Fault prediction method and device for wind turbine generator set

By predicting the operating parameters of the wind turbine and extracting fault characteristics, and combining the first- and second-level models for fault prediction, the problem of insufficient accuracy and universality of the wind turbine fault prediction is solved, and the operation and maintenance costs are reduced.

CN114037127BActive Publication Date: 2024-11-08HUANENG SHEYANG NEW ENERGY POWER GENERATION CO LTD +2
View PDF 9 Cites 0 Cited by

Patent Information

Application Number
CN202111243323.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-25
Publication Date
2024-11-08
Estimated Expiration
2041-10-25

AI Technical Summary

Technical Problem

The prior art is difficult to effectively predict and reduce the failure of wind turbines, resulting in high operation and maintenance costs and insufficient accuracy and versatility in fault prediction.

Method used

By collecting the operating parameters of each subsystem of the wind turbine, determining the target operating parameters, using a first-level model for prediction, and combining a second-level model for fault feature extraction and prediction, the accuracy and universality of fault prediction are enhanced.

Benefits of technology

It realizes high accuracy and versatility prediction of wind turbine failures, reduces unnecessary repairs, and reduces operation and maintenance costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114037127B_ABST
    Figure CN114037127B_ABST
Patent Text Reader

Abstract

The present invention discloses a fault prediction method and device for a wind turbine. The method comprises: collecting operating parameters of each subsystem of the wind turbine, determining target operating parameters in the operating parameters, sequentially taking one parameter in the target operating parameters as the output parameter of the first-level model, taking other parameters except one parameter in the target operating parameters as the input parameters of the first-level model, obtaining the predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model, and obtaining the fault prediction result according to the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, predicts the fault of the wind turbine based on the predicted value of the first-level model through the second-level model, and enhances the accuracy and versatility of the fault prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of energy technology, and in particular to a fault prediction method and device for a wind turbine generator set, a wind turbine generator set, an electronic device and a storage medium. Background Art

[0002] As energy shortages worsen, wind power, as a crucial clean energy source, will play an irreplaceable role in the future low-carbon era. Wind power generation, with its renewable and environmentally friendly advantages, is gaining increasing popularity. Wind turbines, a key component of wind power generation, convert wind energy into AC power. Due to the harsh operating environment and complex terrain, and the fact that wind turbines are large rotating equipment with complex and variable operating conditions, wind turbines are subject to numerous failures and maintenance requirements during operation. Current wind farm operation and maintenance practices primarily rely on reactive and scheduled maintenance, resulting in high O&M costs. To reduce these costs, predictive maintenance has become a key research area. Predictive maintenance aims to shift from reactive maintenance to proactive maintenance, identifying potential problems before they occur and reducing downtime. Furthermore, it reduces unnecessary scheduled maintenance, enables macro-level allocation of maintenance resources, and reduces O&M costs.

[0003] Therefore, how to enhance the accuracy and versatility of wind turbine fault prediction has become an urgent problem to be solved in the field of wind power generation. Summary of the Invention

[0004] The present invention aims to solve one of the technical problems in the above-mentioned technology at least to a certain extent.

[0005] To this end, the first purpose of the present invention is to propose a fault prediction method for a wind turbine, which collects operating parameters of each subsystem of the wind turbine, determines target operating parameters among the operating parameters, sequentially uses one parameter among the target operating parameters as the output parameter of the first-level model, uses the other parameters except one parameter among the target operating parameters as the input parameters of the first-level model, obtains the predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model, and obtains the fault prediction result according to the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0006] The second object of the present invention is to provide a fault prediction device for a wind turbine generator set.

[0007] The third objective of the present invention is to provide a wind turbine generator set.

[0008] A fourth object of the present invention is to provide an electronic device.

[0009] A fifth object of the present invention is to provide a computer-readable storage medium.

[0010] The first aspect of the present invention proposes a fault prediction method for a wind turbine generator set, comprising: collecting operating parameters of each subsystem of the wind turbine generator set; determining target operating parameters among the operating parameters; taking one of the target operating parameters as the output parameter of a first-level model, and taking other parameters of the target operating parameters except the one parameter as the input parameters of the first-level model; obtaining predicted values ​​of the output parameters of the first-level model according to the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model; and obtaining a fault prediction result according to the predicted values ​​of the output parameters of each of the first-level models and the second-level model.

[0011] According to the fault prediction method of a wind turbine generator set according to an embodiment of the present invention, the operating parameters of each subsystem of the wind turbine generator set are collected, the target operating parameters among the operating parameters are determined, one parameter of the target operating parameters is used as the output parameter of the first-level model in turn, and the other parameters except one parameter of the target operating parameters are used as the input parameters of the first-level model. The predicted value of the output parameter of the first-level model is obtained according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model. The fault prediction result is obtained according to the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine generator set based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0012] In addition, the fault prediction method for a wind turbine generator system according to the above embodiment of the present invention may also have the following additional technical features:

[0013] In one embodiment of the present invention, the method further includes: determining the target operating parameters among the operating parameters, including: cleaning abnormal data in the operating parameters; and determining the target operating parameters among the cleaned operating parameters based on sample target operating parameters.

