A method and device for predicting the failure probability of wind turbine tower structures

Abstract

This invention relates to the field of wind power failure prediction technology, specifically a method and device for predicting the failure probability of wind turbine tower structures. The invention uses uncertainties in wind data as random seeds, constructs and adaptively trains multinomial chaotic expansion models under each random seed, effectively balancing explicit and implicit multi-source uncertainties in the tower's dynamic response, and avoids the limitation of traditional deterministic surrogate models that can only represent a single operating condition. A spectral surrogate model is constructed using an ensemble convergent PCE model and convergence verification is completed, significantly improving the model's generalization ability and prediction stability. Monte Carlo simulation is then combined to achieve accurate calculation of the failure probability. Compared to existing technologies, this solution can comprehensively quantify the impact of multi-source uncertainties, significantly improving the accuracy of wind turbine tower failure probability prediction.

Description

Technical Field

[0001] This invention relates to the field of new energy centralized control wind power failure prediction technology, specifically to a method and device for predicting the failure probability of wind turbine tower structure. Background Technology

[0002] In the field of wind power engineering, the wind turbine tower, as a key load-bearing structure supporting the wind turbine unit, is like the "skeleton" of the unit. Its safety and stability are directly related to the normal operation of the wind farm and the safety of personnel. Accurate prediction of the tower failure probability is the core link to ensure the long-term reliable service of the unit and is of great significance to promoting the safe and sustainable development of the wind power industry.

[0003] However, in complex service environments, the probability of tower failure is generally predicted by high-fidelity physical model simulation. Although the above method can guarantee a certain level of accuracy, the computational cost is extremely high, making it difficult to apply to reliability analysis and optimization design that require a large number of repeated calculations.

[0004] Therefore, surrogate models are often introduced to improve the efficiency of predicting tower failure probability. However, this method often relies on deterministic surrogate models and cannot take into account the explicit and implicit multi-source uncertainties faced by the tower dynamic response, resulting in low accuracy of tower failure probability. Summary of the Invention

[0005] This invention provides a method and apparatus for predicting the failure probability of wind turbine tower structures, in order to solve the problem that existing technologies rely heavily on deterministic surrogate models, which cannot take into account both explicit and implicit multi-source uncertainties in the dynamic response of the tower, resulting in low accuracy of tower failure probability.

[0006] In a first aspect, the present invention provides a method for predicting the failure probability of a wind turbine generator tower structure, the method comprising: Each uncertainty factor in the wind power data of the wind turbine generator is set as a random seed, and the design parameters of the wind turbine tower structure and any of the random seeds are used to construct the corresponding polynomial chaotic expansion model under each random seed. Adaptive training and convergence verification are performed on the polynomial chaotic expansion models corresponding to each random seed. The converged polynomial chaotic expansion models are then combined to construct a spectral surrogate model. The spectral proxy model is converged and verified, and Monte Carlo simulation is performed on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

[0007] This invention constructs and adaptively trains a multinomial chaotic expansion model under each random seed by setting uncertainties in wind power data as random seeds. This effectively takes into account both explicit and implicit multi-source uncertainties in the dynamic response of the wind turbine tower, avoiding the limitation of traditional deterministic surrogate models that can only represent a single operating condition. A spectral surrogate model is constructed using an ensemble convergent PCE model, and convergence verification is completed, significantly improving the model's generalization ability and predictive stability. Furthermore, Monte Carlo simulation is combined to achieve accurate calculation of failure probability. Compared to existing technologies, this solution can comprehensively quantify the impact of multi-source uncertainties, significantly improving the accuracy of wind turbine tower failure probability prediction.

[0008] In one optional implementation, the step of setting each uncertainty factor of the wind power data of the wind turbine generator as a random seed, and using the design parameters of the wind turbine tower structure and any of the random seeds to construct the corresponding polynomial chaotic expansion model under each random seed, includes: Obtain wind power data and design parameters for the wind turbine tower structure; Each uncertainty factor in the wind force data is set as multiple random seeds and then fixed. Input the design parameters into a preset design space to generate initial training samples; Using the initial training samples and the fixed random seeds, a multinomial chaotic expansion model of the load response of the wind turbine tower structure under each random seed is constructed.

[0009] This invention extracts wind power data and tower design parameters, sets uncertainties in the data as fixed random seeds, and can stably characterize implicit uncertain environmental conditions. It combines the design space to generate initial training samples and constructs a polynomial chaotic expansion model of the wind turbine tower structure under load response under each random seed. This model can simultaneously accommodate explicit parameters and implicit environmental uncertainties, avoid the interference of random fluctuations on model construction, and ensure that the modeling process is repeatable and the results are stable and reliable.

[0010] In one optional implementation, the step of adaptively training and verifying the convergence of the polynomial chaotic expansion models corresponding to each of the random seeds, and then aggregating the converged polynomial chaotic expansion models to construct a spectral surrogate model includes: Leave-one-out cross-validation is used to sequentially remove individual samples from the initial training samples, and the remaining samples are used to train the polynomial chaotic expansion model corresponding to the fixed random seed. Combine all trained multinomial chaos expansion models to generate a surrogate model set. The misclassification probability of each candidate sample in the design space is evaluated using the surrogate model set. The candidate samples with the highest misclassification probability are selected and added to the initial training samples to generate updated training samples; The updated training samples are used to retrain the trained polynomial chaotic expansion model, and it is determined whether the retrained polynomial chaotic expansion model satisfies the preset first convergence index. If not, the retrained polynomial chaotic expansion model is set as the polynomial chaotic expansion model corresponding to the new fixed random seed, and the process jumps to the step of using leave-one-out cross-validation to sequentially remove individual samples from the initial training samples and using the remaining samples to train the polynomial chaotic expansion model corresponding to the fixed random seed. A spectral proxy model is constructed by using the polynomial chaotic expansion model after all convergences are collected.

