WO2023065580A1

WO2023065580A1 - Fault diagnosis method and apparatus for gearbox of wind turbine generator set

Info

Publication number: WO2023065580A1
Application number: PCT/CN2022/077779
Authority: WO
Inventors: 王青天; 王海明; 张育钧; 张燧; 李小翔; 曾谁飞; 关建越; 陈朝晖; 杨永前; 冯帆; 任鑫; 王�华
Original assignee: 中国华能集团清洁能源技术研究院有限公司; 华能(浙江)能源开发有限公司清洁能源分公司
Priority date: 2021-10-18
Filing date: 2022-02-24
Publication date: 2023-04-27
Also published as: CN114036656A

Abstract

A fault diagnosis method and apparatus for a gearbox of a wind turbine generator set. The method comprises: collecting running parameters of a wind turbine generator set; sequentially taking one gearbox parameter as a gearbox output parameter, and taking the other parameters as gearbox input parameters; inputting a working condition parameter and the gearbox input parameters into a machine learning model corresponding to the gearbox output parameter, so as to obtain a predicted value of the gearbox output parameter; generating a machine learning diagnosis result according to the predicted value of the gearbox output parameter and the gearbox output parameter; inputting the working condition parameter and the gearbox input parameters into a rule model corresponding to the output gearbox parameter, so as to obtain a rule diagnosis result; and performing comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result and a rule set, so as to obtain a target diagnosis result.

Description

Fault diagnosis method and device for wind turbine gearbox

Cross References to Related Applications

This application is based on a Chinese patent application with application number No. 202111211995.2 and a filing date of October 18, 2021, and claims the priority of this Chinese patent application. The entire content of this Chinese patent application is hereby incorporated by reference into this application.

technical field

The present application relates to the field of energy technology, and in particular to a fault diagnosis method and device for a gearbox of a wind turbine, a wind turbine, electronic equipment, and a storage medium.

Background technique

At present, with the aggravation of energy shortage, wind energy, as a very important clean energy, will play an irreplaceable role in the future low-carbon era. Wind power generation has the advantages of renewable and environmental protection and has been widely used. Wind turbine is an important part of wind power generation, which can convert wind energy into AC power. It is a large rotating equipment that operates under variable conditions. Wind turbines are mainly divided into direct-drive units and doubly-fed units. The difference lies in whether there is a transmission process of the gearbox. In the field of wind energy, doubly-fed units occupy an important market share. The gearboxes of doubly-fed units work under alternating loads for a long time state, and the whole system also has complex coupled vibrations, which makes the fault diagnosis of the gearbox extremely complicated.

Therefore, how to accurately diagnose the gearbox on the premise of low cost, effectively improve the reliability of wind turbine operation, and reduce the risk of damage to key components of the wind turbine has become an urgent problem in the field of wind power generation.

Contents of the invention

The first purpose of this application is to propose a fault diagnosis method for wind turbine gearboxes, which can diagnose working condition parameters and gearbox input parameters based on machine learning models and rule models, and diagnose machine learning diagnosis results and rules based on rule sets. The diagnosis results are comprehensively diagnosed to obtain the target diagnosis results, so that the gearbox can be accurately diagnosed at a lower cost, which will effectively improve the reliability of the wind turbine operation and reduce the risk of damage to key components of the unit.

The second purpose of the present application is to propose a fault diagnosis device for a gearbox of a wind turbine.

The third purpose of the present application is to propose a wind turbine.

The fourth object of the present application is to provide an electronic device.

The fifth object of the present application is to provide a computer-readable storage medium.

The embodiment of the first aspect of the present application proposes a fault diagnosis method for the gearbox of a wind turbine, including: collecting the operating parameters of the wind turbine, the operating parameters include working condition parameters and gearbox parameters; One of the parameters in the gearbox is used as an output parameter of the gearbox, and other parameters in the gearbox parameters are used as gearbox input parameters; the working condition parameters and the gearbox input parameters are input to the In the machine learning model corresponding to the output parameter of the gearbox, the predicted value of the output parameter of the gearbox is obtained; according to the predicted value of the output parameter of the gearbox and the output parameter of the gearbox, a machine learning diagnosis result is generated; Input the state parameters and the input parameters of the gearbox into the rule model corresponding to the output parameters of the gearbox to obtain the rule diagnosis result; perform comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result and the rule set to obtain the target diagnostic result.

According to the fault diagnosis method of the wind turbine gearbox in the embodiment of the present application, the operating condition parameters and the gearbox input parameters can be diagnosed based on the machine learning model and the rule model, and the machine learning diagnosis result and the rule diagnosis result can be comprehensively diagnosed based on the rule set , to obtain the target diagnosis result, so that under the premise of lower cost, the accurate diagnosis of the gearbox will effectively improve the reliability of the wind turbine operation and reduce the risk of damage to key components of the unit.

In addition, the fault diagnosis method for the wind turbine gearbox according to the above-mentioned embodiments of the present application may also have the following additional technical features:

In one embodiment of the present application, the method further includes: acquiring sample machine learning diagnosis results, sample rule diagnosis results, and sample target diagnosis results; A sample machine learning diagnosis result and the sample rule diagnosis result; performing a checksum test on the sample machine learning diagnosis result and the sample rule diagnosis result corresponding to different sample target diagnosis results to obtain the rule gather.

