Wind turbine generator high-speed shaft channeling prediction method, device and equipment and medium
Through the combination of data modeling and neural network model, the real-time operating parameters of wind turbines are preprocessed and featured to achieve accurate prediction of high-speed shaft stroke faults of wind turbines, and solve the low accuracy and high cost in the existing technology Problem, reducing failure loss.
Patent Information
- Application Number
- CN202411492742.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2025-02-07
AI Technical Summary
The prior art has low accuracy or high cost when preventing high-speed shaft stroke failure of wind turbine units, and lacks a method to accurately and efficiently prevent failure without adding sensors.
Through data modeling, the real-time operating parameters of the wind turbine are preprocessed and featured, and the high-speed axis fault warning model is input to predict the problem. The neural network model is used to combine historical feature variables and fault status data for training to achieve the prediction of fault status. .
On the basis of not adding sensors, it improves the accuracy of high-speed shaft-stroke warning and reduces fault losses, which has the advantages of low cost, high efficiency and strong interpretation.
Smart Images
Figure CN119397442A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind power generation, and in particular to a method, device, equipment and medium for predicting high-speed shaft misalignment of a wind turbine set. Background Art
[0002] The gearbox is an important component of the transmission chain system of a doubly-fed wind turbine, responsible for converting the low speed of the impeller into the high speed required by the generator. Gearboxes mostly adopt the helical gear design concept, which can effectively reduce impact, vibration and noise, but the disadvantage is that axial force will be generated during operation, so the high-speed shaft generally uses tapered roller bearings to balance the axial force. However, in actual application, the high-speed shaft will have a shaft fault, resulting in long-term unplanned shutdown of the unit, causing huge economic losses.
[0003] There are currently three methods for preventing high-speed shaft misalignment in wind turbines: one is to use gearbox temperature parameters combined with machine learning algorithms for fault warning, and the other is to use vibration acceleration combined with machine learning algorithms for fault warning. In actual application, the above two methods have the problem of low accuracy. Another method is to install a displacement sensor on the brake disc of the high-speed shaft of the gearbox. Although the accuracy is high, the cost is high because of the need to install a sensor. In response to this, the present application needs to propose a method for accurately and efficiently preventing high-speed shaft misalignment without increasing costs. Summary of the invention
[0004] The present application provides a method, device, equipment and medium for predicting high-speed shaft misalignment of a wind turbine set, which uses a data modeling method to perform early warning prediction of high-speed shaft misalignment, thereby improving the accuracy of high-speed shaft misalignment warning without adding sensors and reducing failure losses.
[0005] In a first aspect, the present application provides a method for predicting shaft misalignment of a high-speed shaft of a wind turbine, the method comprising:
[0006] Obtain real-time operating parameters of wind turbines;
[0007] Preprocess the real-time operation parameters to obtain the target data set;
[0008] Construct features for the target data set to obtain target feature variables;
[0009] The target characteristic variables are input into the high-speed shaft misalignment fault warning model for prediction to obtain predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate whether the wind turbine set is in a normal state or a fault state. The high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
[0010] Optionally, preprocessing the real-time operating parameters to obtain a target data set includes:
[0011] Eliminate data in the real-time operating parameters that are in a normal power generation state and whose active power is less than a threshold value to obtain first processed data;
[0012] According to the density-based noise, a spatial clustering DBSCAN algorithm is applied to remove the data with abnormal high-speed shaft speed in the first processed data to obtain the second processed data;
[0013] The second processed data is divided into a plurality of target data groups according to time periods, and target data with a data volume less than a threshold value is eliminated from the plurality of target data groups to obtain a target data set.
[0014] Optionally, construct features for the target data set to obtain target feature variables, including:
[0015] Divide each group of target data in the target data set according to the power interval;
[0016] Solving for extreme values of the high-speed shaft speed and the generator speed in the target data within different power intervals, wherein the extreme values include a maximum value and a minimum value;
[0017] The extreme values of the high-speed shaft speed and the generator speed are smoothed to obtain target characteristic variables, wherein the target characteristic variables include extreme values of different power intervals, extreme values of the high-speed shaft speed within different power intervals, and extreme values of the generator speed.
[0018] Optionally, the training process of the high-speed shaft misalignment fault warning model includes:
[0019] Obtain historical operating parameters of the wind turbine and historical high-speed shaft fault information, and pre-process the historical operating parameters to obtain a historical data set;
[0020] Construct features for historical data sets to obtain historical feature variables;
[0021] The historical data set is annotated according to the historical high-speed shaft misalignment fault information to obtain the historical fault status data;
[0022] The historical data set is divided into a training set and a validation set after being disrupted in time sequence according to a specific ratio;
[0023] The historical feature variables and historical fault label data in the training set are input into the initial model for training to obtain a high-speed shaft runaway fault warning model.
