Wind power coupling slip diagnosis method and device, electronic equipment and storage medium

Abstract

The application discloses a wind power coupling slip diagnosis method and device, electronic equipment and storage medium, which are used to solve the technical problems of low detection efficiency and false detection in the existing wind turbine fault detection method. The method comprises the following steps: receiving a file path from a first user write message queue, responding to an extraction operation of a second user on the file path, parsing a log file corresponding to the file path into structured data, extracting first original features from the structured data through a slip prediction model, constructing corresponding derived features according to the first original features, and further obtaining to-be-tested features, diagnosing the coupling slip condition of a to-be-tested wind turbine based on the to-be-tested features, and determining a target wind turbine in which the coupling slip occurs in the to-be-tested wind turbine according to the diagnosis result, so that the wind turbine log file can be used to quickly determine the wind turbine in which the coupling slip occurs, and the detection efficiency and diagnosis accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of wind turbine abnormality diagnosis technology, and in particular to a method, device, electronic equipment and storage medium for diagnosing slippage of wind turbine couplings. Background Technology

[0002] A wind power generation system converts the kinetic energy of wind into electrical energy; it is also called a wind turbine generator set (hereinafter referred to as a wind turbine). A wind turbine generator set mainly consists of a wind turbine rotor and a generator. The wind turbine gearbox is one of the core components of the wind turbine generator, its function being to convert the low-speed rotation of the wind turbine rotor into high-speed rotation to drive the generator. The coupling is the only mechanical connection between the high-speed rotating gearbox and the generator in the wind turbine generator set. Its main functions include: transmitting torque and speed, compensating for misalignment between the gearbox output shaft and the generator input shaft, mitigating the impact of vibration on the gearbox and generator, providing leakage protection for the wind turbine generator set, and, under abnormal overload conditions, the torque limiter of the coupling can effectively protect the transmission chain of the wind turbine generator set.

[0003] In practice, due to the frequent changes in wind speed in the external wind field, the coupling is subjected to random loads transmitted by the rotor rotation, which can easily cause slippage in the coupling, resulting in significant impact and economic losses on the wind turbine. Therefore, to ensure the normal operation of the wind turbine, abnormal diagnosis is necessary. Currently, the methods for abnormal diagnosis of wind turbines involve manual troubleshooting and adjustment after an abnormality occurs, or monitoring and alarming the entire wind turbine's fault operation. This approach is not only inefficient but also prone to false positives and negatives. Summary of the Invention

[0004] This invention provides a method, device, electronic equipment, and storage medium for diagnosing slippage in wind turbine couplings, which solves or partially solves the technical problems of low detection efficiency and the tendency to make false detections in existing wind turbine fault detection methods.

[0005] This invention provides a method for diagnosing slippage in wind turbine couplings, applied to a wind turbine coupling slippage diagnosis system. The system includes a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. The method includes:

[0006] Receive the file path written to the message queue by the first user, where the file path corresponds to the log file of the wind turbine under test;

[0007] In response to the second user's extraction operation on the file path, the log file is parsed into structured data;

[0008] The slippage prediction model extracts a first original feature from the structured data, constructs a corresponding first derived feature based on the first original feature, and fuses the first original feature with the first derived feature to obtain the feature to be tested. Based on the feature to be tested, the slippage of the coupling of the wind turbine under test is diagnosed.

[0009] Based on the diagnostic results, the target wind turbine unit with coupling slippage was identified among the wind turbine units under test.

[0010] Optionally, the first original feature includes the rotor speed, generator speed, and power generation of the wind turbine under test. The step of extracting the first original feature from the structured data using the slippage prediction model and constructing corresponding first derived features based on the first original feature includes:

[0011] Step S1: Extract the rotor speed, generator speed, and power generation of the wind turbine under test from the structured data using the slippage prediction model;

[0012] Step S2: Calculate the original statistics corresponding to the rotor speed, the generator speed, and the power generation;

[0013] Step S3: First, normalize the original statistics corresponding to the rotor speed and the original statistics corresponding to the generator speed, and then perform the difference calculation to obtain the difference statistics.

[0014] Step S4: Extract the time length corresponding to the first original feature from the log file, divide the first original feature into 4 interval data segments according to the time length, and calculate the rotor speed statistics, generator speed statistics and power generation statistics corresponding to each interval data segment respectively.

[0015] Step S5: First, normalize the rotor speed statistics and generator speed statistics corresponding to the interval data segment, and then perform the difference calculation to obtain the interval difference statistics of the interval data segment.

[0016] Step S6: Repeat step S5 to obtain the interval difference statistics corresponding to each interval data segment;

[0017] Step S7: Combine the difference statistics, rotor speed statistics, generator speed statistics, power generation statistics, interval difference statistics corresponding to each interval data segment, and the original statistics corresponding to the rotor speed, generator speed, and power generation respectively as the first derived feature corresponding to the first original feature;

[0018] The data in the statistics include the mean, standard deviation, maximum value, and minimum value.

[0019] Optionally, the wind turbine coupling slippage diagnostic system also includes an online environment, and the method further includes:

[0020] An initial slippage prediction model based on the random forest algorithm is constructed. A sample set of wind turbines is obtained. The sample set is divided into a training set, a validation set, and a test set according to a 3:1:1 ratio and using stratified sampling. The initial slippage prediction model is then trained based on the training set, the validation set, and the test set.

[0021] Based on the training results, a slippage prediction model for coupling slippage diagnosis is determined, and the slippage prediction model is deployed to the online environment and integrated into a multi-threaded log file extraction module.

[0022] In response to the message queue creation operation, a message queue is set up in the multi-threaded log file extraction module.

[0023] Optionally, training the initial slippage prediction model based on the training set, the validation set, and the test set includes:

[0024] The optimal model parameters are determined by combining grid search, and the initial slippage prediction model is trained and validated using the training set, the validation set, and the test set through a 3-fold cross-validation method.

[0025] Based on the cross-validation results, three different model weights are generated.

[0026] Optionally, the step of determining the slippage prediction model for coupling slippage diagnosis based on the training results includes:

[0027] The target model weight with the highest score is selected from the three model weights, and a slippage prediction model for coupling slippage diagnosis is determined based on the target model weight.

[0028] Optionally, before obtaining the sample set of wind turbine units, the method further includes:

[0029] Acquire historical test data of the wind turbine and multiple historical log files, and perform slip labeling on each of the historical log files based on the historical test data to obtain the corresponding initial labeling samples;

[0030] Extract the second original features of each of the initial labeled samples, construct the corresponding second derived features based on each of the second original features, and fuse each of the second original features with their respective corresponding second derived features to obtain the sample features corresponding to each of the second original features.