In an embodiment of the present application, the method further includes: encoding the diagnosis result of the rule.

In an embodiment of the present application, the method further includes: determining an operable range of the gearbox output parameters; quantifying the gearbox output parameters beyond the operable range to obtain the rule model.

In an embodiment of the present application, the method further includes: quantifying the output parameter of the gearbox according to the overrun threshold of the output parameter of the gearbox provided by the manufacturer, to obtain the rule model.

In an embodiment of the present application, the method further includes: quantifying the output parameters of the gearbox according to the degree of dispersion and fluctuation state of the output parameters of the gearbox at different time periods, to obtain the rule model.

In one embodiment of the present application, the method further includes: determining the target sample gearbox input parameters according to the sample gearbox output parameters; taking the sample working condition parameters and the target sample gearbox input parameters as input, The gear box output parameters are used as output, and the machine learning model to be trained is trained to obtain the machine learning model.

In an embodiment of the present application, the method further includes: calculating a correlation coefficient between a sample gearbox input parameter and the sample gearbox output parameter; The gearbox input parameters are determined as the target sample gearbox input parameters.

In an embodiment of the present application, the method further includes: determining a second preset number of sample gearbox input parameters with greater importance as the target sample gearbox input parameters.

In an embodiment of the present application, the generating a machine learning diagnosis result according to the predicted value of the output parameter of the gearbox and the output parameter of the gearbox includes: calculating the predicted value of the output parameter of the gearbox and the output parameter of the gearbox The gearbox outputs a residual value between parameters; and the machine learning diagnosis result is generated according to the residual value and a residual value threshold.

In one embodiment of the present application, the method further includes: calculating the mean value and standard deviation of the residual value during the machine learning model training process; calculating the residual value threshold according to the mean value and the standard deviation .

The embodiment of the second aspect of the present application proposes a fault diagnosis device for the gearbox of a wind turbine, including: a data acquisition module for collecting operating parameters of the wind turbine, the operating parameters include working condition parameters and gearbox parameters; determine A module for sequentially using one of the gearbox parameters as a gearbox output parameter, and using other parameters in the gearbox parameters except the one parameter as a gearbox input parameter; a prediction module for using The working condition parameters and the input parameters of the gearbox are input into the machine learning model corresponding to the output parameters of the gearbox to obtain the predicted value of the output parameters of the gearbox; The predicted value of the gearbox output parameter and the machine learning diagnosis result are generated; the rule diagnosis module is used to input the working condition parameter and the gearbox input parameter into the rule model corresponding to the gearbox output parameter, Obtaining a rule diagnosis result; a comprehensive diagnosis module, configured to perform a comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result and the rule set, to obtain a target diagnosis result.

The fault diagnosis device for the gearbox of a wind turbine in the embodiment of the present application can diagnose the working condition parameters and the input parameters of the gearbox based on the machine learning model and the rule model, and perform comprehensive diagnosis on the machine learning diagnosis result and the rule diagnosis result based on the rule set, Obtaining the target diagnosis results, so as to accurately diagnose the gearbox at a lower cost, will effectively improve the reliability of the wind turbine operation and reduce the risk of damage to key components of the unit.

In addition, the fault diagnosis device for the wind turbine gearbox according to the above-mentioned embodiments of the present application may also have the following additional technical features:

In one embodiment of the present application, the device further includes: a rule generation module, the rule generation module is used to: obtain sample machine learning diagnosis results, sample rule diagnosis results, and sample target diagnosis results; analyze methods based on association rules Determining the sample machine learning diagnostic results and the sample rule diagnostic results corresponding to different sample target diagnostic results; The rule diagnosis result is checked and tested to obtain the rule set.

In an embodiment of the present application, the rule diagnosis module is further configured to: encode the result of the rule diagnosis.

In an embodiment of the present application, the device further includes: a quantization module, and the training module is used to: determine the operable range of the gearbox output parameters; The parameters are quantized to obtain the rule model.

In an embodiment of the present application, the quantization module is further configured to: quantify the output parameter of the gearbox according to the overrun threshold of the output parameter of the gearbox provided by the manufacturer, so as to obtain the rule model.

In an embodiment of the present application, the quantization module is further configured to: quantify the output parameters of the gearbox according to the degree of dispersion and the fluctuation state of the output parameters of the gearbox at different time periods, so as to obtain the rule model .

In one embodiment of the present application, the device further includes: a training module, the training module includes: a determining unit, configured to determine a target sample gearbox input parameter according to a sample gearbox output parameter; a training unit, used to convert the sample Working condition parameters and the target sample gearbox input parameters are used as input, and the sample gearbox output parameters are used as output to train the machine learning model to be trained to obtain the machine learning model.

In an embodiment of the present application, the determination unit is further configured to: calculate the correlation coefficient between the sample gearbox input parameter and the sample gearbox output parameter; The sample gearbox input parameter is determined as the target sample gearbox input parameter.

In an embodiment of the present application, the determining unit is further configured to: determine a second preset number of sample gearbox input parameters with greater importance as the target sample gearbox input parameters.