[0024] Optionally, the historical data set is annotated according to the historical high-speed shaft crossover fault information to obtain historical fault status data, including:
[0025] According to the high-speed shaft misalignment fault information, the historical data set for the time period from the day the fault was discovered to the day before the bearing was replaced was marked as a fault state, and the historical data set for the time period one month before the fault was discovered and the historical data set for the time period one day after the bearing was replaced was marked as a normal state, so as to obtain historical fault state data.
[0026] Optionally, the historical feature variables and historical fault label data in the training set are input into the initial model for training to obtain a high-speed shaft misalignment fault warning model, including:
[0027] The historical feature variables and historical fault label data in the training set are imported into the gradient boosting decision tree algorithm in the initial model, and the parameters in the initial model are trained using the 10-fold cross-validation method; when the accuracy index of the initial model is greater than or equal to the preset value, a high-speed shaft runaway fault warning model is obtained.
[0028] Optionally, after obtaining the high-speed shaft runaway fault warning model, the method further includes:
[0029] The historical feature variables in the verification set are input into the high-speed shaft misalignment fault warning model for verification, and the evaluation indicators of the high-speed shaft misalignment fault warning model are obtained.
[0030] In the second aspect, a device for predicting shaft misalignment of a high-speed shaft of a wind turbine generator system comprises:
[0031] An acquisition unit, used for acquiring real-time operating parameters of the wind turbine generator set;
[0032] A processing unit, used for preprocessing the real-time operation parameters to obtain a target data set;
[0033] A feature construction unit 403 is used to construct features for a target data set to obtain a target feature variable;
[0034] A prediction unit is used to input the target characteristic variable into the high-speed shaft misalignment fault warning model for prediction, and obtain the predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate that the wind turbine set is in a normal state or a fault state. The high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
[0035] Optionally, the processing unit is specifically configured to:
[0036] Eliminate data in the real-time operating parameters that are in a normal power generation state and whose active power is less than a threshold value to obtain first processed data;
[0037] According to the density-based noise, a spatial clustering DBSCAN algorithm is applied to remove the data with abnormal high-speed shaft speed in the first processed data to obtain the second processed data;
[0038] The second processed data is divided into a plurality of target data groups according to time periods, and target data with a data volume less than a threshold value is eliminated from the plurality of target data groups to obtain a target data set.
[0039] Optionally, the feature construction unit is specifically used for:
[0040] Divide each group of target data in the target data set according to the power interval;
[0041] Solving for extreme values of the high-speed shaft speed and the generator speed in the target data within different power intervals, wherein the extreme values include a maximum value and a minimum value;
[0042] The extreme values of the high-speed shaft speed and the generator speed are smoothed to obtain target characteristic variables, wherein the target characteristic variables include extreme values of different power intervals, extreme values of the high-speed shaft speed within different power intervals, and extreme values of the generator speed.
[0043] Optionally, the device further comprises a training unit, the training unit comprising:
[0044] An acquisition subunit is used to acquire historical operating parameters of the wind turbine and historical high-speed shaft fault information, and pre-process the historical operating parameters to obtain a historical data set;
[0045] Obtaining subunits for constructing features for historical data sets and obtaining historical feature variables;
[0046] A labeling subunit is used to label the historical data set according to the historical high-speed shaft crossover fault information to obtain the historical fault status data;
[0047] A division sub-unit is used to divide the historical data set into a training set and a validation set after disrupting the time sequence according to a specific ratio;
[0048] The training subunit is used to input the historical feature variables and historical fault label data in the training set into the initial model for training to obtain a high-speed shaft runaway fault warning model.
[0049] Optionally, the obtaining subunit is specifically used for:
[0050] According to the high-speed shaft misalignment fault information, the historical data set for the time period from the day the fault was discovered to the day before the bearing was replaced was marked as a fault state, and the historical data set for the time period one month before the fault was discovered and the historical data set for the time period one day after the bearing was replaced was marked as a normal state, so as to obtain historical fault state data.
[0051] Optionally, the training subunit is specifically used to include:
[0052] The historical feature variables and historical fault label data in the training set are imported into the gradient boosting decision tree algorithm in the initial model, and the parameters in the initial model are trained using the 10-fold cross-validation method; when the accuracy index of the initial model is greater than or equal to the preset value, a high-speed shaft runaway fault warning model is obtained.
