A wind turbine generator fault diagnosis method and device
By performing noise reduction and frequency domain conversion on real-time data of wind turbines, and combining vibration prediction models and early warning thresholds, the problem of fault diagnosis under noise interference of wind turbines was solved, and efficient and accurate fault identification and early warning were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING IND BIG DATA INNOVATION CENT CO LTD
- Filing Date
- 2025-07-04
- Publication Date
- 2026-06-30
AI Technical Summary
The operating environment of wind turbines is complex and the noise signal intensity is high, which causes fault characteristics to be submerged, making it difficult for existing technologies to accurately diagnose faults.
By acquiring real-time operating condition data and vibration signal data of wind turbines, noise reduction and frequency domain conversion are performed. A trained vibration prediction model is used for fault diagnosis, and combined with early warning thresholds and a fault scheme library, accurate diagnosis is achieved.
It improves the accuracy and efficiency of wind turbine fault diagnosis, enabling early identification of potential faults and preventing major loss of control.
Smart Images
Figure CN120994959B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind turbine fault diagnosis technology, and also to a method and apparatus for wind turbine fault diagnosis. Background Technology
[0002] Wind turbines, as typical clean energy units, are a crucial component of the power energy system. As rotating machinery rapidly evolves towards larger, more complex, more precise, higher-speed, heavier-load, and more intelligent applications, the application scenarios for wind turbines are becoming increasingly modern and diverse. For power generation companies, this presents both an opportunity to improve power generation efficiency and strengthen reform and innovation, and a challenge to ensure safe production and enhance digital capabilities. With the advancement of technology, companies are increasingly focused on the safety and sustainable operation of their production equipment. Wind turbines operate in complex environments and typically require heavy-load operation. Failures can not only impact production safety and the power grid but may even threaten personnel safety. Therefore, ensuring the safe operation of wind turbines throughout their lifecycle, diagnosing and providing early warnings of mechanical failures, and preventing controllable faults from developing into major uncontrollable failures are crucial for effectively avoiding economic losses and maintaining power grid stability.
[0003] However, the operating environment of wind turbines is complex, and the collected signals often contain noise of unknown intensity. This poses a certain difficulty for extracting fault features of wind power. If the noise signal energy is strong, the fault features are very likely to be submerged in the noise signal, which will lead to inaccurate fault diagnosis. Summary of the Invention
[0004] The technical problem to be solved by the present invention is to provide a method and apparatus for diagnosing wind turbine faults, so as to improve the accuracy of wind turbine fault diagnosis.
[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:
[0006] A first aspect of the present invention provides a method for diagnosing faults in wind turbine generators, comprising:
[0007] Acquire real-time operating condition data and real-time vibration signal data of wind turbine units;
[0008] The real-time vibration signal data is subjected to noise reduction processing to obtain noise-reduced data;
[0009] The noise-reduced data is transformed into frequency domain data.
[0010] Vibration prediction data is obtained based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data.
[0011] The wind turbine is diagnosed based on the frequency domain data and the vibration prediction data to obtain the fault diagnosis results.
[0012] Optionally, real-time operating condition data and real-time vibration signal data of the wind turbine can be acquired, including:
[0013] Acquire real-time operating condition data of the wind turbine generator set; the operating condition data includes temperature parameter data, electrical parameter data, environmental parameter data, and control parameter data;
[0014] Real-time vibration signal data of the wind turbine is acquired by sensors installed at preset locations in the wind turbine.
[0015] Optionally, the real-time vibration signal data is subjected to noise reduction processing to obtain noise-reduced data, including:
[0016] The real-time vibration signal data is decomposed to obtain low-frequency and high-frequency components;
[0017] Thresholding is performed on the high-frequency components to obtain detail coefficients;
[0018] The noise reduction data is obtained based on the low-frequency components and the detail coefficients.
[0019] Optionally, the noise-reduced data is frequency-domain transformed to obtain frequency-domain data, including:
[0020] pass The noise-reduced data is transformed into frequency domain data.
[0021] in, For frequency domain data, The data is denoised, and n is the time-domain index. N is the total number of sampling points, and k is the frequency domain index. , is the rotation factor.
