Wind driven generator bearing state analysis method and system and medium
By using real-time data acquisition from sensor arrays and optimizing feature selection algorithms, combined with a dynamic early warning mechanism, the adaptability and accuracy issues of bearing condition assessment in existing technologies have been resolved, enabling efficient equipment condition assessment and predictive maintenance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- RUIHU ZHIKE DATA (SUZHOU) CO LTD
- Filing Date
- 2025-12-29
- Publication Date
- 2026-05-12
AI Technical Summary
The existing CMS system cannot adapt to the differences between different models and individual bearings, making it difficult to assess the overall deterioration trend of bearings and to automatically provide a quantitative division of 'break-in - normal - early failure - late failure', which affects predictive maintenance in the field.
The operating parameters of the wind turbine bearings are collected in real time by a sensor array, preprocessed and feature extracted, and then optimized by a feature selection algorithm for dimensionality reduction. The data are then input into the bearing condition identification model and combined with a dynamic early warning mechanism for condition assessment and anomaly warning.
It enables automatic assessment and identification of the bearing condition of wind turbines, reduces the technical requirements for maintenance personnel, and improves the reliability and predictability of equipment condition assessment.
Smart Images
Figure CN122020331A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of bearing condition analysis technology, and more specifically, to a method, system, and medium for analyzing the condition of wind turbine bearings. Background Technology
[0002] Wind turbine bearings undergo several stages after being put into operation, including break-in, normal operation, and failure. Existing CMS systems, which use fixed thresholds or spectral feature comparison methods, have the following shortcomings:
[0003] 1) Fixed thresholds cannot adapt to the individual differences between different machine models and bearings;
[0004] 2) Spectrum comparison only focuses on the failure frequency of components, making it difficult to assess the overall deterioration trend of the bearing;
[0005] 3) It cannot automatically provide a quantitative division of "break-in - normal - early failure - late failure", which is not conducive to predictive maintenance on site. Summary of the Invention
[0006] The purpose of this application is to provide a method, system, and medium for analyzing the condition of wind turbine bearings. By automatically assessing and identifying the condition of wind turbine bearings, artificial intelligence technology is used to achieve direct identification of the results, reducing the technical requirements of wind farm maintenance personnel, realizing high-density equipment condition assessment, and improving the reliability of equipment operation and maintenance and assessment.
[0007] This application also provides a method for analyzing the condition of wind turbine bearings, including:
[0008] Based on the real-time acquisition of operating parameters of wind turbine bearings by sensor array, multi-dimensional operating parameter information is obtained;
[0009] The multidimensional operating parameter information is preprocessed to obtain the preprocessed multidimensional operating parameter information.
[0010] Based on the preprocessed multidimensional operating parameter information, time-domain features and frequency-domain features are extracted respectively to obtain a feature set;
[0011] A feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset.
[0012] The optimal feature subset is input into the preset bearing condition recognition model, and the real-time condition information of the wind turbine bearing is output based on the bearing condition recognition model.
[0013] Based on the dynamic early warning mechanism, the real-time status information of wind turbine bearings is used to perform status assessment and anomaly warning, and the status analysis results are obtained.
[0014] Optionally, in the wind turbine bearing condition analysis method described in the embodiments of this application, the sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor;
[0015] The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration.
[0016] The bearing housing temperature signal is acquired based on a temperature sensor;
[0017] Bearing speed signals are acquired using a speed sensor;
[0018] During the data acquisition process, the data acquisition timestamp and the operating condition information of the wind turbine are recorded simultaneously. The operating condition information includes wind speed, unit output power and yaw angle, thus obtaining multi-dimensional operating parameter information.
[0019] Optionally, in the wind turbine bearing condition analysis method described in the embodiments of this application, the preprocessing operation includes outlier removal, data smoothing and noise reduction, signal resampling and data standardization in sequence. The outlier removal adopts the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation.
[0020] Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals;
[0021] Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency;
[0022] Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
[0023] Optionally, in the wind turbine bearing condition analysis method described in the embodiments of this application, the time-domain features include at least one of peak value, peak factor, kurtosis, skewness, root mean square value and waveform factor;
[0024] The frequency domain features include at least one of the centroid frequency, mean square frequency, frequency variance, and peak frequency.