[0053] Optionally, the device further comprises:
[0054] The verification unit is used to input the historical characteristic variables in the verification set into the high-speed shaft misalignment fault warning model for verification, so as to obtain the evaluation index of the high-speed shaft misalignment fault warning model.
[0055] In a third aspect, the present application provides an electronic device, the electronic device comprising a memory and a processor:
[0056] The memory is used to store computer programs;
[0057] The processor is used to execute the method provided by the first aspect according to the computer program.
[0058] In a fourth aspect, the present application further provides a computer-readable storage medium, which is used to store a computer program, and the computer program is used to execute the method provided in the first aspect above.
[0059] It can be seen that this application has the following beneficial effects:
[0060] The present application provides a method for predicting high-speed shaft misalignment of a wind turbine. First, after obtaining the real-time operating parameters of the wind turbine, the real-time operating parameters are preprocessed to obtain a target data set, and then the target data set is feature constructed to obtain a target feature variable. Finally, the target feature variable is input into a high-speed shaft misalignment fault warning model for prediction to obtain predicted fault state data of the wind turbine. The predicted fault state data is used to indicate that the wind turbine is in a normal state or a fault state, wherein the high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are historical feature variables of the wind turbine and corresponding historical fault state data. In this process, the feature variables and fault state data in the real-time operating parameters are learned and trained using a data modeling method to establish a corresponding high-speed shaft misalignment fault warning model, which has the advantages of low cost, high efficiency, and strong interpretability, that is, the accuracy of high-speed shaft misalignment warning can be improved without adding sensors, and the failure loss can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.
[0062] Figure 1 This is a flow chart of an embodiment of a method for predicting shaft misalignment of a high-speed shaft of a wind turbine in an embodiment of the present application;
[0063] Figure 2 It is a flow chart of another embodiment of a method for predicting shaft misalignment of a high-speed shaft of a wind turbine in the embodiment of the present application;
[0064] Figure 3 This is a high-speed shaft speed abnormal value marking result in an embodiment of the present application;
[0065] Figure 4 This is a schematic diagram of the structure of a device for predicting shaft misalignment of a high-speed shaft of a wind turbine in an embodiment of the present application;
[0066] Figure 5 This is a schematic diagram of the structure of an electronic device in an embodiment of the present application;
[0067] Figure 6 This is a schematic diagram of the specific structure of an electronic device in an embodiment of the present application. DETAILED DESCRIPTION
[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0069] The "first" in the name of the "first processed data" mentioned in the embodiment of the present application is only used as a name identifier and does not represent the first in order. The same rule applies to "second", "third", etc.
[0070] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0071] At present, the methods for preventing high-speed shaft shifting of wind turbines in the prior art are divided into three categories: one is to perform fault warning based on gearbox temperature parameters combined with machine learning algorithms, but since the axial movement of the high-speed shaft basically does not cause an increase in friction torque and thus increase the bearing temperature, the accuracy is relatively low; one is to perform fault warning based on vibration acceleration combined with machine learning algorithms, but since most of the current vibration sensors are acceleration sensors that are not sensitive to axial displacement, the accuracy is also relatively low; one is to install a displacement sensor on the brake disc of the high-speed shaft of the gearbox to measure the axial displacement of the high-speed shaft in real time. This method has a high accuracy rate, but the disadvantage is that it is relatively expensive.
[0072] In the embodiment of the present application, a data modeling method is used to perform early warning prediction for high-speed shaft misalignment, which can improve the accuracy of high-speed shaft misalignment warning and reduce failure losses without adding sensors.
[0073] In specific implementation, the method may include, for example: after obtaining the real-time operating parameters of the wind turbine, preprocessing the real-time operating parameters to obtain a target data set, and then constructing features for the target data set to obtain target feature variables; finally, inputting the target feature variables into a high-speed shaft misalignment fault warning model for prediction to obtain predicted fault status data of the wind turbine, wherein the predicted fault status data is used to indicate that the wind turbine is in a normal state or a fault state, wherein the high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are historical feature variables of the wind turbine and corresponding historical fault status data.
[0074] It can be seen that the method provided by the implementation of this application utilizes the data modeling method to learn and train the characteristic variables and fault status data in the real-time operating parameters, and establishes a corresponding high-speed shaft misalignment fault warning model, which has the advantages of low cost, high efficiency, and strong interpretability. That is, it can improve the accuracy of high-speed shaft misalignment warning and reduce fault losses without adding sensors.
[0075] To facilitate understanding of the specific implementation of the method for predicting shaft misalignment of a high-speed shaft of a wind turbine set provided in the embodiment of the present application, it will be described below with reference to the accompanying drawings.