[0022] Optionally, based on the real-time operating condition data and the trained vibration prediction model, vibration prediction data is obtained, including:
[0023] The real-time operating condition data is input into the input layer of the vibration prediction model to obtain the first output result;
[0024] The first output result is input into at least one processing layer of the vibration prediction model to obtain the second output result;
[0025] The second output result is input into the output layer of the vibration prediction model to obtain vibration prediction data.
[0026] Optionally, the training process of the vibration prediction model includes:
[0027] Acquire historical operating condition data and corresponding historical vibration signal data of wind turbines under preset conditions;
[0028] The historical operating condition data and the historical vibration signal data are preprocessed to obtain preprocessed historical data.
[0029] Historical features are extracted from the preprocessed historical data to obtain historical feature data;
[0030] The preset network model is trained based on the historical feature data to obtain the initial training results;
[0031] The preset network model is optimized based on the initial training results to obtain the vibration prediction model.
[0032] Optionally, fault diagnosis of the wind turbine is performed based on the frequency domain data and the vibration prediction data to obtain fault diagnosis results, including:
[0033] The residual is obtained based on the frequency domain data and the vibration prediction data;
[0034] The warning result is determined based on the residual and the preset warning threshold;
[0035] Based on the warning results and the preset fault solution library, a fault solution is obtained;
[0036] Based on the frequency domain data, the vibration prediction data, the early warning results, and the fault diagnosis scheme, the wind turbine is diagnosed to obtain the fault diagnosis results.
[0037] A second aspect of the present invention provides a wind turbine fault diagnosis device, comprising:
[0038] The acquisition module is used to acquire real-time operating condition data and real-time vibration signal data of the wind turbine.
[0039] The processing module is used to perform noise reduction processing on the real-time vibration signal data to obtain noise-reduced data; to perform frequency domain conversion on the noise-reduced data to obtain frequency domain data; to obtain vibration prediction data based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data; and to perform fault diagnosis on the wind turbine based on the frequency domain data and the vibration prediction data to obtain fault diagnosis results.
[0040] A third aspect of the present invention provides a computing device, comprising: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described in the first aspect.
[0041] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method as described in the first aspect.
[0042] The above-described solution of the present invention has at least the following beneficial effects:
[0043] The above-mentioned solution of the present invention acquires real-time operating condition data and real-time vibration signal data of wind turbines, then performs noise reduction processing on the real-time vibration signal data to obtain noise-reduced data, then performs frequency domain conversion on the noise-reduced data to obtain frequency domain data, and obtains vibration prediction data based on the real-time operating condition data and the trained vibration prediction model. The frequency domain data and vibration prediction data are compared to diagnose faults in wind turbines and obtain fault diagnosis results. This can effectively improve the accuracy of fault diagnosis of wind turbines and improve the operating efficiency of wind turbines. Attached Figure Description
[0044] Figure 1 This is a flowchart illustrating the wind turbine fault diagnosis method in an embodiment of the present invention;
[0045] Figure 2 This is a schematic diagram of the structure of the wind turbine fault diagnosis device in an embodiment of the present invention. Detailed Implementation
[0046] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
[0047] like Figure 1 As shown, an embodiment of the present invention proposes a method for diagnosing wind turbine faults, including the following steps:
[0048] Step 101: Obtain real-time operating condition data and real-time vibration signal data of the wind turbine.
[0049] Step 102: Perform noise reduction processing on the real-time vibration signal data to obtain noise-reduced data;
[0050] Step 103: Perform frequency domain transformation on the noise reduction data to obtain frequency domain data;
[0051] Step 104: Obtain vibration prediction data based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data.
[0052] Step 105: Perform fault diagnosis on the wind turbine based on the frequency domain data and the vibration prediction data to obtain the fault diagnosis result.
[0053] The wind turbine fault diagnosis method proposed in this invention acquires real-time operating condition data and real-time vibration signal data of the wind turbine, then performs noise reduction processing on the real-time vibration signal data to obtain noise-reduced data, and then performs frequency domain conversion on the noise-reduced data to obtain frequency domain data. Based on the real-time operating condition data and the trained vibration prediction model, vibration prediction data is obtained. The frequency domain data and vibration prediction data are compared to diagnose the wind turbine fault and obtain the fault diagnosis result. This method can effectively improve the accuracy of wind turbine fault diagnosis and improve the operating efficiency of the wind turbine.