[0011] This invention constructs a surrogate model set through leave-one-out cross-validation, accurately selecting candidate samples with the highest misclassification probability to expand the training set. This enables adaptive iterative optimization of the multinomial chaotic expansion model, effectively improving the model's fitting accuracy and generalization ability. After convergence is determined by the first convergence metric, a spectral surrogate model is constructed by combining multiple models, which can more comprehensively characterize the impact of multi-source uncertainties.

[0012] In one optional implementation, the polynomial chaotic expansion model after all sets converge, constructing a spectral surrogate model, includes: Extract the model coefficients and structural performance response data corresponding to the load response of the fully converged polynomial chaotic expansion model; Based on the Caronan-Louis method, using all the model coefficients and all the structural performance response data, the mean term, eigenvalues, characteristic functions, and random coefficients of the response are calculated; By integrating the response mean, the eigenvalues, the feature functions, and the random coefficients, a spectral proxy model is constructed.

[0013] This invention extracts the coefficients and response data of a convergent PCE model, uses the Caronan-Louis method to calculate key parameters, and integrates them to construct a spectral surrogate model, enabling unified representation and dimensionality reduction of uncertainty information under multiple operating conditions. This model effectively integrates model information under various random seeds, taking into account both explicit and implicit uncertainties, and significantly improves the model's generalization ability and prediction stability.

[0014] In one optional implementation, the convergence verification of the spectral surrogate model and the execution of Monte Carlo simulation on the verified spectral surrogate model to obtain the predicted failure probability of the wind turbine tower structure include: Determine whether the spectral proxy model satisfies the preset second convergence index; If not, then proceed to the step of setting each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and using the design parameters of the wind turbine tower structure and any of the random seeds to construct the corresponding polynomial chaotic expansion model under each random seed; If so, Monte Carlo simulation will be performed on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

[0015] This invention achieves closed-loop verification of model accuracy by judging the spectral surrogate model using a second convergence criterion, ensuring that the model meets reliability requirements. If the requirements are not met, iterative optimization is returned, effectively avoiding prediction errors caused by model bias; Monte Carlo simulation is then performed after the convergence requirements are met, significantly improving the stability and accuracy of failure probability calculation.

[0016] In one alternative implementation, the method further includes: Determine whether the predicted failure probability of the wind turbine tower structure is greater than or equal to a preset failure threshold; If so, an early warning message will be generated and sent to the operations and maintenance personnel.

[0017] This invention can determine the safety risk level of the wind turbine tower structure in real time by comparing the predicted failure probability with a preset threshold. When the threshold is exceeded, an early warning message is automatically generated and pushed to maintenance personnel, enabling proactive risk perception and timely response, effectively avoiding structural failure accidents, improving the intelligence and safety of wind turbine tower operation and maintenance, and ensuring the long-term stable operation of the unit.

[0018] Secondly, the present invention provides a device for predicting the failure probability of a wind turbine tower structure, the device comprising: The module is used to set various uncertainties in the wind power data of the wind turbine generator set as random seeds, and to construct the corresponding polynomial chaotic expansion model under each random seed using the design parameters of the wind turbine tower structure and any of the random seeds. The training module is used to adaptively train and verify the convergence of the polynomial chaotic expansion model corresponding to each random seed, and to collect the converged polynomial chaotic expansion models to construct the spectral surrogate model. The prediction module is used to perform convergence verification on the spectral proxy model and execute Monte Carlo simulation on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

[0019] Thirdly, the present invention provides an electronic device, comprising: 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 computer instructions to perform the wind turbine tower structure failure probability prediction method of the first aspect or any corresponding embodiment described above.

[0020] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the wind turbine tower structure failure probability prediction method of the first aspect or any corresponding embodiment thereof.

[0021] Fifthly, the present invention provides a computer program product, including computer instructions, which are used to cause a computer to execute the method for predicting the failure probability of a wind turbine tower structure as described in the first aspect or any corresponding embodiment. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the first step in the method for predicting the failure probability of a wind turbine tower structure according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the second process of the method for predicting the failure probability of wind turbine tower structure according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the third process of the method for predicting the failure probability of wind turbine tower structure according to an embodiment of the present invention; Figure 4 These are constraints according to embodiments of the present invention. Response diagram; Figure 5 These are constraints according to embodiments of the present invention. Response diagram; Figure 6 These are constraints according to embodiments of the present invention. Response diagram; Figure 7 This is a schematic diagram of the tower structure model parameters and stress conditions according to an embodiment of the present invention; Figure 8 According to an embodiment of the present invention The probability density curve of the corresponding random prediction; Figure 9 According to an embodiment of the present invention The probability density curve of the corresponding random prediction; Figure 10 According to an embodiment of the present invention The probability density curve of the corresponding random prediction; Figure 11 This is a structural block diagram of a wind turbine tower structure failure probability prediction device according to an embodiment of the present invention; Figure 12 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying 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. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] This invention provides a method for predicting the failure probability of wind turbine tower structures. By setting uncertainties in wind data as random seeds, a multinomial chaotic expansion model is constructed and adaptively trained under each random seed. This effectively takes into account both explicit and implicit multi-source uncertainties in the tower's dynamic response, avoiding the limitation of traditional deterministic surrogate models that can only represent a single operating condition. A spectral surrogate model is constructed using an ensemble convergent PCE model and convergence verification is completed, significantly improving the model's generalization ability and prediction stability. Furthermore, Monte Carlo simulation is combined to achieve accurate calculation of the failure probability. Compared to existing technologies, this solution can comprehensively quantify the impact of multi-source uncertainties, significantly improving the accuracy of wind turbine tower failure probability prediction.

[0028] According to an embodiment of the present invention, a method for predicting the failure probability of a wind turbine tower structure is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] This embodiment provides a method for predicting the failure probability of wind turbine tower structures. Figure 1 This is a flowchart of a method for predicting the failure probability of a wind turbine tower structure according to an embodiment of the present invention, as shown below. Figure 1 As shown, the process includes the following steps: Step S101: Set each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and use the design parameters of the wind turbine tower structure and any random seed to construct the corresponding polynomial chaotic expansion model under each random seed.

[0030] It should be noted that a wind turbine generator set refers to a system consisting of a wind turbine generator, a control system, and energy storage equipment.