[0076] It should be noted that the subject implementing the wind turbine high-speed shaft misalignment prediction method may be a wind turbine high-speed shaft misalignment prediction device provided in the embodiment of the present application, and the wind turbine high-speed shaft misalignment prediction device may be carried in an electronic device or a functional module of an electronic device. The electronic device in the embodiment of the present application may be any device capable of implementing the wind turbine high-speed shaft misalignment prediction method in the embodiment of the present application, for example, an Internet of Things (IoT) device.
[0077] Figure 1A flow chart of a method for predicting high-speed shaft misalignment of a wind turbine set provided in an embodiment of the present application. The method can be applied to a device for predicting high-speed shaft misalignment of a wind turbine set, which device can be, for example, Figure 4 The wind turbine high-speed shaft misalignment prediction device 400 is shown, or the wind turbine high-speed shaft misalignment prediction device can also be integrated in Figure 5 Functional modules in the electronic device 500 are shown.
[0078] like Figure 1 As shown, the method includes the following S101 to S104:
[0079] S101: Acquire real-time operating parameters of the wind turbine generator system.
[0080] In order to predict the fault state of the wind turbine, it is first necessary to obtain the real-time operating parameters of the wind turbine, pre-process the real-time operating parameters, obtain the target data set, then construct the features of the target data set to obtain the target feature variables, and finally input the target feature variables into the high-speed shaft fault warning model for prediction to obtain the predicted fault state data of the wind turbine. Therefore, the embodiment of the present application obtains the real-time operating parameters of the wind turbine through S101, making preliminary preparations for obtaining the target data set.
[0081] As an example, S101 may include: obtaining the real-time operating parameters of the wind turbine including the unit number, the unit state machine, the unit active power, the high-speed shaft speed, and the generator speed. That is, in order to predict the fault state of the wind turbine, it is necessary to combine the automatic operating parameters of the wind turbine to make a judgment and prediction, so it is necessary to first obtain the real-time operating parameters of the wind turbine.
[0082] S102: Preprocess the real-time operation parameters to obtain a target data set.
[0083] As an example, S102 may include: S1021, eliminating data in real-time operating parameters that are in normal power generation status and whose active power is less than a threshold value, to obtain first processed data; S1022, eliminating data with abnormal high-speed shaft speed in the first processed data according to the density-based noise spatial clustering DBSCAN algorithm, to obtain second processed data; S1023, dividing the second processed data into multiple groups of target data according to time periods, and eliminating target data in the multiple groups of target data whose data volume is less than a threshold value, to obtain a target data set.
[0084] In the above S1021, since the data of the wind turbine generator set fluctuates greatly when it is just started (active power is less than the threshold), it will affect the accuracy of the subsequent model. Therefore, during preprocessing, it is necessary to eliminate the data in the real-time operating parameters that are in normal power generation state and whose active power is less than the threshold.
[0085] In the above S1022, since the high-speed shaft speed uses a proximity switch to count the speed, high speeds are prone to occur. These measurement errors will affect the accuracy of the algorithm. Therefore, it is necessary to eliminate data with abnormal high-speed shaft speed.
[0086] In order to obtain the target data set in the above S1023, the second processed data needs to be divided into multiple groups of target data according to time periods, that is, it can be divided according to days, for example, the second processed data on the 12th is divided into one group of target data. And in order to improve the accuracy of the model, it is necessary to eliminate the target data with a data volume less than a threshold in the multiple groups of target data, that is, if the data volume of a group of target data is 1, the data volume is too small, which will affect the accuracy of the model, so this group of target data needs to be eliminated.
[0087] In this process, in order to improve the prediction accuracy of the model, the real-time operating parameters need to be preprocessed to obtain the target data set.
[0088] S103: Construct features for the target data set to obtain target feature variables.
[0089] The high-speed shaft fault warning model in the embodiment of the present application is to learn the relationship between the training feature variables and the fault status data. Therefore, in order to obtain the fault status data, it is necessary to first obtain the feature variables corresponding to the target data set.
[0090] As an example, S103 may include: S1031, dividing each group of target data in the target data set according to power intervals; S1032, solving the extreme values of the high-speed shaft speed and the generator speed in the target data within different power intervals, and the extreme values include the maximum value and the minimum value; S1033, smoothing the extreme values of the high-speed shaft speed and the generator speed within different power intervals to obtain target characteristic variables, and the target characteristic variables include the extreme values of different power intervals and the extreme values of the high-speed shaft speed and the extreme values of the generator speed within different power intervals.