[0031] The initial labeled samples corresponding to each of the sample features are all updated to target labeled samples, where the target labeled samples are either slipping samples or non-slipping samples;

[0032] If the target labeled sample is a slipping sample, then the slipping sample is augmented and then the data is smoothed using the sliding window method to obtain the corresponding augmented slipping sample.

[0033] Each of the extended slippage samples and each of the non-slippage samples are stored as a sample set for the wind turbine.

[0034] Optionally, the step of performing sample augmentation on the slippage sample and then smoothing the data using the sliding window method to obtain the corresponding augmented slippage sample includes:

[0035] From the sample set, a slip sample is drawn without replacement as the first target slip sample, and a target labeled sample is drawn with replacement as a random sample;

[0036] The random sample is used to replace the data within 20% of the interval before and after the first target slip sample to obtain the first replaced slip sample;

[0037] The first replacement slip sample is smoothed using a sliding window method to obtain the expanded slip sample corresponding to the first target slip sample.

[0038] Optionally, the step of performing sample augmentation on the slippage sample and then smoothing the data using the sliding window method to obtain the corresponding augmented slippage sample includes:

[0039] One slip sample is extracted from the sample set as the second target slip sample;

[0040] A preset amount of random white noise with a mean of 0 is added to the second target slip sample to form a second replacement slip sample;

[0041] The second replacement slip sample is smoothed using the sliding window method to obtain the expanded slip sample corresponding to the second target slip sample.

[0042] Optionally, the wind turbine coupling slippage diagnostic system further includes a database, and the method further includes:

[0043] The diagnostic results are written into the database;

[0044] After the slippage prediction model has been run, the diagnostic results are extracted from the database, and the slippage prediction model is reviewed and analyzed based on the diagnostic results to screen out the error detection samples.

[0045] The misdetected samples are labeled to obtain corresponding misdetected labeled samples, and the misdetected labeled samples are added to the sample set of the wind turbine to obtain an incremental sample set;

[0046] The slippage prediction model is incrementally trained using the incremental sample set, and then updated based on the incremental training results.

[0047] This invention also provides a wind turbine coupling slippage diagnostic device, applied to a wind turbine coupling slippage diagnostic system. The wind turbine coupling slippage diagnostic system includes a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. The device includes:

[0048] The file path receiving module is used to receive the file path written to the message queue by the first user, wherein the file path corresponds to the log file of the wind turbine under test.

[0049] The structured data parsing module is used to parse the log file into structured data in response to the second user's extraction operation on the file path;

[0050] The slippage diagnosis and processing module is used to extract a first original feature from the structured data through the slippage prediction model, construct a corresponding first derived feature based on the first original feature, fuse the first original feature and the first derived feature to obtain the feature to be tested, and diagnose the slippage of the coupling of the wind turbine under test based on the feature to be tested.

[0051] The target wind turbine identification module is used to identify the target wind turbines in the test wind turbines that have experienced coupling slippage based on the diagnostic results.

[0052] The present invention also provides an electronic device, the device comprising a processor and a memory:

[0053] The memory is used to store program code and transmit the program code to the processor;

[0054] The processor is used to execute the wind turbine coupling slippage diagnosis method as described above, according to the instructions in the program code.

[0055] The present invention also provides a computer-readable storage medium for storing program code for executing the wind turbine coupling slippage diagnosis method as described in any of the preceding claims.

[0056] As can be seen from the above technical solutions, the present invention has the following advantages: it provides a method for diagnosing slippage in wind turbine couplings, and a corresponding wind turbine coupling slippage diagnosis system. First, a sample set of wind turbine units is constructed, and the sample set is processed and divided into training sets. Then, a slippage prediction model based on a random forest algorithm is constructed, and the model is trained using the sample training set through a training module. The trained model is then deployed to an online environment. Simultaneously, a multi-threaded log file extraction module is integrated. First, a user can upload the file path corresponding to the log file of the wind turbine unit under test to the message queue of the multi-threaded log file extraction module. Second, a user can extract the file from the message queue. The system parses the log files corresponding to the specified file paths into structured data and imports them into the prediction module of the slippage prediction model for coupling slippage anomaly diagnosis. This enables the rapid identification of wind turbine units experiencing coupling slippage through their log files, improving detection efficiency and diagnostic accuracy. The diagnostic results can then be input into a database. The training module of the slippage prediction model can retrieve the diagnostic results from the database and perform incremental training to optimize the model based on these results. This forms a virtuous cycle of "model training → model use → model optimization training → model use," continuously improving the accuracy of coupling slippage anomaly diagnosis, reducing detection time, and increasing response speed. Attached Figure Description

[0057] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1 A flowchart illustrating the steps of a wind turbine coupling slippage diagnosis method provided in this embodiment of the invention;

[0059] Figure 2 This is a schematic diagram of a wind turbine coupling slippage diagnostic system provided in an embodiment of the present invention;

[0060] Figure 3 A flowchart illustrating the steps for constructing a derived feature, as provided in an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of feature engineering for a feature to be detected, provided in an embodiment of the present invention.

[0062] Figure 5 A flowchart illustrating the steps for constructing a wind turbine sample set according to an embodiment of the present invention;

[0063] Figure 6A flowchart illustrating the steps for constructing and training a slippage prediction model, as provided in an embodiment of the present invention;

[0064] Figure 7 This is a schematic diagram of the overall process of a wind turbine coupling slippage diagnosis method provided in an embodiment of the present invention;

[0065] Figure 8 This is a structural block diagram of a wind turbine coupling slippage diagnostic device provided in an embodiment of the present invention. Detailed Implementation

[0066] This invention provides a method, device, electronic equipment, and storage medium for diagnosing slippage in wind turbine couplings, which solves or partially solves the technical problems of low detection efficiency and the tendency to make false detections in existing wind turbine fault detection methods.

[0067] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0068] As an example, with the continuous development of various technologies and the continuous improvement of people's living standards, the demand for electricity is constantly increasing, thus placing higher and higher demands on the power system. Wind turbines are a very important form of power system, providing power guarantees for people's daily life and work needs. A wind turbine mainly consists of a wind rotor and a generator. The wind turbine gearbox is one of the core components of the wind turbine, its function being to convert the low-speed rotation generated by the wind rotor into high-speed rotation to drive the generator. The coupling is the only mechanical connection between the high-speed rotating gearbox and the generator of the wind turbine, playing a crucial role. In reality, due to the frequent changes in wind speed in the external wind field, the coupling is subjected to random loads transmitted by the rotor rotation, easily causing coupling slippage, which has a significant impact on the wind turbine and causes economic losses. Therefore, in order to ensure the normal operation of the wind turbine, it is necessary to perform abnormality diagnosis on the wind turbine. In related technologies, the current method for diagnosing anomalies in wind turbines is to manually troubleshoot and adjust the turbine after an anomaly occurs, or to detect and alarm on the fault operation of the entire wind turbine. This method is not only inefficient, but also prone to false positives and false negatives.