[0031] Wind data refers to datasets that include parameters such as wind speed, wind direction, and turbulence intensity.

[0032] Uncertain factors refer to variables such as wind strength, wind direction changes, and material loss that cannot be accurately predicted in advance.

[0033] Random seeds are used to characterize the uncertainty in wind data. They can be used to generate specific working conditions and simulate load scenarios under different environments.

[0034] The wind turbine tower structure refers to the structure consisting of the tower body, foundation, and connecting components.

[0035] Design parameters refer to core design indicators such as the geometric dimensions, material specifications, and connection methods of the wind turbine tower.

[0036] The polynomial chaotic expansion model refers to a surrogate model based on polynomial expansion, which can establish a mapping relationship between input parameters (such as wind force and material properties) and structural performance (such as tower stability).

[0037] In this embodiment of the invention, after obtaining the wind power data of the wind turbine generator set and the design parameters of the wind turbine tower, various uncertain factors in the wind power data are set as random seeds. Any random seed is combined with the design parameters of the wind turbine tower to construct a polynomial chaotic expansion model of the load response of the wind turbine tower structure under each random seed, thereby obtaining multiple polynomial chaotic expansion models.

[0038] Step S102: Adaptively train and verify the convergence of the polynomial chaotic expansion model corresponding to each random seed, and then combine the converged polynomial chaotic expansion models to construct the spectral surrogate model.

[0039] It should be noted that adaptive training refers to the training process of dynamically adjusting model parameters based on model operation feedback, so that the model can adapt to the working conditions corresponding to different random seeds and improve prediction accuracy.

[0040] Convergence verification refers to the verification process of comparing the model's prediction results with actual performance data to determine whether the model's output meets the preset standards, thereby ensuring the model's stability and prediction reliability.

[0041] The spectral surrogate model refers to a predictive model that integrates multiple convergent polynomial chaotic expansion models, is compatible with multi-source uncertainties, and covers different operating conditions.

[0042] In this embodiment of the invention, after obtaining the wind power data of the wind turbine generator set, the design parameters of the wind turbine tower and each random seed, the polynomial chaotic expansion model under each random seed is first adaptively trained, the model parameters are adjusted according to the actual operation scenario, and convergence verification is carried out simultaneously to determine whether the model output meets the preset performance requirements. After all the models corresponding to all random seeds have reached the convergence criteria, all converged polynomial chaotic expansion models are collected, and the core parameters and performance data of each model are integrated to construct a spectral surrogate model that can be compatible with multiple operating conditions and cover multiple uncertain factors.

[0043] Step S103: Perform convergence verification on the spectral surrogate model, and execute Monte Carlo simulation on the verified spectral surrogate model to obtain the predicted failure probability of the wind turbine tower structure.

[0044] It should be noted that Monte Carlo simulation refers to the process of simulating the operating state of wind turbine towers under different environments through a large number of random simulation calculations, and accurately calculating the probability of structural failure.

[0045] Predicted failure probability refers to the probability that a wind turbine tower may experience structural failure during long-term operation, as determined through simulation calculations.

[0046] In this embodiment of the invention, the constructed spectral proxy model is rigorously converged and verified. After confirming that the model output meets the preset performance standards and has no obvious deviation, the verified spectral proxy model is subjected to Monte Carlo simulation. Through multi-scenario simulation calculation, the predicted failure probability of the wind turbine tower structure is accurately calculated.

[0047] This embodiment provides a method for predicting the failure probability of wind turbine tower structures. Figure 2 This is a flowchart of a method for predicting the failure probability of a wind turbine tower structure according to an embodiment of the present invention, as shown below. Figure 2 As shown, the process includes the following steps: Step S201: Set each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and use the design parameters of the wind turbine tower structure and any random seed to construct the corresponding polynomial chaotic expansion model under each random seed.

[0048] In some optional implementations, step S201 above includes: Step S2011: Obtain wind power data and design parameters of the wind turbine tower structure.

[0049] In this embodiment of the invention, the wind power data of the wind turbine generator is mainly obtained from wind power sensors, generator operation logs, publicly available wind energy monitoring data in the industry, and on-site measured data. At the same time, it is combined with the generator's historical operation records to ensure the authenticity and completeness of the data. The design parameters of the wind turbine tower structure are derived from the previous engineering design scheme, tower material test reports, industry design standards, and actual construction parameters, covering core information such as geometric dimensions, material specifications, and connection methods.

[0050] Step S2012: Set each uncertainty factor of the wind data as multiple random seeds and fix them.

[0051] In embodiments of the present invention, such as Figure 3 As shown, a random seed for a wind model is generated randomly. And fix it. Specifically, for various uncertainties in wind data, such as wind speed fluctuations and environmental changes, each type of uncertainty is set as an independent random seed. Then, through a preset parameter locking mechanism, the value range and corresponding working conditions of each random seed are fixed to ensure that each random seed always corresponds to a specific environmental load scenario, such as the load state under different wind speeds and different ambient temperatures, and there will be no parameter fluctuations.

[0052] Specifically, random seed This refers to the random seed of the wind model. The specific engineering background is: when simulating a wind turbine, it is necessary to generate a wind model. If the wind model is a turbulent wind, it is necessary to specify a random seed for it. The random seed represents a certain turbulent wind time series. The random seed can be drawn by taking any integer.

[0053] Step S2013: Input the design parameters into the preset design space to generate initial training samples.

[0054] It should be noted that the design space refers to the range of parameters that includes all possible combinations of design parameters and operating scenarios, used to generate training samples that meet actual needs.

[0055] Initial training samples refer to basic data samples generated based on design parameters and random seeds.

[0056] In this embodiment of the invention, initial training sample points are generated in the design space according to the design parameters of the wind turbine tower structure.

[0057] Step S2014: Using the initial training samples and fixed random seeds, construct a multinomial chaotic expansion model of the load response of the wind turbine tower structure under each random seed.

[0058] It should be noted that load response refers to the structural mechanical feedback generated by the wind turbine tower structure under external loads such as wind load, its own weight, and operational vibration. It mainly includes quantifiable structural response results such as stress, strain, displacement, deformation, and internal forces.