[0014] In one embodiment of the present invention, the method further includes: collecting sample operating parameters of each subsystem of the wind turbine generator set; cleaning abnormal data in the sample operating parameters; performing preliminary screening on the cleaned sample operating parameters based on expert experience; performing variance screening on the sample operating parameters after the preliminary screening, and eliminating the sample operating parameters whose variance is less than a preset variance threshold; performing correlation screening on the sample operating parameters after the variance screening, and determining the sample operating parameters whose correlation coefficient is greater than a preset correlation coefficient threshold as the sample target operating parameters.

[0015] In one embodiment of the present invention, the method further includes: sequentially taking one of the sample target operating parameters as the sample output parameter of the first-level model, and taking the other parameters of the sample target operating parameters except the one parameter as the sample input parameters of the first-level model; performing feature extraction on the sample input parameters of the first-level model to obtain the sample input parameter features of the first-level model; taking the sample input parameter features as input and the sample output parameters as output, training the first-level model to be trained to obtain the first-level model.

[0016] In one embodiment of the present invention, the method further includes: obtaining a predicted value of the output parameter of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, including: performing feature extraction on the input parameters of the first-level model to obtain the input parameter features of the first-level model; inputting the input parameter features of the first-level model into the first-level model corresponding to the output parameters of the first-level model to obtain the predicted value of the output parameter of the first-level model.

[0017] In one embodiment of the present invention, the method further includes: obtaining a fault prediction result based on the predicted values ​​of the output parameters of each of the first-level models and the second-level model, including: calculating the difference between the output parameters of the first-level model and the predicted values ​​of the output parameters of the first-level model to obtain a residual feature; generating a residual state feature based on the residual feature and time information; inputting the residual feature, the residual state feature and the operating condition data into the second-level model to obtain the fault prediction result.

[0018] In one embodiment of the present invention, the method further includes: calculating the difference between the sample output parameters of the first-level model and the predicted values ​​of the sample output parameters of the first-level model within a preset number of time periods before the sample fault occurs, to obtain a sample residual feature; generating a sample residual state feature based on the sample residual feature and sample time information; taking the sample residual feature, the sample residual state feature and sample operating condition data as input, and taking the sample fault as output, to train the second-level model to be trained, to obtain the second-level model.

[0019] The second aspect of the present invention proposes a fault prediction device for a wind turbine generator set, comprising: an acquisition module for acquiring operating parameters of each subsystem of the wind turbine generator set; a first determination module for determining target operating parameters among the operating parameters; a second determination module for sequentially using one of the target operating parameters as the output parameter of a first-level model, and using other parameters of the target operating parameters except the one parameter as the input parameters of the first-level model; a first prediction module for obtaining a predicted value of the output parameter of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model; and a second prediction module for obtaining a fault prediction result based on the predicted values ​​of the output parameters of each of the first-level models and the second-level model.

[0020] The fault prediction device for a wind turbine according to an embodiment of the present invention collects operating parameters of each subsystem of the wind turbine, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of a first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains predicted values ​​of the output parameters of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, and obtains fault prediction results based on the predicted values ​​of the output parameters of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes secondary extraction of fault characteristics of each subsystem, and performs fault prediction on the wind turbine through the second-level model based on the predicted values ​​of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0021] In addition, the fault prediction device for a wind turbine generator system according to the above embodiment of the present invention may also have the following additional technical features:

[0022] In one embodiment of the present invention, the first determining module is specifically configured to: clean abnormal data in the operating parameters; and determine the target operating parameters in the cleaned operating parameters based on sample target operating parameters.

[0023] In one embodiment of the present invention, the first determination module is further used to: collect sample operating parameters of each subsystem of the wind turbine generator set; clean abnormal data in the sample operating parameters; preliminarily screen the cleaned sample operating parameters based on expert experience; perform variance screening on the sample operating parameters after the preliminary screening, and eliminate the sample operating parameters whose variance is less than a preset variance threshold; perform correlation screening on the sample operating parameters after the variance screening, and determine the sample operating parameters whose correlation coefficient is greater than a preset correlation coefficient threshold as the sample target operating parameters.

[0024] In one embodiment of the present invention, the device also includes: a first training module, which is used to: sequentially use one parameter of the sample target operating parameters as the sample output parameter of the first-level model, and use the other parameters of the sample target operating parameters except the one parameter as the sample input parameters of the first-level model; perform feature extraction on the sample input parameters of the first-level model to obtain the sample input parameter features of the first-level model; use the sample input parameter features as input and the sample output parameters as output to train the first-level model to obtain the first-level model.

[0025] In one embodiment of the present invention, the first prediction module is specifically used to: perform feature extraction on the input parameters of the first-level model to obtain the input parameter features of the first-level model; input the input parameter features of the first-level model into the first-level model corresponding to the output parameters of the first-level model to obtain the predicted values ​​of the output parameters of the first-level model.