[0025] Optionally, in the wind turbine bearing condition analysis method described in this application embodiment, a feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, followed by a feature validity verification step:
[0026] Input the optimal feature subset into the validation model;
[0027] Calculate the state recognition accuracy, recall, and F1 score corresponding to the feature subset;
[0028] Determine whether the state recognition accuracy, recall, and F1 score are greater than preset thresholds;
[0029] If the state recognition accuracy, recall, and F1 score are all greater than the preset thresholds, then the feature optimization is deemed effective.
[0030] If any of the state recognition accuracy, recall, or F1 score is less than or equal to a preset threshold, feature extraction will be performed again or the feature selection algorithm parameters will be adjusted.
[0031] Optionally, in the wind turbine bearing condition analysis method described in the embodiments of this application, a bearing condition level threshold is set based on a dynamic early warning mechanism, and a bearing health assessment report is generated based on real-time condition information and the bearing condition level threshold. When the real-time condition information is greater than or equal to the bearing condition level threshold, an early warning signal is triggered.
[0032] The bearing health assessment report includes the bearing's real-time status level, real-time values and trend curves of operating parameters, the degree of abnormality of characteristic parameters, and the predicted value of the bearing's remaining service life. The predicted value of the bearing's remaining service life is calculated by a life prediction model constructed based on the optimal feature subset. The life prediction model is an LSTM neural network model or a grey prediction model.
[0033] Secondly, embodiments of this application provide a wind turbine bearing condition analysis system. The system includes a memory and a processor. The memory includes a program for a wind turbine bearing condition analysis method. When the program for the wind turbine bearing condition analysis method is executed by the processor, it performs the following steps:
[0034] Based on the real-time acquisition of operating parameters of wind turbine bearings by sensor array, multi-dimensional operating parameter information is obtained;
[0035] The multidimensional operating parameter information is preprocessed to obtain the preprocessed multidimensional operating parameter information.
[0036] Based on the preprocessed multidimensional operating parameter information, time-domain features and frequency-domain features are extracted respectively to obtain a feature set;
[0037] A feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset.
[0038] The optimal feature subset is input into the preset bearing condition recognition model, and the real-time condition information of the wind turbine bearing is output based on the bearing condition recognition model.
[0039] Based on the dynamic early warning mechanism, the real-time status information of wind turbine bearings is used to perform status assessment and anomaly warning, and the status analysis results are obtained.
[0040] Optionally, in the wind turbine bearing condition analysis system described in the embodiments of this application, the sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor;
[0041] The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration.
[0042] The bearing housing temperature signal is acquired based on a temperature sensor;
[0043] Bearing speed signals are acquired using a speed sensor;
[0044] During the data acquisition process, the data acquisition timestamp and the operating condition information of the wind turbine are recorded simultaneously. The operating condition information includes wind speed, unit output power and yaw angle, thus obtaining multi-dimensional operating parameter information.
[0045] Optionally, in the wind turbine bearing condition analysis system described in the embodiments of this application, the preprocessing operation includes outlier removal, data smoothing and noise reduction, signal resampling and data standardization in sequence. The outlier removal adopts the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation.
[0046] Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals;
[0047] Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency;
[0048] Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
[0049] Thirdly, embodiments of this application also provide a computer-readable storage medium, which includes a wind turbine bearing condition analysis method program. When the wind turbine bearing condition analysis method program is executed by a processor, it implements the steps of the wind turbine bearing condition analysis method as described in any of the above claims.