[0069] Therefore, one of the core inventive points of this invention is: providing a method for diagnosing slippage in wind turbine couplings, and a corresponding wind turbine coupling slippage diagnosis system. First, a sample set of wind turbine units is constructed, and the sample set is processed and divided into training sets. Then, a slippage prediction model based on a random forest algorithm is constructed and trained using the sample training set through a training module. The trained model is then deployed to an online environment. Simultaneously, a multi-threaded log file extraction module is integrated. A first user can upload the file path corresponding to the log file of the wind turbine unit under test to the message queue of the multi-threaded log file extraction module. A second user can extract the file path from the message queue. The system parses the log files corresponding to the file path into structured data and imports them into the prediction module of the slippage prediction model for coupling slippage anomaly diagnosis. This allows for the rapid identification of wind turbine units experiencing coupling slippage through their log files, improving detection efficiency and diagnostic accuracy. The diagnostic results can then be input into a database. The training module of the slippage prediction model can retrieve the diagnostic results from the database and perform incremental training to optimize the model based on these results. This forms a virtuous cycle of "model training → model use → model optimization training → model use," continuously improving the accuracy of coupling slippage anomaly diagnosis, reducing detection time, and increasing response speed.

[0070] Reference Figure 1 This diagram illustrates a flowchart of a wind turbine coupling slippage diagnosis method provided by an embodiment of the present invention. Applied to a wind turbine coupling slippage diagnosis system, the system may include a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. The method specifically includes the following steps:

[0071] Step 101: Receive the file path written to the message queue by the first user, where the file path corresponds to the log file of the wind turbine under test;

[0072] Couplings are a crucial mechanical connection component in wind turbine generators. During operation, coupling slippage can significantly impact the normal operation of the wind turbine, leading to unnecessary economic losses. Therefore, detecting coupling slippage can be a key diagnostic factor when assessing wind turbine anomalies. Users responsible for data collection can upload the wind turbine log files to the wind turbine coupling slippage diagnostic system. Users seeking the diagnostic results can then extract the files and use the slippage prediction model within the system to analyze the data and output the diagnostic results.

[0073] Specifically, the wind turbine coupling slippage diagnostic system can receive file paths written to the message queue by a first user, where the file path corresponds to the log file of the wind turbine under test, and the first user can be the user who uploaded the file path.

[0074] For better illustration, refer to, by way of example, Figure 2 The diagram shows a schematic of a wind turbine coupling slippage diagnostic system provided in an embodiment of the present invention.

[0075] As shown in the figure, the wind turbine coupling slippage diagnosis system 200 mainly includes a multi-threaded log file extraction module 201, a slippage prediction model based on the random forest algorithm 202, an online environment 203, and a database 204.

[0076] The multi-threaded log file extraction module 201 may include a producer module 2011 and a consumer module 2012. The producer module 2011 and the consumer module 2012 communicate with each other through a message queue. Therefore, the producer module can be understood as the receiving module corresponding to the user who receives the uploaded file path, and the consumer module can be understood as the extraction module corresponding to the user who extracts the file path. For ease of distinction and explanation, the user corresponding to the producer module is set as the first user, and the user corresponding to the consumer module is set as the second user.

[0077] The slippage prediction model 202 may include a training module 2021 for model training optimization and a prediction module 2022 for anomaly diagnosis.

[0078] The online environment 203 is the release environment, that is, the environment accessed by real users. The slip prediction model selected after training through the training module 2021 can be deployed to the online environment 203, and integrated into the multi-threaded log file extraction module 201 for user use.

[0079] Database 204 can be used to store the diagnostic results output after anomaly diagnosis, or it can be used to store other related data, such as the results of random forest model predictions and related output indicators.

[0080] In the wind turbine coupling slippage diagnosis system 200, a slippage prediction model based on the random forest algorithm can be constructed and trained through the training module 2021. The trained model is then deployed to the online environment 203. Simultaneously, a multi-threaded log file extraction module 201 is integrated. The first user can upload the file path corresponding to the log file of the wind turbine under test to the message queue through the producer module 2011. The second user can extract the file path from the message queue through the consumer module 2012 and parse the corresponding log file into structured data. This data is then imported into the prediction module 2022 of the slippage prediction model 202 for coupling slippage anomaly diagnosis. The diagnosis results are input into the database 204. The training module 2021 of the slippage prediction model 202 can obtain the diagnosis results from the database 204, optimize the model training based on the diagnosis results, and update the slippage prediction model. This forms a virtuous cycle of "model training → model use → model optimization training → model use," continuously improving the accuracy of coupling slippage anomaly diagnosis, reducing detection time, and increasing response speed.

[0081] Step 102: In response to the second user's extraction operation on the file path, the log file is parsed into structured data;

[0082] Then, in response to the second user's extraction operation on the file path, the log file corresponding to the file path can be parsed into structured data that can be used by the slippage prediction model.

[0083] Step 103: Extract the first original feature from the structured data through the slippage prediction model, construct the corresponding first derived feature based on the first original feature, and fuse the first original feature and the first derived feature to obtain the feature to be tested. Diagnose the slippage of the coupling of the wind turbine under test based on the feature to be tested.

[0084] When processing structured data, the data processing flow during model training can be referenced. Specifically, a slippage prediction model can be used to extract the first original feature from the structured data, then construct the corresponding first derived feature based on the first original feature, and fuse the first original feature and the first derived feature to obtain the feature to be tested. Then, the slippage of the coupling of the wind turbine under test can be diagnosed based on the feature to be tested.

[0085] As an optional embodiment, the first original features may include the rotor speed, generator speed, and power generation of the wind turbine under test. For better illustration, refer to... Figure 3The diagram illustrates a flowchart of the steps for constructing derived features according to an embodiment of the present invention. Specifically, the steps of extracting a first original feature from structured data using a slippage prediction model and constructing a corresponding first derived feature based on the first original feature may include:

[0086] Step S1: Extract the rotor speed, generator speed, and power generation of the wind turbine under test from the structured data using the slippage prediction model;

[0087] Step S2: Calculate the original statistics corresponding to the rotor speed, the generator speed, and the power generation;

[0088] Step S3: First, normalize the original statistics corresponding to the rotor speed and the original statistics corresponding to the generator speed, and then perform the difference calculation to obtain the difference statistics.