[0026] In one embodiment of the present invention, the second prediction module is specifically used to: calculate the difference between the output parameter of the first-level model and the predicted value of the output parameter of the first-level model to obtain the residual feature; generate the residual state feature based on the residual feature and time information; input the residual feature, the residual state feature and the operating condition data into the second-level model to obtain the fault prediction result.

[0027] In one embodiment of the present invention, the device also includes: a second training module, which is used to: calculate the difference between the sample output parameters of the first-level model and the predicted values ​​of the sample output parameters of the first-level model within a preset number of time periods before the sample fault occurs, to obtain sample residual features; generate sample residual state features based on the sample residual features and sample time information; use the sample residual features, the sample residual state features and sample operating condition data as input, and use the sample fault as output to train the second-level model to obtain the second-level model.

[0028] A third embodiment of the present invention provides a wind turbine generator set, comprising: a fault prediction device for the wind turbine generator set as described in the second embodiment of the present invention.

[0029] The wind turbine generator set of the embodiment of the present invention collects operating parameters of each subsystem of the wind turbine generator set, determines target operating parameters among the operating parameters, sequentially uses one parameter among the target operating parameters as the output parameter of the first-level model, uses the other parameters except one parameter among the target operating parameters as the input parameters of the first-level model, obtains the predicted value of the output parameter of the first-level model based on the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model, and obtains the fault prediction result based on the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine generator set based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of the fault prediction.

[0030] The fourth embodiment of the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the fault prediction method for a wind turbine as described in the first embodiment of the present invention is implemented.

[0031] The electronic device of the embodiment of the present invention executes a computer program stored in a memory through a processor, collects operating parameters of each subsystem of the wind turbine, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of the first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains predicted values ​​of the output parameters of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, and obtains fault prediction results based on the predicted values ​​of the output parameters of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes secondary extraction of fault characteristics of each subsystem, and performs fault prediction on the wind turbine based on the predicted values ​​of the first-level model through the second-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0032] The fifth embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the fault prediction method for a wind turbine generator system as described in the first embodiment of the present application is implemented.

[0033] The computer-readable storage medium of an embodiment of the present invention, by storing a computer program and being executed by a processor, collects operating parameters of each subsystem of a wind turbine generator set, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of a first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains predicted values ​​of the output parameters of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, and obtains fault prediction results based on the predicted values ​​of the output parameters of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of a wind turbine generator set based on the first-level model, realizes secondary extraction of fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted values ​​of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0034] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0036] Figure 1 A schematic flow chart of a fault prediction method for a wind turbine generator system according to an embodiment of the present invention;

[0037] Figure 2 A schematic diagram of a flow chart of determining target operating parameters in a fault prediction method for a wind turbine generator system according to an embodiment of the present invention;

[0038] Figure 3 A schematic diagram of a flow chart of determining sample target operating parameters in a fault prediction method for a wind turbine generator system according to an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of a process for generating predicted values ​​of output parameters of a primary model in a fault prediction method for a wind turbine generator system according to an embodiment of the present invention;

[0040] Figure 5 A schematic flow chart of a training process of a primary model in a fault prediction method for a wind turbine generator system according to a specific example of the present invention;

[0041] Figure 6 A schematic diagram of a flow chart of generating a fault prediction result in a fault prediction method for a wind turbine generator system according to a specific example of the present invention;

[0042] Figure 7Schematic diagram of the flow of the training process of the secondary model in the fault prediction method of the wind turbine generator according to a specific example of the present invention;

[0043] Figure 8 2 is a schematic structural diagram of a fault prediction device for a wind turbine generator system according to an embodiment of the present invention;

[0044] Figure 9 is a structural diagram of a wind turbine generator set according to an embodiment of the present invention; and

[0045] Figure 10 FIG. 1 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0047] The following describes a fault prediction method and device for a wind turbine generator set, a wind turbine generator set, an electronic device, and a storage medium according to embodiments of the present invention with reference to the accompanying drawings.

[0048] Figure 1 FIG. 1 is a flow chart of a fault prediction method for a wind turbine generator system according to an embodiment of the present invention.

[0049] like Figure 1 As shown, the fault prediction method of a wind turbine generator system according to an embodiment of the present invention includes:

[0050] S101, collecting operating parameters of each subsystem of the wind turbine.

[0051] In an embodiment of the present invention, a wind turbine is divided into multiple subsystems based on expert experience, such as a generator, a gearbox, a main bearing, a pitch system, and a yaw system. Operating parameters of each subsystem are collected, for example, using a Supervisory Control and Data Acquisition (Scada) system.

[0052] S102: Determine target operating parameters among the operating parameters.

[0053] In some embodiments, target operating parameters corresponding to each subsystem of the wind turbine generator set are determined from the collected operating parameters of each subsystem of the wind turbine generator set, so as to perform fault prediction on each subsystem of the wind turbine generator set based on the target operating parameters, thereby realizing fault prediction of the wind turbine generator set, wherein the target operating parameters may be parameters related to the normal working state of each subsystem of the wind turbine generator set, for example, the target operating parameters corresponding to the gearbox may be the high-speed shaft gearbox inlet temperature, gearbox oil temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, gearbox outlet oil pressure, etc.