[0091] Since the extreme values of high-speed shaft speed and generator speed in different power intervals have certain differences, it is necessary to first divide the target data in different power intervals, that is, the target characteristic variables obtained later need to be combined with the active power of the unit and the corresponding extreme values of high-speed shaft speed and generator speed. It should be noted that since the extreme values of high-speed shaft speed and generator speed may have sudden changes, in order to avoid affecting the accuracy of the model, the extreme values of high-speed shaft speed and generator speed need to be smoothed, and finally the extreme values of different power intervals and the extreme values of high-speed shaft speed and generator speed in different power intervals are obtained as the target characteristic variables.
[0092] S104: Input the target characteristic variable into the high-speed shaft misalignment fault warning model for prediction, and obtain the predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate that the wind turbine set is in a normal state or a fault state. The high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
[0093] As an example, S104 may include: inputting the target characteristic variables obtained in the above process into the high-speed shaft runaway fault warning model, and obtaining the predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate that the wind turbine set is in a normal state or a fault state, that is, when the predicted fault status data is in a normal state, it indicates that the predicted state of the wind turbine set is a normal state; similarly, when the predicted fault status data is in a fault state, it indicates that the predicted state of the wind turbine set is a fault state.
[0094] The embodiment of the present application also includes a process for training a high-speed shaft misalignment fault warning model, and the specific process is as follows: (1) obtaining historical operating parameters and historical high-speed shaft misalignment fault information of the wind turbine, and preprocessing the historical operating parameters to obtain a historical data set; (2) constructing features for the historical data set to obtain historical feature variables; (3) labeling the historical data set according to the high-speed shaft misalignment fault information of the wind turbine to obtain historical fault status data; (4) disrupting the time sequence of the historical data set according to a specific ratio and dividing it into a training set and a validation set; (5) inputting the historical feature variables and historical fault label data in the training set into the initial model for training to obtain a high-speed shaft misalignment fault warning model.
[0095] As an example, in the above training process (3), the process of labeling the historical data set according to the high-speed shaft misalignment fault information of the wind turbine to obtain the historical fault status data may include: according to the high-speed shaft misalignment fault information, labeling the historical data set of the time period from the day when the fault is discovered to the day before the bearing replacement is completed as a fault state, and labeling the historical data set of the time period one month before the fault is discovered and the historical data set of the time period one day after the bearing replacement is completed as a normal state, to obtain the historical fault label data. For example, if the fault time of the wind turbine is 2024-05-01 and the replacement completion time is 2024-05-16, then the historical data set of the time period from 2024-05-01 to 2024-05-15 is labeled as a fault state, and the historical data set of the time period from 2024-05-17 to the current time period is labeled as a normal state, and the historical data set of the time period from before to 2024-04-01 is labeled as a fault state. It should be noted that the embodiment of the present application records the normal state as label 0 and the fault state as label 1.
[0096] As an example, in the above-mentioned training process (5), the historical feature variables and historical fault label data in the training set are input into the initial model for training to obtain a high-speed shaft misalignment fault warning model, which can include: importing the historical feature variables and historical fault label data in the training set into the gradient boosting decision tree algorithm in the initial model, and using a 10-fold cross-validation method to train the parameters in the initial model; when the accuracy index of the initial model is greater than or equal to a preset value, a high-speed shaft misalignment fault warning model is obtained.
[0097] It should be noted that in the above training process, the accuracy index F1 score of the initial model is set equal to the preset value as the target for training, where F1 is an indicator used in statistics to measure the accuracy of a binary classification model. The F1 score can be regarded as a weighted average of the model's accuracy and recall rate, with a maximum value of 1 and a minimum value of 0.
[0098] In addition, after obtaining the high-speed shaft misalignment fault warning model, the embodiment of the present application can also use the verification set obtained above to verify the high-speed shaft misalignment fault warning model. The specific process may include, for example: inputting the historical feature variables in the verification set into the high-speed shaft misalignment fault warning model for verification, and obtaining the evaluation index of the high-speed shaft misalignment fault warning model.
[0099] Among them, the evaluation indicators may include machine learning performance indicators such as calculation accuracy, precision, and recall. The accuracy rate indicates the ability of the model to predict correctly, the precision rate indicates the ability of the model to correctly predict faults, and the recall rate indicates the ability of the model to correctly predict faults.
[0100] It can be seen that the embodiment of the present application uses a data modeling method to learn and train the characteristic variables and fault status data in the real-time operating parameters, and establishes a corresponding high-speed shaft misalignment fault warning model, so as to improve the accuracy of the high-speed shaft misalignment warning and reduce fault losses without adding sensors.
[0101] In order to make the method provided in the embodiment of the present application clearer and easier to understand, Figure 2 A specific example of training the model in this method is explained.