[0054] In an optional embodiment of the present invention, step 101 includes:
[0055] Step 1011: Obtain real-time operating condition data of the wind turbine; the operating condition data includes temperature parameter data, electrical parameter data, environmental parameter data, and control parameter data;
[0056] Specifically, real-time operating data of wind turbines forms the basis for subsequent prediction of mechanical vibration signals and provides a diagnostic basis for subsequent fault diagnosis. This data can be collected via temperature sensors (gearbox oil temperature, generator winding temperature, and bearing temperature), power analyzers or current transformers (power, voltage, current, frequency, etc.), anemometers (wind speed, direction, air density, etc.), and temperature, humidity, and air pressure sensors or barometers (environmental parameters including wind speed, direction, air density, temperature, humidity, and air pressure). Control parameters such as blade pitch angle, yaw angle, and rotational speed can also be collected via encoders or angle sensors. Depending on the specific application, real-time operating data from other types of wind turbines can be used; the above is just one example.
[0057] Step 1012: Acquire real-time vibration signal data of the wind turbine by using sensors installed at preset positions in the wind turbine.
[0058] Specifically, the preset locations may include at least one key component such as the main bearing, gearbox input shaft, gearbox output shaft, or generator drive / non-drive end. Sensors are installed on these key components to collect mechanical vibration signal data from at least one of the main bearing, gearbox input shaft, gearbox output shaft, or generator drive / non-drive end, providing a data basis for subsequent fault diagnosis.
[0059] In an optional embodiment of the present invention, step 102 includes:
[0060] Step 1021: Decompose the real-time vibration signal data to obtain low-frequency components and high-frequency components;
[0061] Specifically, through The low-frequency component is obtained by... The high-frequency components are obtained, among which, Low-frequency components, These are the coefficients of the low-pass filter. This is part of the downsampling and filtering operation on real-time vibration signal data. For high-frequency components, Here are the coefficients of the high-pass filter, where n is the data index and m is the index of the filter coefficient.
[0062] Step 1022: Perform thresholding on the high-frequency components to obtain detail coefficients;
[0063] Specifically, through or Thresholding is performed on the high-frequency components to obtain detail coefficients, wherein, For detail coefficients, For high-frequency components, For the preset threshold, , Let M be the standard deviation of the noise, and M be determined by the length of the signal. It is a symbolic function.
[0064] Step 1023: Obtain noise reduction data based on the low-frequency component and the detail coefficient.
[0065] Specifically, through The noise reduction data was obtained, among which, For the low-frequency components at level b-1, To reconstruct the coefficients of the low-pass filter, For the low-frequency components of the b-th order, To reconstruct the coefficients of the high-pass filter, This is the floor function, used to implement upsampling operations. Here, m represents the detail coefficients at level b, m is the index of the filter coefficients, and n is the data index.
[0066] The noise reduction process described above can yield highly accurate mechanical vibration signal data, which is beneficial for improving the accuracy of subsequent fault diagnosis.
[0067] In an optional embodiment of the present invention, step 103 includes:
[0068] pass The noise-reduced data is transformed into frequency domain data.
[0069] in, For frequency domain data, The data is denoised, and n is the time-domain index. N is the total number of sampling points, and k is the frequency domain index. , is the rotation factor.
[0070] Specifically, frequency domain data, including frequency and amplitude, is the foundational data for subsequent fault diagnosis.
[0071] In an optional embodiment of the present invention, step 104, obtaining vibration prediction data based on the real-time operating condition data and the trained vibration prediction model, includes:
[0072] Step 10411: Input the real-time operating condition data into the input layer of the vibration prediction model to obtain the first output result;
[0073] Specifically, the input layer of the vibration prediction model normalizes the real-time operating condition data to eliminate the influence of dimensions and improve processing efficiency. Here, the input layer adopts... The real-time operating condition data is normalized to obtain the first output result, in which... The first output is denoted as x, which represents real-time operating condition data. This represents the minimum value of real-time operating condition data. This represents the maximum value of real-time operating condition data.