[0050] As can be seen from the above, the wind turbine bearing condition analysis method, system, and medium provided in this application collect multi-dimensional operating parameter information of the wind turbine bearing in real time based on a sensor array; preprocess the multi-dimensional operating parameter information to obtain preprocessed multi-dimensional operating parameter information; extract time-domain features and frequency-domain features based on the preprocessed multi-dimensional operating parameter information to obtain a feature set; use a feature selection algorithm to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain an optimal feature subset; input the optimal feature subset into a preset bearing condition recognition model, and output the real-time condition information of the wind turbine bearing based on the bearing condition recognition model; perform condition assessment and anomaly warning based on the real-time condition information of the wind turbine bearing based on a dynamic early warning mechanism to obtain the condition analysis result; the present invention automatically assesses and identifies the condition of the wind turbine bearing, uses artificial intelligence technology to achieve direct identification of the results, reduces the technical requirements of wind farm maintenance personnel, achieves high-density equipment condition assessment, and improves the reliability of equipment operation and maintenance and assessment.
[0051] Other features and advantages of this application will be set forth in the following description, and the advantages of this application will be apparent in part from the description, or may be learned by practicing the embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description
[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0053] Figure 1 A flowchart of a wind turbine bearing condition analysis method provided in this application embodiment;
[0054] Figure 2 A flowchart illustrating the algorithm for a wind turbine bearing condition analysis method provided in this application embodiment;
[0055] Figure 3 A schematic diagram illustrating the prediction results of a wind turbine bearing condition analysis method provided in this application embodiment;
[0056] Figure 4 This is a schematic block diagram of the hardware connection of a wind turbine bearing condition analysis system provided in an embodiment of this application. Detailed Implementation
[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application 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 this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0058] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0059] Please refer to Figures 1-4 As shown, this wind turbine bearing condition analysis method is used in terminal equipment. The method includes the following steps:
[0060] S101, based on the sensor group, collects the operating parameters of the wind turbine bearing in real time to obtain multi-dimensional operating parameter information;
[0061] S102, perform preprocessing on the multidimensional operating parameter information to obtain preprocessed multidimensional operating parameter information;
[0062] S103, based on the preprocessed multidimensional operating parameter information, extract time-domain features and frequency-domain features respectively to obtain a feature set;
[0063] S104. The feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset.
[0064] S105, Input the optimal feature subset into the preset bearing condition recognition model, and output the real-time condition information of the wind turbine bearing based on the bearing condition recognition model;
[0065] It should be noted that the bearing condition recognition model is a machine learning-based classification model. The classification model includes at least a support vector machine model, a random forest model, a gradient boosting tree model, or a convolutional neural network model. The classification model is trained in the following way: historical operating parameters of wind turbine bearings under different conditions are collected, and after preprocessing and feature optimization, training datasets and test datasets are constructed. The initial classification model is trained using the training dataset, and the model performance is verified and the model hyperparameters are adjusted using the test dataset until the model accuracy reaches a preset threshold.
[0066] S106, based on the dynamic early warning mechanism, the real-time status information of the wind turbine bearing is used to perform status assessment and anomaly early warning, and the status analysis results are obtained.
[0067] According to an embodiment of the present invention, in a method for analyzing the condition of a wind turbine bearing, the sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor.
[0068] The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration.
[0069] The bearing housing temperature signal is acquired based on a temperature sensor;
[0070] Bearing speed signals are acquired using a speed sensor;
[0071] During the data acquisition process, the data acquisition timestamp and the wind turbine's operating condition information are recorded simultaneously. The operating condition information includes wind speed, unit output power, and yaw angle, thus obtaining multi-dimensional operating parameter information.
[0072] It should be noted that the vibration sensor is a piezoelectric accelerometer, which is installed on the bearing housing corresponding to the outer ring of the bearing and the shaft end corresponding to the inner ring, respectively, with a sampling frequency set to 10kHz-50kHz; the temperature sensor is a platinum resistance temperature sensor, which is embedded in the temperature measuring hole reserved in the bearing housing, with a temperature measuring range of -40℃ to 150℃; the speed sensor is a Hall effect speed sensor, which is installed on the side of the gear disk coaxial with the bearing, and the sampling frequency is synchronized with that of the vibration sensor.
[0073] According to an embodiment of the present invention, in a wind turbine bearing condition analysis method of this application embodiment, the preprocessing operation includes outlier removal, data smoothing and noise reduction, signal resampling and data standardization in sequence. The outlier removal adopts the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation.