[0059] In this embodiment of the invention, a random seed is generated randomly. This random seed is then fixed, ensuring that the performance response at each design point in the design space is uniquely determined. Then, using a Sobol pseudo-random sequence, initial training samples are generated in the global design space, and a PCE model (i.e., a multinomial chaotic expansion model) of the load response of the wind turbine tower structure under this random seed is constructed. The PCE model structure is represented as follows:

[0060] In the formula, A It is the set of multiple indices that define the polynomial basis in PCE. a It is a multivariate index used to identify terms in a multivariate polynomial. This represents the orthogonal basis of the PCE model.

[0061] It is worth mentioning that Sobol pseudo-random sequences are samples of input variables for the performance response function, such as the geometric design variables of the tower. Under a certain wind random seed, it is possible to conduct research on the response under different performance function inputs. To explore the performance response in the global design space, Sobol sampling is used to generate initial samples in the global space.

[0062] Step S202: Adaptively train and verify the convergence of the polynomial chaotic expansion model corresponding to each random seed, and then combine the converged polynomial chaotic expansion models to construct the spectral surrogate model.

[0063] In some optional implementations, step S202 above includes: Step S2021: Leave-one-out cross-validation is used to sequentially remove individual samples from the initial training samples, and the remaining samples are used to train the multinomial chaotic expansion model corresponding to the fixed random seed.

[0064] It should be noted that leave-one-out cross-validation is a model validation method that optimizes model accuracy by sequentially removing individual samples from the initial training samples, using the remaining samples to train the model, and then validating the model performance with the removed samples.

[0065] In this embodiment of the invention, a surrogate model set lPCE is constructed based on the current training sample set using leave-one-out cross-validation. Specifically, for the current... n The training sample set consisting of 10 samples is used to train the PCE model by removing one training sample at a time and using the remaining samples.

[0066] Step S2022: Combine all trained polynomial chaotic expansion models to generate a surrogate model set.

[0067] It should be noted that the proxy model set refers to a model cluster formed by integrating all trained multinomial chaotic expansion models corresponding to different random seeds.

[0068] In this embodiment of the invention, all trained polynomial chaotic expansion models are aggregated to form a surrogate model set. Among them, superscript ( i () indicates removing the first training sample from the set. i The PCE model was trained after a number of samples.

[0069] Step S2023: Evaluate the misclassification probability of each candidate sample in the design space using a set of surrogate models.

[0070] It should be noted that candidate samples refer to various parameter combinations within the design space used for model validation and performance evaluation.

[0071] The probability of misclassification refers to the probability that a candidate sample is misclassified in the model's prediction.

[0072] In this embodiment of the invention, the lPCE set is used to evaluate the misclassification probability of global candidate samples. The misclassification probability of candidate samples is determined by the learning function shown in the following formula:

[0073] In the formula, A learning function metric representing the probability of misclassification. Represents candidate sample points, Indicates the determination of candidate samples The number of lPCE models in the safe region, and conversely, the number of lPCE models in the failure region. And satisfy the relation .

[0074] Step S2024: Select the candidate sample with the highest misclassification probability and add it to the initial training sample to generate an updated training sample.

[0075] It should be noted that updating the training samples refers to adding the candidate samples with the highest misclassification probability to the initial training samples to form a new sample set.

[0076] In this embodiment of the invention, the sample point with the highest misclassification probability is selected and added to the training set to obtain updated training samples.

[0077] Specifically, when U PM =1 means that all lPCE models determine that the candidate sample is in the safe region or the failure region, and the current PCE model can correctly classify whether the response to the candidate sample is failed; otherwise, when U PM When = 0, it means that the current PCE model has the highest probability of misclassifying the response at that candidate sample. Therefore, U is selected from the global candidate samples. PM The candidate sample with the smallest value is added to the training sample set.

[0078] Step S2025: Retrain the trained polynomial chaotic expansion model using updated training samples, and determine whether the retrained polynomial chaotic expansion model satisfies the preset first convergence index.

[0079] It should be noted that the first convergence metric refers to a preset standard used to judge whether the model is qualified after retraining. It usually includes requirements such as prediction error range and model stability. Meeting this metric means that the model has achieved the expected training effect.

[0080] In this embodiment of the invention, the fixed random seed is updated. The PCE model is then evaluated, and its convergence criterion is determined. If the convergence criterion is met, the next step is executed. Specifically, the generalization ability of the PCE model is determined using the mean squared error as the convergence criterion, as shown in the following formula:

[0081] In the formula, y (l) Indicates the first i training samples x i The actual response at the location express In the i training samples The predicted value at that location, n This indicates the number of training sample points.

[0082] Step S2026: If not, set the retrained polynomial chaotic expansion model as the polynomial chaotic expansion model corresponding to the new fixed random seed, and jump to the step of using leave-one-out cross-validation to remove individual samples from the initial training samples in turn, and using the remaining samples to train the polynomial chaotic expansion model corresponding to the fixed random seed.

[0083] In this embodiment of the invention, if the first convergence criterion is not met, step S2021 is executed, and the leave-one-out cross-validation technique is re-executed to sequentially remove individual samples from the initial training samples, train the retrained polynomial chaotic expansion model, and construct a surrogate model set, etc.

[0084] Step S2027: Collect all converged polynomial chaotic expansion models and construct a spectral proxy model.

[0085] Specifically, step S2027 above includes: Step a1: Extract the model coefficients and structural performance response data corresponding to the load response of the fully converged polynomial chaotic expansion model.

[0086] It should be noted that model coefficients refer to the core parameters used to construct the polynomial expansion in a polynomial chaotic expansion model, which determine the mapping relationship between input and performance response (output).

[0087] Structural performance response data refers to the data (such as load and stability) output by a polynomial chaotic expansion model under specific input parameters (such as random seeds and design parameters) that are related to the structural performance of the wind turbine tower.

[0088] In this embodiment of the invention, model coefficients and structural performance response data calculated by load response are extracted from PCE models trained under different random seeds.