[0102] like Figure 2 As shown, the implementation may include:
[0103] S201: Acquire historical operating parameters of the wind turbine and historical high-speed shaft misalignment fault information.
[0104] As an example, S201 may include: obtaining historical operating parameters of the wind turbine including: unit number, unit state machine, unit active power, high-speed shaft speed, generator speed; obtaining historical high-speed shaft misalignment fault information of the wind turbine including: unit operation file and fault troubleshooting table.
[0105] S202: Preprocess historical operating parameters to obtain a historical data set.
[0106] like Figure 3 As shown, since the data with abnormal high-speed shaft speed is eliminated, it is necessary to eliminate the data with abnormal high-speed shaft speed. As an example, S202 may include: first, eliminate the data with the unit power generation state of 34 and the active power less than 1000KW; then eliminate the data with abnormal high-speed shaft speed according to the DBSCAN algorithm, and determine (0.7, 100) as the neighborhood parameter of the DBSCAN algorithm through multiple experiments; finally, divide the data into multiple groups of target data according to the time period, and in order to improve the accuracy of the model, it is necessary to eliminate the target data with less than 69120 data in the multiple groups of target data.
[0107] S203: Construct features for the historical data set to obtain historical feature variables.
[0108] As an example, S203 may include: S2031, dividing each group of target data in the target data set according to the 100KW power interval; S1032, solving the high-speed shaft speed_maximum value, high-speed shaft speed_minimum value, generator speed_maximum value and generator speed_minimum value in the target data within different power intervals; S1033, smoothing the high-speed shaft speed_maximum value, high-speed shaft speed_minimum value, generator speed_maximum value and generator speed_minimum value within different power intervals to obtain target characteristic variables, wherein the target characteristic variables include extreme values of different power intervals and extreme values of the high-speed shaft speed and the generator speed within different power intervals.
[0109] S204: Label the historical data set according to the high-speed shaft misalignment fault information of the wind turbine generator set to obtain historical fault label data.
[0110] As an example, S204 may include: based on the high-speed shaft misalignment fault information, labeling the historical data set for the time period from the day the fault was discovered to the day before the bearing was replaced as a fault state, and labeling the historical data set for the time period one month before the fault was discovered and the historical data set for the time period one day after the bearing was replaced as a normal state, to obtain historical fault label data. In the embodiment of the present application, the label for the normal state is recorded as label 0, and the label for the fault state is recorded as label 1.
[0111] S205: The historical data set is disrupted in time sequence according to a specific ratio and then divided into a training set and a validation set.
[0112] As an example, S205 may include: disrupting the time sequence of the historical data set and dividing it into a training set and a validation set in a ratio of 7:3.
[0113] S206: Inputting historical feature variables and historical fault label data in the training set into the initial model for training to obtain a high-speed shaft runaway fault warning model.
[0114] As an example, S206 may include: importing the historical feature variables and historical fault label data in the training set into the gradient boosting decision tree algorithm in the initial model, and using the 10-fold cross validation method to train the parameters in the initial model; taking the F1 of the training model greater than or equal to 0.85 as the target value, training and learning the relationship between the feature variables and the fault labels, and obtaining the high-speed shaft crossover fault warning model when the accuracy index of the initial model is greater than or equal to 0.85. In order to reduce the overfitting of the model, the 10-fold cross validation method is used for parameter training when constructing the high-speed shaft crossover fault warning model.
[0115] S207: Inputting the historical characteristic variables in the verification set into the high-speed shaft misalignment fault warning model for verification, and obtaining evaluation indicators of the high-speed shaft misalignment fault warning model.
[0116] As an example, S207 may include: importing the test set obtained above into the high-speed shaft misalignment fault warning model to obtain evaluation indicators of the high-speed shaft misalignment fault warning model, which may be specifically 88.7% accuracy, 86.7% precision, 89.6% recall, and 60% of the cases can be warned 7 to 16 days in advance, 24% of the cases can be warned 1 to 3 days in advance, and 16% of the cases are triggered simultaneously, where the time of the fault triggered simultaneously as the model warning is the same day as the time when the actual fault occurs.
[0117] This embodiment provides a method for predicting high-speed shaft misalignment in a wind turbine set, which uses a data modeling method to learn and train characteristic variables and fault status data in real-time operating parameters, and establishes a corresponding high-speed shaft misalignment fault warning model, which has the advantages of low cost, high efficiency and strong interpretability.