[0054] S103 , sequentially taking one of the target operating parameters as an output parameter of the first-level model, and taking the other parameters except one of the target operating parameters as input parameters of the first-level model.

[0055] In some embodiments, for a subsystem of a wind turbine generator, one of the multiple target operating parameters corresponding to the subsystem is sequentially used as an output parameter of the first-level model, and the other parameters of the multiple target operating parameters corresponding to the subsystem, excluding the one parameter determined as the output parameter, are used as input parameters of the model. That is, each of the multiple target operating parameters corresponding to a subsystem is used as an output parameter of the first-level model for sequential prediction. In this way, the multiple target operating parameters corresponding to each subsystem of the wind turbine generator are predicted using the first-level model.

[0056] For example, the multiple target operating parameters corresponding to the gearbox are the target operating parameters corresponding to the gearbox, which can be the high-speed shaft gearbox inlet temperature, gearbox oil temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure. The high-speed shaft gearbox inlet temperature is used as the output parameter of the first-level model, and the gearbox oil temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure are used as input parameters of the gearbox model; once again, the gearbox oil temperature is used as the output parameter of the first-level model, and the high-speed shaft gearbox inlet temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure are used as input parameters of the first-level model, and the prediction of each target operating parameter is realized in turn.

[0057] S104, obtaining a predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model.

[0058] In some embodiments, the output parameters of each first-level model determined in the above step S103 can correspond to a first-level model as the state characteristics of the wind turbine subsystem, and the first-level model corresponding to the parameter is obtained according to the output parameters of the first-level model. According to the input parameters of the first-level model corresponding to the output parameters of the first-level model and the first-level model corresponding to the output parameters of the first-level model, the output parameters of the first-level model are predicted to obtain the corresponding predicted values.

[0059] In some embodiments, the above process is looped to sequentially obtain predictions of the output parameters of each primary model determined in step S103 and obtain corresponding prediction values, thereby obtaining prediction values ​​of target operating parameters corresponding to each subsystem of the wind turbine.

[0060] In some embodiments, the first-level model corresponding to the output parameters of the first-level model can be obtained from a hard disk or a database, and can be a pre-trained first-level model.

[0061] In some embodiments, the predicted values ​​of the output parameters of multiple primary models may be saved in a database so that the predicted values ​​of the target operating parameters corresponding to each subsystem of the wind turbine generator set can be obtained in a timely manner when predicting faults of the wind turbine generator set.

[0062] S105 , obtaining a fault prediction result based on the predicted values ​​of the output parameters of each first-level model and the second-level model.

[0063] In some implementations, fault prediction is performed on the wind turbine generator system based on the predicted values ​​of the output parameters of each primary model and the secondary model to obtain a fault prediction result.

[0064] In summary, according to the fault prediction method of a wind turbine generator set in an embodiment of the present invention, the operating parameters of each subsystem of the wind turbine generator set are collected, the target operating parameters among the operating parameters are determined, one parameter among the target operating parameters is used as the output parameter of the first-level model in turn, and the other parameters except one parameter among the target operating parameters are used as the input parameters of the first-level model. The predicted value of the output parameter of the first-level model is obtained according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model. The fault prediction result is obtained according to the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine generator set based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0065] exist Figure 1 Based on the embodiment shown, Figure 2 As shown, in step S102, "determining target operating parameters in the operating parameters" may specifically include the following steps:

[0066] S201, cleaning abnormal data in the operating parameters.

[0067] In some embodiments, the collected operating parameters can be pre-processed by data cleaning, for example, cleaning abnormal data in the operating parameters: judging the validity of sensor data and eliminating abnormal sensor data; eliminating wind turbine shutdown data and fault shutdown data based on wind turbine operating condition information; eliminating abnormal power generation data of wind turbines such as power limit data based on wind turbine operating status information, so as to enhance the accuracy of fault prediction.

[0068] S202: Determine target operating parameters among the cleaned operating parameters according to the sample target operating parameters.

[0069] In the embodiment of the present invention, the target operating parameters may be determined from the operating parameters after cleaning according to the pre-obtained target operating parameters of the sample.

[0070] Based on any of the above embodiments, Figure 3 As shown, the fault prediction method for a wind turbine generator system according to an embodiment of the present invention may further include a process for determining sample target operating parameters, which may specifically include the following steps:

[0071] S301, collecting sample operating parameters of each subsystem of the wind turbine generator system.

[0072] In some implementations, historical operating data of each subsystem of the wind turbine generator system is collected based on Scada as sample operating parameters.

[0073] S302, cleaning abnormal data in the sample operation parameters.

[0074] In some implementations, abnormal data in the sample operating parameters are cleaned to remove the abnormal data and retain valid data. The specific implementation is similar to the above embodiment and will not be repeated here.