[0089] Step a2: Based on the Caronan-Louis method, using all model coefficients and all structural performance response data, calculate the response mean, eigenvalues, characteristic functions, and random coefficients.

[0090] It should be noted that the Carlo Nelson-Louis method refers to a data analysis method based on orthogonal decomposition. Its core function is to integrate and extract features from multiple sets of data, such as model coefficients and structural performance response data, to achieve data dimensionality reduction and core information mining.

[0091] The mean response term refers to the average value of the structural performance response data calculated using the Caronan-Louis method.

[0092] Eigenvalues ​​refer to the core feature parameters obtained after orthogonally decomposing relevant data based on the Karonan-Louis method.

[0093] The characteristic function refers to the function obtained by orthogonal decomposition using the Caronanello method.

[0094] Random coefficients refer to parameters related to operating condition fluctuations, calculated based on the Caronanello method.

[0095] In this embodiment of the invention, the response mean, eigenvalues, eigenfunctions, and random coefficients required to construct the spectral proxy model are calculated.

[0096] The performance response of a wind turbine tower structure under implicit multi-source uncertainties is characterized as a spatial random field using the Karhunen–Loève expansion. The mathematical formula for the random field characterized by the Karhunen–Loève expansion is as follows:

[0097] In the formula, Represents the mean response function. and To construct the eigenvalues ​​and eigenfunctions required for the spectral surrogate model, These are the random coefficients required to construct the spectral surrogate model. Since the covariance function in the KL expansion is difficult to obtain from high-dimensional problems through integration, further analysis of the coefficients of each part of the KL expansion using other surrogate modeling techniques is necessary. By controlling different random seeds ω, structural response surfaces under corresponding random seeds are formed, and PCE models of the corresponding response surfaces are constructed. The coefficients of the KL expansion are analyzed using PCE models of multiple PCE trajectories, thereby constructing the spectral surrogate model.

[0098] Specifically, the mean response function The proxy is calculated using the following formula:

[0099] In the formula, R This represents the number of PCE models constructed through steps S201 to S2026, and also the number of random seeds extracted.

[0100] Specifically, eigenvalues With characteristic function The solution to the integral eigenvalue problem is as follows:

[0101] In the formula, x 'Indicates and x A different set of input variables. The covariance c(x,x') of the discrete training samples is expressed as follows:

[0102] in, Indicates the center PCE, Indicates the first r ( r =1,…, R ) PCE models. In the Karhunen-Loève expansion, the eigenvalues ​​{ All values ​​are non-negative and arranged in descending order. This sequence quantifies the contribution of each corresponding eigenfunction to the total variance of the random field. The rapid decay of eigenvalues ​​means that the main features of the random field can be captured by the first few dominant modes. Therefore, a small threshold parameter ε = 0.001 is usually set, and the expansion is placed in the first few modes. K The term is truncated as shown in the following formula:

[0103] Specifically, random coefficients This is the projection of the stochastic performance response of the tower structure onto the characteristic function, and the projection formula is shown below:

[0104] in, It is the probability density function of the input variables. Based on the obtained discrete random coefficients... Reconstructing random coefficients using Copula or kernel density estimation Probability density statistical distribution When the spectral proxy model is invoked to predict the performance response output of the wind turbine tower structure, it will be from... A random coefficient is randomly selected from the data. Then, the performance response output is predicted based on the mean response function and the eigenvalues ​​and eigenfunctions. When under the same design input... x When the spectral proxy model is repeatedly invoked, the spectral proxy model outputs... y Due to the random coefficient The changes follow a certain statistical distribution .

[0105] Step a3: Integrate the response mean, eigenvalues, eigenfunctions, and random coefficients to construct a spectral surrogate model.

[0106] In this embodiment of the invention, the mean term of the response, eigenvalues, eigenfunctions and random coefficients are integrated and optimized to construct a spectral proxy model that can cover multi-source uncertainties and adapt to multiple operating conditions of wind turbine towers, providing efficient and accurate model support for subsequent failure probability calculation and safety assessment of wind turbine tower structures.

[0107] Step S203: Perform convergence verification on the spectral surrogate model, and execute Monte Carlo simulation on the verified spectral surrogate model to obtain the predicted failure probability of the wind turbine tower structure.

[0108] In some optional implementations, step S203 above includes: Step S2031: Determine whether the spectral surrogate model satisfies the preset second convergence index.

[0109] It should be noted that the second convergence metric refers to a pre-defined standard used to judge whether the spectral surrogate model is qualified. It is different from the convergence metric used in the early model training and focuses on verifying the stability and prediction accuracy of the spectral surrogate model.

[0110] In this embodiment of the invention, the second-order Watson-Stokes distance output by the spectral proxy model before and after the update is calculated, and it is determined whether the distance is less than a preset convergence threshold (e.g., 0.05). The specific formula is as follows:

[0111] In the formula, Denotes the second-order Watson-Standard distance. and These represent the spectral proxy models before and after the update, respectively. and They are and The inverse cumulative distribution function.

[0112] It is worth mentioning that by integrating the response mean, eigenvalues, eigenfunctions, and random coefficients, and optimizing the parameters, the first spectral surrogate model constructed is the updated spectral surrogate model. This is achieved by subtracting the second-order Watson-Stokes distance of the first spectral surrogate model from the second-order Watson-Stokes distance (denoted as 0) of the previous spectral surrogate model (before it was constructed). Similarly, if a second spectral surrogate model is reconstructed, the updated second spectral surrogate model is subtracted from the first spectral surrogate model.

[0113] Step S2032, if not, then proceed to the step of setting each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and using the design parameters of the wind turbine tower structure and any random seed to construct the corresponding polynomial chaotic expansion model under each random seed.

[0114] In this embodiment of the invention, if the second-order Watson-Stokes distance output by the spectral proxy model before and after the update is greater than 0.05, step S201 needs to be executed again.