[0118] See also Figure 4 The embodiment of the present application provides a device 400 for predicting shaft misalignment of a high-speed shaft of a wind turbine, the device comprising:
[0119] An acquisition unit 401 is used to acquire real-time operating parameters of a wind turbine;
[0120] The processing unit 402 is used to pre-process the real-time operation parameters to obtain a target data set;
[0121] A feature construction unit 403 is used to construct features for a target data set to obtain a target feature variable;
[0122] The prediction unit 404 is used to input the target characteristic variable into the high-speed shaft misalignment fault warning model for prediction, and obtain the predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate that the wind turbine set is in a normal state or a fault state. The high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
[0123] Optionally, the processing unit 402 is specifically configured to:
[0124] Eliminate data in the real-time operating parameters that are in a normal power generation state and whose active power is less than a threshold value to obtain first processed data;
[0125] According to the density-based noise, a spatial clustering DBSCAN algorithm is applied to remove the data with abnormal high-speed shaft speed in the first processed data to obtain the second processed data;
[0126] The second processed data is divided into a plurality of target data groups according to time periods, and target data with a data volume less than a threshold value is eliminated from the plurality of target data groups to obtain a target data set.
[0127] Optionally, the feature construction unit 403 is specifically used for:
[0128] Divide each group of target data in the target data set according to the power interval;
[0129] Solving for extreme values of the high-speed shaft speed and the generator speed in target data within different power intervals, wherein the extreme values include a maximum value and a minimum value;
[0130] The extreme values of the high-speed shaft speed and the generator speed are smoothed to obtain target characteristic variables, wherein the target characteristic variables include extreme values of different power intervals, extreme values of the high-speed shaft speed within different power intervals, and extreme values of the generator speed.
[0131] Optionally, the apparatus 400 further includes a training unit, and the training unit includes:
[0132] An acquisition subunit is used to acquire historical operating parameters of the wind turbine and historical high-speed shaft fault information, and pre-process the historical operating parameters to obtain a historical data set;
[0133] Obtaining subunits for constructing features for historical data sets and obtaining historical feature variables;
[0134] A labeling subunit is used to label the historical data set according to the historical high-speed shaft crossover fault information to obtain the historical fault status data;
[0135] A division sub-unit is used to divide the historical data set into a training set and a validation set after disrupting the time sequence according to a specific ratio;
[0136] The training subunit is used to input the historical feature variables and historical fault label data in the training set into the initial model for training to obtain a high-speed shaft runaway fault warning model.
[0137] Optionally, the obtaining subunit is specifically used for:
[0138] According to the high-speed shaft misalignment fault information, the historical data set for the time period from the day the fault was discovered to the day before the bearing was replaced was marked as a fault state, and the historical data set for the time period one month before the fault was discovered and the historical data set for the time period one day after the bearing was replaced was marked as a normal state, so as to obtain historical fault state data.
[0139] Optionally, the training subunit is specifically used to include:
[0140] The historical feature variables and historical fault label data in the training set are imported into the gradient boosting decision tree algorithm in the initial model, and the parameters in the initial model are trained using the 10-fold cross-validation method; when the accuracy index of the initial model is greater than or equal to the preset value, a high-speed shaft runaway fault warning model is obtained.
[0141] Optionally, the device 400 further includes:
[0142] The verification unit is used to input the historical characteristic variables in the verification set into the high-speed shaft misalignment fault warning model for verification, so as to obtain the evaluation index of the high-speed shaft misalignment fault warning model.
[0143] It should be noted that the specific implementation method and effect of the wind turbine high-speed shaft misalignment prediction device 400 can be found in the above Figure 1 or Figure 2 The relevant descriptions in the provided methods will not be repeated here.
[0144] The present application embodiment also provides an electronic device 500, such as Figure 5 As shown, the device 500 includes a memory 501 and a processor 502:
[0145] The memory 501 is used to store computer programs;
[0146] The processor 502 is used to execute the above Figure 1 or Figure 2 The method provided.
[0147] Specifically, Figure 6 As shown, an electronic device 500 includes a power supply device 9, a processor 7, a memory 6, a network interface 12, a display device 13, an input device 14 and a storage medium 15 connected via a system bus 8. Among them, the processor 7 is used to provide computing and control capabilities. The storage medium 15 stores an operating system 16, a computer program 10 and data 11. The memory 6 provides an operating environment for the operating system 16 and the computer program 10 in the storage medium 15. The network interface 12 can be a wireless or wired network interface for network connection communication. The display device 13 can be various types of displays, including LCD displays, CRT displays, LED displays, etc. The input device 14 can be a keyboard, a mouse, a touch layer of a display screen, etc.
[0148] In addition, the present application also provides a computer-readable storage medium, the computer-readable storage medium is used to store a computer program, and the computer program is used to execute Figure 1 or Figure 2 The method provided.