[0074] Step 10412: Input the first output result into at least one processing layer of the vibration prediction model to obtain the second output result;
[0075] Specifically, the processing layer of the vibration prediction model performs a nonlinear transformation on the first output result to extract higher-order features. Here, the processing layer... A nonlinear transformation is performed on the first output result to obtain the second output result, where, This represents the output of the j-th neuron in the k-th processing layer. For activation function, For the kth The outputs of all neurons in layer 1 are weighted and summed. For the kth The connection weights from the i-th neuron in layer 1 to the j-th neuron in layer k. For the kth The output of the i-th neuron in layer 1, where p is the number of neurons in the (k-1)-th processing layer. This is the bias term for the j-th neuron in the k-th layer. It should be noted that the output of the last processing layer is the second output result. If the vibration prediction model has k processing layers, then the output results of all neurons in the k-th processing layer are the second output result.
[0076] Step 10413: Input the second output result into the output layer of the vibration prediction model to obtain vibration prediction data.
[0077] Specifically, the output layer of the vibration prediction model is through Vibration prediction data were obtained, among which, For vibration prediction data, This is the second output result. is the connection weight from the k-th processing layer neuron to the output layer, M is the number of neurons in the k-th processing layer, and b is the bias term of the output layer.
[0078] In an optional embodiment of the present invention, the training process of the vibration prediction model in step 104 includes:
[0079] Step 10421: Obtain historical operating condition data and corresponding historical vibration signal data of the wind turbine under preset conditions;
[0080] Specifically, historical operating condition data and corresponding historical vibration signal data of wind turbines under normal operating conditions (i.e., preset conditions without failure) can be obtained from the wind turbine management database or platform and used as training samples to train the preset network model.
[0081] Step 10422: Preprocess the historical operating condition data and the historical vibration signal data to obtain preprocessed historical data;
[0082] Specifically, preprocessing methods can include at least one of data cleaning, outlier removal, and normalization to improve the accuracy of historical data and subsequent model training.
[0083] Step 10423: Extract historical features from the preprocessed historical data to obtain historical feature data;
[0084] Specifically, historical frequency domain data can be extracted from historical vibration signal data in preprocessed historical data using methods such as steps 102 and 103, and then combined with historical operating condition data in preprocessed historical data to form historical feature data.
[0085] Step 10424: Train the preset network model based on the historical feature data to obtain the initial training results;
[0086] Specifically, the number of neurons in the processing layers of the preset network model is set according to the number of features in the historical operating condition data. For example, if the number of features in the historical operating condition data is 3, then each processing layer of the preset network model contains 3 neurons, and the number of processing layers can be 1 to 3. The historical feature data is divided into training, validation, and test sets according to a ratio of 70%, 15%, and 15%. The training set is input into the preset network model for training. During training, various parameters of the preset network model, such as the weights and biases of the processing layers and the weights and biases of the output layer, are adjusted and optimized. After training, a pre-trained preset network model is obtained. Then, the performance of the pre-trained preset network model is verified using a validation set. During the verification process, the loss function is monitored. If the loss function starts to rise, it indicates that the model may be overfitting, and training should be stopped to prevent overfitting. For loss function, Historical vibration signal data, To predict vibration signal data, the model's performance is validated using a validation set, resulting in a pre-trained network model, i.e., the initial training result.
[0087] Step 10425: Optimize the preset network model based on the initial training results to obtain the vibration prediction model.
[0088] Specifically, the test set is input into the preset network model in the initial training results for testing. If the loss function is lower than the preset value (e.g., 0.01), it indicates that the model performance is good and can be used as a vibration prediction model. Otherwise, data needs to be collected again for training.
[0089] In an optional embodiment of the present invention, step 105 includes:
[0090] Step 1051: Obtain the residual based on the frequency domain data and the vibration prediction data;
[0091] Specifically, according to The residuals are calculated, where e is and y is . The purpose of calculating the residuals for vibration prediction data is to determine the difference between the frequency domain data and the vibration prediction data, providing a basis for subsequent diagnosis of whether the wind turbine has a fault.