In one embodiment of the present application, the diagnosis module is further used to: calculate the residual value between the predicted value of the output parameter of the gearbox and the output parameter of the gearbox; A difference threshold generates the machine learning diagnosis.

In one embodiment of the present application, the diagnosis module is further used to: calculate the mean and standard deviation of the residual value during the training process of the machine learning model; calculate the residual value according to the mean and the standard deviation Difference threshold.

The embodiment of the third aspect of the present application provides a wind turbine, including: the fault diagnosis device for the gearbox of the wind turbine as described in the embodiment of the second aspect of the application.

The wind turbine in the embodiment of the present application can diagnose the operating condition parameters and gearbox input parameters based on the machine learning model and the rule model, and perform comprehensive diagnosis on the machine learning diagnosis result and the rule diagnosis result based on the rule set to obtain the target diagnosis result, thereby On the premise of lower cost, accurate diagnosis of the gearbox will effectively improve the reliability of wind turbine operation and reduce the risk of damage to key components of the unit.

The embodiment of the fourth aspect of the present application proposes an electronic device, including: a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the program, it realizes the The method for diagnosing the fault of the wind turbine gearbox described in the embodiment of the first aspect.

The electronic equipment of the embodiment of the present application, through the processor executing the computer program stored in the memory, can diagnose the working condition parameters and the gearbox input parameters based on the machine learning model and the rule model, and diagnose the machine learning diagnosis results and rules based on the rule set The diagnosis results are comprehensively diagnosed to obtain the target diagnosis results, so that the gearbox can be accurately diagnosed at a lower cost, which will effectively improve the reliability of the wind turbine operation and reduce the risk of damage to key components of the unit.

The embodiment of the fifth aspect of the present application proposes a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the fault of the wind turbine gearbox as described in the embodiment of the first aspect of the present application is realized. diagnosis method.

The computer-readable storage medium of the embodiment of the present application stores computer programs and is executed by a processor, and can diagnose working condition parameters and gearbox input parameters based on machine learning models and rule models, and diagnose machine learning diagnosis results and gearbox input parameters based on rule sets. Based on the comprehensive diagnosis of the rule diagnosis results, the target diagnosis results can be obtained, so that the gearbox can be accurately diagnosed at a lower cost, which will effectively improve the reliability of the wind turbine operation and reduce the risk of damage to key components of the unit.

The embodiment of the sixth aspect of the present application provides a computer program product, the computer program product includes computer program code, when the computer program code is run on the computer, to execute the computer program described in the embodiment of the first aspect of the application Fault diagnosis method of wind turbine gearbox.

The embodiment of the seventh aspect of the present application provides a computer program, the computer program includes computer program code, and when the computer program code is run on the computer, the computer executes the computer program described in the embodiment of the first aspect of the present application. Fault diagnosis method of wind turbine gearbox.

Description of drawings

FIG. 1 is a schematic flowchart of a fault diagnosis method for a wind turbine gearbox according to an embodiment of the present application;

2 is a schematic flow chart of the training process of the machine learning model in the fault diagnosis method for the gearbox of a wind turbine according to an embodiment of the present application;

Fig. 3 is a schematic flow chart of determining input parameters of a target sample gearbox in a fault diagnosis method for a gearbox of a wind turbine according to an embodiment of the present application;

Fig. 4 is a schematic flowchart of abnormal self-diagnosis of a machine learning model in a fault diagnosis method for a gearbox of a wind turbine according to an embodiment of the present application;

Fig. 5 is a schematic flowchart of determining a residual value threshold in a fault diagnosis method for a wind turbine gearbox according to an example of the present application;

6 is a schematic diagram of generating a rule model in a fault diagnosis method for a wind turbine gearbox according to an example of the present application;

FIG. 7 is a schematic flow diagram of generating a rule model based on an operating mechanism in a fault diagnosis method for a wind turbine gearbox according to an example of the present application;

Fig. 8 is a schematic flowchart of generating a rule set in a fault diagnosis method for a gearbox of a wind turbine according to an embodiment of the present application;

FIG. 9 is a schematic diagram of a scene of a fault diagnosis method for a wind turbine gearbox according to an example of the present application;

10 is a schematic structural diagram of a fault diagnosis device for a wind turbine gearbox according to an embodiment of the present application;

Fig. 11 is a schematic structural diagram of a wind turbine according to an embodiment of the present application;

FIG. 12 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

Embodiments of the present application are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary, and are intended to explain the present application, and should not be construed as limiting the present application.

The fault diagnosis method, device, wind turbine, electronic equipment and storage medium of the wind turbine gearbox according to the embodiments of the present application will be described below with reference to the accompanying drawings.

Fig. 1 is a schematic flowchart of a fault diagnosis method for a gearbox of a wind turbine according to an embodiment of the present application.

As shown in Figure 1, the fault diagnosis method of the wind turbine gearbox in the embodiment of the present application includes:

S101, collecting the operating parameters of the wind turbine, the operating parameters include working condition parameters and gearbox parameters.

In the embodiment of the present application, the operating parameters of the wind turbines are collected, for example, the operating parameters of the wind turbines are acquired through a supervisory control and data acquisition (Scada) system.

