uncertainty parameters of the wind turbine blade; thereby improving the accuracy of setting the data range of the uncertainty parameters.

[0047] 102. Based on the aforementioned uncertainty parameters, determine the structural properties of the wind turbine blades.

[0048] Here, by analyzing the uncertainty parameters, the wind turbine blades are designed as a multi-segment structure; and the Young's modulus property of each segment in the multi-segment structure is determined to obtain the structural properties.

[0049] In some possible implementations, the wind turbine blade adopts a segmented model, assuming that the blade consists of n segments, where n=4, and is composed of four structural segments. , , , ), and each segment of the structure has a different Young's modulus distribution, ( , , , Meanwhile, an external stimulus F(t) affects... Excitement is applied. The prediction objective is to obtain [the desired result] in the frequency domain. The acceleration response of the wind turbine blades is shown in Table 1.

[0050] Table 1. Statistical Table of Frequency Values ​​of Wind Turbine Blades

[0051]

[0052] 103. Based on the structural properties, generate model samples for vibration response analysis.

[0053] Here, the optimal Latin hypercube is used to design the range of parameters to generate model samples. These model samples are used for vibration response analysis to ensure reasonable coverage of all possible uncertainties in the parameter space.

[0054] Based on the range of uncertainty variables in Table 2, the optimal Latin hypercube design is used to generate model samples.

[0055] Table 2. Statistical table of sampling intervals for Young's modulus

[0056]

[0057] Generate 1000 samples, including 700 training samples, 150 validation samples, and 150 test samples.

[0058] In some possible implementations, the data range of the structural properties is defined by using the optimal Latin hypercube; and a model sample for vibration response analysis is generated based on the data range of the structural properties.

[0059] Here, after determining the uncertainty parameters of the system, sample data collection is carried out, including the following process:

[0060] First, the optimal Latin hypercube is used to design the range of parameters, generating model samples. These model samples are used for vibration response analysis, ensuring reasonable coverage of all possible uncertainties within the parameter space.

[0061] Secondly, based on the generated model samples, vibration analysis is performed to extract vibration response values ​​at different frequency points. These response values ​​provide real response data for subsequent prediction models.

[0062] Finally, the vibration response data was divided into training, validation, and test sets. The training set was used to train the model, the validation set was used for model tuning, and the test set was used to evaluate the model's performance. This division of the datasets ensured the scientific use of data and the effectiveness of model evaluation.

[0063] 104. Based on the model samples and the physical information neural network model trained by the model samples, determine the vibration analysis results of the wind turbine blades.

[0064] Here, a preliminary prediction is made using the PINN model to obtain the vibration analysis results of the wind turbine blades.

[0065] In some possible implementations, firstly, based on the model samples, the residual matrix of the physical information neural network model is determined; secondly, the preset standardized features and the residual matrix are used as inputs to the physical information neural network model, and Gaussian process regression is performed on the physical information neural network model. The kernel hyperparameters of the physical information neural network model are then adjusted using the maximum marginal likelihood algorithm to obtain a trained physical information neural network model (i.e., the PINN-GPR model); finally, the trained physical information neural network model is used to perform vibration analysis on the model samples to obtain the vibration analysis results of the wind turbine blades.

[0066] Here, the mean function for GPR is provided through the PINN preliminary prediction model (i.e., the physical information neural network model), which is divided into the following steps:

[0067] The first step is to standardize the training set data to ensure that all input features are calculated at the same scale, thus preventing errors caused by different data scales during model training.

[0068] The second step involves inputting the standardized training data into PINN. PINN's input layer accepts uncertainty parameters, the hidden layers undergo a nonlinear transformation using the Tanh activation function, and the final output layer predicts the system's response. This process yields preliminary predictions of the vibration response.

[0069] The third step is to calculate the total loss function, which includes data loss, physical loss, and conditional loss.

[0070] The fourth step is to use the Optuna automated hyperparameter optimization framework to adjust the hyperparameters of PINN (such as learning rate, number of network layers, etc.) to minimize the total loss function and obtain the optimal model.

