Wind turbine fault identification method, device and terminal equipment

Abstract

The application provides a wind turbine fault identification method device and a terminal device, relates to the technical field of wind turbine fault identification, and the method comprises the following steps: acquiring the real-time temperature value of a to-be-monitored component of a wind turbine, the real-time operation parameter of the wind turbine and the real-time vibration data of the to-be-monitored component; taking the real-time operation parameter of the wind turbine as input, outputting the predicted temperature value of the to-be-monitored component through a pre-trained temperature prediction model, and determining the temperature residual error between the predicted temperature value and the actual temperature value; extracting the time domain features of the real-time vibration data and the frequency domain features of the real-time vibration data respectively, taking the temperature residual error, the time domain features of the real-time vibration data and the frequency domain features of the real-time vibration data as input, and outputting the fault prediction result of the to-be-monitored component through a pre-trained fault identification model. The application can effectively improve the fault identification accuracy of the wind turbine.

Description

Technical Field

[0001] This application relates to the field of wind turbine fault identification technology, specifically to a wind turbine fault identification method, a wind turbine fault identification device, and a terminal device. Background Technology

[0002] Gearbox failure is a major cause of power generation efficiency problems in wind turbines, directly impacting the overall power generation performance of the unit. Therefore, fault warning systems for wind turbine gearboxes are crucial for ensuring the safe operation of wind turbines and improving the economic benefits of wind farms. However, existing fault monitoring methods suffer from large diagnostic errors and low accuracy, making it difficult to accurately identify the operating status of the wind turbine gearbox.

[0003] Application content

[0004] The purpose of this application is to provide a method, device, and terminal equipment for identifying faults in wind turbine generator sets, so as to solve the problem of low accuracy in identifying the operating status of wind turbine gearboxes in existing methods.

[0005] To achieve the above objectives, the first aspect of this application provides a method for fault identification of wind turbine generators, comprising:

[0006] The real-time temperature value of the component to be monitored in the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the component to be monitored are obtained.

[0007] Using the real-time operating parameters of the wind turbine as input, the pre-trained temperature prediction model outputs the predicted temperature value of the component to be monitored, and determines the temperature residual between the predicted temperature value and the actual temperature value.

[0008] The time-domain features and frequency-domain features of the real-time vibration data are extracted respectively. Using the temperature residual, the time-domain features and frequency-domain features of the real-time vibration data as inputs, the pre-trained fault identification model outputs the fault prediction results of the monitored component.

[0009] Optionally, the component to be monitored is the front bearing of the gearbox of the wind turbine.

[0010] Optionally, the temperature prediction model is obtained by training the GRU neural network with the historical operating parameters of the wind turbine, and the fault identification model is obtained by training the XGBoost algorithm with the historical temperature residuals, the time-domain features of historical vibration data, and the frequency-domain features of historical vibration data.

[0011] Optionally, the real-time operating parameters of the wind turbine generator set include at least:

[0012] The wind turbine's gearbox inlet oil temperature, active power, generator front and rear bearing temperatures, gearbox filter inlet pressure, and gearbox filter outlet pressure are one or more of the following:

[0013] Optionally, after obtaining the real-time operating parameters of the wind turbine, the method further includes:

[0014] Single-point threshold filtering is performed on the obtained real-time operating parameters of the wind turbine.

[0015] Outlier removal is performed on real-time operating parameters after single-point threshold filtering based on the Raida criterion.

[0016] The real-time operating parameters after outlier removal are normalized.

[0017] Optionally, after acquiring real-time vibration data, the method further includes: performing wavelet packet denoising on the real-time vibration data.

[0018] Optionally, the time-domain characteristics of the real-time vibration data include:

[0019] RMS acceleration, peak-to-peak acceleration, acceleration waveform index, peak acceleration index, acceleration pulse index, acceleration margin index, acceleration kurtosis index, RMS velocity, peak-to-peak velocity, velocity waveform index, peak velocity index, velocity pulse index, velocity margin index, and velocity kurtosis index.