In some implementations, the operating parameters may include, but not limited to, operating condition parameters, gearbox parameters, and the like. For example, the working condition parameters can be power, speed, wind speed, and pitch angle, etc., and the gearbox parameters can be key characteristic parameters related to gearbox faults, such as gearbox oil temperature, gearbox inlet temperature, and gearbox bearing temperature. The content contained in the operating parameters can be set according to needs, and this application does not make too many limitations.

In some implementations, preprocessing such as data cleaning can be performed on the collected operating parameters, such as eliminating shutdown data, power limit data, sensor abnormal data and power abnormal data, etc., to enhance the accuracy of fault diagnosis.

S102, sequentially using one of the gearbox parameters as a gearbox output parameter, and taking other parameters except one of the gearbox parameters as a gearbox input parameter.

In some embodiments, one of the multiple gearbox parameters collected is sequentially used as a gearbox output parameter, and other parameters among the above-mentioned multiple gearbox parameters except one parameter determined as a gearbox output parameter are used as The input parameters of the gearbox, that is, make each of the multiple collected parameters of the gearbox respectively serve as the output parameters of the gearbox for fault diagnosis in sequence.

For example, the collected gearbox parameters include gearbox inlet temperature, gearbox oil temperature, gearbox bearing driving end temperature, gearbox bearing non-driving end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure. The gearbox inlet temperature is used as the gearbox output parameter, and the gearbox oil temperature, gearbox bearing driving end temperature, gearbox bearing non-driving end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure are used as gearbox input parameters. Again, the gearbox oil temperature is used as the gearbox output parameter, and the gearbox inlet temperature, gearbox bearing driving end temperature, gearbox bearing non-driving end temperature, gearbox inlet oil pressure, and gearbox outlet oil pressure are used as gearbox Input parameters. Diagnosis of each gearbox parameter is realized in turn.

S103. Input the working condition parameters and the input parameters of the gearbox into the machine learning model corresponding to the output parameters of the gearbox to obtain the predicted value of the output parameters of the gearbox.

In the embodiment of the present application, each gearbox output parameter may correspond to a machine learning model as a state feature. For a gearbox output parameter, obtain the machine learning model corresponding to the parameter, and input the working condition parameter obtained in step S101 and the gearbox input parameter corresponding to the gearbox output parameter determined in step S102 into the obtained machine learning model, The predicted values of the output parameters of the gearbox are obtained.

In some implementations, the above process is repeated to complete the prediction of each gearbox output parameter determined in step S102 in order to obtain the corresponding predicted value.

In some implementations, the machine learning model corresponding to the output parameters of the gearbox can be obtained from a hard disk, server or cloud, and can be a pre-trained machine learning model.

S104. Generate a machine learning diagnosis result according to the predicted value of the output parameter of the gearbox and the output parameter of the gearbox.

In the embodiment of the present application, abnormal self-diagnosis is performed according to the output parameters of the gearbox and the predicted values of the corresponding output parameters of the gearbox, and a machine learning diagnosis result is generated.

As a feasible implementation, since the training data used in the machine learning training is the data of the wind turbine in a fault-free state, it can be compared with the residual error of the output parameter of the gearbox to be diagnosed and the corresponding predicted value. Significant differences in residual values are used to generate machine learning diagnostics.

In some implementations, multiple residual value thresholds can be set according to the residual value in the training phase to perform different degrees of abnormal self-diagnosis.

S105, inputting the working condition parameters and the input parameters of the gearbox into the rule model corresponding to the output parameters of the gearbox to obtain a rule diagnosis result.

In the embodiment of the present application, the working condition parameters and the input parameters of the gearbox are input into the rule model corresponding to the output parameters of the gearbox, and the rule diagnosis is performed to obtain the rule diagnosis results, such as: oil temperature exceeds category 1 or oil temperature exceeds the limit Class 2 and other abnormal levels.

In some implementations, the rule diagnostic result output by the rule model can be coded, for example, the first type of overrun is coded as R001, that is, the text is characterized to be stored in the database, saving storage space.

S106. Perform a comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result, and the rule set to obtain a target diagnosis result.

In the embodiment of the present application, the rule set may include rules expressed in the form of If-Then. Different rules represent different machine learning diagnosis results and the corresponding relationship between rule diagnosis results and failure modes. According to machine learning diagnosis results, rule diagnosis results and rule Aggregate for comprehensive diagnosis and get the target diagnosis result. For example:

If: high speed condition, oil temperature over limit category 2, driving end bearing temperature is normal, oil temperature fluctuation is abnormal category 1, oil pressure-oil temperature feasible range is normal,

Then: The cooling system of the gearbox is faulty.

To sum up, according to the fault diagnosis method of the wind turbine gearbox in the embodiment of the present application, the working condition parameters and the input parameters of the gearbox can be diagnosed based on the machine learning model and the rule model, and the machine learning diagnosis result and the rule diagnosis result can be analyzed based on the rule set. Carrying out comprehensive diagnosis and obtaining the target diagnosis results, so as to accurately diagnose the gearbox under the premise of lower cost, will effectively improve the reliability of wind turbine operation and reduce the risk of damage to key components of the unit.

