Wind turbine generator transmission chain system fatigue load prediction method and related device

Abstract

The invention discloses a wind turbine generator transmission chain system fatigue load prediction method and a related device. The method comprises the steps of obtaining a load time sequence of a wind turbine generator transmission chain system; calculating the fatigue load of the wind turbine generator transmission chain system according to the load time sequence of the wind turbine generator transmission chain system; decomposing the fatigue load of the wind turbine generator transmission chain system to obtain a plurality of subsequences; inputting each subsequence into a corresponding model to obtain a plurality of decomposition component predicted values; and integrating the predicted values of the decomposition components to obtain a prediction result of the fatigue load of the transmission chain system of the wind turbine generator, and the method and the related device can accurately predict the fatigue load of the transmission chain system of the wind turbine generator.

Description

Technical Field

[0001] This invention belongs to the field of fatigue load prediction technology for wind turbine drive train systems, and relates to a method and related device for predicting fatigue loads in wind turbine drive train systems. Background Technology

[0002] During long-term operation, the drivetrain system of a wind turbine is subject to various external environmental and operating conditions, leading to fatigue load exceeding limits and causing transmission system failures and damage, severely affecting the normal operation of the wind turbine. Reducing the fatigue load during turbine operation can effectively improve turbine reliability and extend turbine life.

[0003] In the field of wind turbine load calculation and analysis, the Blade Element Momentum (BEM) model is more efficient than computational fluid dynamics and vortex wake methods, making it the most commonly used load analysis method in various studies. FU calculated and analyzed the influence characteristics of different wind conditions, such as wind shear, on the maximum, minimum, and equivalent fatigue loads of various parts of the wind turbine based on BEM. GHANDOUR proposed a model combining potential flow with BEM to further investigate the impact of the tower shadow effect on the wind turbine blade root load. However, traditional load calculation methods are relatively complex, leading to time-consuming and costly real-time load control strategies for in-service wind turbines, hindering large-scale implementation. Therefore, employing machine learning methods, which have good performance in data prediction and data fitting, to accurately predict the fatigue load of the wind turbine drive system is of great significance.

[0004] In recent years, artificial intelligence-based data analysis techniques have also been applied by scholars to the calculation and analysis of wind turbine loads, further reducing the computational load required for load analysis and improving its efficiency. By predicting the fatigue load of the transmission system, potential problems can be identified in a timely manner, while improving the reliability and stability of wind turbine units. WANG Wei proposed an equivalent fatigue load prediction model for wind turbine blade roots and yaw points based on the Adam algorithm optimization; however, the prediction model is prone to overfitting. ZHANG B proposed a data-driven modeling method for wind turbine fatigue loads based on DNN, establishing the relationship between easily measurable parameters such as wind speed and power and fatigue loads. However, when the operating conditions of the training data become significantly incomplete, the model accuracy decreases to some extent. WANG Xiao proposed a transmission system fatigue load prediction model based on a convolutional bidirectional long short-term memory neural network; however, as fatigue load data is a typical one-dimensional time-series variable, its inherent data and temporal characteristics have not been deeply explored, and there is considerable room for improvement in prediction accuracy. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and related device for predicting fatigue loads in wind turbine drive train systems. This method and related device can accurately predict the fatigue loads in wind turbine drive train systems.

[0006] To achieve the above objectives, this invention discloses a method for predicting fatigue loads in a wind turbine drivetrain system, comprising:

[0007] Obtain the load timing sequence of the wind turbine drivetrain system;

[0008] Calculate the fatigue load of the wind turbine drive train system based on the load timing of the wind turbine drive train system.

[0009] The fatigue load of the wind turbine drive train system is decomposed to obtain several sub-sequences;

[0010] Each subsequence is input into the corresponding model to obtain several decomposed component prediction values;

[0011] The predicted values ​​of each decomposed component are integrated to obtain the predicted results of fatigue load of the wind turbine drive train system.

[0012] A further improvement of the fatigue load prediction method for wind turbine drive train system described in this invention is as follows:

[0013] Furthermore, the process of decomposing the fatigue load of the wind turbine drivetrain system to obtain several sub-sequences is as follows:

[0014] VMD was used to decompose the fatigue load data of the wind turbine drive train system to obtain high-frequency subsequence, medium-frequency subsequence and low-frequency subsequence;

[0015] Furthermore, the process of inputting each subsequence into the corresponding model to obtain several decomposed component prediction values ​​is as follows:

[0016] The high-frequency subsequences are input into the long short-term memory network model;

[0017] The mid-frequency subsequence is input into a deep neural network model;

[0018] The low-frequency subsequence is input into the GBRT model;

[0019] Several predicted values ​​of the decomposed components were obtained.