[0071] The fifth step involves iterating through steps two through four to gradually optimize the model and obtain accurate prediction results. Each iteration updates the weights and hyperparameters of PINN to obtain a trained physical information neural network model, ultimately yielding the preliminary vibration response prediction results of PINN (i.e., the vibration analysis results of the wind turbine blades).

[0072] In some embodiments, an initial response value is obtained by performing preliminary predictions on the initial physical information neural network based on the model samples, and the parameters of the initial physical information neural network are adjusted based on the preliminary response value to obtain the physical information neural network model.

[0073] Here, the model samples are standardized to obtain processed samples; for example, the model samples are filtered and re-extracted for features. Then, based on the processed samples, the initial physical information neural network (PINN) is used for forward prediction to obtain the preliminary response value; for example, the training set data is standardized to obtain processed samples, and forward prediction is performed using PINN, with parameters as the input layer and the hidden layer using the Tanh activation function to obtain the output layer's response value (i.e., the preliminary response value). The data loss, physical loss, and conditional loss of the preliminary response value are determined in the initial physical information neural network. Based on the data loss, physical loss, and conditional loss, the total loss is calculated. The hyperparameters of the initial physical information neural network are adjusted and iteratively trained using the total loss to obtain the physical information neural network model. The total loss can be obtained by weighted summation of the data loss, physical loss, and conditional loss, where the conditional loss includes boundary conditions and initial conditions. The hyperparameters are updated using the Optuna automated hyperparameter optimization framework, and through multiple iterations, the training of the PINN model is completed, thus obtaining the physical information neural network model.

[0074] Here, the PINN model is used as the mean function for GPR training: First, the residuals are calculated, representing the difference between the standardized output of PINN and the standardized true labels of the training set, thus obtaining the residual matrix. Next, a composite kernel function is constructed, using the standardized features and training residuals as inputs to GPR training, and the kernel hyperparameters are optimized using the marginal likelihood maximization method. Finally, the kernel hyperparameters are updated iteratively multiple times to complete the training of the PINN model, thus obtaining the physical information neural network model.

[0075] 105. Based on the vibration analysis results, the response of the wind turbine blade is predicted to obtain the target prediction result of the wind turbine blade.

[0076] Here, after obtaining the vibration analysis results, the resonant frequency of the vibration response and the response value within the resonant frequency region are extracted; based on the resonant frequency and the response value within the resonant frequency region, the response of the wind turbine blade is predicted, and the target prediction result of the wind turbine blade is obtained.

[0077] In some possible implementations, the wind turbine blades are predicted based on their resonant frequency and the response values ​​within the resonant frequency region to obtain an initial prediction result. Then, the initial prediction result is corrected point by point based on residual compensation to obtain a corrected prediction result. The corrected prediction result is then compared with the actual response result to obtain the predicted value and error index for each frequency point. Finally, based on the predicted value and error index for each frequency point, the target prediction result for the wind turbine blades is obtained.

[0078] Here, after obtaining the vibration analysis results, the residual matrix of the vibration analysis results in the physical information neural network model is calculated, and a composite kernel function is constructed. The standardized features and training residuals are used as inputs for GPR training, and the kernel hyperparameters are optimized by maximizing marginal likelihood. Then, the residual compensation is added point by point to the standardized prediction results of PINN to obtain the corrected standardized results, and the original acceleration units are converted back by the label standardized parameters. Finally, the corrected prediction results are compared with the actual responses for each sample in the test set, and the predicted value and error index for each frequency point are output. The updated prediction model (i.e., the trained physical information neural network model) is validated by using the test set.

[0079] Here, the residual matrix reflects the prediction error of the PINN model. A composite kernel function is designed using standardized features and calculated residuals. This composite kernel function combines physical mechanisms and residual information, serving as input to the GPR model (i.e., the Physical Information Neural Network model). The kernel hyperparameters are optimized by maximizing marginal likelihood to improve the prediction accuracy of the GPR. After training the GPR model (i.e., the trained Physical Information Neural Network model), the residual compensation value and its standard deviation vector for each frequency point are obtained. These residual compensations are added point-by-point to the standardized prediction results of PINN to obtain the "corrected" standardized prediction values. Then, the corrected prediction results are converted back to the original acceleration units using label-standardized parameters. By comparing the corrected prediction results with the actual response for each sample in the test set, the predicted value and error index for each frequency point are output to evaluate the model's accuracy. The updated prediction model is validated using the test set to ensure it has good generalization ability and can effectively predict vibration responses under different operating conditions.