[0092] Step 1052: Determine the warning result based on the residual and the preset warning threshold;
[0093] Specifically, if the residual is greater than the preset warning threshold, it indicates that there is a fault in the wind turbine and a warning is required. The warning result can include the comparison result between the residual and the preset warning threshold and the corresponding warning method.
[0094] Step 1053: Based on the warning results and the preset fault solution library, obtain the fault solution;
[0095] Specifically, the preset fault solution library is set according to historical experience, including fault types, fault mechanisms, and characteristic phenomena corresponding to the early warning results. It may also include corresponding historical fault solutions to facilitate rapid fault diagnosis and solution determination for wind turbine units.
[0096] Step 1054: Perform fault diagnosis on the wind turbine based on the frequency domain data, the vibration prediction data, the early warning results, and the fault scheme to obtain the fault diagnosis results.
[0097] Specifically, the fault diagnosis results include frequency domain data, vibration prediction data, early warning results, and fault solutions, which allows managers to intuitively and quickly understand the faults of wind turbine units and improve the efficiency of fault resolution based on the fault diagnosis results.
[0098] A specific embodiment of the wind turbine fault diagnosis method of the present invention includes:
[0099] Step 111: Obtain real-time operating condition data and real-time vibration signal data of the wind turbine.
[0100] Real-time operating condition data of wind turbines, such as temperature, electrical, environmental, and control parameters, are acquired to provide a data foundation for subsequent prediction of mechanical vibration signals. Real-time vibration signal data of wind turbines are also acquired through sensors installed at preset locations within the wind turbines to provide a data foundation for subsequent fault diagnosis.
[0101] Step 112: Denoise the real-time vibration signal data to obtain denoised data;
[0102] Noise reduction processing of real-time vibration signal data can reduce noise interference and improve data accuracy. This can be achieved through wavelet denoising and ensemble empirical mode decomposition threshold denoising. Abnormal interference during the data acquisition process can cause jumps in the time-domain waveform, leading to excessively high peak-to-peak values. In such cases, the waveform signal needs to be filtered to remove these jumps, i.e., to identify the periodic impact signal before noise reduction.
[0103] Step 113: Perform frequency domain transformation on the noise-reduced data to obtain frequency domain data;
[0104] Feature extraction (frequency domain transformation) is performed on the denoised vibration data. Empirical mode decomposition and its optimization algorithms (such as periodic impact signal extraction) are used to extract early signals that reflect the mechanical fault characteristics of the wind turbine. Feature extraction is performed on the filtered high-frequency waveform signal, combining fault type to extract characteristic frequencies in the spectrum and envelope spectrum, such as 1X amplitude, 2X amplitude, BPFI (rolling outer ring fault frequency), BPFO (rolling inner ring fault frequency), meshing frequency, frequency band noise floor, and vibration phase. The denoised data is a one-dimensional array. Frequency domain data consists of frequency and amplitude arrays, such as [[0,0.01],[0.75,0.11],[1.5,0.13],……]. By combining the frequency domain data with the wind turbine equipment parameters and real-time rotational speed, the desired characteristic components (i.e., amplitudes) can be extracted from the transformation results. For example, if the real-time rotational speed is 600 rpm, then the frequency corresponding to 1x amplitude is 60 Hz. Therefore, the amplitude corresponding to the 60 Hz frequency index is the first harmonic characteristic component.
[0105] Step 114: Obtain vibration prediction data based on real-time operating condition data and the trained vibration prediction model;
[0106] First, a pre-set network model is trained based on historical operating data of the wind turbine to obtain a vibration prediction model. The vibration prediction model then generates vibration prediction data based on real-time operating condition data.
[0107] Step 115: Perform fault diagnosis on the wind turbine based on the frequency domain data and the vibration prediction data to obtain the fault diagnosis result.