[0075] S303: Preliminary screening of the running parameters of the cleaned samples is performed based on expert experience.

[0076] In some implementations, the cleaned sample operating parameters are preliminarily screened based on expert experience to determine operating parameters in the sample operating parameters that are related to the fault state of the corresponding wind turbine subsystem.

[0077] S304: Perform variance screening on the sample operating parameters after the preliminary screening, and eliminate the sample operating parameters whose variance is less than a preset variance threshold.

[0078] In some embodiments, the variance screening can be continued for the sample operating parameters after the preliminary screening, and the variance σ of each sample operating parameter can be calculated using the following variance calculation formula:x :

[0079]

[0080] Among them, x is the sample operation parameter, i is the number of sample operation parameters, is the mean of the sample running parameters.

[0081] According to the variance of each sample operating parameter, the sample operating parameters with a variance less than a preset variance threshold are eliminated, wherein the preset variance threshold can be set as needed and is not limited in the present invention.

[0082] S305 , performing correlation screening on the sample operating parameters after variance screening, and determining the sample operating parameters with correlation coefficients greater than a preset correlation coefficient threshold as sample target operating parameters.

[0083] In some embodiments, the correlation coefficient of the sample operating parameters after variance screening is calculated, and the sample operating parameters after variance screening are subjected to correlation screening based on the correlation coefficient, and the sample operating parameters whose correlation coefficient is greater than a preset correlation coefficient threshold are determined as sample target operating parameters, wherein the preset correlation coefficient threshold can be set as needed, and the present invention does not limit it.

[0084] The correlation coefficient calculation formula is as follows:

[0085]

[0086] Wherein, x represents the sample operation parameter, and y represents the remaining sample operation parameters except x.

[0087] Based on any of the above embodiments, Figure 4 As shown, in step S104, "obtaining a predicted value of the output parameter of the first-level model according to the first-level model input parameter and the first-level model output parameter corresponding to the first-level model" may specifically include the following steps:

[0088] S401, performing feature extraction on the input parameters of the first-level model to obtain input parameter features of the first-level model.

[0089] In some embodiments, a convolutional neural network (CNN) may be used to extract features of the input parameters of the primary model to obtain input parameter features of the primary model.

[0090] In some embodiments, different input parameter features can be obtained through multiple convolution-pooling operations.

[0091] In some embodiments, the input parameters of the primary model may be normalized before feature extraction is performed on the input parameters.

[0092] S402: Input the input parameter features of the first-level model into the first-level model corresponding to the output parameters of the first-level model to obtain the predicted values ​​of the output parameters of the first-level model.

[0093] In some embodiments, the extracted input parameter features of the first-level model are input into the first-level model corresponding to the output parameters of the first-level model to perform output parameter prediction to obtain predicted values ​​of the output parameters of the first-level model.

[0094] Based on any of the above embodiments, Figure 5 As shown, the fault prediction method for a wind turbine generator system according to an embodiment of the present invention may further include a training process of a primary model, which may specifically include the following steps:

[0095] S501 , sequentially taking one parameter of the sample target operating parameters as a sample output parameter of the first-level model, and taking the other parameters of the sample target operating parameters except one parameter as sample input parameters of the first-level model.

[0096] It should be noted here that the method for determining the sample output parameters and input parameters of the first-level model is similar to that in the above embodiment and will not be repeated here.

[0097] S502: Extract features of the sample input parameters of the first-level model to obtain features of the sample input parameters of the first-level model.

[0098] In some embodiments, feature extraction is performed on the sample input parameters of the first-level model based on a convolutional neural network to obtain the sample input parameter features of the first-level model.

[0099] S503 , taking the sample input parameter features as input and the sample output parameters as output, and training the first-level model to be trained to obtain a first-level model.

[0100] In some embodiments, a first-level model can be constructed based on a random forest model, with the sample input parameter features as the input of the model and the sample output parameters as the output of the model. The model hyperparameters can be optimized using a grid search method to obtain the optimal training model as the first-level model for target operating parameter prediction.

[0101] Among them, the mean square error can be used as a discrimination criterion for hyperparameter optimization.

[0102] Therefore, the above process can generate trained first-level models corresponding to multiple sample output parameters, and store multiple first-level models on a hard disk or in a database so that the required first-level models can be obtained in time to predict the target operating parameters.

[0103] Based on any of the above embodiments, Figure 6 As shown, in step S105, "obtaining a fault prediction result according to the predicted values ​​of the output parameters of each first-level model and the second-level model" may specifically include the following steps:

[0104] S601, calculating the difference between the output parameter of the first-level model and the predicted value of the output parameter of the first-level model to obtain a residual feature.

[0105] In some embodiments, the difference between the output parameter of each first-level model and the predicted value of the output parameter of the first-level model is calculated to obtain the residual feature of the output parameter of each first-level model.

[0106] For example, a residual threshold is set, that is, a difference threshold between the output parameter of the first-level model and the predicted value of the output parameter of the first-level model, and the state where the residual is greater than or equal to the residual threshold is set to 1, and the state where the residual is less than the residual threshold is set to 0, thereby obtaining a residual feature.