[0089] Step S4: Extract the time length corresponding to the first original feature from the log file, divide the first original feature into 4 interval data segments according to the time length, and calculate the rotor speed statistics, generator speed statistics and power generation statistics corresponding to each interval data segment respectively.

[0090] Specifically, the first original feature is divided into original feature segment 1, original feature segment 2, original feature segment 3 and original feature segment 4 according to the time length. The rotor speed statistics, generator speed statistics and power generation statistics corresponding to original feature segments 1 to 4 are calculated respectively. In total, the rotor speed statistics, generator speed statistics and power generation statistics corresponding to 4 different interval data segments can be obtained, for a total of 12 interval statistics.

[0091] Step S5: First, normalize the rotor speed statistics and generator speed statistics corresponding to the interval data segment, and then perform the difference calculation to obtain the interval difference statistics of the interval data segment.

[0092] Step S6: Repeat step S5 to obtain the interval difference statistics corresponding to each interval data segment;

[0093] Similarly, through this step, we can obtain the interval difference statistics corresponding to the four different interval data segments, for a total of four interval difference statistics.

[0094] Step S7: Combine the difference statistics, rotor speed statistics, generator speed statistics, power generation statistics, interval difference statistics corresponding to each interval data segment, and the original statistics corresponding to the rotor speed, generator speed, and power generation respectively as the first derived feature corresponding to the first original feature;

[0095] After the above steps, the first derived feature includes a total of 20 statistics, namely the 3 original statistics obtained in step S1, the 1 difference statistic obtained in step S2, the 12 interval statistics obtained in step S4, and the 4 interval difference statistics obtained through steps S5 to S6. The data in the statistics may include the mean, standard deviation, maximum value, and minimum value.

[0096] For example, refer to Figure 4 The diagram illustrates a feature engineering schematic of a feature to be detected according to an embodiment of the present invention.

[0097] As shown in the figure, the entire feature engineering process mainly includes two parts: the extraction of original features and the construction of derived features from the original features. The original features include data such as the rotor speed, generator speed, and power generation of the wind turbine under test. The derived features mainly include five parts: the original rotor speed statistics, generator speed statistics, power generation statistics, and the normalized difference statistics between generator speed and rotor speed obtained after processing the original features; the segmented features obtained after segmentation; and the segmented features, which can further include the rotor speed statistics, generator speed statistics, power generation statistics, and the normalized difference statistics between generator speed and rotor speed after data segmentation.

[0098] After obtaining the first derived feature, the first original feature and the first derived feature can be fused to obtain the feature to be tested. Then, based on the feature to be tested, the slippage of the coupling of the wind turbine unit under test can be diagnosed, and the corresponding diagnostic results can be generated.

[0099] As an example, suppose there are multiple wind turbines under test that need slippage diagnosis. Each wind turbine under test corresponds to a feature to be tested. The feature to be tested is input into each decision tree in the slippage prediction model to determine whether the feature belongs to the slippage or non-slippage category. Based on the judgment results of each decision tree, it is determined whether the slippage category or the non-slippage category is selected most often. Finally, it is determined which category the feature to be tested belongs to. If most decision trees determine that the feature to be tested belongs to the slippage category, then the category corresponding to the feature to be tested is determined to be slippage, the wind turbine under test corresponding to the feature to be tested has slipped, and the corresponding diagnosis result is generated.

[0100] Step 104: Based on the diagnostic results, identify the target wind turbine unit in the wind turbine unit under test that has experienced coupling slippage.

[0101] Then, based on the output diagnostic results, the target wind turbine unit with coupling slippage can be identified, enabling rapid diagnosis of coupling slippage, quickly identifying wind turbine units with coupling slippage, and timely maintenance or adjustment, ensuring the normal operation of the wind turbine unit and avoiding greater losses.

[0102] As can be seen from the foregoing embodiments, the wind power coupling slippage diagnosis system may also include a database for storing data. After the slippage prediction model outputs the diagnosis results, the diagnosis results can be written into the database. After each run of the model to perform slippage diagnosis, the diagnosis results are input into the training module of the slippage prediction model. When there are abnormal data with false positives or false negatives in the diagnosis results, these abnormal data are identified as false positive samples. Then, based on the false positive samples, a corresponding incremental sample set is established to optimize and train the slippage prediction model.

[0103] In the specific implementation, after the slippage prediction model has run, the diagnostic results can be extracted from the database, and the slippage prediction model can be reviewed and analyzed based on the diagnostic results. Error detection samples are then filtered out, and the error detection samples are labeled to obtain the corresponding error-labeled samples. The error-labeled samples are then added to the wind turbine's sample set to obtain an incremental sample set. The incremental sample set is then used to incrementally train the slippage prediction model, and the slippage prediction model is updated based on the incremental training results. In this way, the slippage prediction model is optimized and trained through the diagnostic results corresponding to the slippage diagnosis, which enhances the robustness of the wind turbine coupling slippage diagnosis system. This makes the optimized slippage prediction model more accurate and faster in the next slippage diagnosis.

[0104] As an optional embodiment, before obtaining the sample set of wind turbines, a sample set of wind turbines can be constructed based on the historical detection data of the wind turbines, and used to conduct preliminary training on the slippage prediction model based on the constructed random forest algorithm.

[0105] Reference Figure 5 The diagram illustrates a flowchart of the steps for constructing a wind turbine sample set according to an embodiment of the present invention, which may specifically include the following steps:

[0106] Step 501: Obtain historical test data of the wind turbine and multiple historical log files; perform slip labeling on each of the historical log files based on the historical test data to obtain their respective initial labeling samples.

[0107] Log files are records or collections of files used to record system operation events. They play an important role in processing historical data, diagnosing problems, and understanding system activities. Therefore, before data processing, historical detection data of wind turbines and multiple historical log files can be extracted. Based on the historical detection data, each historical log file can be labeled with slippage to obtain its corresponding initial label samples. For example, log files on the day when coupling slippage occurred can be labeled as 1, and those on the other hand can be labeled as 0. By labeling the initially collected log files with slippage, it is easier to distinguish whether the coupling has slipped, thereby improving the accuracy of slippage diagnosis during subsequent model training.

[0108] Step 502: Extract the second original features of each of the initial labeled samples, construct the corresponding second derived features based on each of the second original features, and fuse each of the second original features with their respective corresponding second derived features to obtain the sample features corresponding to each of the second original features.