On the basis of the embodiment shown in Figure 1, as shown in Figure 2, the fault diagnosis method of the wind turbine gearbox in the embodiment of the present application includes the training process of the machine learning model, which may specifically include the following steps:

S201. Determine a target sample gearbox input parameter according to the sample gearbox output parameter.

In the embodiment of the present application, the training of the machine learning model is carried out based on the full working condition data, the parameters of the sample gearbox parameters are analyzed except the sample gearbox output parameters, and the target sample gearbox input parameters are determined according to the sample gearbox output parameters , so that the determined target sample gearbox input parameters have a high correlation with the sample gearbox output parameters.

In order to further enhance the accuracy of model training, you can filter all working condition data before model training, for example, you can filter in the following two ways:

Method 1: Select multiple variable parameters from the working condition data, such as power, rotational speed and pitch angle, etc., and perform working condition clustering to obtain data groups under different working conditions, and perform data sampling in each working condition. Multiple sets of full working condition data are obtained as training data for machine learning.

Method 2: The working conditions can be divided into finer granularity according to the power, such as 0 ~ 100kW is a range corresponding to a working condition, sampling is carried out in each power range, and the data under each power range is obtained to form multiple groups of full working conditions training data.

S202, using the sample operating condition parameters and the target sample gearbox input parameters as input, and taking the sample gearbox output parameters as output, to train the machine learning model to be trained to obtain the machine learning model.

In the embodiment of the present application, one model can be used as the machine model to be trained, or a family of models can be selected as the machine model to be trained, and an optimal model can be selected from the family of models.

For example, random forest model, neural network model and Xgboost model can be selected as a family of models, and the training data is divided into three parts: training set, verification set and test set, and the sample working condition parameters and target sample gearbox input parameters are used as Input, with sample gearbox output parameters as output, to train the machine learning model to be trained. For example, train multiple models separately on the training set, and optimize the hyperparameters of each model; use the verification set to evaluate the performance of each model under different hyperparameter combinations, so as to determine the optimal parameters, and then determine the corresponding model of each model. Candidate models; use the test set to compare the performance of various candidate models, and then determine the best model, that is, determine a model in a family of models, such as the Xgboost model.

In some implementations, after the training of the machine learning model is completed, the trained model is persisted, such as saving the model to a hard disk, a storage device of a server, or the cloud, etc., to provide model preparation for fault diagnosis.

On the basis of the above-mentioned embodiments, as shown in FIG. 3 , in step S201 "determining the input parameters of the target sample gearbox according to the output parameters of the sample gearbox" may include the following steps:

S301. Calculating a correlation coefficient between a sample gearbox input parameter and a sample gearbox output parameter.

In the embodiment of the present application, for each sample gearbox output parameter, the correlation coefficient between the remaining sample gearbox parameters (that is, the sample gearbox input parameters) and the sample gearbox output parameters can be calculated respectively, and the sample gearbox input parameters can be compared. The correlation between the parameters and the output parameters of the sample gearbox, where the calculation formula of the correlation coefficient is as follows:

Among them, x represents the sample gearbox output parameters, y represents the sample gearbox input parameters,

represents the mean corresponding to the sample gearbox output parameters,

Indicates the mean value corresponding to the input parameters of the sample gearbox, Cov(x,y) is the covariance corresponding to the output parameters of the sample gearbox and the input parameters of the sample gearbox, σ _x is the standard corresponding to the output parameters of the sample gearbox and the input parameters of the sample gearbox Difference.

S302. Determine a first preset number of sample gearbox input parameters with relatively large correlation coefficients as target sample gearbox input parameters.

In the embodiment of the present application, by calculating the correlation coefficient of each sample gearbox input parameter corresponding to a specific sample gearbox output parameter, the first predetermined number is selected among the multiple sample gearbox input parameters with relatively large correlation coefficients The sample gearbox input parameters of , which are determined as the target sample gearbox input parameters. For example, according to the correlation coefficient, the input parameters of the sample gearbox are sorted according to the correlation, and the first predetermined number of sample gearbox input parameters with larger correlation coefficients are selected from large to small, wherein the first predetermined number can be set according to needs.

In addition, in addition to the method for determining the input parameters of the target sample gearbox shown in FIG. 3 , in the embodiment of the present application, the input parameters of the second preset number of sample gearboxes with greater importance can also be determined as the target gearbox Input parameters.

In some embodiments, a random forest model can be used to construct a machine learning model of a certain sample gearbox output parameter and corresponding multiple sample gearbox input parameters, and by sorting the importance of multiple sample gearbox input parameters, select A second preset number of sample gearbox input parameters with greater importance are determined as target gearbox input parameters. Wherein, the second preset quantity can be set as required.

On the basis of any of the above-mentioned embodiments, as shown in FIG. 4, in step S104, "generating machine learning diagnosis results according to the predicted value of the gearbox output parameters and the gearbox output parameters" may specifically include the following steps:

S401. Calculate the predicted value of the output parameter of the gearbox and the residual value between the output parameter of the gearbox.

In the embodiment of the present application, the predicted value of the output parameter of the gearbox is compared with the real value, and the residual value between the predicted value of the output parameter of the gearbox and the output parameter of the gearbox is calculated. The larger the residual value, the more the predicted result The greater the difference from the actual result.