[0115] Step S2033: If so, perform Monte Carlo simulation on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

[0116] In this embodiment of the invention, if the second-order Watson-Stokes distance output by the spectral proxy model before and after the update is less than 0.05, the specific method for performing Monte Carlo simulation on the validated spectral proxy model is as follows: Based on the validated spectral proxy model, and combined with the actual operating scenario of the wind turbine tower, a large number of random samples are used to generate operating condition samples corresponding to different environmental loads (such as wind speed and wind direction changes). These samples are substituted into the spectral proxy model for repeated simulation calculations. By statistically analyzing all simulation results, the predicted failure probability of the wind turbine tower structure is finally accurately calculated. The specific formula is as follows:

[0117] In the formula, The predicted failure probability, N The number of Monte Carlo simulation samples. This is an indicator function used to count the number of failed samples in a Monte Carlo simulation.

[0118] Step S204: Determine whether the predicted failure probability of the wind turbine tower structure is greater than or equal to the preset failure threshold.

[0119] It should be noted that the preset failure threshold refers to the critical value preset based on the structural strength of the wind turbine tower, industry safety standards, and operational requirements.

[0120] In this embodiment of the invention, a preset failure threshold is defined (set in conjunction with the safety standards and structural bearing limits of the wind turbine tower), and the predicted failure probability of the wind turbine tower structure is compared with the failure threshold.

[0121] Step S205: If yes, generate an early warning message and send it to the maintenance personnel.

[0122] It should be noted that the warning information refers to the prompt information generated when the predicted failure probability reaches a preset threshold.

[0123] In this embodiment of the invention, if the predicted failure probability of the wind turbine tower structure is greater than or equal to the failure threshold, a corresponding early warning message is generated, which includes key information such as the failure risk level, the predicted failure probability value, and potential failure hazards. The early warning message is then sent to the operation and maintenance personnel, providing clear guidance for them to conduct timely hazard investigation and adjust the operation and maintenance plan.

[0124] In specific embodiments, two examples, Example (1) and Example (2), are provided to further illustrate the implementation of the present invention. The specific examples are as follows: Example (1) To verify the effectiveness of the method of this invention, this implementation case uses the model and data given by Wu (Wu Z, Chen Z, Chen G, et al. A probability feasible region enhanced important boundary sampling method for reliability-based design optimization[J]. Structural and Multidisciplinary Optimization, 2021, 63: 341-355.) and substitutes them into the method of this invention. The experimental case includes two independent design variables that follow a normal distribution and three performance functions with implicit random variables, and implicit random variables are added to the analysis model. For experimental setup, the implicit random variables are made to follow a normal distribution, but the number and distribution of implicit random variables are unknown when actually calculating the performance response. The performance function with implicit random variables in implementation case (1) is expressed as follows:

[0125] In the formula, This is the performance response function; These are random design variables, and they each follow a mean of 1 / 2. , The distribution follows a normal distribution with a standard deviation of 0.3. Let ω be an implicit random variable, and let ω be a random seed. When evaluating the performance response, each randomly selected seed ω is considered a random realization of the implicit random variable, i.e., there exists... Relationship. For the experimental setup, assume that the random distributions of these implicit random parameters satisfy the following: , , , .

[0126] A "failure event" is defined as occurring when the performance function response G < 0. (Selection) , and , Two sets of design values ​​are used to predict the failure probability under the corresponding design. In this embodiment, the PCE model is constructed using the open-source code UQ Lab, the PCE coefficients are calculated using LARS, and the PCE is adaptively truncated using the q-norm method. First, 10 Sobol sequence samples are generated based on the 2D input variables, and the PCE model is adaptively trained under a fixed random seed ω until the model training converges. Then, a spectral surrogate model of the performance function is constructed. To ensure the continuity of the convergence of the spectral surrogate model, it is recommended to use a batch training method to train the spectral surrogate model, that is, every 5 PCE models are trained as a batch, and then the spectral surrogate model is updated, and this process is repeated until the spectral surrogate model training converges. Finally, when using the spectral surrogate model for MCS reliability analysis, 106 samples are used for analysis.

[0127] The spectral surrogate model trained using the method of this invention can effectively reproduce the randomness of performance response with implicit random parameters, such as... Figure 4 , Figure 5 , Figure 6 As shown in Table 1, the reliability analysis results are as follows. This represents the failure probability. Compared to direct MCS methods, the method of this invention overcomes the limitation that traditional deterministic surrogate models cannot surrogate performance response systems with implicit random variables, and significantly reduces the computational cost of reliability analysis. This is mainly because this method uses training samples based on Karhunen–Loève expansion and adaptive sampling, employs a PCE model and kernel density estimation (or Copula function) to perform global statistical inference on the randomness of the random field, and performs MCS based on the statistical characteristics of the inference, rather than blindly performing MCS directly. Furthermore, due to the random coefficients of the spectral surrogate model... It is obtained through global training sample analysis, therefore the trained spectral surrogate model can also predict different global designs. The probability of failure.

[0128] Table 1 Comparison of Calculation Results for Implementation Case 1

[0129] Example (2) This embodiment uses a case study of predicting the failure probability of a 5MW wind turbine tower structure under operating condition 1.3 (turbulent wind condition) as specified in IEC 61400-9 to test the effectiveness of the method of this invention. The external shape of the tower structure is determined by the outer diameter of the top of the tower structure. d and bottom outer diameter D A cone is formed by linear interpolation. Tower structures designed with two sets of parameters—d=2.38m and D=4.98m, and d=3.87m and D=6.00m—are used as the objects of reliability analysis. The performance function is derived from the ultimate stress. and the twist angle at the top of the tower The composition, tower structure model parameters, and stress conditions are as follows: Figure 7 As shown in Table 2, the performance function is defined as follows:

[0130] Table 2 Parameters for predicting the failure probability of 5MW wind turbine tower structure

[0131] in, D and d These are the diameters of the bottom and top of the tower structure, respectively, and are due to manufacturing errors. d and D It will follow a specific probability distribution. In this implementation case, it is assumed to follow a normal distribution, as shown in Table 2. and These are the ultimate stress constraints and the torsional angle constraints at the top of the tower structure, respectively. This is the allowable ultimate stress value. This is the allowable value for the ultimate torsion angle at the top of the tower; , , These are the gravity load at the base of the tower, the bending moment at the base of the tower, and the torque at the top of the tower, respectively. They characterize the load response of the tower structure under the combined action of multiple random excitations from wind, waves, and current. Here, ω is the random seed set in the simulation. The cross-sectional area of ​​the tower base is... The section modulus of bending. The moment of inertia at the top of the tower is given. The load response in the tower structure performance function is obtained by high-fidelity time-domain simulation of the wind turbine system using OpenFAST, based on the turbulent wind generated by TurbSim.