[0080] In some embodiments, after constructing the PINN-GPR model, uncertainty propagation is performed on the pre-trained initial physical information neural network model. Based on the distribution characteristics of the uncertainty parameters, a Monte Carlo simulation (MCS) sampling method is used to generate the cumulative distribution function (CDF) of the maximum vibration response, achieving rapid uncertainty propagation from input parameters to output response. By quantifying the physical mechanism of the parameters and the impact of uncertainty on them, and combining the PINN and GPR models, an accurate vibration response prediction surrogate model is generated. Through multiple optimizations and iterative training, standardization, and corrections, the accuracy of vibration response prediction can be effectively improved.

[0081] Here, the specific steps for evaluating the distribution of vibration response under uncertain parameters are as follows:

[0082] The first step is to determine the uncertainty range of the input parameters based on the statistical distribution characteristics of each uncertainty parameter (e.g., normal or uniform distribution):

[0083] The second step involves performing extensive random sampling to account for the uncertainty of the input parameters. This sampling generates vibration response outputs under different parameter combinations, yielding the final output.

[0084] The third step is to generate the Cumulative Distribution Function (CDF) of the maximum vibration response based on the output results. This function describes the possible distribution of the maximum vibration response of the system under different input parameters.

[0085] The fourth step, through CDF (Contingency Design), enables rapid uncertainty propagation from input parameters to output response. This process provides a fast and accurate tool for uncertainty analysis, helping designers assess the risks and uncertainties under different operating conditions.

[0086] In a specific example, the uncertainty transfer of the vibration response is completed based on the uncertainty parameter distribution in Table 3:

[0087] Table 3. Parameter distribution of Young's modulus

[0088]

[0089] Based on the uncertainty parameter distribution in Table 3, 200 samples were generated to complete the CDF distribution of the maximum vibration response.

[0090] The implementation process of the wind turbine blade vibration response prediction method provided in this embodiment of the invention can be achieved through... Figure 3Implementation: The entire implementation process can be divided into four parts, namely, Part 1: data collection; Part 2: PINN prediction; Part 3: GPR model correction; and Part 4: uncertainty propagation. In Part 1, optimal hypercube design is used to obtain sample data, and vibration response analysis is performed to obtain training, validation, and test sets. Then, in Part 2, PINN forward prediction is used, combined with multi-objective loss functions (including data, physical damage, frequency domain constraints, and peak constraints) to update hyperparameters and iterate multiple times to obtain the PINN prediction model (i.e., the physical information neural network model). Then, in Part 3, the PINN prediction model is used as the mean function m(x) for GPR training. Kernel function design, GPR training, residual prediction / error analysis, calibration / inverse normalization, prediction visualization, and response prediction are performed sequentially to achieve GPR model correction and obtain the trained physical information neural network model. Finally, in the fourth part, after obtaining the uncertainty parameter distribution, the input sample parameters are further generated. The results of the response prediction in the third part are used as the input to the trained physical information neural network model (i.e., the PINN-GPR model) to output the sample response, thereby responding to the uncertainty distribution.

[0091] To compare with methods in some embodiments, GPR was used for response prediction, where the numerical solution was used as the actual value, and the results are as follows: Figure 4 and Figure 5 As shown, the predicted and actual value curves of the sample do not match. Here, Mean represents the mean, Median represents the median, Std Dev represents the standard deviation, P90 represents the value at the 90th percentile after sorting the data from smallest to largest, P95 represents the value at the 95th percentile after sorting the data from smallest to largest, and P99 represents the value at the 99th percentile after sorting the data from smallest to largest. When the MAPE value (mean absolute percentage error) is 30%, the sample proportion reaches more than 10%.