[0020] Optionally, the frequency domain characteristics of the real-time vibration data include:

[0021] Acceleration spectrum characteristics, acceleration spectrum characteristics, acceleration spectrum characteristics, acceleration frequency characteristics, velocity spectrum characteristics, velocity spectrum characteristics, velocity spectrum characteristics, velocity spectrum characteristics, acceleration envelope spectrum characteristics, acceleration envelope spectrum characteristics, acceleration envelope spectrum characteristics and acceleration envelope spectrum characteristics.

[0022] A second aspect of this application provides a wind turbine fault identification device, comprising:

[0023] The data acquisition module is configured to acquire the real-time temperature value of the monitored component of the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the monitored component.

[0024] The residual calculation module is configured to take the real-time operating parameters of the wind turbine as input, output the predicted temperature value of the component to be monitored through a pre-trained temperature prediction model, and determine the temperature residual between the predicted temperature value and the actual temperature value.

[0025] The fault prediction module is configured to extract the time-domain features and frequency-domain features of the real-time vibration data respectively, and output the fault prediction results of the monitored component by taking the temperature residual, the time-domain features and the frequency-domain features of the real-time vibration data as inputs and the pre-trained fault identification model.

[0026] A third aspect of this application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the aforementioned wind turbine fault identification method.

[0027] The embodiments provided in this application have the following beneficial effects:

[0028] This application uses the real-time operating parameters of the wind turbine to output the predicted temperature value of the component to be monitored through a pre-trained temperature prediction model. By calculating the temperature residual between the predicted temperature value and the actual temperature value, the application predicts the fault of the component to be monitored based on the temperature residual and real-time vibration data, thereby effectively improving the fault identification accuracy of the wind turbine.

[0029] Other features and advantages of the embodiments or implementations of this application will be described in detail in the following detailed description section. Attached Figure Description

[0030] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings:

[0031] Figure 1 This illustration schematically shows a flowchart of a wind turbine fault identification method according to an embodiment of the present application;

[0032] Figure 2 A logic diagram of the wind turbine fault identification method according to an embodiment of this application is illustrated schematically.

[0033] Figure 3 A schematic diagram of the GRU neural network structure according to an embodiment of this application is shown.

[0034] Figure 4 A schematic block diagram of a wind turbine fault identification device according to an embodiment of this application is shown.

[0035] Figure 5 The schematic diagram illustrates a terminal device structure according to an embodiment of this application.

[0036] Explanation of reference numerals in the attached figures

[0037] 10 - Terminal device, 100 - Processor, 101 - Memory, 102 - Computer program. Detailed Implementation

[0038] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the embodiments of this application.

[0039] In this application, the component to be monitored is the front bearing of the gearbox in a wind turbine. In a wind turbine, the gearbox is a critical component in the drivetrain, and its role is irreplaceable. If a major failure such as internal gear damage occurs, the entire wind turbine needs to be disassembled and the entire drivetrain replaced. Therefore, gearbox failure is one of the factors causing prolonged turbine downtime. The gearbox in the drivetrain functions to transmit the torque of the impeller to the generator, transferring wind energy from the turbine's main shaft to the gearbox via the rotor blades, and finally to the generator system via a flexible coupling. Simultaneously, the gearbox also increases the rotor speed to the generator's synchronous speed, driving the generator to rotate rapidly and generate electricity. Gear and bearing failures account for the highest proportion in the gearbox structure. Their normal and effective operation directly affects not only the overall gearbox operating status but also the operation and maintenance of the entire equipment. Gearboxes may experience various forms of failure during long-term operation under alternating loads. Among these, the front bearing of the gearbox, due to its high speed and relatively large torque, is prone to overheating and triggering alarms or failures that lead to turbine shutdowns during prolonged continuous operation. Existing fault monitoring methods for gearbox front bearings typically monitor their real-time temperature. However, single monitoring data often only reflects changes in a single physical parameter and cannot reflect the interaction and coupling relationships between components or subsystems, resulting in low accuracy of the monitoring results.