[0108] Residual analysis is performed on frequency domain data and vibration prediction data. Based on the 3σ rule, if the deviation exceeds 3σ at multiple times, the wind turbine is considered abnormal, and an early warning signal is issued. Based on the early warning signal and combined with the dynamic mechanism of wind power equipment, accurate location and early warning of common mechanical faults in wind turbines are achieved. Expert knowledge in the wind power industry is organized and packaged into a mechanism model (i.e., a preset fault solution library), as shown in Table 1. After identifying an abnormal equipment status in the early warning signal, the preset fault solution library allows for accurate location of the wind turbine equipment fault.
[0109] Table 1 Preset Fault Solution Library
[0110]
[0111] The wind turbine fault diagnosis method of this invention preprocesses the collected real-time vibration waveform data to extract the frequency domain data from the vibration signal; then, it obtains vibration prediction data based on the vibration prediction model, performs residual feature identification on the frequency domain data and the vibration prediction data, and considers the equipment to be abnormal if the residual exceeds a threshold. It calculates the current characteristic amplitude under the operating condition based on statistical methods, and finds the specific fault mode and degree of the equipment according to the preset fault scheme library, so as to provide support for the stable operation of the wind turbine equipment.
[0112] like Figure 2 As shown, an embodiment of the present invention provides a wind turbine fault diagnosis device 200, comprising:
[0113] The acquisition module 201 is used to acquire real-time operating condition data and real-time vibration signal data of the wind turbine.
[0114] The processing module 202 is used to perform noise reduction processing on the real-time vibration signal data to obtain noise-reduced data; to perform frequency domain conversion on the noise-reduced data to obtain frequency domain data; to obtain vibration prediction data based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data; and to perform fault diagnosis on the wind turbine based on the frequency domain data and the vibration prediction data to obtain fault diagnosis results.
[0115] Optionally, real-time operating condition data and real-time vibration signal data of the wind turbine can be acquired, including:
[0116] Acquire real-time operating condition data of the wind turbine generator set; the operating condition data includes temperature parameter data, electrical parameter data, environmental parameter data, and control parameter data;
[0117] Real-time vibration signal data of the wind turbine is acquired by sensors installed at preset locations in the wind turbine.
[0118] Optionally, the real-time vibration signal data is subjected to noise reduction processing to obtain noise-reduced data, including:
[0119] The real-time vibration signal data is decomposed to obtain low-frequency and high-frequency components;
[0120] Thresholding is performed on the high-frequency components to obtain detail coefficients;
[0121] The noise reduction data is obtained based on the low-frequency components and the detail coefficients.
[0122] Optionally, the noise-reduced data is frequency-domain transformed to obtain frequency-domain data, including:
[0123] pass The noise-reduced data is transformed into frequency domain data.
[0124] in, For frequency domain data, The data is denoised, and n is the time-domain index. N is the total number of sampling points, and k is the frequency domain index. , is the rotation factor.
[0125] Optionally, based on the real-time operating condition data and the trained vibration prediction model, vibration prediction data is obtained, including:
[0126] The real-time operating condition data is input into the input layer of the vibration prediction model to obtain the first output result;
[0127] The first output result is input into at least one processing layer of the vibration prediction model to obtain the second output result;
[0128] The second output result is input into the output layer of the vibration prediction model to obtain vibration prediction data.
[0129] Optionally, the training process of the vibration prediction model includes:
[0130] Acquire historical operating condition data and corresponding historical vibration signal data of wind turbines under preset conditions;
[0131] The historical operating condition data and the historical vibration signal data are preprocessed to obtain preprocessed historical data.
[0132] Historical features are extracted from the preprocessed historical data to obtain historical feature data;
[0133] The preset network model is trained based on the historical feature data to obtain the initial training results;
[0134] The preset network model is optimized based on the initial training results to obtain the vibration prediction model.
[0135] Optionally, fault diagnosis of the wind turbine is performed based on the frequency domain data and the vibration prediction data to obtain fault diagnosis results, including:
[0136] The residual is obtained based on the frequency domain data and the vibration prediction data;
[0137] The warning result is determined based on the residual and the preset warning threshold;
[0138] Based on the warning results and the preset fault solution library, a fault solution is obtained;
[0139] Based on the frequency domain data, the vibration prediction data, the early warning results, and the fault diagnosis scheme, the wind turbine is diagnosed to obtain the fault diagnosis results.