[0092] To further illustrate the advantages of the embodiments of the present invention, the following process will be used. Figure 3 The method shown (PINN-GPR model) is used for prediction, and the results are as follows: Figure 6 and Figure 7 As shown, the predicted and actual value curves of the samples match well. When the MAPE value exceeds 30%, the sample proportion is less than 1%. Accuracy is significantly improved. Based on the PINN-GPR model (i.e., the trained physical information neural network model), according to the range in Table 3, 200 samples were generated using MCS. The CDF of the maximum vibration response is shown below. Figure 8 As shown: the change in maximum vibration response is mainly at 2 m / s. 2-2.2 m / s 2 (For example, the change in the maximum vibration response is mainly in Q1 (2.0069 m / s) 2 -Q3 (2.1870m / s 2 Between )). For example Figure 8 As shown, the maximum vibration response range is 0-2.6 m / s². 2 .

[0093] In this embodiment of the invention, by analyzing the uncertainty parameters of the wind turbine blades and determining their structural properties based on these parameters, a model sample for vibration response analysis is generated based on these structural properties. This approach, by introducing uncertainty parameters for uncertainty propagation analysis, more realistically reflects the uncertainties present in actual engineering projects, thus generating a more accurate model sample. Finally, using this model sample and a physical information neural network model, the vibration analysis results of the wind turbine blades are determined. Based on these vibration analysis results, the response of the wind turbine blades is predicted, yielding the target prediction result. Thus, combining physical information with a neural network model ensures that the prediction results not only consider the statistical characteristics of the data but also incorporate the influence of physical laws and system uncertainties, further improving the accuracy of vibration response prediction.

[0094] This invention provides a wind turbine blade vibration response prediction system. Please refer to [link / reference]. Figure 9 The diagram illustrates the structural composition of a wind turbine blade vibration response prediction system according to an embodiment of the present invention. The system 900 includes:

[0095] The first determining module 901 is used to determine the uncertainty parameters of the wind turbine blades;

[0096] The second determining module 902 is used to determine the structural properties of the wind turbine blades based on the uncertainty parameters.

[0097] The generation module 903 is used to generate model samples for vibration response analysis based on the structural properties.

[0098] The third determining module 904 is used to determine the vibration analysis results of the wind turbine blades based on the model samples and the physical information neural network model trained by the model samples.

[0099] The prediction module 905 is used to predict the response of the wind turbine blade based on the vibration analysis results, and obtain the target prediction result of the wind turbine blade.

[0100] In some possible implementations, the first determining module 901 is further configured to determine the modal frequency data of each order of the wind turbine blade; set the data range of the modal frequency data of each order using the optimal Latin hypercube; and determine the uncertainty parameters of the wind turbine blade based on the data range.

[0101] In some possible implementations, the second determining module 902 is further configured to design the wind turbine blade as a multi-segment structure based on the uncertainty parameters; and to determine the Young's modulus property of each segment in the multi-segment structure to obtain the structural properties.

[0102] In some possible implementations, the generation module 903 is used to set the data range of the structural properties using the optimal Latin hypercube; and to generate model samples for vibration response analysis based on the data range of the structural properties.

[0103] In some possible implementations, the third determining module 904 is used to determine the residual matrix of the physical information neural network model based on the model samples; use the preset standardized features and the residual matrix as inputs to the physical information neural network model, perform Gaussian process regression training on the physical information neural network model, and use the maximum marginal likelihood algorithm to adjust the kernel hyperparameters of the physical information neural network model to obtain a trained physical information neural network model; use the trained physical information neural network model to perform vibration analysis on the model samples to obtain the vibration analysis results of the wind turbine blades.

[0104] In some possible implementations, the third determining module 904 is used to perform preliminary predictions of the initial physical information neural network based on the model samples to obtain preliminary response values;

[0105] The parameters of the initial physical information neural network are adjusted based on the preliminary response value to obtain the physical information neural network model.