[0040] To solve the above problems, such as Figure 1 and Figure 2 As shown, the first aspect of this application provides a method for fault identification of wind turbine generators, including:

[0041] S100: Acquire the real-time temperature value of the components to be monitored in the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the components to be monitored.

[0042] S200: Using the real-time operating parameters of the wind turbine as input, the pre-trained temperature prediction model outputs the predicted temperature value of the component to be monitored, and determines the temperature residual between the predicted temperature value and the actual temperature value.

[0043] S300: Extract the time-domain features and frequency-domain features of the real-time vibration data respectively. Using the temperature residual, the time-domain features and frequency-domain features of the real-time vibration data as input, the pre-trained fault identification model outputs the fault prediction results of the monitored component.

[0044] Thus, this application uses the real-time operating parameters of the wind turbine to output the predicted temperature value of the component to be monitored through a pre-trained temperature prediction model. By calculating the temperature residual between the predicted temperature value and the actual temperature value, the application predicts the fault of the component to be monitored based on the temperature residual and real-time vibration data, thereby effectively improving the fault identification accuracy of the wind turbine.

[0045] In this application, the temperature prediction model is obtained by training the GRU neural network with the historical operating parameters of the wind turbine, and the fault identification model is obtained by training the XGBoost algorithm with the historical temperature residuals, the time-domain features of historical vibration data, and the frequency-domain features of historical vibration data.

[0046] The real-time and historical operating parameters of the wind turbine generator include at least one or more of the following: gearbox inlet oil temperature, active power, front and rear bearing temperatures of the generator, gearbox filter inlet pressure, and gearbox filter outlet pressure. Operating parameters of the wind turbine generator can be obtained by acquiring SCADA data. However, using all SCADA data as input to the temperature prediction model would significantly increase the complexity and computation time of the model. Furthermore, the inclusion of too many parameters that do not affect the predicted quantity would lead to low prediction accuracy. Therefore, this application uses parameters from the SCADA data of the wind turbine generator that have a grey relational relationship with the front bearing temperature of the gearbox as input to the temperature prediction model. Grey relational analysis determines the degree of correlation between factors by comparing the similarity between curves. Using grey relational analysis, SCADA parameters with a high correlation to a monitoring quantity characterizing the operating state of the front bearing of the gearbox can be effectively selected as input to the model. Grey relational analysis is existing technology and will not be elaborated upon here.

[0047] After obtaining the real-time temperature value of the front bearing of the gearbox of the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the components to be monitored, the SCADA data contains a lot of abnormal noise due to factors such as turbine failure, sensor damage, SCADA system crash, and data transmission network issues caused by the wind turbine's operating environment. Therefore, it is necessary to preprocess the operating parameters obtained through SCADA data.

[0048] First, single-point threshold filtering is applied to the acquired real-time operating parameters of the wind turbines, deleting data points exceeding the threshold range and removing data that cannot reflect temperature and vibration characteristics under conditions such as wind turbine shutdown due to faults or curtailment. Second, outlier removal is performed on the real-time operating parameters after single-point threshold filtering based on the Laida criterion (3σ criterion), eliminating outliers and large-amplitude pulse interference, and then median filtering is applied until the data meets the requirements. Finally, the real-time operating parameters after outlier removal are normalized. The data used in training and testing are normalized using the following formula:

[0049] Similarly, after acquiring real-time vibration data, the method also includes: performing wavelet packet denoising on the real-time vibration data. It is understood that the vibration data of the gearbox front bearing can be obtained from the CMS data of the wind turbine's online mechanical condition monitoring system (CMS system). After outlier removal through the above steps, since the vibration signal during wind turbine operation is mixed with a large amount of interference and noise, this application also uses wavelet packets for denoising during the preprocessing of the vibration data. First, a wavelet basis is selected and the decomposition level is determined, and the vibration signal is decomposed into wavelet packets. Second, the optimal wavelet packet basis is determined through the entropy standard. Then, for each wavelet packet decomposition coefficient, an appropriate threshold is selected to quantize the coefficient. Finally, the vibration signal is reconstructed using wavelet packets.