[0020] Furthermore, it also includes:

[0021] Load timing samples of wind turbine drive train system based on OpenFAST software simulation calculation;

[0022] The fatigue load sample of the wind turbine drivetrain system is calculated based on the load time sequence sample of the wind turbine drivetrain system.

[0023] The correlation between the measurement data of wind turbine generators and the fatigue load samples of wind turbine generator drivetrain systems was analyzed. Feature data that is compatible with fatigue loads was selected, and a database for predicting fatigue loads of wind turbine generator drivetrain systems was constructed based on this.

[0024] The database was used to train the Long Short-Term Memory Network model, the Deep Neural Network model, and the GBRT model.

[0025] Furthermore, the high-frequency subsequence is input into a long short-term memory network model to obtain the decomposed component prediction value y. t for:

[0026] y t =w y ·h t +b

[0027] Among them, h t This is the output of the hidden layer.

[0028] Furthermore, the low-frequency subsequence is input into the GBRT model to obtain the decomposed component prediction value L(yf(x)):

[0029]

[0030] Where x is the input variable; y is the output variable; i is the number of iterations; f0(x) is the initial predicted value; and f(x) is the model predicted value.

[0031] This invention discloses a fatigue load prediction system for a wind turbine drivetrain system, comprising:

[0032] The acquisition module is used to acquire the load timing of the wind turbine drive train system;

[0033] The calculation module is used to calculate the fatigue load of the wind turbine drive train system based on the load timing of the wind turbine drive train system.

[0034] The decomposition module is used to decompose the fatigue load of the wind turbine drive train system to obtain several sub-sequences;

[0035] The prediction module is used to input each subsequence into the corresponding model to obtain several decomposed component prediction values;

[0036] An integration module is used to integrate the predicted values ​​of each decomposed component to obtain the predicted results of the fatigue load of the wind turbine drive train system.

[0037] A further improvement of the fatigue load prediction system for the wind turbine drive train system described in this invention is as follows:

[0038] Furthermore, the prediction module includes:

[0039] The first prediction unit is used to input the high-frequency subsequence into the long short-term memory network model;

[0040] The second prediction unit is used to input the mid-frequency sub-sequence into the deep neural network model;

[0041] The third prediction unit is used to input the low-frequency subsequence into the GBRT model;

[0042] The fourth prediction unit is used to obtain the predicted values ​​of several decomposed components.

[0043] This invention discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the fatigue load prediction method for the wind turbine drive train system.

[0044] The present invention discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the fatigue load prediction method for the wind turbine drive train system.

[0045] The present invention has the following beneficial effects:

[0046] In practical operation, the fatigue load prediction method and related device for wind turbine drivetrain systems described in this invention establish a data-driven "feature selection-decomposition-integration prediction" framework for predicting fatigue loads in wind turbine drivetrain systems. To address the problem of prediction errors caused by the complexity of fatigue load data in wind turbine drivetrain systems, VMD is applied to the fatigue load prediction of wind turbine drivetrain systems. Different models are then used for prediction, and finally integrated to ensure the reliability and stability of the prediction results.

[0047] Furthermore, this invention adopts a decomposition-integration prediction mode. First, the fatigue load data of the wind turbine drive train system is decomposed by VMD. Then, a series of subsequences obtained after decomposition are predicted to obtain prediction components. Finally, the prediction components are integrated to obtain the final prediction result. Attached Figure Description

[0048] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0049] Figure 1This is a flowchart of the method of the present invention;

[0050] Figure 2 A schematic diagram of the feature selection method for MIC-CNN;

[0051] Figure 3 This is a schematic diagram of the VMD decomposition algorithm. Detailed Implementation

[0052] 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, not all, of the embodiments of the present invention. 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.

[0053] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0054] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0055] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0056] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0057] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0058] 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0060] Example 1

[0061] refer to Figure 1 , Figure 2 and Figure 3 The fatigue load prediction method for wind turbine drivetrain system of the present invention includes the following steps:

[0062] 1) Simulation calculation of load timing of wind turbine drive train system based on OpenFAST software;

[0063] 2) The fatigue load of the wind turbine drive train system obtained through calculation;

[0064] In engineering applications, the DEL (Delayed Elasticity Scale) is the most commonly used method for calculating fatigue loads. The calculation of DEL requires specific information from the original signal. Based on the rainflow counting method, the load time sequence of the complex transmission chain system is converted into a load inversion set affecting fatigue, and the calculation is based on the Palmgren-Miner rule and the SN curve. DEL is the amplitude of the sinusoidal stress that produces the same damage as the original signal at a constant frequency f and time T, as shown in the formula:

[0065]

[0066] Among them, DEL TS The equivalent fatigue load of the transmission system; T is the original time history, f is the frequency, Tf = 1, m = 4; N i n is the total number of cycles until the material fails. i For the i-th stress range Next work cycle number.