[0106] In some possible implementations, the third determining module 904 is used to standardize the model samples to obtain processed samples; based on the processed samples, perform forward prediction using the initial physical information neural network to obtain the preliminary response value; in the initial physical information neural network, determine the data loss, physical loss, and conditional loss of the preliminary response value; based on the data loss, physical loss, and conditional loss, adjust the hyperparameters of the initial physical information neural network and perform iterative training to obtain the physical information neural network model.

[0107] In some possible implementations, the prediction module 905 is used to extract the resonant frequency of the vibration response and the response value within the resonant frequency region based on the vibration analysis results; and to predict the response of the wind turbine blade based on the resonant frequency and the response value within the resonant frequency region, thereby obtaining the target prediction result of the wind turbine blade.

[0108] In some possible implementations, the prediction module 905 is used to predict the response of the wind turbine blade based on the resonant frequency and the response value within the resonant frequency region to obtain an initial prediction result for the wind turbine blade; to correct the initial prediction result point by point based on residual compensation to obtain a corrected prediction result; to compare the corrected prediction result with the actual response result to obtain a predicted value and error index for each frequency point; and to obtain a target prediction result for the wind turbine blade based on the predicted value and error index for each frequency point.

[0109] Optionally, the transmission medium can be a wired link (e.g., but not limited to, coaxial cable, optical fiber, and Digital Subscriber Line (DSL)) or a wireless link (e.g., but not limited to, Wireless Fidelity (WIFI), Bluetooth, and mobile device networks). It should be noted that the control device provided in the above embodiments is only an example illustrating the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the method embodiments provided in the above embodiments belong to the same concept, and their specific implementation processes are detailed in the method embodiments, and will not be repeated here.

[0110] Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. For example, as shown... Figure 10 As shown, the computer device 1000 includes: a memory 1001, a processor 1002, and a computer program 1003 stored in the memory 1001 and running on the processor 1002, wherein when the processor 1002 executes the computer program 1003, the computer device can execute any of the wind turbine blade vibration response prediction methods described above.

[0111] Furthermore, this embodiment of the invention also protects a control device, which may include a memory and a processor. The memory stores executable program code, and the processor is used to call and execute the executable program code to perform a wind turbine blade vibration response prediction method provided by this embodiment of the invention. This embodiment can divide the control device into functional modules based on the above method example. For example, each module can correspond to a specific function, or two or more functions can be integrated into a single processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents a logical functional division; other division methods may exist in actual implementation. It should also be noted that all relevant content of each step involved in the above method embodiment can be referenced to the functional description of the corresponding functional module, and will not be repeated here.

[0112] It should be understood that the control device provided in this embodiment is used to execute the above-described method for predicting the vibration response of wind turbine blades, and therefore can achieve the same effect as the above-described implementation method. When using an integrated unit, the control device may include a processing module and a storage module. When the control device is applied to a device, the processing module can be used to control and manage the device's actions. The storage module can be used to support the device in executing mutual program code, etc. The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this invention. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of Digital Signal Processing (DSP) and a microprocessor, etc., and the storage module may be a memory.

[0113] Furthermore, the control device provided in the embodiments of the present invention may specifically be a chip, component, or module. The chip may include a connected processor and a memory. The memory stores instructions, and when the processor calls and executes the instructions, the chip can execute the wind turbine blade vibration response prediction method provided in the above embodiments. This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the aforementioned method steps to implement the wind turbine blade vibration response prediction method provided in the above embodiments.

[0114] This embodiment also provides a computer program product. When the computer program product is run on a computer, it causes the computer to execute the aforementioned related steps to realize the wind turbine blade vibration response prediction method provided in the above embodiment. The control device, computer-readable storage medium, computer program product, or chip provided in this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here. Through the description of the above embodiments, those skilled in the art can understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the control device can be divided into different functional modules to complete all or part of the functions described above. In the embodiments provided by this invention, it should be understood that the disclosed control device and method can be implemented in other ways. For example, the control device embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another control device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, control device or unit, and can be electrical, mechanical or other forms.

[0115] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multiple task processing and parallel processing are possible or may be advantageous. The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. The above content is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be covered within the protection scope of the present invention.