[0074] Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals;
[0075] Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency;
[0076] Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
[0077] According to an embodiment of the present invention, in a wind turbine bearing condition analysis method of this application, the time-domain characteristics include at least one of peak value, peak factor, kurtosis, skewness, root mean square value and waveform factor.
[0078] Frequency domain characteristics include at least one of the centroid frequency, mean square frequency, frequency variance, and peak frequency.
[0079] According to an embodiment of the present invention, in a wind turbine bearing condition analysis method of this application, a feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, followed by a feature validity verification step:
[0080] Input the optimal feature subset into the validation model;
[0081] Calculate the state recognition accuracy, recall, and F1 score corresponding to the feature subset;
[0082] Determine whether the state recognition accuracy, recall, and F1 score are greater than preset thresholds;
[0083] If the state recognition accuracy, recall, and F1 score are all greater than the preset thresholds, then the feature optimization is deemed effective.
[0084] If any of the state recognition accuracy, recall, or F1 score is less than or equal to a preset threshold, feature extraction will be performed again or the feature selection algorithm parameters will be adjusted.
[0085] It should be noted that the feature selection algorithm is one or more of the following combinations: ReliefF algorithm, Random Forest feature importance ranking algorithm, or L1 regularization algorithm. The specific implementation process is as follows: the feature selection algorithm is used to calculate the importance score of each feature in the feature set, an importance score threshold is set, features with scores higher than the threshold are retained, and the optimal feature subset is obtained. When multiple algorithms are used in combination, a voting mechanism is used to determine the final retained features.
[0086] According to an embodiment of the present invention, in a wind turbine bearing condition analysis method of this application, a bearing condition level threshold is set based on a dynamic early warning mechanism. Based on real-time condition information and the bearing condition level threshold, a bearing health assessment report is generated. When the real-time condition information is greater than or equal to the bearing condition level threshold, an early warning signal is triggered. The early warning signal includes graded early warnings, namely, Level 1, Level 2, and Level 3 early warnings, corresponding to the slight wear, moderate wear, and severe wear states of the bearing. Different levels of early warning signals correspond to different early warning methods. Level 1 early warnings use local audio-visual prompts, Level 2 early warnings use remote platform message pushes, and Level 3 early warnings use remote platform message pushes combined with staff SMS reminders.
[0087] The bearing health assessment report includes the bearing's real-time status level, real-time values and trend curves of operating parameters, the degree of abnormality of characteristic parameters, and the predicted value of the bearing's remaining service life. The predicted value of the bearing's remaining service life is calculated by a life prediction model constructed based on the optimal feature subset. The life prediction model is an LSTM neural network model or a grey prediction model.
[0088] According to an embodiment of the present invention, the virtual simulation of a 3 MW direct-drive onshore turbine unit specifically includes:
[0089] 1. Scenarios and Hardware;
[0090] Unit: Assuming a 3 MW onshore direct-drive permanent magnet wind turbine, main bearing model SKF 239 / 800 CAK / W33, rated dynamic load 1.5 MN, operating wind speed 6–8 m / s, and ambient temperature 20 ℃.
[0091] Sensor: Virtual model V-123, sensitivity 100 mV / g, bandwidth 0.5 Hz–25 kHz, resonant frequency ≥30kHz.
[0092] Edge gateway: ARM Cortex-A72 1.5 GHz, 4 GB RAM, Ubuntu 20.04, Python 3.8.
[0093] Cloud-based: 8 vCPUs / 16 GB, scikit-learn 1.0.
[0094] 2. Parameter point values;
[0095] Sampling frequency fs = 25.6 kHz (fixed value, not range);
[0096] Sampling length 10 s → 256,000 points;
[0097] Similarity threshold S_alarm=0.85 (initial setting, to be updated online later);
[0098] K-means: k = 4 (run-in - normal - minor failure - major failure), random seed = 42, maximum iterations 500;
[0099] SVC: RBF kernel, C = 10, γ = 0.01, 5-fold cross-validation.
[0100] The entire process is calculated (key intermediate quantities are disclosed).