[0140] The wind turbine fault diagnosis device proposed in this invention acquires real-time operating condition data and real-time vibration signal data of the wind turbine, then performs noise reduction processing on the real-time vibration signal data to obtain noise-reduced data, and then performs frequency domain conversion on the noise-reduced data to obtain frequency domain data. Based on the real-time operating condition data and the trained vibration prediction model, vibration prediction data is obtained. The frequency domain data and vibration prediction data are compared to diagnose faults in the wind turbine and obtain fault diagnosis results. This can effectively improve the accuracy of fault diagnosis of wind turbines and improve the operating efficiency of wind turbines.
[0141] It should be noted that this device corresponds to the method described above, and all implementations in the method embodiments described above are applicable to the embodiments of this device and can achieve the same technical effect. Further details are omitted in this embodiment.
[0142] This invention also provides a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any of the above embodiments. All implementations in the above method embodiments are applicable to the embodiments of this device and can achieve the same technical effects. Further details are omitted in this embodiment.
[0143] This invention also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method as described in any of the above embodiments. All implementations in the above method embodiments are applicable to the embodiments of this device and can achieve the same technical effects. Further details are omitted in this embodiment.
[0144] It should be noted that in the apparatus and method of the present invention, the components or steps can obviously be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of the present invention. Furthermore, the steps for performing the above series of processes can naturally be performed in the order described and in chronological order, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel, overlapping, or independently of each other.
[0145] It should be noted that in the above embodiments, 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 a process, method, article, or apparatus. Without further limitations, 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. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments described above is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
[0146] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for diagnosing faults in wind turbine generators, characterized in that, include: Acquire real-time operating condition data and real-time vibration signal data of wind turbine units; The real-time vibration signal data is subjected to noise reduction processing to obtain noise-reduced data; The noise-reduced data is transformed into frequency domain data. Vibration prediction data is obtained based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data. Based on the frequency domain data and the vibration prediction data, fault diagnosis is performed on the wind turbine to obtain fault diagnosis results; The noise reduction data is subjected to frequency domain transformation to obtain frequency domain data, including: pass The noise-reduced data is transformed into frequency domain data. in, For frequency domain data, The data is denoised, and n is the time-domain index. N is the total number of sampling points, and k is the frequency domain index. , It is the rotation factor; The vibration prediction data, obtained based on the real-time operating condition data and the trained vibration prediction model, includes: The real-time operating condition data is input into the input layer of the vibration prediction model to obtain the first output result; specifically, the input layer of the vibration prediction model performs normalization processing on the real-time operating condition data, and the input layer adopts... The real-time operating condition data is normalized to obtain the first output result, in which... The first output is 'x', where 'x' represents real-time operating condition data. This represents the minimum value of real-time operating condition data. This represents the maximum value of the real-time operating condition data. The first output result is input into at least one processing layer of the vibration prediction model to obtain a second output result; specifically, the processing layer of the vibration prediction model performs a nonlinear transformation on the first output result, and the processing layer... A nonlinear transformation is performed on the first output result to obtain the second output result, where, This represents the output of the j-th neuron in the k-th processing layer. For activation function, For the kth The outputs of all neurons in layer 1 are weighted and summed. For the kth The connection weights from the i-th neuron in layer 1 to the j-th neuron in layer k. For the kth The output of the i-th neuron in layer 1, where p is the number of neurons in the (k-1)-th processing layer. This is the bias term for the j-th neuron in the k-th layer; The second output result is input into the output layer of the vibration prediction model to obtain vibration prediction data; specifically, the output layer of the vibration prediction model is processed through... Vibration prediction data were obtained, among which, For vibration prediction data, This is the second output result. is the connection weight from the k-th processing layer neuron to the output layer, M is the number of neurons in the k-th processing layer, and b is the bias term of the output layer.
2. The wind turbine fault diagnosis method according to claim 1, characterized in that, Acquire real-time operating condition data and real-time vibration signal data of wind turbine units, including: Acquire real-time operating condition data of the wind turbine generator set; the operating condition data includes temperature parameter data, electrical parameter data, environmental parameter data, and control parameter data; Real-time vibration signal data of the wind turbine is acquired by sensors installed at preset locations in the wind turbine.