[0132] As a limiting constraint, the IEC 61400-1 standard specifies that the maximum value of the time-series load simulated over 10 minutes should be used as the ultimate load response. Under turbulent wind conditions, even with the same simulation input settings, the presence of the random seed ω will result in a random variable for the simulated extreme load. Since the random seed only represents a specific turbulent wind time series and does not directly participate in the simulation calculation of the tower structure load, the random seed ω cannot be used as a valid uncertainty input parameter. Certain implicit random parameters exist in the process of simulating extreme loads, making... , , The randomness of the implementation is difficult to predict accurately. Therefore, ordinary deterministic proxy models such as Kriging and PCE cannot directly proxy the performance function of the implementation case (2).

[0133] The load data obtained from 1063 simulations using the method of this invention is input into the performance function, and a spectral surrogate model of the performance function is trained. The prediction results of the trained spectral surrogate model are compared with 10,000 sets of OpenFast simulation samples. The spectral surrogate model can better reproduce the randomness of extreme loads, as shown in the following figures. Figure 8 , Figure 9 , Figure 10 As shown.

[0134] The prediction results of wind turbine failure probability using the method of this invention are shown in Table 3. With only a small number of simulations, the method of this invention achieves more accurate and reliable optimization results, demonstrating its contribution to solving failure probabilities containing multi-source implicit random parameters. Furthermore, due to the random coefficients of the spectral surrogate model... It is obtained through global training sample analysis, therefore the trained spectral surrogate model can also realize the tower structure under the 1.3 working condition in other different designs. d and D Failure probability prediction.

[0135] Table 3 Summary of Optimization Results for Wind Turbine Tower Structure

[0136] To address the challenge of accurately and efficiently predicting failure probabilities in wind turbines due to the high uncertainty of their service environment, this invention combines Karhunen-Loève expansion, PCE models, statistical modeling methods, and adaptive sampling techniques to construct a spectral surrogate model for the performance response of wind turbine tower structures. Unlike existing deterministic surrogate models, this spectral surrogate model can analyze the stochastic characteristics of the wind turbine tower structure performance response system with implicit random variables based on training sample data. Ultimately, this invention uses this spectral surrogate model to predict the failure probability of wind turbine tower structures of different designs. It effectively considers the impact of multi-source implicit uncertainties such as wind, waves, and currents on the performance response uncertainty of wind turbine tower structures, thus overcoming the difficulties of ordinary surrogate models being unable to directly surrogate the stochastic performance response system of wind turbine tower structures and the low efficiency of direct MCS, achieving accurate and efficient prediction of the failure probability of wind turbine tower structures.

[0137] This embodiment also provides a wind turbine tower structure failure probability prediction device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0138] This embodiment provides a device for predicting the failure probability of wind turbine tower structures, such as... Figure 11 As shown, it includes: Module 301 is used to set various uncertainties in the wind power data of the wind turbine generator set as random seeds, and to construct the corresponding polynomial chaotic expansion model under each random seed using the design parameters of the wind turbine tower structure and any random seed. Training module 302 is used to adaptively train and verify the convergence of the corresponding polynomial chaotic expansion model under each random seed, and to collect the converged polynomial chaotic expansion model to construct the spectral surrogate model. The prediction module 303 is used to perform convergence verification on the spectral surrogate model and execute Monte Carlo simulation on the verified spectral surrogate model to obtain the predicted failure probability of the wind turbine tower structure.

[0139] In some alternative implementations, the construction module 301 includes: The acquisition unit is used to acquire wind power data and design parameters of the wind turbine tower structure. The setting unit is used to set various uncertainties in wind data as multiple random seeds and fix them; The input unit is used to input design parameters into a preset design space to generate initial training samples. Model units are constructed to build a multinomial chaotic expansion model of the load response of the wind turbine tower structure under each random seed, using initial training samples and fixed random seeds.

[0140] In some alternative implementations, training module 302 includes: The training unit is used to sequentially remove individual samples from the initial training samples using leave-one-out cross-validation, and then use the remaining samples to train the multinomial chaotic expansion model corresponding to the fixed random seed. The combining unit is used to combine all trained multinomial chaotic expansion models to generate a surrogate model set. Evaluation unit, used to evaluate the misclassification probability of each candidate sample in the design space through a set of surrogate models; Add a unit to filter the candidate sample with the highest misclassification probability and add it to the initial training sample to generate an updated training sample; The retraining unit is used to retrain the trained polynomial chaotic expansion model using updated training samples, and to determine whether the retrained polynomial chaotic expansion model satisfies the preset first convergence index. The jump execution unit is used to, if not, set the retrained polynomial chaotic expansion model as the new polynomial chaotic expansion model corresponding to the fixed random seed, and jump to execute the step of using leave-one-out cross-validation to remove individual samples from the initial training samples in turn, and using the remaining samples to train the polynomial chaotic expansion model corresponding to the fixed random seed. The set unit is used to collect all converged polynomial chaotic expansion models and construct spectral surrogate models.

[0141] In some alternative implementations, the collection unit includes: Extract sub-units to extract the model coefficients and structural performance response data corresponding to the load response of the fully converged polynomial chaotic expansion model; The computational sub-unit is used to calculate the mean term, eigenvalues, characteristic functions, and random coefficients of the response based on the Caronanello method, using all model coefficients and all structural performance response data. The integration subunit is used to integrate the response mean term, eigenvalues, characteristic functions, and random coefficients to construct a spectral surrogate model.