For example, taking temperature as the current output parameter of the gearbox to be diagnosed as an example, the predicted value and the actual value of temperature can be calculated according to the following formula:

rec=y _i -F(x _i )

Among them, y _i is the real value of the temperature, and F(xi ₎ is the predicted value of the temperature by the machine learning model.

S402. Generate a machine learning diagnosis result according to the residual value and the residual value threshold.

In the embodiment of the present application, the residual value corresponding to the output parameter of the gearbox is compared with the residual value threshold to generate a machine learning diagnosis result.

On the basis of any of the above embodiments, as shown in Figure 5, the "residual value threshold" in step S403 can be obtained according to the following steps:

S501. Calculate the mean and standard deviation of the residual values during the machine learning model training process.

In some implementations, the mean mean and standard deviation σ of the residuals between the gearbox output parameters and the corresponding predicted values during machine learning model training are calculated.

S502. Calculate the residual value threshold according to the mean value and the standard deviation.

In some implementations, the sum of the mean mean and K times the standard deviation σ (mean+kσ) is used as the residual value threshold to perform abnormal self-diagnosis and generate a machine learning diagnosis result. The multiple K can be set according to needs, and is not limited in this application.

On the basis of any of the above embodiments, the embodiment of the present application also includes the generation process of the rule model, as shown in Figure 6, which can be obtained in three ways:

Method 1: Determine the operable domains of multiple gearbox output parameters based on the operating mechanism of the wind turbine to determine diagnostic rules to generate a rule model. As shown in Figure 7, generating a rule model based on the operating mechanism of the wind turbine may specifically include the following steps :

S701. Determine the operable domain of the output parameters of the gearbox.

In the embodiment of the present application, according to the change characteristics corresponding to the output parameters of the gearbox, the operable domain of the output parameters of the gearbox is determined, that is, the data range of the output parameters of the gearbox to ensure normal operation of the gearbox.

S702. Quantify the output parameters of the gearbox beyond the operable domain to obtain a rule model.

In the embodiment of this application, the output parameters of the gearbox beyond the operable range are quantified, and different quantification intervals can be used to characterize various levels of abnormalities in the state of the gearbox, which can be used as the diagnostic rule of the rule model to output parameters for diagnosis.

The second way is to generate a rule model based on expert experience, that is, quantify the output parameters of the gearbox according to the overrun threshold of the output parameters of the gearbox provided by the manufacturer, so as to determine the diagnosis rules and obtain the rule model.

For example, taking temperature as an output parameter of a gearbox, the temperature is quantified according to the temperature overrun threshold provided by the equipment manufacturer. In some implementations, multiple different thresholds can be set according to the overrun threshold provided by the manufacturer, and then the temperature High and low risk quantification, such as:

Wherein, x represents the parameter value of the temperature, and a and b represent different temperature overrun thresholds.

The third way is to generate a rule model based on expert experience. It is also possible to quantify the output parameters of the gearbox according to the degree of dispersion and fluctuation of the output parameters of the gearbox in different periods of time, so as to determine the diagnosis rules and obtain the rule model.

On the basis of any of the above-mentioned embodiments, as shown in FIG. 8, the embodiment of the present application also includes the generation process of the "rule set" in step S106, which may specifically include the following steps:

S801. Obtain a sample machine learning diagnosis result, a sample rule diagnosis result, and a sample target diagnosis result.

In the embodiment of the present application, the accumulated failure case set is used as the sample set, which includes sample operating condition parameters, sample gearbox parameters, and sample target diagnosis results, etc., and the above machine learning diagnosis is performed on the sample set based on the machine learning model to obtain the sample machine Learn the diagnosis result; perform the above rule diagnosis on the sample set based on the rule model, and obtain the sample rule diagnosis result.

S802. Determine sample machine learning diagnosis results and sample rule diagnosis results corresponding to different sample target diagnosis results based on an association rule analysis method.

In the embodiment of this application, there are many types of sample machine learning diagnosis results and sample rule diagnosis results obtained on the fault case set. For example, the sample machine learning diagnosis results can include oil temperature residual exceeding the limit and inlet temperature residual exceeding the limit. category, driving end or non-driving end bearing temperature residual overrun category, etc.; sample rule diagnosis results can include oil temperature overrun category, bearing temperature overrun category, oil pressure-oil temperature feasible range category, etc., which need to be based on different types The association between the sample machine learning diagnosis results and sample rule diagnosis results and different sample target diagnosis results determines the gearbox failure modes corresponding to different types of sample machine learning diagnosis results and sample rule diagnosis results.

In some embodiments, the sample machine learning diagnosis results and the sample rule diagnosis results can be uniformly identified as an item set X, the sample target diagnosis results in the failure case set can be represented as an item set Y, and X and Y can be identified based on the association rule analysis method. Association to determine the item set X' corresponding to different sample target results, where the item set X' includes sample machine learning diagnosis results and sample rule diagnosis results corresponding to different sample target results.

S803. Perform verification and testing on sample machine learning diagnosis results and sample rule diagnosis results corresponding to different sample target diagnosis results to obtain a rule set.