Claims

1. A method of predicting the vibration response of a wind turbine blade, characterized by, The method comprises: determining an uncertainty parameter of a fan blade; designing the fan blade into a multi-section structure based on the uncertainty parameter; determining a Young's modulus attribute of each section of the multi-section structure to obtain a structural attribute of the fan blade; setting a data range of the structural attribute by using optimal Latin hypercubes; generating a model sample for vibration response analysis based on the data range of the structural attribute; determining a vibration analysis result of the fan blade based on the model sample and a physical information neural network model trained by the model sample; determining a residual matrix of the physical information neural network model based on the model sample; taking preset standardized features and the residual matrix as inputs of the physical information neural network model, performing Gaussian process regression training on the physical information neural network model, and adjusting kernel hyperparameters of the physical information neural network model by using a maximum marginal likelihood algorithm to obtain a trained physical information neural network model; performing vibration analysis on the model sample by using the trained physical information neural network model to obtain the vibration analysis result of the fan blade.

2. The method of claim 1, wherein, The determination of the uncertainty parameter of the fan blade comprises: determining modal frequency data of each order of the fan blade; setting a data range of the modal frequency data of each order by using optimal Latin hypercubes; determining the uncertainty parameter of the fan blade based on the data range.

3. The method of claim 1, wherein, The method further comprises: performing preliminary prediction on an initial physical information neural network based on the model sample to obtain a preliminary response value; adjusting parameters of the initial physical information neural network based on the preliminary response value to obtain the physical information neural network model.

4. The method of claim 3, wherein, The preliminary prediction on the initial physical information neural network based on the model sample to obtain the preliminary response value comprises: performing standardization processing on the model sample to obtain processed samples; performing forward prediction on the initial physical information neural network based on the processed samples to obtain the preliminary response value. Correspondingly, the adjustment of the parameters of the initial physical information neural network based on the preliminary response value to obtain the physical information neural network model comprises: determining data loss, physical loss and conditional loss of the preliminary response value in the initial physical information neural network; adjusting and iteratively training hyperparameters of the initial physical information neural network based on the data loss, the physical loss and the conditional loss to obtain the physical information neural network model.

5. The method of claim 1, wherein, The response prediction on the fan blade based on the vibration analysis result to obtain a target prediction result of the fan blade comprises: extracting a resonance frequency and a response value in a resonance frequency region of the vibration response based on the vibration analysis result; performing response prediction on the fan blade based on the resonance frequency and the response value in the resonance frequency region to obtain the target prediction result of the fan blade.

6. The method of claim 5, wherein, The response prediction on the fan blade based on the resonance frequency and the response value in the resonance frequency region to obtain the target prediction result of the fan blade comprises: Based on the response value in the resonance frequency and the resonance frequency region, a response prediction is performed on the fan blade to obtain an initial prediction result of the fan blade; Based on residual compensation, a point-by-point correction is performed on the initial prediction result to obtain a corrected prediction result; The corrected prediction result is compared with a real response result to obtain a prediction value and an error index of each frequency point; Based on the prediction value and the error index of each frequency point, a target prediction result of the fan blade is obtained.

7. A wind turbine blade vibration response prediction system, characterized by, The system comprises: A first determination module configured to determine an uncertainty parameter of a fan blade; A second determination module configured to design the fan blade into a multi-segment structure based on the uncertainty parameter, and determine a Young's modulus attribute of each segment of the multi-segment structure to obtain a structure attribute; A generation module configured to set a data range of the structure attribute by using an optimal Latin hypercube, and generate a model sample for vibration response analysis based on the data range of the structure attribute; A third determination module configured to determine a vibration analysis result of the fan blade based on the model sample and a physical information neural network model trained by the model sample; A prediction module configured to determine a residual matrix of the physical information neural network model based on the model sample, use a preset standardized feature and the residual matrix as inputs of the physical information neural network model, perform a Gaussian process regression training on the physical information neural network model, adjust kernel hyperparameters of the physical information neural network model by using a maximum marginal likelihood algorithm, obtain a trained physical information neural network model, and perform a vibration analysis on the model sample by using the trained physical information neural network model to obtain the vibration analysis result of the fan blade.

Images