[0050] After preprocessing the acquired real-time operating parameters and real-time vibration data, the real-time operating parameters are used as input to the temperature prediction model, which then outputs the predicted temperature value of the front bearing of the gearbox.

[0051] In this application, the temperature prediction model is obtained by training a GRU neural network (i.e., a gated recurrent unit neural network) using historical operating parameters of the wind turbine. Understandably, before training the GRU neural network using historical operating parameters of the wind turbine, the historical operating parameters also need to be preprocessed using the method described above.

[0052] In wind turbine fault diagnosis, abnormal temperature is an important indicator of malfunctions in key equipment. Gearbox bearings can overheat during operation due to wear, poor lubrication, or inadequate shielding. Excessively high temperatures can affect the normal operation of the equipment. Therefore, this application utilizes temperature characteristics to establish a temperature prediction model for the front bearing of the gearbox based on a gated recurrent unit (GRU) neural network. This model can calculate the temperature residual between the real-time temperature value of the front bearing and the predicted temperature value output by the temperature prediction model. The real-time temperature value can be obtained through a pre-set temperature sensor.

[0053] Among them, the GRU neural network is an enhanced version of the LSTM neural network. Its principle is similar to LSTM, utilizing gating mechanisms to control input, memory, and other information, and predicting the output at the current time step. For example... Figure 3 As shown, the GRU structure has a reset gate and an update gate. The reset gate determines how new input information is combined with previous memory information, while the update gate defines the amount of previous memory information stored at the current time step. These two gating vectors determine which information is ultimately output as the gated recurrent unit, enabling it to retain information from long-term sequences without removing it over time or due to irrelevance in the prediction.

[0054] For each GRU unit, if there is an input sequence x = (x1, x2, ... x... t The two gate signals are obtained using the following formula:

[0055] z t =σ(W (z) x t +U (z) h t-1 );

[0056] r t =σ(W (r) x t +U (r) h t-1 );

[0057] Among them, z t For updating the gate; r t To reset the door; x t U is the input vector at time t; (z) U (r) W (z) W (r) h is the weight matrix; t-1 σ represents the information stored at time t-1; σ is the Sigmoid activation function.

[0058] Use the reset gate to obtain new memory information h t ', for h t =tanh(Wx t +r t * Uh t-1 );

[0059] Where "*" indicates element-wise multiplication in the matrix; W and U are weight matrices; and tanh is the hyperbolic tangent function.

[0060] Finally, calculate the final GRU output memory information h. t :

[0061] ht =z t * h t-1 +(1-z t )*h t ';

[0062] y t =σ(W0h t );

[0063] Where W0 is the weight matrix, y t This is the output of the output layer.

[0064] After obtaining the temperature residual between the predicted and actual temperatures of the front bearing of the gearbox, the time-domain and frequency-domain features of the real-time vibration data are further extracted. Time-domain feature extraction is the most direct and convenient feature extraction method in vibration signal analysis. Seven time-domain feature indices of the gearbox front bearing vibration signal are used: RMS value, peak-to-peak value, peak value index, waveform index, impulse index, margin index, and kurtosis index. Among the time-domain feature indices, the RMS value reflects the intensity and energy of the vibration signal and is suitable for vibrations caused by irregular continuous defects resulting from bearing wear; the peak-to-peak value index reflects the magnitude of impact vibrations generated by local fault points in the bearing; the peak value index reflects the degree of peaks in the vibration signal waveform and can reflect faults such as scratches and scoring in the wind turbine gearbox bearing; the waveform index reflects faults such as pitting in the gearbox bearing; the impulse index and margin index show a significant increasing trend in the early stages of the fault and are more sensitive to impact-type faults; the kurtosis index reflects the degree to which the signal deviates from a normal distribution, and the larger the kurtosis value, the more severe the impact on the bearing. The time-domain features extracted in this application include 14 time-domain features across 7 time-domain indices, including acceleration vibration signals and velocity vibration signals. Specifically, the time-domain features of the real-time vibration data in this application include: effective acceleration value, peak-to-peak acceleration value, acceleration waveform index, peak acceleration index, acceleration pulse index, acceleration margin index, acceleration kurtosis index, effective velocity value, peak-to-peak velocity value, velocity waveform index, peak velocity index, velocity pulse index, velocity margin index, and velocity kurtosis index.