[0107] S602: Generate residual state features according to the residual features and time information.

[0108] In some embodiments, based on time information, the time corresponding to the target operating parameter is divided into N time periods, and the residual features in each time period are accumulated to obtain the window cumulative features. By setting a threshold V, the residual state features are generated based on the window cumulative features. For example, the state in which the window cumulative features in a time window are greater than or equal to the threshold V is set to 1, and the state in which the window cumulative features are less than the threshold V is set to 0, thereby obtaining the residual state features.

[0109] S603: Input the residual characteristics, residual state characteristics and working condition data into the secondary model to obtain the fault prediction result.

[0110] In some embodiments, residual features, residual state features, and operating condition data are input into a secondary model as model input data to obtain a fault prediction result. A trained secondary model can be obtained from a database or hard disk. The operating condition parameters may include power, rotational speed, wind speed, and pitch angle.

[0111] Based on any of the above embodiments, Figure 7 As shown, the fault prediction method for a wind turbine generator system according to an embodiment of the present invention may further include a training process of a secondary model, which may specifically include the following steps:

[0112] S701 , calculating the difference between the sample output parameter of the first-level model and the predicted value of the sample output parameter of the first-level model within a preset number of time periods before the sample fault occurs, and obtaining the sample residual feature.

[0113] In some embodiments, a time period T is set, n time periods are preset, the output parameters of the first-level model within n time periods T before the sample failure occurs are selected, and the predicted values ​​of the sample output parameters of the corresponding first-level model are obtained from a database or hard disk, and the difference between the sample output parameters of the first-level model and the predicted values ​​of the sample output parameters of the first-level model is calculated to obtain the sample residual characteristics within n time periods before the sample failure occurs.

[0114] S702: Generate a sample residual state feature according to the sample residual feature and the sample time information.

[0115] It should be noted here that the method for generating the sample residual state feature is similar to that in the above embodiment and will not be described again here.

[0116] S703 , taking the sample residual features, the sample residual state features and the sample operating condition data as inputs and the sample faults as outputs, and training the secondary model to be trained to obtain a secondary model.

[0117] In some embodiments, a secondary model is constructed based on deep neural networks (DNNs) and expert experience, combining the different failure modes of each wind turbine subsystem. A secondary model for wind turbine fault prediction is generated based on the prediction results of the primary model. Sample residual features, sample residual state features, and sample operating condition data are used as model inputs, and corresponding sample faults are used as model outputs. The secondary model to be trained is trained to obtain optimal model parameters, resulting in a secondary model. The model is then saved to a database or hard disk for fault prediction based on the secondary model.

[0118] In order to implement the above embodiment, the present invention further provides a fault prediction device for a wind turbine generator set.

[0119] Figure 8 2 is a schematic structural diagram of a fault prediction device for a wind turbine generator system according to an embodiment of the present invention.

[0120] like Figure 8 As shown, the fault prediction device 1000 of a wind turbine generator system according to an embodiment of the present invention includes: a collection module 1001 , a first determination module 1002 , a second determination module 1003 , a first prediction module 1004 and a second prediction module 1005 .

[0121] The acquisition module 1001 is used to acquire the operating parameters of each subsystem of the wind turbine.

[0122] The first determining module 1002 is configured to determine a target operating parameter among the operating parameters.

[0123] The second determining module 1003 is configured to sequentially use one of the target operating parameters as an output parameter of a first-level model, and use the other parameters of the target operating parameters except the one parameter as input parameters of the first-level model.

[0124] The first prediction module 1004 is configured to obtain a predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model.

[0125] The second prediction module 1005 is used to obtain a fault prediction result based on the predicted values ​​of the output parameters of each of the first-level models and the second-level model.

[0126] In one embodiment of the present invention, the first determining module 1002 is specifically configured to: clean abnormal data in the operating parameters; and determine the target operating parameters in the cleaned operating parameters according to sample target operating parameters.

[0127] In one embodiment of the present invention, the first determination module 1002 is further used to: collect sample operating parameters of each subsystem of the wind turbine generator set; clean abnormal data in the sample operating parameters; preliminarily screen the cleaned sample operating parameters based on expert experience; perform variance screening on the sample operating parameters after the preliminary screening, and eliminate the sample operating parameters whose variance is less than a preset variance threshold; perform correlation screening on the sample operating parameters after the variance screening, and determine the sample operating parameters whose correlation coefficient is greater than a preset correlation coefficient threshold as the sample target operating parameters.

[0128] In one embodiment of the present invention, the fault prediction device of the wind turbine group also includes: a first training module, which is used to: sequentially use one parameter of the sample target operating parameters as the sample output parameter of the first-level model, and use other parameters of the sample target operating parameters except the one parameter as the sample input parameters of the first-level model; perform feature extraction on the sample input parameters of the first-level model to obtain the sample input parameter features of the first-level model; use the sample input parameter features as input and the sample output parameters as output to train the first-level model to obtain the first-level model.