In the embodiment of the present application, domain experts verify the item set X′ including the sample machine learning diagnosis results and sample rule diagnosis results corresponding to different sample target diagnosis results to ensure the rationality of the generated rules. After passing the rationality verification , the corresponding relationship between different sample target diagnosis results and different sample machine learning diagnosis results and sample rule diagnosis results in item set X′ and item set Y can be expressed in the form of if X′ then Y.

In some implementations, if X' then Y is used as a candidate rule for accuracy testing, for example, verifying whether the accuracy rate of the candidate rule in the failure case set is stable, and testing the false positive rate of the candidate rule in the healthy data set. In addition, in order to ensure that any failure mode in the itemset Y has one and only one rule corresponding to it, the candidate rules and the existing rules can also be redundantly processed to obtain a rule set.

In order for those skilled in the art to understand this application more clearly, Fig. 9 is a schematic diagram of a scenario of a fault diagnosis method for a wind turbine gearbox according to an example of this application. As shown in Fig. 9, the fault diagnosis method may include an offline part and an online part , where the offline part is mainly used to extract knowledge from the fault case set through machine learning diagnosis and rule diagnosis, analyze the machine learning diagnosis results and rule diagnosis results based on expert experience, and generate a rule set; the online part is mainly used for rule set-based, Working condition parameters and gearbox parameters for gearbox fault diagnosis: input working condition parameters and gearbox parameters into machine learning model and rule model respectively, output machine learning diagnosis results and rule diagnosis results, and analyze machine learning diagnosis results and Based on the diagnosis results of the rules, comprehensive diagnosis is carried out to obtain the target diagnosis results.

In order to realize the above embodiments, the present application also proposes a fault diagnosis device for a gearbox of a wind turbine.

Fig. 10 is a schematic structural diagram of a fault diagnosis device for a gearbox of a wind turbine according to an embodiment of the present application.

As shown in FIG. 10 , the fault diagnosis device 1000 for the wind turbine gearbox of the embodiment of the present application includes: a data acquisition module 1001 , a determination module 1002 , a prediction module 1003 , a diagnosis module 1004 , a rule diagnosis module 1005 and a comprehensive diagnosis module 1006 .

The data collection module 1001 is used to collect the operating parameters of the wind turbine, and the operating parameters include working condition parameters and gearbox parameters;

The determining module 1002 is configured to sequentially use one of the gearbox parameters as a gearbox output parameter, and use other parameters except one of the gearbox parameters as gearbox input parameters.

The prediction module 1003 is configured to input the working condition parameters and the input parameters of the gearbox into the machine learning model corresponding to the output parameters of the gearbox to obtain the predicted value of the output parameters of the gearbox.

The diagnosis module 1004 is configured to generate a machine learning diagnosis result according to the predicted value of the output parameter of the gearbox and the output parameter of the gearbox.

The rule diagnosis module 1005 is used to input the operating condition parameters and the gearbox input parameters into the rule model corresponding to the gearbox output parameters to obtain the rule diagnosis result.

The comprehensive diagnosis module 1006 is configured to perform comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result and the rule set, and obtain the target diagnosis result.

In an embodiment of the present application, the device further includes: a rule generation module, and the rule generation module is used to: obtain sample machine learning diagnosis results, sample rule diagnosis results, and sample target diagnosis results; The sample machine learning diagnosis results and sample rule diagnosis results corresponding to the diagnosis results; the checksum test is performed on the sample machine learning diagnosis results and sample rule diagnosis results corresponding to different sample target diagnosis results to obtain a rule set.

In the embodiment of the present application, the rule diagnosis module 1005 is further configured to: code the result of the rule diagnosis.

In the embodiment of the present application, the device further includes: a quantization module, and the training module is used to: determine the operable range of the gearbox output parameters; quantify the gearbox output parameters beyond the operable range to obtain a rule model.

In the embodiment of the present application, the quantization module is further configured to: quantify the output parameters of the gearbox according to the overrun threshold of the output parameters of the gearbox provided by the manufacturer to obtain a rule model.

In the embodiment of the present application, the quantization module is further used to: quantify the output parameters of the gearbox according to the degree of dispersion and the fluctuation state of the output parameters of the gearbox at different time periods to obtain a rule model.

In the embodiment of the present application, the device further includes: a training module, and the training module includes: a determination unit, used to determine the input parameters of the target sample gearbox according to the output parameters of the sample gearbox; a training unit, used to combine the sample working condition parameters and the target The input parameters of the sample gearbox are taken as input, and the output parameters of the sample gearbox are taken as output, and the machine learning model to be trained is trained to obtain the machine learning model.

In the embodiment of the present application, the determination unit is further used to: calculate the correlation coefficient between the sample gearbox input parameters and the sample gearbox output parameters; determine the first preset number of sample gearbox input parameters with relatively large correlation coefficients as Target sample gearbox input parameters.

In the embodiment of the present application, the determining unit is further configured to: determine a second preset number of sample gearbox input parameters with greater importance as target sample gearbox input parameters.

In the embodiment of the present application, the diagnosis module 1004 is also used to: calculate the predicted value of the gearbox output parameter and the residual value between the gearbox output parameter; generate a machine learning diagnosis result according to the residual value and the residual value threshold .