[0109] Next, the second original features of each initially labeled sample can be extracted, the corresponding second derived features can be constructed based on each second original feature, and each second original feature can be fused with its corresponding second derived feature to obtain the sample features corresponding to each second original feature.

[0110] Similar to the first original feature, the second original feature may include the rotor speed, generator speed and power generation of the wind turbine in the historical detection data. The steps for constructing the second derived feature based on the second original feature are similar to the steps for constructing the first derived feature based on the first original feature in the foregoing embodiments, and will not be repeated here. Those skilled in the art can refer to the relevant steps in the foregoing embodiments.

[0111] Next, each second original feature can be fused with its corresponding second derived feature to obtain the sample features corresponding to each second original feature.

[0112] Step 503: Update the initial labeled samples corresponding to each of the sample features to target labeled samples, wherein the target labeled samples are slipping samples or non-slipping samples;

[0113] When the log label corresponding to the target labeled sample is 1, the target labeled sample is a slip sample, also known as a positive sample, indicating that the coupling has slipped. Conversely, when the log label corresponding to the target labeled sample is 0, the target labeled sample is a non-slip sample, also known as a negative sample, indicating that the coupling has not slipped.

[0114] Step 504: If the target labeled sample is a slipping sample, then the slipping sample is subjected to sample augmentation processing, and then the data is smoothed by the sliding window method to obtain the corresponding augmented slipping sample.

[0115] As can be seen from the actual situation, in order to ensure normal power generation and supply, wind turbines are in normal working condition most of the time. Compared with the normal operating conditions, the slippage of the coupling is relatively rare, resulting in far fewer slippage samples than non-slippage samples. That is, the number of positive and negative samples is unbalanced. At this time, it is necessary to perform sample expansion processing on the positive samples (i.e. slippage samples) and use the window sliding method to perform data smoothing processing to achieve further processing of the target labeled samples, so that the data used for training is more referential.

[0116] As an optional embodiment, the steps of expanding the slippage sample and then smoothing the data using the sliding window method to obtain the corresponding expanded slippage sample can be as follows: First, a slippage sample is drawn from the sample set without replacement as the first target slippage sample, and a target labeled sample is drawn with replacement as a random sample. Then, the random sample is used to replace the data in the 20% interval before and after the first target slippage sample to obtain the first replacement slippage sample. Then, the first replacement slippage sample is smoothed using the sliding window method to obtain the expanded slippage sample corresponding to the first target slippage sample.

[0117] As another optional embodiment, the steps of performing sample augmentation on the slipping sample and then performing data smoothing through the sliding window method to obtain the corresponding augmented slipping sample may include: firstly, extracting a slipping sample from the sample set as the second target slipping sample; then adding a preset amount of random white noise with a mean of 0 to the second target slipping sample to form a second replacement slipping sample; and then performing data smoothing on the second replacement slipping sample through the sliding window method to obtain the augmented slipping sample corresponding to the second target slipping sample.

[0118] By augmenting the slipping samples and smoothing the data using a window sliding method, the quantitative difference between slipping and non-slipping samples is balanced, further achieving data smoothing for the target labeled samples, making the training data more reliable.

[0119] Step 505: Store each of the extended slippage samples and each of the non-slippage samples as a sample set for the wind turbine.

[0120] Next, each extended slippage sample and each non-slippage sample can be stored as a sample set of the wind turbine for subsequent model training.

[0121] Reference Figure 6The diagram illustrates a flowchart of the steps for constructing and training a slippage prediction model according to an embodiment of the present invention, which may specifically include the following steps:

[0122] Step 601: Construct an initial slippage prediction model based on the random forest algorithm, obtain a sample set of wind turbines, divide the sample set into a training set, a validation set, and a test set according to a 3:1:1 ratio and using stratified sampling, and train the initial slippage prediction model based on the training set, the validation set, and the test set.

[0123] In layman's terms, a random forest is a classifier that uses multiple trees to train and predict samples. A random forest is an algorithmic forest built randomly, consisting of numerous decision trees. There is no correlation between the individual decision trees in a random forest. After the random forest is built, each decision tree in the random forest judges the sample individually to determine which class it belongs to. The class most frequently chosen is then predicted as the class of the sample. For example, if most decision trees determine that a given sample belongs to the slippage class, then the sample is classified as slippage.

[0124] Therefore, this embodiment of the invention constructs an initial slippage prediction model based on the random forest algorithm, and the model parameters of the initial slippage prediction model can be determined according to the actual situation, such as the number of decision trees n_estimators, the maximum depth of the decision trees max_depth, and the node splitting criterion.

[0125] Next, the sample set of the wind turbines constructed in the previous embodiment can be obtained, and the sample set can be divided into training set, validation set and test set according to the ratio of 3:1:1 and by using stratified sampling.

[0126] Then, the initial slip prediction model can be trained based on the training set, validation set, and test set. Specifically, the model parameters are determined by combining grid search, and the initial slip prediction model is trained and validated using the training set, validation set, and test set through a 3-fold cross-validation method. Then, based on the cross-validation results, such as the different model evaluation index results that can be output through cross-validation, three different model weights are generated accordingly.

[0127] The model evaluation index used in this embodiment of the invention is F1-score (a statistical index used to measure the accuracy of binary classification models), and the calculation formula is: F1 = 2TP / (2TP + FN + FP), where TP is the number of coupling slippage samples in the sample set that are diagnosed as slippage, FN is the number of samples in the sample set that are diagnosed as normal but actually slippage, and FP is the number of samples in the sample set that are diagnosed as coupling slippage but are actually normal.

[0128] Furthermore, by training and validating the initial slippage prediction model using a 3-fold cross-validation method, the following can be achieved:

[0129] The initial slip prediction model is trained using a training set and a validation set, and the trained initial slip prediction model is validated using a test set to generate the first model weights.

[0130] The initial slip prediction model is trained using a training set and a test set, and the trained initial slip prediction model is validated using a validation set to generate the weights of the second model.

[0131] The initial slip prediction model is trained using a validation set and a test set, and the trained initial slip prediction model is validated using a training set to generate the weights of the third model.

[0132] Step 602: Based on the training results, determine the slippage prediction model for coupling slippage diagnosis, deploy the slippage prediction model to the online environment, and integrate it into the multi-threaded log file extraction module.

[0133] In the specific implementation, the slippage prediction model for coupling slippage diagnosis is determined based on the training results. This can be done by selecting the target model weight with the highest score from the three model weights, and determining the slippage prediction model for coupling slippage diagnosis based on the target model weight. In this way, the model with the best slippage diagnosis effect can be selected through model training as the slippage prediction model for wind turbine coupling, thereby maximizing the accuracy of slippage diagnosis.