[0067] 3) Analyze the correlation between measurable data of wind turbine generators and fatigue loads in the wind turbine generator drivetrain system, select characteristic data that are compatible with fatigue loads, and construct a database for predicting fatigue loads in the wind turbine generator drivetrain system. The steps are as follows:

[0068] 31) Select the average value of electromagnetic power The standard deviations of electromagnetic power (std(P)), electromagnetic power variation (std(ΔP)), pitch angle (std(P)), and rotor speed (std(ω)) r ), average wind turbine speed and the average value of the pitch angle As measurable data for wind turbine units.

[0069] 32) The Maximum Information Factor (MIC) is used to measure the correlation between the measurable data of the wind turbine and the fatigue load of the wind turbine drivetrain system. The formula is as follows:

[0070]

[0071] Where Ω(x,y) represents a two-dimensional grid set of size xy, X(G) and Y(G) represent discrete variables based on the grid Ω(x,y), X is the measurable data of the wind turbine, Y is the fatigue load of the wind turbine drivetrain system, I(X(G),Y(G)) represents their mutual information, and B is the maximum number of grids, which is recommended to be B = n. 0.6 .

[0072] 33) Based on the correlation diagram between the load and various influencing factors in the MIC correlation analysis, a convolutional neural network (CNN) model is used to extract the features between the fatigue load of the wind turbine drive chain system and the measurable data of the wind turbine, which is used to construct a feature dataset that is conducive to improving the prediction accuracy.

[0073] The CNN layer mainly consists of convolutional layers, batch normalization layers, nonlinear activation layers, and max pooling layers. Among these, the convolutional layer is the most crucial, extracting features from the wind turbine's operating data through convolutional operations. The batch normalization layer normalizes the extracted features such as the turbine pitch angle, generator electromagnetic power, and rotor speed, improving the model's learning rate and reducing the network's initialization sensitivity. After batch normalization, the nonlinear activation layer facilitates feature mapping, thereby improving the network's expressive power. The max pooling layer compresses the network parameters, reducing the dimensionality of the feature data. The CNN convolutional layer first uses convolutional operations to extract features from the wind turbine's operating data, including speed, power, and pitch angle. Then, it uses the max pooling layer for sampling, reducing the data size while simplifying computation.

[0074] 3.4) Use a wrapper-style feature selection method based on MIC-CNN, such as... Figure 2 As shown, feature selection typically includes two stages and four components: feature subset generation, subset evaluation, termination criteria, and result validation. Stage 1 generates candidate feature subsets from the feature subset components based on a search strategy. Each candidate feature subset is measured using a defined evaluation method and compared with the previously best candidate feature subset. If the new feature subset performs better, it replaces the original best feature subset. The generation and evaluation process is repeated until the termination criteria are met. In Stage 2, the selected feature subsets are structurally validated, and a database for predicting fatigue loads in wind turbine drivetrain systems is constructed based on the learning accuracy (ACC) as feature data positively correlated with fatigue loads.

[0075] 4) Use VMD to decompose the fatigue load data of the wind turbine drivetrain system, such as... Figure 3 As shown, a constrained variational model is established:

[0076]

[0077] Among them, u k Let f(t) be the original input data, k be the iterative signal, K be the number of decomposed modes, δ(t) be the impulse function, and ω be the modal components. k For each component center frequency, It is represented as the partial derivative of the function with respect to t, and st represents the constraint condition.

[0078] 41) Introducing the augmented Lagrangian function transforms the constrained variational problem into an unconstrained variational problem. The expression for the augmented Lagrangian function is as follows:

[0079]

[0080] Where α represents the bandwidth parameter; λ(t) represents the Lagrange multiplier.

[0081] 42) Iterative solution of IMF using the alternating direction method of multiplication operators, {u k The updated formula is:

[0082]

[0083] 43){ω k The updated formula is:

[0084]

[0085] 44) The update formula for λ is:

[0086]

[0087] 45) The iteration termination condition is:

[0088]

[0089] 46) Start iterating from n=0, the number of iterations is n=n+1, k starts from 1, until the preset number of decomposition modes K. The value of K is determined by observing whether the center frequency iteration curves overlap. When the center frequency iteration curves overlap, the number of decomposition modes K is taken as K-1 as the preset number of decomposition modes.