[0065] The vibration spectrum of a bearing reflects the amplitude distribution of the signal at different frequencies. The spectral structure of gearbox bearings in normal and abnormal units differs significantly. Spectral analysis can determine whether a bearing has failed and, if so, the location of the failure. Based on Fourier series theory, assuming the discretized vibration signal time series has a sampling frequency of f... s The number of sampling points is N, and the spectrum is s. kk = 1, 2, ..., k, where k is the number of spectral lines. This application uses four frequency domain indicators: F1, the magnitude of vibration energy; F2, the deviation of the signal spectrum from the mean of the signal spectrum; F3, the degree of asymmetry of the signal spectrum relative to the mean; and F4, the magnitude of the peak value of the signal at the mean. Besides the spectrum effectively reflecting fault information during wind turbine operation, the envelope spectrum can also effectively reflect equipment impact-related fault information. The envelope spectrum is often used to detect bearing defects. When the bearing surface peels or is damaged due to fatigue or concentrated stress, it generates periodic impact vibration signals. This impact vibration can be roughly divided into low-frequency pulses generated by repeated impacts on bearing components during operation and inherent vibrations generated by impact vibration. Therefore, envelope spectrum analysis can accurately diagnose typical bearing faults. Thus, the frequency domain features extracted in this application include four frequency domain indicators (F1, F2, F3, F4), totaling 12 frequency domain features, including acceleration spectrum, velocity spectrum, and acceleration envelope spectrum. Specifically, the frequency domain characteristics of the real-time vibration data in this application include: acceleration spectrum characteristics, acceleration spectrum characteristics, acceleration frequency characteristics, velocity spectrum characteristics, velocity spectrum characteristics, velocity spectrum characteristics, velocity spectrum characteristics, acceleration envelope spectrum characteristics, acceleration envelope spectrum characteristics, and acceleration envelope spectrum characteristics.

[0066] After obtaining the temperature residual, time-domain characteristics of real-time vibration data, and frequency-domain characteristics of real-time vibration data of the front bearing of the gearbox, the fault identification model obtained by training the XGBoost algorithm with the above data as input and using the time-domain characteristics of historical temperature residual, historical vibration data, and frequency-domain characteristics of historical vibration data can output the fault category of the front bearing of the gearbox.

[0067] In this application, the input feature vectors of the gearbox front bearing fault identification model based on XGBoost are: temperature residual of the gearbox front bearing; effective acceleration value, peak-to-peak acceleration value, acceleration waveform index, peak acceleration index, acceleration pulse index, acceleration margin index, acceleration kurtosis index, effective velocity value, peak-to-peak velocity value, velocity waveform index, peak velocity index, velocity pulse index, velocity margin index, velocity kurtosis index; acceleration spectrum features, acceleration spectrum features, acceleration frequency features, velocity spectrum features, velocity spectrum features, velocity spectrum features, velocity spectrum features, velocity spectrum features, acceleration envelope spectrum features, acceleration envelope spectrum features, acceleration envelope spectrum features, acceleration envelope spectrum features.

[0068] Ensemble algorithms complete the learning task by constructing and combining multiple learners. The XGBoost algorithm is an optimization of the adaptive boosting (adaBoost) and gradient boosting decision tree (GBDT) algorithms. It achieves accurate classification prediction by continuously fitting and optimizing the objective function. XGBoost adds a regularization term to the original function, reducing the possibility of overfitting and accelerating convergence. XGBoost is a tree ensemble model that sums the results of K trees to obtain the final prediction value. Because XGBoost uses a gradient boosting algorithm based on decision trees, its advantage lies in automatically acquiring the importance of features during the construction of the boosting tree. Generally, the more times a feature is used as a splitting attribute in all trees, the more important that feature is. Therefore, XGBoost can automatically calculate the score of each feature in the input feature vector during training, thus effectively selecting features.