[0101] Construct the "normal reference" vector B0 (23-dimensional point values): T1 (mean) = 0 m / s², T7 (kurtosis) = 3.00, F1 (frequency domain centroid) = 120 Hz, F2 (mean square frequency) = 180 Hz ….
[0102] Virtual waveform at time t: A 0.1 mm² peel is introduced into the outer ring, corresponding to the feature offsets: T7 = 4.20, F1 = 135Hz. The remaining features are linearly amplified by 1.05–1.30 times according to engineering experience to obtain B_t.
[0103] Improved cosine similarity calculation:
[0104] molecular
[0105]
[0106] denominator
[0107] ;
[0108] S_t = 1.38 / 1.68 = 0.821 < 0.85, triggering an exception.
[0109] After standardization, the data is sent to a pre-set cluster center. The data with the smallest Euclidean distance is labeled as "minor fault".
[0110] If S_t < 0.85 and the label ≠ “normal” for 12 consecutive hours, the system pushes a predictive maintenance suggestion to SCADA.
[0111] 4. Virtual results;
[0112] Early warning time: 14 days (only 2 days for the traditional fixed threshold method);
[0113] False alarm rate: 0‰ (1000 consecutive virtual runs, no false alarms).
[0114] The single calculation time at the edge is 0.73 s, the peak memory usage is <1.2 GB, and the sampling interval is 1 h.
[0115] This virtual embodiment fully reproduces all the technical features of claims 1-5 and achieves the beneficial effects described in the specification.
[0116] The virtual simulation of a 5 MW doubly-fed turbine generator set at sea is as follows:
[0117] 1. Scenario: 5 MW offshore doubly-fed induction generator unit, bearing model FAG 230 / 600-B-MB, average load 2.2 MN, ambient humidity 95%.
[0118] 2. The sampling rate was increased to fs = 51.2 kHz, and the rest of the hardware was the same as in Example 1.
[0119] 3. Construct a three-segment virtual feature sequence: "break-in → normal → minor fault":
[0120] Break-in period: T7=3.40, F1=140 Hz;
[0121] Normal period: T7 = 3.00, F1 = 120 Hz;
[0122] Minor fault period: T7=4.50, F1=155 Hz.
[0123] 4. Calculations show that the SVC output labels match the preset stage 100% accurately.
[0124] 5. Virtual inspection results: The outer ring spalling length was 5 mm, consistent with the prediction; the false alarm rate was 0%, and the early warning was issued 15 days in advance.
[0125] According to an embodiment of the present invention, the low-temperature virtual simulation of a 1.5 MW unit on a plateau specifically includes:
[0126] 1. Scenario: Altitude 3800 m, ambient temperature -20 ℃, bearing model ZWZ 230 / 500 CA / W33.
[0127] 2. Introduce a temperature correction factor α = 0.97 for T1-T10, while keeping the rest unchanged.
[0128] 3. Virtual characteristics: At low temperature, the T7 baseline rises to 3.15, and the failure period T7 = 4.35.
[0129] 4. The calculated S_t = 0.829 < 0.85, the system still accurately outputs "minor fault", with a false alarm rate of <1.5%.
[0130] The method of proof is also effective for low-temperature operating conditions.
[0131] Secondly, embodiments of this application provide a wind turbine bearing condition analysis system. The system includes a memory and a processor. The memory includes a program for a wind turbine bearing condition analysis method. When the program for the wind turbine bearing condition analysis method is executed by the processor, it performs the following steps:
[0132] Based on the real-time acquisition of operating parameters of wind turbine bearings by sensor array, multi-dimensional operating parameter information is obtained;
[0133] The multidimensional operating parameter information is preprocessed to obtain the preprocessed multidimensional operating parameter information.
[0134] Based on the preprocessed multidimensional operating parameter information, time-domain features and frequency-domain features are extracted respectively to obtain a feature set;
[0135] A feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset.
[0136] The optimal feature subset is input into the preset bearing condition recognition model, and the real-time condition information of the wind turbine bearing is output based on the bearing condition recognition model.