[0142] In some alternative implementations, the prediction module 303 includes: The judgment unit is used to determine whether the spectral proxy model satisfies the preset second convergence index. The jump unit is used to jump to the execution of the steps of setting each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and using the design parameters of the wind turbine tower structure and any random seed to construct the corresponding polynomial chaotic expansion model under each random seed. The simulation unit is used to perform Monte Carlo simulation on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

[0143] In some alternative embodiments, the device further includes: The threshold judgment unit is used to determine whether the predicted failure probability of the wind turbine tower structure is greater than or equal to the preset failure threshold. The sending unit is used to generate an early warning message and send it to the operation and maintenance personnel if the condition is met.

[0144] The wind turbine tower structure failure probability prediction device provided in this embodiment of the invention can execute the wind turbine tower structure failure probability prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0145] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0146] The following is a detailed reference. Figure 12 This diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 401, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 402 or a program loaded from memory 408 into random access memory (RAM) 403. RAM 403 also stores various programs and data required for the operation of the electronic device. The processor 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0147] Typically, the following devices can be connected to I / O interface 405: input devices 406 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 407 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 408 including, for example, magnetic tapes, hard disks, etc.; and communication devices 409. Communication device 409 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 12 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0148] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 409, or installed from a memory 408, or installed from a ROM 402. When the computer program is executed by the processor 401, it performs the functions defined in the wind turbine tower structure failure probability prediction method of the embodiments of the present invention.

[0149] Figure 12 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0150] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the wind turbine tower structure failure probability prediction method shown in the above embodiments is implemented.

[0151] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0152] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method for predicting the failure probability of a wind turbine generator tower structure, characterized in that, The method includes: Each uncertainty factor in the wind power data of the wind turbine generator is set as a random seed, and the design parameters of the wind turbine tower structure and any of the random seeds are used to construct the corresponding polynomial chaotic expansion model under each random seed. Adaptive training and convergence verification are performed on the polynomial chaotic expansion models corresponding to each random seed. The converged polynomial chaotic expansion models are then combined to construct a spectral surrogate model. The spectral proxy model is converged and verified, and Monte Carlo simulation is performed on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

2. The method according to claim 1, characterized in that, The step of setting various uncertainties in the wind power data of the wind turbine generator as random seeds, and using the design parameters of the wind turbine tower structure and any of the random seeds to construct the corresponding polynomial chaotic expansion model under each random seed, includes: Obtain wind power data and design parameters for wind turbine tower structures; Each uncertainty factor in the wind force data is set as multiple random seeds and then fixed. Input the design parameters into a preset design space to generate initial training samples; Using the initial training samples and the fixed random seeds, a multinomial chaotic expansion model of the load response of the wind turbine tower structure under each random seed is constructed.

3. The method according to claim 2, characterized in that, The process of adaptively training and verifying the convergence of the polynomial chaotic expansion models corresponding to each random seed, and then combining the converged polynomial chaotic expansion models to construct a spectral surrogate model includes: Leave-one-out cross-validation is used to sequentially remove individual samples from the initial training samples, and the remaining samples are used to train the polynomial chaotic expansion model corresponding to the fixed random seed. Combine all trained multinomial chaos expansion models to generate a surrogate model set. The misclassification probability of each candidate sample in the design space is evaluated using the surrogate model set. The candidate samples with the highest misclassification probability are selected and added to the initial training samples to generate updated training samples; The updated training samples are used to retrain the trained polynomial chaotic expansion model, and it is determined whether the retrained polynomial chaotic expansion model satisfies the preset first convergence index. If not, the retrained polynomial chaotic expansion model is set as the polynomial chaotic expansion model corresponding to the new fixed random seed, and the process jumps to the step of using leave-one-out cross-validation to sequentially remove individual samples from the initial training samples and using the remaining samples to train the polynomial chaotic expansion model corresponding to the fixed random seed. A spectral proxy model is constructed by using the polynomial chaotic expansion model after all convergences are collected.

4. The method according to claim 3, characterized in that, The polynomial chaotic expansion model after all sets converge, constructing a spectral proxy model, includes: Extract the model coefficients and structural performance response data corresponding to the load response of the fully converged polynomial chaotic expansion model; Based on the Caronan-Louis method, using all the model coefficients and all the structural performance response data, the mean term, eigenvalues, characteristic functions, and random coefficients of the response are calculated; By integrating the response mean, the eigenvalues, the feature functions, and the random coefficients, a spectral proxy model is constructed.

5. The method according to claim 1, characterized in that, The process of performing convergence verification on the spectral surrogate model and executing Monte Carlo simulation on the verified spectral surrogate model to obtain the predicted failure probability of the wind turbine tower structure includes: Determine whether the spectral proxy model satisfies the preset second convergence index; If not, then proceed to the step of setting each uncertainty factor of the wind power data of the wind turbine generator set as a random seed, and using the design parameters of the wind turbine tower structure and any of the random seeds to construct the corresponding polynomial chaotic expansion model under each random seed; If so, Monte Carlo simulation will be performed on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

6. The method according to claim 1, characterized in that, This method also includes: Determine whether the predicted failure probability of the wind turbine tower structure is greater than or equal to a preset failure threshold; If so, an early warning message will be generated and sent to the operations and maintenance personnel.

7. A device for predicting the failure probability of a wind turbine tower structure, characterized in that, The device includes: The module is used to set various uncertainties in the wind power data of the wind turbine generator set as random seeds, and to construct the corresponding polynomial chaotic expansion model under each random seed using the design parameters of the wind turbine tower structure and any of the random seeds. The training module is used to adaptively train and verify the convergence of the polynomial chaotic expansion model corresponding to each random seed, and to collect the converged polynomial chaotic expansion models to construct the spectral surrogate model. The prediction module is used to perform convergence verification on the spectral proxy model and execute Monte Carlo simulation on the verified spectral proxy model to obtain the predicted failure probability of the wind turbine tower structure.

8. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the method for predicting the failure probability of a wind turbine tower structure as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the method for predicting the failure probability of the wind turbine tower structure as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The method includes computer instructions for causing a computer to execute the method for predicting the failure probability of a wind turbine tower structure as described in any one of claims 1 to 6.

Images