[0129] In one embodiment of the present invention, the first prediction module 1004 is specifically used to: perform feature extraction on the input parameters of the first-level model to obtain the input parameter features of the first-level model; input the input parameter features of the first-level model into the first-level model corresponding to the output parameters of the first-level model to obtain the predicted values ​​of the output parameters of the first-level model.

[0130] In one embodiment of the present invention, the second prediction module 1005 is specifically used to: calculate the difference between the output parameter of the first-level model and the predicted value of the output parameter of the first-level model to obtain the residual feature; generate the residual state feature according to the residual feature and time information; input the residual feature, the residual state feature and the operating condition data into the second-level model to obtain the fault prediction result.

[0131] In one embodiment of the present invention, the fault prediction device of a wind turbine generator system further includes: a second training module, which is used to: calculate the difference between the sample output parameters of the first-level model and the predicted values ​​of the sample output parameters of the first-level model within a preset number of time periods before the sample fault occurs, to obtain a sample residual feature; generate a sample residual state feature based on the sample residual feature and sample time information; use the sample residual feature, the sample residual state feature and sample operating condition data as input, and use the sample fault as output to train the second-level model to obtain the second-level model.

[0132] It should be noted that for details not disclosed in the fault prediction device for a wind turbine generator set according to an embodiment of the present invention, please refer to the details disclosed in the fault prediction method for a wind turbine generator set according to an embodiment of the present invention, which will not be repeated here.

[0133] In summary, the fault prediction device for a wind turbine according to an embodiment of the present invention collects operating parameters of each subsystem of the wind turbine, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of the first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains the predicted value of the output parameter of the first-level model based on the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model, and obtains the fault prediction result based on the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0134] In order to implement the above embodiment, the present invention further proposes a wind turbine generator set.

[0135] Figure 9 FIG. 1 is a structural diagram of a wind turbine generator set according to an embodiment of the present invention.

[0136] like Figure 9 As shown, a wind turbine generator set 1100 according to an embodiment of the present invention includes the above-mentioned fault prediction device 1000 for the wind turbine generator set.

[0137] The wind turbine generator set of the embodiment of the present invention can determine the target operating parameters among the operating parameters by collecting the operating parameters of each subsystem of the wind turbine generator set, sequentially use one parameter among the target operating parameters as the output parameter of the first-level model, use the other parameters except one parameter among the target operating parameters as the input parameters of the first-level model, obtain the predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model, and obtain the fault prediction result according to the predicted value of the output parameter of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine generator set based on the first-level model, realizes the secondary extraction of the fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted value of the first-level model, thereby enhancing the accuracy and versatility of the fault prediction.

[0138] In order to implement the above embodiment, Figure 10 As shown, an embodiment of the present invention proposes an electronic device 1200, including: a memory 1201, a processor 1202 and a computer program stored in the memory 1201 and executable on the processor 1202. When the processor 1202 executes the program, the above-mentioned wind turbine fault prediction method is implemented.

[0139] The electronic device of the embodiment of the present invention executes a computer program stored in a memory through a processor, collects operating parameters of each subsystem of the wind turbine, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of the first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains predicted values ​​of the output parameters of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, and obtains fault prediction results based on the predicted values ​​of the output parameters of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of the wind turbine based on the first-level model, realizes secondary extraction of fault characteristics of each subsystem, and performs fault prediction on the wind turbine based on the predicted values ​​of the first-level model through the second-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0140] To implement the above embodiment, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the above-mentioned fault prediction method for a wind turbine is implemented.

[0141] The computer-readable storage medium of an embodiment of the present invention, by storing a computer program and being executed by a processor, collects operating parameters of each subsystem of a wind turbine generator set, determines target operating parameters among the operating parameters, sequentially uses one of the target operating parameters as the output parameter of a first-level model, uses the other parameters except one of the target operating parameters as the input parameters of the first-level model, obtains predicted values ​​of the output parameters of the first-level model based on the first-level model corresponding to the input parameters of the first-level model and the output parameters of the first-level model, and obtains fault prediction results based on the predicted values ​​of the output parameters of each first-level model and the second-level model. The embodiment of the present invention predicts the target operating parameters of each subsystem of a wind turbine generator set based on the first-level model, realizes secondary extraction of fault characteristics of each subsystem, and performs fault prediction on the wind turbine generator set through the second-level model based on the predicted values ​​of the first-level model, thereby enhancing the accuracy and versatility of fault prediction.