[0090] 47) Repeat steps 42) to 44) until the iteration termination condition is met.

[0091] 5) Considering different data volumes and complexity, and taking into account prediction accuracy and prediction time, the most suitable neural network is selected as the base learner. Different neural networks are trained and learned. High-frequency subsequences are predicted using a Long Short-Term Memory (LSTM) model, mid-frequency subsequences using a Deep Neural Network (DNN) model, and low-frequency subsequences using a GBRT model. The calculation methods for each prediction model are as follows:

[0092] 51) The LSTM calculation method is as follows:

[0093] First, a forget gate is applied to help the LSTM decide what information to discard from the cell state. Here, f represents the activation function of the forget gate, which is:

[0094] f t=f(w f [h t-1 ,x t ]+b f )

[0095] Then the output gate filters and the output cell state is O. t Output gate:

[0096] O t =f(w o [h t-1 ,x t ]+b o )

[0097] Furthermore, the activation function acts on the new cell state and is then multiplied by the output O. t To provide the necessary information, the hidden layer outputs:

[0098] h t =O t ·tanh(C t )

[0099] Finally, the predicted value y is obtained from the output layer. t :

[0100] y t =w y ·h t +b

[0101] 52) The calculation method for neurons in a multilayer neural network (DNN) is as follows:

[0102] z=∑w i x i +b

[0103] Where z is the neuron value; wi is the input weight; xi is the input; and b is the bias value. Simultaneously, to enhance the expressive power of the neural network, a nonlinear function (tanh) is introduced as the activation function, calculated as follows:

[0104]

[0105] 53) The GBRT algorithm mainly consists of three parts: CART, the GB algorithm, and the reduction concept. For continuous data in the algorithm, the sum of squared errors is used to characterize the loss function. The prediction result of the GBRT model after M iterations is:

[0106]

[0107] Where x is the input variable; y is the output variable; i is the number of iterations; f0(x) is the initial predicted value; and f(x) is the model predicted value.

[0108] 6) Train the prediction model for each decomposition component to obtain the prediction output. The goal is to minimize the root mean square error (RMSE) of the predicted values. During training, the network parameters are adjusted. The formula for calculating RMSE is:

[0109]

[0110] 7) Finally, the data is reduced in dimensionality by a flattening layer, and then transformed and output by a fully connected layer to obtain the predicted values ​​of each decomposed component of the fatigue load. The predicted components are then integrated to obtain the final prediction result.

[0111] Example 2

[0112] The fatigue load prediction system for wind turbine drive train system of the present invention includes:

[0113] The acquisition module is used to acquire the load timing of the wind turbine drive train system;

[0114] The calculation module is used to calculate the fatigue load of the wind turbine drive train system based on the load timing of the wind turbine drive train system.

[0115] The decomposition module is used to decompose the fatigue load of the wind turbine drive train system to obtain several sub-sequences;

[0116] The prediction module is used to input each subsequence into the corresponding model to obtain several decomposed component prediction values;

[0117] An integration module is used to integrate the predicted values ​​of each decomposed component to obtain the predicted results of the fatigue load of the wind turbine drive train system.

[0118] In this embodiment, the prediction module includes:

[0119] The first prediction unit is used to input the high-frequency subsequence into the long short-term memory network model;

[0120] The second prediction unit is used to input the mid-frequency sub-sequence into the deep neural network model;

[0121] The third prediction unit is used to input the low-frequency subsequence into the GBRT model;

[0122] The fourth prediction unit is used to obtain the predicted values ​​of several decomposed components.

[0123] The module division in this embodiment is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in each embodiment of this application can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0124] Example 3

[0125] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a method for predicting fatigue loads on a wind turbine drivetrain system. For example, the method includes: acquiring the load timing sequence of the wind turbine drivetrain system; calculating the fatigue load of the wind turbine drivetrain system based on the load timing sequence; decomposing the fatigue load of the wind turbine drivetrain system into several sub-sequences; inputting each sub-sequence into a corresponding model to obtain several decomposed component prediction values; and integrating the decomposed component prediction values ​​to obtain a prediction result for the fatigue load of the wind turbine drivetrain system. The memory may include main memory, such as high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device. The processor, network interface, and memory are interconnected via an internal bus, which may be an industry standard architecture bus, a peripheral component interconnection standard bus, an extended industry standard architecture bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory stores the program; specifically, the program may include program code, which includes computer operation instructions. Memory can include main memory and non-volatile memory, and provides instructions and data to the processor.