[0069] The fault identification model in this application outputs five states: normal, damage to the inner ring of the front gearbox bearing, damage to the outer ring of the front gearbox bearing, gearbox shaft imbalance, and damage to the rolling elements of the gearbox, with corresponding data labels of 0, 1, 2, 3, and 4, respectively. The judgment criterion for identifying the front gearbox bearing fault is: the recognition accuracy of the five states output by the classifier is used, and the class with the highest percentage of classifier output is selected as the final decision basis for identification. Understandably, the specific training process of the XGBoost algorithm is existing technology and will not be elaborated here.

[0070] like Figure 4 As shown, the second aspect of this application provides a wind turbine fault identification device, comprising:

[0071] The data acquisition module is configured to acquire the real-time temperature value of the components to be monitored in the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the components to be monitored.

[0072] The residual calculation module is configured to take the real-time operating parameters of the wind turbine as input, output the predicted temperature value of the component to be monitored through a pre-trained temperature prediction model, and determine the temperature residual between the predicted temperature value and the actual temperature value.

[0073] The fault prediction module is configured to extract the time-domain features and frequency-domain features of real-time vibration data respectively. Taking the temperature residual, the time-domain features and frequency-domain features of real-time vibration data as input, the pre-trained fault identification model outputs the fault prediction results of the monitored component.

[0074] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0075] A third aspect of this application provides a terminal 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 aforementioned wind turbine fault identification method.

[0076] like Figure 5 The diagram shown is a schematic representation of a terminal device provided in an embodiment of this application. Figure 5 As shown, the terminal device 10 of this embodiment includes a processor 100, a memory 101, and a computer program 102 stored in the memory 101 and executable on the processor 100. When the processor 100 executes the computer program 102, it implements the steps in the above method embodiments. Alternatively, when the processor 100 executes the computer program 102, it implements the functions of each module / unit in the above device embodiments.

[0077] For example, computer program 102 may be divided into one or more modules / units, one or more of which are stored in memory 101 and executed by processor 100 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 102 in terminal device 10.

[0078] Terminal device 10 may be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. Terminal device 10 may include, but is not limited to, a processor 100 and a memory 101. Those skilled in the art will understand that... Figure 5 This is merely an example of terminal device 10 and does not constitute a limitation on terminal device 10. It may include more or fewer components than shown, or combine certain components, or different components. For example, terminal device may also include input / output devices, network access devices, buses, etc.

[0079] The processor 100 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0080] The memory 101 can be an internal storage unit of the terminal device 10, such as a hard disk or RAM of the terminal device 10. The memory 101 can also be an external storage device of the terminal device 10, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device 10. Furthermore, the memory 101 can include both internal and external storage units of the terminal device 10. The memory 101 is used to store computer programs and other programs and data required by the terminal device 10. The memory 101 can also be used to temporarily store data that has been output or will be output.

[0081] 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.

[0082] In summary, this application takes the front bearing of a wind turbine gearbox as the research object and proposes a time-frequency domain combined modeling method that combines SCADA data and vibration data. A temperature prediction model for the front bearing of the wind turbine gearbox is established based on a GRU neural network, and the temperature residual characteristics are calculated. These characteristics are combined with the time-frequency domain characteristics of the vibration signal to establish an XGBoost-based fault identification model for the front bearing, thereby identifying the fault state and category of the front bearing and enabling precise fault location to support emergency response. This application constructs a temperature prediction model for the front bearing based on a gated recurrent unit neural network. The GRU neural network, with the front bearing temperature as a feature, can better track and identify long-term series data of the gradual deterioration process of the fault, improving the accuracy and efficiency of temperature modeling. Furthermore, the XGBoost-based fault identification method for the front bearing in this application can better identify five states of the generator front bearing compared to decision trees, random forests, and SVM algorithms: normal operation, inner ring damage, outer ring damage, shaft imbalance, and rolling element damage. Meanwhile, this application combines the temperature residual characteristics of the front bearing of the gearbox with the time-frequency domain characteristics of vibration data, thereby improving the modeling accuracy and generalization of fault identification of the front bearing of the gearbox.