[0134] As mentioned above, the wind turbine coupling slippage diagnosis system also includes an online environment. Once the slippage prediction model for coupling slippage prediction is determined, it can be deployed to the online environment and integrated into a multi-threaded log file extraction module. This will facilitate subsequent user access to the slippage prediction model for wind turbine coupling slippage diagnosis via the multi-threaded log file extraction module.

[0135] Step 603: In response to the message queue creation operation, a message queue is set up in the multi-threaded log file extraction module.

[0136] Before the wind turbine coupling slippage diagnostic system is put into use, a message queue can be set up in the multi-threaded log file extraction module to cache the file path of the wind turbine log file, and it can also serve as a communication relay station between the first user (the user who uploads the file path) and the second user (the user who extracts the file path).

[0137] In this embodiment of the invention, a method for diagnosing slippage in wind turbine couplings and a corresponding wind turbine coupling slippage diagnosis system are provided. First, a sample set of wind turbine generators is constructed, and the sample set is processed and divided into training sets. Then, a slippage prediction model based on a random forest algorithm is constructed and trained using the sample training set through a training module. The trained model is then deployed to an online environment. Simultaneously, a multi-threaded log file extraction module is integrated. A first user can upload the file path corresponding to the log file of the wind turbine generator under test to the message queue of the multi-threaded log file extraction module. A second user can extract the file path from the message queue and... The log files corresponding to the slippage prediction model are parsed into structured data and imported into the prediction module to diagnose coupling slippage anomalies. This enables the rapid identification of wind turbine units experiencing coupling slippage through their log files, improving detection efficiency and diagnostic accuracy. The diagnostic results can then be input into a database. The training module of the slippage prediction model can retrieve the diagnostic results from the database and perform incremental training to optimize the model based on these results. This forms a virtuous cycle of "model training → model use → model optimization training → model use," continuously improving the accuracy of coupling slippage anomaly diagnosis, reducing detection time, and increasing response speed.

[0138] For ease of understanding, the embodiments of the present invention are described below through specific examples.

[0139] Reference Figure 7 This diagram illustrates the overall process of a wind turbine coupling slippage diagnosis method provided by an embodiment of the present invention. It should be noted that this wind turbine coupling slippage diagnosis method is applied to a wind turbine coupling slippage diagnosis system. This example only provides a simple explanation of the overall process from constructing the dataset to optimizing and training the model based on the diagnosis results. The details involved in each step can be referred to the relevant parts of the foregoing embodiments, and will not be repeated here. It is understood that the present invention does not impose any limitations on this.

[0140] 1. Obtain historical testing data of wind turbine units and construct a wind turbine unit sample set;

[0141] 2. Process and divide the wind turbine sample set to obtain a training set (training set / validation set / test set) for training the model;

[0142] 3. Construct a slip prediction model based on the random forest algorithm, and use the training set obtained in step 2 to train the slip prediction model using 3-fold cross-validation.

[0143] 4. Obtain the log file of the wind turbine to be tested and parse it into structured data that can be recognized and processed by the model;

[0144] 5. Use a slippage prediction model to perform online prediction on structured data, output diagnostic results, and store the diagnostic results in the database;

[0145] 6. If the diagnostic results indicate that there are misdetected samples in this online prediction, the misdetected samples are marked and added to the wind turbine sample set to obtain an incremental sample set. The slippage prediction model is incrementally trained based on the incremental sample set to enhance the robustness of the wind turbine coupling slippage diagnostic system.

[0146] Reference Figure 8 This diagram illustrates a structural block diagram of a wind turbine coupling slippage diagnostic device according to an embodiment of the present invention. The device is applied to a wind turbine coupling slippage diagnostic system, which includes a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. Specifically, the device may include:

[0147] The file path receiving module 801 is used to receive a file path written to the message queue by the first user, wherein the file path corresponds to the log file of the wind turbine under test.

[0148] The structured data parsing module 802 is used to parse the log file into structured data in response to the extraction operation of the second user on the file path;

[0149] The slippage diagnosis and processing module 803 is used to extract a first original feature from the structured data through the slippage prediction model, construct a corresponding first derived feature based on the first original feature, fuse the first original feature and the first derived feature to obtain the feature to be tested, and diagnose the slippage of the coupling of the wind turbine under test based on the feature to be tested.

[0150] The target wind turbine determination module 804 is used to determine the target wind turbine in the wind turbine under test that has coupling slippage based on the diagnostic results.

[0151] In one optional embodiment, the first original features include the rotor speed, generator speed, and power generation of the wind turbine under test, and the slippage diagnosis processing module 803 includes:

[0152] The wind turbine data extraction module is used to perform step S1: extract the rotor speed, generator speed and power generation of the wind turbine under test from the structured data through the slippage prediction model;

[0153] The raw statistics calculation module is used to execute step S2: calculate the raw statistics corresponding to the rotor speed, the generator speed, and the power generation, respectively.

[0154] The difference statistics calculation module is used to execute step S3: first normalize the original statistics corresponding to the rotor speed and the original statistics corresponding to the generator speed, and then perform the difference calculation to obtain the difference statistics.

[0155] The interval data segment calculation module is used to perform step S4: extract the time length corresponding to the first original feature from the log file, divide the first original feature into 4 interval data segments according to the time length, and calculate the rotor speed statistics, generator speed statistics and power generation statistics corresponding to each interval data segment respectively.

[0156] The interval difference statistics calculation module is used to execute step S5: first normalize the rotor speed statistics and generator speed statistics corresponding to the interval data segment, and then perform the difference calculation to obtain the interval difference statistics of the interval data segment.

[0157] The interval difference statistics repetition processing module is used to execute step S6: repeat step S5 to obtain the interval difference statistics corresponding to each interval data segment;

[0158] The first derived feature construction module is used to execute step S7: taking the difference statistics, the rotor speed statistics, the generator speed statistics, the power generation statistics, the interval difference statistics corresponding to each interval data segment, and the original statistics corresponding to the rotor speed, the generator speed, and the power generation respectively as the first derived feature corresponding to the first original feature;

[0159] The data in the statistics include the mean, standard deviation, maximum value, and minimum value.

[0160] In one alternative embodiment, the wind turbine coupling slippage diagnostic system further includes an online environment, and the device further includes:

[0161] The initial slippage prediction model training module is used to construct an initial slippage prediction model based on the random forest algorithm, obtain a sample set of wind turbines, divide the sample set into a training set, a validation set, and a test set according to a 3:1:1 ratio and using a stratified sampling method, and train the initial slippage prediction model based on the training set, the validation set, and the test set.