[0126] Example 4

[0127] A computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of a method for predicting fatigue loads on a wind turbine drivetrain system. For example, the method includes: acquiring the load timing sequence of the wind turbine drivetrain system; calculating the fatigue load of the wind turbine drivetrain system based on the load timing sequence; decomposing the fatigue load of the wind turbine drivetrain system into several sub-sequences; inputting each sub-sequence into a corresponding model to obtain several decomposed component prediction values; and integrating the decomposed component prediction values ​​to obtain a prediction result for the fatigue load of the wind turbine drivetrain system. Specifically, the computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0128] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0130] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0132] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and disclosure of the invention. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0133] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

[0134] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any simple modifications, alterations, or equivalent structural changes made to the above embodiments based on the technical essence of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for predicting fatigue loads in a wind turbine drivetrain system, characterized in that, include: Obtain the load timing sequence of the wind turbine drivetrain system; Calculate the fatigue load of the wind turbine drive train system based on the load timing of the wind turbine drive train system. The fatigue load of the wind turbine drive train system is decomposed to obtain several sub-sequences; Each subsequence is input into the corresponding model to obtain several decomposed component prediction values; The predicted values ​​of each decomposed component are integrated to obtain the predicted results of fatigue load of the wind turbine drive train system.

2. The fatigue load prediction method for wind turbine drive train system according to claim 1, characterized in that, The process of decomposing the fatigue load of the wind turbine drivetrain system to obtain several sub-sequences is as follows: VMD was used to decompose the fatigue load data of the wind turbine drive train system to obtain high-frequency subsequences, mid-frequency subsequences and low-frequency subsequences.

3. The fatigue load prediction method for wind turbine drive train system according to claim 2, characterized in that, The process of inputting each subsequence into the corresponding model to obtain several decomposed component prediction values ​​is as follows: The high-frequency subsequences are input into the long short-term memory network model; The mid-frequency subsequence is input into a deep neural network model; The low-frequency subsequence is input into the GBRT model; Several predicted values ​​of the decomposed components were obtained.

4. The fatigue load prediction method for wind turbine drive train system according to claim 3, characterized in that, Also includes: Load timing samples of wind turbine drive train system based on OpenFAST software simulation calculation; The fatigue load sample of the wind turbine drivetrain system is calculated based on the load time sequence sample of the wind turbine drivetrain system. The correlation between the measurement data of wind turbine generators and the fatigue load samples of wind turbine generator drivetrain systems was analyzed. Feature data that is compatible with fatigue loads was selected, and a database for predicting fatigue loads of wind turbine generator drivetrain systems was constructed based on this. The database was used to train the Long Short-Term Memory Network model, the Deep Neural Network model, and the GBRT model.

5. The fatigue load prediction method for wind turbine drive train system according to claim 3, characterized in that, The high-frequency subsequence is input into the Long Short-Term Memory network model to obtain the decomposed component prediction value y. t for: y t =w y ·h t +b Among them, h t This is the output of the hidden layer.

6. The fatigue load prediction method for wind turbine drive train system according to claim 3, characterized in that, The low-frequency subsequence is input into the GBRT model to obtain the decomposed component prediction value L(yf(x)): Where x is the input variable; y is the output variable; i is the number of iterations; f0(x) is the initial predicted value; and f(x) is the model predicted value.

7. A fatigue load prediction system for a wind turbine drive train system, characterized in that, include: The acquisition module is used to acquire the load timing of the wind turbine drive train system; The calculation module is used to calculate the fatigue load of the wind turbine drive train system based on the load timing of the wind turbine drive train system. The decomposition module is used to decompose the fatigue load of the wind turbine drive train system to obtain several sub-sequences; The prediction module is used to input each subsequence into the corresponding model to obtain several decomposed component prediction values; An integration module is used to integrate the predicted values ​​of each decomposed component to obtain the predicted results of the fatigue load of the wind turbine drive train system.

8. The fatigue load prediction system for wind turbine drive train system according to claim 7, characterized in that, The prediction module includes: The first prediction unit is used to input the high-frequency subsequence into the long short-term memory network model; The second prediction unit is used to input the mid-frequency sub-sequence into the deep neural network model; The third prediction unit is used to input the low-frequency subsequence into the GBRT model; The fourth prediction unit is used to obtain the predicted values ​​of several decomposed components.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the fatigue load prediction method for the wind turbine drive train system as described in any one of claims 1-6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the fatigue load prediction method for the wind turbine drive train system as described in any one of claims 1-6.

Images