[0137] Based on the dynamic early warning mechanism, the real-time status information of wind turbine bearings is used to perform status assessment and anomaly warning, and the status analysis results are obtained.
[0138] It should be noted that the bearing health assessment report supports visualization, showing the historical trend of operating parameters through line graphs, the abnormal distribution of characteristic parameters through heat maps, and the percentage duration of each bearing condition through pie charts. It also supports the export function of the report, with export formats including PDF, Excel and CSV, which facilitates offline analysis by staff.
[0139] According to an embodiment of the present invention, in a wind turbine bearing condition analysis system of this application, the sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor;
[0140] The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration.
[0141] The bearing housing temperature signal is acquired based on a temperature sensor;
[0142] Bearing speed signals are acquired using a speed sensor;
[0143] During the data acquisition process, the data acquisition timestamp and the wind turbine's operating condition information are recorded simultaneously. The operating condition information includes wind speed, unit output power, and yaw angle, thus obtaining multi-dimensional operating parameter information.
[0144] According to an embodiment of the present invention, in a wind turbine bearing condition analysis system of this application, the preprocessing operation includes outlier removal, data smoothing and noise reduction, signal resampling and data standardization in sequence. The outlier removal adopts the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation.
[0145] Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals;
[0146] Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency;
[0147] Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
[0148] A third aspect of the present invention provides a computer-readable storage medium including a wind turbine bearing condition analysis method program, wherein when the wind turbine bearing condition analysis method program is executed by a processor, it implements the steps of a wind turbine bearing condition analysis method as described in any of the above claims.
[0149] This invention discloses a method, system, and medium for analyzing the condition of wind turbine bearings. It acquires multi-dimensional operating parameter information by real-time acquisition of wind turbine bearing operating parameters using a sensor array. This multi-dimensional operating parameter information is preprocessed to obtain preprocessed multi-dimensional operating parameter information. Time-domain and frequency-domain features are extracted from the preprocessed multi-dimensional operating parameter information to obtain a feature set. A feature selection algorithm is used to optimize the feature set by dimensionality reduction, eliminating redundant and invalid features to obtain an optimal feature subset. This optimal feature subset is input into a preset bearing condition recognition model, which outputs real-time condition information of the wind turbine bearing. A dynamic early warning mechanism is used to assess the real-time condition information of the wind turbine bearing and provide anomaly warnings, resulting in a condition analysis result. This invention automatically assesses and identifies the condition of wind turbine bearings, using artificial intelligence technology to achieve direct identification of results, reducing the technical requirements for wind farm maintenance personnel, achieving high-density equipment condition assessment, and improving the reliability of equipment operation and maintenance and assessment.
[0150] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.
[0151] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.
[0152] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.
[0153] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0154] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.
Claims
1. A method for analyzing the condition of wind turbine bearings, characterized in that, include: Based on the real-time acquisition of operating parameters of wind turbine bearings by sensor array, multi-dimensional operating parameter information is obtained; The multidimensional operating parameter information is preprocessed to obtain the preprocessed multidimensional operating parameter information. Based on the preprocessed multidimensional operating parameter information, time-domain features and frequency-domain features are extracted respectively to obtain a feature set; A feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset. The optimal feature subset is input into the preset bearing condition recognition model, and the real-time condition information of the wind turbine bearing is output based on the bearing condition recognition model. Based on the dynamic early warning mechanism, the real-time status information of wind turbine bearings is used to perform status assessment and anomaly warning, and the status analysis results are obtained.
2. The method for analyzing the condition of wind turbine bearings according to claim 1, characterized in that, The sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor; The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration. The bearing housing temperature signal is acquired based on a temperature sensor; Bearing speed signals are acquired using a speed sensor; During the data acquisition process, the data acquisition timestamp and the operating condition information of the wind turbine are recorded simultaneously. The operating condition information includes wind speed, unit output power and yaw angle, thus obtaining multi-dimensional operating parameter information.