[0162] The slippage prediction model deployment module is used to determine the slippage prediction model for coupling slippage diagnosis based on the training results, deploy the slippage prediction model to the online environment, and integrate it with the multi-threaded log file extraction module.

[0163] The message queue setting module is used to set up the message queue in the multi-threaded log file extraction module in response to the message queue creation operation.

[0164] In one optional embodiment, the initial slippage prediction model training module includes:

[0165] The 3-fold cross-validation execution module is used to combine grid search to determine the optimized model parameters, and to train and validate the initial slippage prediction model using the training set, the validation set, and the test set through 3-fold cross-validation.

[0166] The model weight generation module is used to generate three different model weights based on the cross-validation results.

[0167] In one optional embodiment, the slippage prediction model deployment module package is specifically used for:

[0168] The target model weight with the highest score is selected from the three model weights, and a slippage prediction model for coupling slippage diagnosis is determined based on the target model weight.

[0169] In one alternative embodiment, the device further includes:

[0170] The initial annotation sample generation module is used to acquire historical detection data of wind turbines and multiple historical log files, and to perform slip annotation on each of the historical log files based on the historical detection data to obtain the corresponding initial annotation samples.

[0171] The sample feature generation module is used to extract the second original features of each of the initial labeled samples, construct the corresponding second derived features based on each of the second original features, and fuse each of the second original features with their respective corresponding second derived features to obtain the sample features corresponding to each of the second original features.

[0172] The target annotation sample update module is used to update the initial annotation samples corresponding to each of the sample features to target annotation samples, wherein the target annotation samples are slipping samples or non-slipping samples;

[0173] An expanded slip sample generation module is used to expand the slip sample if the target labeled sample is a slip sample, and then perform data smoothing processing through the sliding window method to obtain the corresponding expanded slip sample.

[0174] The sample set storage module is used to store each of the extended slippage samples and each of the non-slippage samples as the sample set of the wind turbine.

[0175] In one optional embodiment, the expanded slippage sample generation module includes:

[0176] The first target slippage sample extraction module is used to extract a slippage sample from the sample set without replacement as the first target slippage sample and to extract a target labeled sample with replacement as a random sample.

[0177] The first replacement slip sample generation module is used to replace the data in the 20% range before and after the first target slip sample with the random sample to obtain the first replacement slip sample;

[0178] The first data smoothing module is used to perform data smoothing on the first replacement slipping sample using the sliding window method to obtain the expanded slipping sample corresponding to the first target slipping sample.

[0179] In one optional embodiment, the expanded slippage sample generation module includes:

[0180] The second target slippage sample extraction module is used to extract a slippage sample from the sample set as the second target slippage sample;

[0181] The second replacement slip sample generation module is used to add a preset amount of random white noise with a mean of 0 to the second target slip sample to form a second replacement slip sample.

[0182] The second data smoothing module is used to perform data smoothing on the second replacement slipping sample using the sliding window method to obtain the expanded slipping sample corresponding to the second target slipping sample.

[0183] In one alternative embodiment, the wind turbine coupling slippage diagnostic system further includes a database, and the device further includes:

[0184] A database writing module is used to write the diagnostic results into the database;

[0185] The false detection sample screening module is used to extract the diagnostic results from the database after the slippage prediction model has finished running, and to perform a review analysis of the slippage prediction model based on the diagnostic results to screen out false detection samples.

[0186] The incremental sample set acquisition module is used to mark the misdetected samples as misdetected, obtain the corresponding misdetected marked samples, and add the misdetected marked samples to the sample set of the wind turbine to obtain the incremental sample set.

[0187] The incremental training module is used to incrementally train the slippage prediction model using the incremental sample set, and update the slippage prediction model based on the incremental training results.

[0188] As the device embodiment is basically similar to the method embodiment, it is described in a relatively simple way. For relevant details, please refer to the description of the method embodiment above.

[0189] This invention also provides an electronic device, which includes a processor and a memory:

[0190] The memory is used to store program code and transfer the program code to the processor;

[0191] The processor is used to execute the wind turbine coupling slippage diagnosis method of any embodiment of the present invention according to the instructions in the program code.

[0192] This invention also provides a computer-readable storage medium for storing program code for executing the wind turbine coupling slippage diagnosis method of any embodiment of this invention.

[0193] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0194] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0195] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0196] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0197] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer 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 steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0198] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for diagnosing slippage in wind turbine couplings, characterized in that, An application is made to a wind turbine coupling slippage diagnosis system. The system includes a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. The method includes: Receive the file path written to the message queue by the first user, where the file path corresponds to the log file of the wind turbine under test; In response to the second user's extraction operation on the file path, the log file is parsed into structured data; The slippage prediction model extracts a first original feature from the structured data, constructs a corresponding first derived feature based on the first original feature, and fuses the first original feature with the first derived feature to obtain the feature to be tested. Based on the feature to be tested, the slippage of the coupling of the wind turbine under test is diagnosed. Based on the diagnostic results, the target wind turbine unit in the wind turbine unit under test was identified as having coupling slippage. The first original feature includes the rotor speed, generator speed, and power generation of the wind turbine under test. The step of extracting the first original feature from the structured data using the slippage prediction model and constructing corresponding first derived features based on the first original feature includes: Step S1: Extract the rotor speed, generator speed, and power generation of the wind turbine under test from the structured data using the slippage prediction model; Step S2: Calculate the original statistics corresponding to the rotor speed, the generator speed, and the power generation; Step S3: First, normalize the original statistics corresponding to the rotor speed and the original statistics corresponding to the generator speed, and then perform the difference calculation to obtain the difference statistics. Step S4: Extract the time length corresponding to the first original feature from the log file, divide the first original feature into 4 interval data segments according to the time length, and calculate the rotor speed statistics, generator speed statistics and power generation statistics corresponding to each interval data segment respectively. Step S5: First, normalize the rotor speed statistics and generator speed statistics corresponding to the interval data segment, and then perform the difference calculation to obtain the interval difference statistics of the interval data segment. Step S6: Repeat step S5 to obtain the interval difference statistics corresponding to each interval data segment; Step S7: Combine the difference statistics, rotor speed statistics, generator speed statistics, power generation statistics, interval difference statistics corresponding to each interval data segment, and the original statistics corresponding to the rotor speed, generator speed, and power generation respectively as the first derived feature corresponding to the first original feature; The data in the statistics include the mean, standard deviation, maximum value, and minimum value.