3. The wind turbine fault diagnosis method according to claim 1, characterized in that, The real-time vibration signal data is denoised to obtain denoised data, including: The real-time vibration signal data is decomposed to obtain low-frequency and high-frequency components; Thresholding is performed on the high-frequency components to obtain detail coefficients; The noise reduction data is obtained based on the low-frequency components and the detail coefficients.
4. The wind turbine fault diagnosis method according to claim 1, characterized in that, The training process of the vibration prediction model includes: Acquire historical operating condition data and corresponding historical vibration signal data of wind turbines under preset conditions; The historical operating condition data and the historical vibration signal data are preprocessed to obtain preprocessed historical data. Historical features are extracted from the preprocessed historical data to obtain historical feature data; The preset network model is trained based on the historical feature data to obtain the initial training results; The preset network model is optimized based on the initial training results to obtain the vibration prediction model.
5. The wind turbine fault diagnosis method according to claim 1, characterized in that, Based on the frequency domain data and the vibration prediction data, fault diagnosis is performed on the wind turbine to obtain fault diagnosis results, including: The residual is obtained based on the frequency domain data and the vibration prediction data; The warning result is determined based on the residual and the preset warning threshold; Based on the warning results and the preset fault solution library, a fault solution is obtained; Based on the frequency domain data, the vibration prediction data, the early warning results, and the fault diagnosis scheme, the wind turbine is diagnosed to obtain the fault diagnosis results.
6. A wind turbine fault diagnosis device, characterized in that, include: The acquisition module is used to acquire real-time operating condition data and real-time vibration signal data of the wind turbine. The processing module is used to perform noise reduction processing on the real-time vibration signal data to obtain noise-reduced data; to perform frequency domain conversion on the noise-reduced data to obtain frequency domain data; to obtain vibration prediction data based on the real-time operating condition data and the trained vibration prediction model; the vibration prediction model is obtained by training a preset network model based on the historical operating condition data of the wind turbine and the corresponding historical vibration signal data; and to perform fault diagnosis on the wind turbine based on the frequency domain data and the vibration prediction data to obtain fault diagnosis results. The noise reduction data is subjected to frequency domain transformation to obtain frequency domain data, including: pass The noise-reduced data is transformed into frequency domain data. in, For frequency domain data, The data is denoised, and n is the time-domain index. N is the total number of sampling points, and k is the frequency domain index. , It is the rotation factor; The vibration prediction data, obtained based on the real-time operating condition data and the trained vibration prediction model, includes: The real-time operating condition data is input into the input layer of the vibration prediction model to obtain the first output result; specifically, the input layer of the vibration prediction model performs normalization processing on the real-time operating condition data, and the input layer adopts... The real-time operating condition data is normalized to obtain the first output result, in which... The first output is 'x', where 'x' represents real-time operating condition data. This represents the minimum value of real-time operating condition data. This represents the maximum value of the real-time operating condition data. The first output result is input into at least one processing layer of the vibration prediction model to obtain a second output result; specifically, the processing layer of the vibration prediction model performs a nonlinear transformation on the first output result, and the processing layer... A nonlinear transformation is performed on the first output result to obtain the second output result, where, This represents the output of the j-th neuron in the k-th processing layer. For activation function, For the kth The outputs of all neurons in layer 1 are weighted and summed. For the kth The connection weights from the i-th neuron in layer 1 to the j-th neuron in layer k. For the kth The output of the i-th neuron in layer 1, where p is the number of neurons in the (k-1)-th processing layer. This is the bias term for the j-th neuron in the k-th layer; The second output result is input into the output layer of the vibration prediction model to obtain vibration prediction data; specifically, the output layer of the vibration prediction model is processed through... Vibration prediction data were obtained, among which, For vibration prediction data, This is the second output result. is the connection weight from the k-th processing layer neuron to the output layer, M is the number of neurons in the k-th processing layer, and b is the bias term of the output layer.
7. A computing device, characterized in that, include: A processor, a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The system stores instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 5.