3. The method for analyzing the bearing condition of a wind turbine generator according to claim 2, characterized in that, The preprocessing operations include outlier removal, data smoothing and denoising, signal resampling, and data standardization. Outlier removal uses the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation. Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals; Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency; Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
4. The method for analyzing the condition of wind turbine bearings according to claim 3, characterized in that, The time-domain features include at least one of peak value, peak factor, kurtosis, skewness, root mean square value, and waveform factor; The frequency domain features include at least one of the centroid frequency, mean square frequency, frequency variance, and peak frequency.
5. The method for analyzing the condition of a wind turbine bearing according to claim 4, characterized in that, The feature set is optimized by dimensionality reduction using a feature selection algorithm, followed by a feature validity verification step. Input the optimal feature subset into the validation model; Calculate the state recognition accuracy, recall, and F1 score corresponding to the feature subset; Determine whether the state recognition accuracy, recall, and F1 score are greater than preset thresholds; If the state recognition accuracy, recall, and F1 score are all greater than the preset thresholds, then the feature optimization is deemed effective. If any of the state recognition accuracy, recall, or F1 score is less than or equal to a preset threshold, feature extraction will be performed again or the feature selection algorithm parameters will be adjusted.
6. The method for analyzing the condition of a wind turbine bearing according to claim 5, characterized in that, The bearing status level threshold is set based on the dynamic early warning mechanism. Based on the real-time status information and the bearing status level threshold, a bearing health assessment report is generated. When the real-time status information is greater than or equal to the bearing status level threshold, an early warning signal is triggered. The bearing health assessment report includes the bearing's real-time status level, real-time values and trend curves of operating parameters, the degree of abnormality of characteristic parameters, and the predicted value of the bearing's remaining service life. The predicted value of the bearing's remaining service life is calculated by a life prediction model constructed based on the optimal feature subset. The life prediction model is an LSTM neural network model or a grey prediction model.
7. A wind turbine bearing condition analysis system, characterized in that, The system includes a memory and a processor. The memory contains a program for a wind turbine bearing condition analysis method. When the program for the wind turbine bearing condition analysis method is executed by the processor, it performs the following steps: Based on the real-time acquisition of operating parameters of wind turbine bearings by sensor array, multi-dimensional operating parameter information is obtained; The multidimensional operating parameter information is preprocessed to obtain the preprocessed multidimensional operating parameter information. Based on the preprocessed multidimensional operating parameter information, time-domain features and frequency-domain features are extracted respectively to obtain a feature set; A feature selection algorithm is used to perform dimensionality reduction optimization on the feature set, eliminating redundant and invalid features to obtain the optimal feature subset. The optimal feature subset is input into the preset bearing condition recognition model, and the real-time condition information of the wind turbine bearing is output based on the bearing condition recognition model. Based on the dynamic early warning mechanism, the real-time status information of wind turbine bearings is used to perform status assessment and anomaly warning, and the status analysis results are obtained.
8. The wind turbine bearing condition analysis system according to claim 7, characterized in that, The sensor group includes at least a vibration sensor, a temperature sensor, and a speed sensor; The operating parameters collected by the vibration sensor include at least the vibration signal of the outer ring of the bearing, the vibration signal of the inner ring, and the bearing vibration acceleration. The bearing housing temperature signal is acquired based on a temperature sensor; Bearing speed signals are acquired using a speed sensor; During the data acquisition process, the data acquisition timestamp and the operating condition information of the wind turbine are recorded simultaneously. The operating condition information includes wind speed, unit output power and yaw angle, thus obtaining multi-dimensional operating parameter information.
9. A wind turbine bearing condition analysis system according to claim 8, characterized in that, The preprocessing operations include outlier removal, data smoothing and denoising, signal resampling, and data standardization. Outlier removal uses the Grubbs criterion to identify and remove outlier data that exceeds the range of 3σ, where σ represents the standard deviation. Data smoothing and denoising uses a wavelet threshold denoising algorithm to process vibration signals; Signal resampling unifies parameters at different sampling frequencies to a preset sampling frequency; Data standardization uses the Z-score standardization method to transform parameters to the same order of magnitude.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a wind turbine bearing condition analysis method program, which, when executed by a processor, implements the steps of a wind turbine bearing condition analysis method as described in any one of claims 1 to 6.