2. The method for diagnosing slippage in wind turbine couplings according to claim 1, characterized in that, The wind turbine coupling slippage diagnostic system also includes an online environment, and the method further includes: An initial slippage prediction model based on the random forest algorithm is constructed. A sample set of wind turbines is obtained. The sample set is divided into a training set, a validation set, and a test set according to a 3:1:1 ratio and using stratified sampling. The initial slippage prediction model is then trained based on the training set, the validation set, and the test set. Based on the training results, a slippage prediction model for coupling slippage diagnosis is determined, and the slippage prediction model is deployed to the online environment and integrated into a multi-threaded log file extraction module. In response to the message queue creation operation, a message queue is set up in the multi-threaded log file extraction module.

3. The method for diagnosing slippage in wind turbine couplings according to claim 2, characterized in that, Training the initial slip prediction model based on the training set, the validation set, and the test set includes: The optimal model parameters are determined by combining grid search, and the initial slippage prediction model is trained and validated using the training set, the validation set, and the test set through a 3-fold cross-validation method. Based on the cross-validation results, three different model weights are generated.

4. The method for diagnosing slippage in wind turbine couplings according to claim 3, characterized in that, The slippage prediction model determined based on training results for coupling slippage diagnosis includes: The target model weight with the highest score is selected from the three model weights, and a slippage prediction model for coupling slippage diagnosis is determined based on the target model weight.

5. The method for diagnosing slippage in wind turbine couplings according to any one of claims 2 to 4, characterized in that, Before obtaining the sample set of wind turbine units, the method further includes: Acquire historical test data of the wind turbine and multiple historical log files, and perform slip labeling on each of the historical log files based on the historical test data to obtain the corresponding initial labeling samples; Extract the second original features of each of the initial labeled samples, construct the corresponding second derived features based on each of the second original features, and fuse each of the second original features with their respective corresponding second derived features to obtain the sample features corresponding to each of the second original features. The initial labeled samples corresponding to each of the sample features are all updated to target labeled samples, where the target labeled samples are either slipping samples or non-slipping samples; If the target labeled sample is a slipping sample, then the slipping sample is augmented and then the data is smoothed using the sliding window method to obtain the corresponding augmented slipping sample. Each of the extended slippage samples and each of the non-slippage samples are stored as a sample set for the wind turbine.

6. The method for diagnosing slippage in wind turbine couplings according to claim 5, characterized in that, The process of augmenting the slippage samples and then smoothing the data using a sliding window method to obtain the corresponding augmented slippage samples includes: From the sample set, a slip sample is drawn without replacement as the first target slip sample, and a target labeled sample is drawn with replacement as a random sample; The random sample is used to replace the data within 20% of the interval before and after the first target slip sample to obtain the first replaced slip sample; The first replacement slip sample is smoothed using a sliding window method to obtain the expanded slip sample corresponding to the first target slip sample.

7. The method for diagnosing slippage in wind turbine couplings according to claim 5, characterized in that, The process of augmenting the slippage samples and then smoothing the data using a sliding window method to obtain the corresponding augmented slippage samples includes: One slip sample is extracted from the sample set as the second target slip sample; A preset amount of random white noise with a mean of 0 is added to the second target slip sample to form a second replacement slip sample; The second replacement slip sample is smoothed using the sliding window method to obtain the expanded slip sample corresponding to the second target slip sample.

8. The method for diagnosing slippage in wind turbine couplings according to claim 1, characterized in that, The wind turbine coupling slippage diagnostic system also includes a database, and the method further includes: The diagnostic results are written into the database; After the slippage prediction model has been run, the diagnostic results are extracted from the database, and the slippage prediction model is reviewed and analyzed based on the diagnostic results to screen out the error detection samples. The misdetected samples are labeled to obtain corresponding misdetected labeled samples, and the misdetected labeled samples are added to the sample set of the wind turbine to obtain an incremental sample set; The slippage prediction model is incrementally trained using the incremental sample set, and then updated based on the incremental training results.

9. A wind turbine coupling slippage diagnostic device, characterized in that, An application is made to a wind turbine coupling slippage diagnosis system. The system includes a multi-threaded log file extraction module and a slippage prediction model based on a random forest algorithm. The multi-threaded log file extraction module is equipped with a message queue. The device includes: The file path receiving module is used to receive the file path written to the message queue by the first user, wherein the file path corresponds to the log file of the wind turbine under test. The structured data parsing module is used to parse the log file into structured data in response to the second user's extraction operation on the file path; The slippage diagnosis and processing module is used to extract a first original feature from the structured data through the slippage prediction model, construct a corresponding first derived feature based on the first original feature, fuse the first original feature and the first derived feature to obtain the feature to be tested, and diagnose the slippage of the coupling of the wind turbine under test based on the feature to be tested. The target wind turbine identification module is used to identify the target wind turbines in the wind turbines under test that have experienced coupling slippage based on the diagnostic results. The first original feature includes the rotor speed, generator speed, and power generation of the wind turbine under test, and the slippage diagnosis and processing module includes: The wind turbine data extraction module is used to perform step S1: extract the rotor speed, generator speed and power generation of the wind turbine under test from the structured data through the slippage prediction model; The raw statistics calculation module is used to perform step S2: calculate the raw statistics corresponding to the rotor speed, the generator speed and the power generation respectively; The difference statistics calculation module is used to execute step S3: first normalize the original statistics corresponding to the rotor speed and the original statistics corresponding to the generator speed, and then perform the difference calculation to obtain the difference statistics. The interval data segment calculation module is used to perform step S4: extract the time length corresponding to the first original feature from the log file, divide the first original feature into 4 interval data segments according to the time length, and calculate the rotor speed statistics, generator speed statistics and power generation statistics corresponding to each interval data segment respectively. The interval difference statistics calculation module is used to execute step S5: first normalize the rotor speed statistics and generator speed statistics corresponding to the interval data segment, and then perform the difference calculation to obtain the interval difference statistics of the interval data segment. The interval difference statistics repetition processing module is used to execute step S6: repeat step S5 to obtain the interval difference statistics corresponding to each interval data segment; The first derived feature construction module is used to execute step S7: taking the difference statistics, the rotor speed statistics, the generator speed statistics, the power generation statistics, the interval difference statistics corresponding to each interval data segment, and the original statistics corresponding to the rotor speed, the generator speed, and the power generation respectively as the first derived feature corresponding to the first original feature; The data in the statistics include the mean, standard deviation, maximum value, and minimum value.

10. An electronic device, characterized in that, The device includes a processor and a memory: The memory is used to store program code and transmit the program code to the processor; The processor is used to execute the wind turbine coupling slippage diagnosis method according to any one of claims 1-8 according to the instructions in the program code.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store program code for executing the wind turbine coupling slippage diagnosis method according to any one of claims 1-8.

Images