[0142] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0143] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0144] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediary. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0145] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0146] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A fault prediction method for a wind turbine generator system, characterized in that: include: Collect the operating parameters of each subsystem of the wind turbine, including the generator, gearbox, main bearing, pitch system and yaw system. The operating parameters corresponding to the gearbox include the high-speed shaft gearbox inlet temperature, gearbox oil temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure; determining a target operating parameter among the operating parameters; Sequentially taking one of the target operating parameters as an output parameter of a primary model, and taking other parameters of the target operating parameters except the one parameter as input parameters of the primary model; Obtaining a predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model; Obtaining a fault prediction result according to the predicted values ​​of the output parameters of each of the first-level models and the second-level model; The fault prediction result is obtained according to the predicted value of the output parameter of each of the first-level models and the second-level model, including: Calculating the difference between the output parameter of the primary model and the predicted value of the output parameter of the primary model to obtain a residual feature; According to the time information, the time corresponding to the target operating parameters is divided into N time periods, and the residual features in each time period are accumulated to obtain the window cumulative features. By setting the threshold V, the residual state features are generated based on the window cumulative features; The residual characteristics, the residual state characteristics and the operating condition data are input into the secondary model to obtain the fault prediction result, and the operating condition parameters include power, rotation speed, wind speed and pitch angle.

2. The fault prediction method according to claim 1, characterized in that: The determining of the target operating parameters among the operating parameters includes: Cleaning abnormal data in the operating parameters; The target operating parameters among the operating parameters after cleaning are determined according to the sample target operating parameters.

3. The fault prediction method according to claim 2, characterized in that: Also includes: Collect sample operating parameters of each subsystem of the wind turbine; Cleaning abnormal data in the sample operation parameters; Conducting preliminary screening of the running parameters of the cleaned samples based on expert experience; Performing variance screening on the sample operating parameters after preliminary screening, and eliminating the sample operating parameters whose variance is less than a preset variance threshold; The sample operating parameters after the variance screening are subjected to correlation screening, and the sample operating parameters whose correlation coefficients are greater than a preset correlation coefficient threshold are determined as the sample target operating parameters.

4. The fault prediction method according to claim 3, characterized in that: Also includes: Sequentially taking one of the sample target operating parameters as a sample output parameter of a primary model, and taking other parameters of the sample target operating parameters except the one parameter as sample input parameters of the primary model; Performing feature extraction on the sample input parameters of the first-level model to obtain features of the sample input parameters of the first-level model; The sample input parameter features are taken as input, and the sample output parameters are taken as output, and the first-level model to be trained is trained to obtain the first-level model.

5. The fault prediction method according to claim 1, characterized in that: The step of obtaining a predicted value of the output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model comprises: Performing feature extraction on the input parameters of the primary model to obtain input parameter features of the primary model; The input parameter features of the first-level model are input into the first-level model corresponding to the output parameters of the first-level model to obtain the predicted values ​​of the output parameters of the first-level model.

6. The fault prediction method according to claim 1, characterized in that: Also includes: Calculate the difference between the sample output parameter of the first-level model and the predicted value of the sample output parameter of the first-level model within a preset number of time periods before the sample failure occurs, and obtain the sample residual feature; Generate sample residual state features according to sample residual features and sample time information; The sample residual feature, the sample residual state feature and the sample operating condition data are taken as input, and the sample fault is taken as output, and the secondary model to be trained is trained to obtain the secondary model.

7. A fault prediction device for a wind turbine generator set, characterized in that: include: The acquisition module is used to collect the operating parameters of each subsystem of the wind turbine, including the generator, gearbox, main bearing, pitch system and yaw system. The operating parameters corresponding to the gearbox include the high-speed shaft gearbox inlet temperature, gearbox oil temperature, gearbox bearing drive end temperature, gearbox bearing non-drive end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure; A first determining module, used to determine a target operating parameter among the operating parameters; A second determination module is used to sequentially use one of the target operating parameters as an output parameter of a primary model, and use other parameters of the target operating parameters except the one parameter as input parameters of the primary model; A first prediction module, configured to obtain a predicted value of an output parameter of the first-level model according to the first-level model corresponding to the input parameter of the first-level model and the output parameter of the first-level model; A second prediction module, used to obtain a fault prediction result according to the predicted value of the output parameter of each of the first-level models and the second-level model; The second prediction module is further used for: Calculating the difference between the output parameter of the primary model and the predicted value of the output parameter of the primary model to obtain a residual feature; According to the time information, the time corresponding to the target operating parameters is divided into N time periods, and the residual features in each time period are accumulated to obtain the window cumulative features. By setting the threshold V, the residual state features are generated based on the window cumulative features; The residual characteristics, the residual state characteristics and the operating condition data are input into the secondary model to obtain the fault prediction result, and the operating condition parameters include power, rotation speed, wind speed and pitch angle.

8. A wind turbine generator set, characterized in that: include: The fault prediction device for a wind turbine generator set as claimed in claim 7.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the fault prediction method for a wind turbine set as claimed in any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the fault prediction method for a wind turbine set as claimed in any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Wind turbine generator state parameter abnormity identification method based on combination prediction

    CN105719002A

  • Calculation method applied to multi-parameter fault prediction and judgment indexes of wind generating set

    CN105760617A

  • Recurrent neural network model construction method and gearbox fault detection method and device

    CN105787561A

  • Generator bearing fault prediction method

    CN107977508A

  • Bearing temperature model based wind turbine fault prediction method

    CN108680358A