[0149] Through the description of the above implementation methods, it can be known that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a general hardware platform. Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium, such as a read-only memory (ROM) / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network communication device such as a router) to execute the methods described in each embodiment of the present application or some parts of the embodiments.
[0150] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is only schematic, in which the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the goals of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.
[0151] The above description is merely an exemplary embodiment of the present application and is not intended to limit the protection scope of the present application.
Claims
1. A method for predicting high-speed shaft misalignment of a wind turbine, characterized in that: The method comprises: Obtain real-time operating parameters of wind turbines; Preprocessing the real-time operating parameters to obtain a target data set; Performing feature construction on the target data set to obtain target feature variables; The target characteristic variable is input into the high-speed shaft misalignment fault warning model for prediction to obtain predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate whether the wind turbine set is in a normal state or a fault state. The high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
2. The method according to claim 1, characterized in that The preprocessing of the real-time operating parameters to obtain a target data set includes: Eliminate data in the real-time operating parameters that are in a normal power generation state and whose active power is less than a threshold value to obtain first processed data; Eliminate data with abnormal high-speed shaft speed in the first processed data according to the density-based noise spatial clustering DBSCAN algorithm to obtain second processed data; The second processed data is divided into a plurality of groups of target data according to time periods, and target data with a data volume less than a threshold value in the plurality of groups of target data is removed to obtain the target data set.
3. The method according to claim 2, characterized in that The step of constructing features of the target data set to obtain target feature variables includes: Dividing each group of target data in the target data set according to power intervals; Solving for extreme values of the high-speed shaft speed and the generator speed in the target data within different power intervals, wherein the extreme values include a maximum value and a minimum value; The extreme values of the high-speed shaft speed and the generator speed are smoothed to obtain the target characteristic variable, wherein the target characteristic variable includes the extreme values of the different power intervals, the extreme values of the high-speed shaft speed within the different power intervals, and the extreme values of the generator speed.
4. The method according to claim 1, characterized in that: The training process of the high-speed shaft misalignment fault warning model includes: Acquire historical operating parameters and historical high-speed shaft misalignment fault information of the wind turbine set, and pre-process the historical operating parameters to obtain a historical data set; Performing feature construction on the historical data set to obtain the historical feature variables; Marking the historical data set according to the historical high-speed shaft misalignment fault information to obtain the historical fault status data; The historical data set is divided into a training set and a validation set after disrupting the time sequence according to a specific ratio; The historical feature variables and historical fault label data in the training set are input into the initial model for training to obtain the high-speed shaft runaway fault warning model.
5. The method according to claim 4, characterized in that The step of labeling the historical data set according to the historical high-speed shaft misalignment fault information to obtain the historical fault status data includes: According to the high-speed shaft misalignment fault information, the historical data set for the time period from the day the fault was discovered to the day before the bearing was replaced is marked as a fault state, and the historical data set for the time period one month before the fault was discovered and the historical data set for the time period one day after the bearing was replaced are marked as normal states, so as to obtain the historical fault state data.
6. The method according to claim 4, characterized in that The inputting of the historical characteristic variables and historical fault label data in the training set into the initial model for training to obtain the high-speed shaft misalignment fault warning model includes: Importing the historical feature variables and historical fault label data in the training set into the gradient boosting decision tree algorithm in the initial model, and training the parameters in the initial model using a 10-fold cross validation method; When the accuracy index of the initial model is greater than or equal to a preset value, the high-speed shaft runaway fault warning model is obtained.
7. The method according to claim 4, characterized in that After obtaining the high-speed shaft misalignment fault warning model, the method further includes: The historical characteristic variables in the verification set are input into the high-speed shaft misalignment fault warning model for verification, so as to obtain the evaluation index of the high-speed shaft misalignment fault warning model.
8. A device for predicting high-speed shaft misalignment of a wind turbine, characterized in that: The device comprises: An acquisition unit, used for acquiring real-time operating parameters of the wind turbine generator set; A processing unit, used for preprocessing the real-time operating parameters to obtain a target data set; A feature construction unit, used to construct features for the target data set to obtain target feature variables; A prediction unit is used to input the target characteristic variable into a high-speed shaft misalignment fault warning model for prediction, and obtain predicted fault status data of the wind turbine set, wherein the predicted fault status data is used to indicate that the wind turbine set is in a normal state or a fault state, and the high-speed shaft misalignment fault warning model is a trained neural network model, and the training samples of the high-speed shaft misalignment fault warning model are the historical characteristic variables of the wind turbine set and the corresponding historical fault status data.
9. An electronic device, characterized in that: The device comprises a memory and a processor, and the electronic device is used to execute a program stored in the memory to run the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the method according to any one of claims 1 to 7.