[0083] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0084] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for fault identification of wind turbine generators, characterized in that, include: The method involves acquiring real-time temperature values ​​of the components to be monitored in a wind turbine, real-time operating parameters of the wind turbine, and real-time vibration data of the components to be monitored. Wavelet packet denoising is then performed on the real-time vibration data, including: selecting a wavelet basis and determining the decomposition level; performing wavelet packet decomposition on the vibration signal; determining the optimal wavelet packet basis using an entropy standard; quantizing each wavelet packet decomposition coefficient by selecting a corresponding threshold; and reconstructing the vibration signal using wavelet packets. Using the real-time operating parameters of the wind turbine as input, the pre-trained temperature prediction model outputs the predicted temperature value of the component to be monitored, and determines the temperature residual between the predicted temperature value and the actual temperature value. The time-domain features and frequency-domain features of the real-time vibration data are extracted respectively. Using the temperature residual, the time-domain features and frequency-domain features of the real-time vibration data as input, the pre-trained fault identification model outputs the fault prediction result of the monitored component. The frequency-domain features of the real-time vibration data include: acceleration spectrum features, acceleration spectrum features, acceleration frequency features, velocity spectrum features, velocity spectrum features, velocity spectrum features, velocity spectrum features, acceleration envelope spectrum features, acceleration envelope spectrum features, and acceleration envelope spectrum features. The component to be monitored is the front bearing of the gearbox of the wind turbine, and the real-time operating parameters of the wind turbine include at least the following: The wind turbine's gearbox inlet oil temperature, active power, generator rear bearing temperature, wind turbine gearbox filter inlet pressure, and wind turbine gearbox filter outlet pressure are one or more of the following: The real-time operating parameters are parameters that have a gray correlation with the temperature of the front bearing of the gearbox; The temperature prediction model is obtained by training the GRU neural network with the historical operating parameters of the wind turbine, and the fault identification model is obtained by training the XGBoost algorithm with the historical temperature residual, the time domain characteristics of historical vibration data, and the frequency domain characteristics of historical vibration data.

2. The wind turbine fault identification method according to claim 1, characterized in that, After obtaining the real-time operating parameters of the wind turbine, the method further includes: Single-point threshold filtering is performed on the obtained real-time operating parameters of the wind turbine. Outlier removal is performed on real-time operating parameters after single-point threshold filtering based on the Raida criterion. The real-time operating parameters after outlier removal are normalized.

3. The wind turbine fault identification method according to claim 1, characterized in that, The time-domain characteristics of the real-time vibration data include: RMS acceleration, peak-to-peak acceleration, acceleration waveform index, peak acceleration index, acceleration pulse index, acceleration margin index, acceleration kurtosis index, RMS velocity, peak-to-peak velocity, velocity waveform index, peak velocity index, velocity pulse index, velocity margin index, and velocity kurtosis index.

4. A wind turbine fault identification device, employing the wind turbine fault identification method as described in any one of claims 1 to 3, characterized in that, include: The data acquisition module is configured to acquire the real-time temperature value of the monitored component of the wind turbine, the real-time operating parameters of the wind turbine, and the real-time vibration data of the monitored component. The residual calculation module is configured to take the real-time operating parameters of the wind turbine as input, output the predicted temperature value of the component to be monitored through a pre-trained temperature prediction model, and determine the temperature residual between the predicted temperature value and the actual temperature value. The fault prediction module is configured to extract the time-domain features and frequency-domain features of the real-time vibration data respectively, and output the fault prediction results of the monitored component by taking the temperature residual, the time-domain features and the frequency-domain features of the real-time vibration data as inputs and the pre-trained fault identification model.

5. A terminal 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 wind turbine fault identification method according to any one of claims 1 to 3.

Images