In the embodiment of the present application, the diagnostic module 1004 is further configured to: calculate the mean and standard deviation of the residual value during the machine learning model training process; calculate the residual value threshold according to the mean and standard deviation.

It should be noted that for details not disclosed in the fault diagnosis device for wind turbine gearboxes in the embodiments of the present application, please refer to the details disclosed in the fault diagnosis method for wind turbine gearboxes in the embodiments of the present application, and will not be repeated here.

To sum up, the fault diagnosis device for the wind turbine gearbox of the embodiment of the present application can diagnose the working condition parameters and the input parameters of the gearbox based on the machine learning model and the rule model, and can diagnose the machine learning diagnosis results and the rule diagnosis results based on the rule set. Comprehensive diagnosis can obtain the target diagnosis results, so that the gearbox can be accurately diagnosed at a lower cost, which will effectively improve the reliability of wind turbine operation and reduce the risk of damage to key components of the unit.

In order to realize the above embodiments, the present application also proposes a wind turbine.

Fig. 11 is a schematic structural diagram of a wind turbine according to an embodiment of the present application.

As shown in FIG. 11 , a wind turbine 1100 according to an embodiment of the present application includes the above-mentioned fault diagnosis device 1000 for a gearbox of a wind turbine.

In order to implement the above embodiments, as shown in FIG. 12 , the embodiment of the present application proposes an electronic device 1200, including: a memory 1201, a processor 1202, and a computer program stored in the memory 1201 and operable on the processor 1202, When the processor 1202 executes the program, the above-mentioned fault diagnosis method for the gearbox of the wind turbine is implemented.

In order to realize the above-mentioned embodiments, the embodiment of the present application proposes a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned method for fault diagnosis of a gearbox of a wind turbine is realized.

In order to realize the above-mentioned embodiments, the embodiment of the present application proposes a computer program product, the computer program product includes computer program code, when the computer program code is run on the computer, to execute the above-mentioned wind turbine gearbox Fault diagnosis method.

In order to realize the above-mentioned embodiments, the embodiment of the present application proposes a computer program, the computer program includes computer program code, when the computer program code is run on the computer, so that the computer executes the above-mentioned failure of the wind turbine gearbox diagnosis method.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, a feature defined as "first" and "second" may explicitly or implicitly include one or more of these features. In the description of the present application, "plurality" means two or more, unless otherwise specifically defined.

In this application, terms such as "installation", "connection", "connection" and "fixation" should be interpreted in a broad sense, for example, it can be a fixed connection or a detachable connection, unless otherwise clearly specified and limited. , or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediary, and it can be the internal communication of two components or the interaction relationship between two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in this application according to specific situations.

In the description of this specification, descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the described specific features, structures, materials or characteristics may be combined in any suitable manner in any one or more embodiments or examples.

In addition, those skilled in the art can combine and combine different embodiments or examples and features of different embodiments or examples described in this specification without conflicting with each other.

Claims

A fault diagnosis method for a wind turbine gearbox, characterized in that it includes:

Collecting the operating parameters of the wind turbine, the operating parameters include working condition parameters and gearbox parameters;

Taking one of the gearbox parameters as a gearbox output parameter in turn, and using other parameters in the gearbox parameters except the one parameter as a gearbox input parameter;

Inputting the working condition parameters and the gearbox input parameters into a machine learning model corresponding to the gearbox output parameters to obtain a predicted value of the gearbox output parameters;

generating a machine learning diagnosis result based on the predicted value of the gearbox output parameter and the gearbox output parameter;

inputting the operating condition parameters and the gearbox input parameters into a rule model corresponding to the gearbox output parameters to obtain a rule diagnosis result;

A comprehensive diagnosis is performed according to the machine learning diagnosis result, the rule diagnosis result and the rule set to obtain a target diagnosis result.
The fault diagnosis method according to claim 1, further comprising:

Get sample machine learning diagnosis results, sample rule diagnosis results and sample target diagnosis results;

determining the sample machine learning diagnostic results and the sample rule diagnostic results corresponding to different sample target diagnostic results based on an association rule analysis method;

Perform a checksum test on the sample machine learning diagnosis results and the sample rule diagnosis results corresponding to different sample target diagnosis results to obtain the rule set.
The fault diagnosis method according to claim 1 or 2, further comprising:

The rule diagnosis result is encoded.
The fault diagnosis method according to any one of claims 1 to 3, further comprising:

determining an operational domain of said gearbox output parameters;

The output parameters of the gearbox beyond the operable domain are quantified to obtain the rule model.
The fault diagnosis method according to any one of claims 1 to 4, further comprising:

The output parameter of the gearbox is quantified according to the overrun threshold of the output parameter of the gearbox provided by the manufacturer to obtain the rule model.
The fault diagnosis method according to any one of claims 1 to 5, further comprising:

According to the degree of dispersion and fluctuation state of the output parameters of the gearbox at different time periods, the output parameters of the gearbox are quantified to obtain the rule model.
The fault diagnosis method according to any one of claims 1 to 6, further comprising:

determining a target sample gearbox input parameter based on the sample gearbox output parameter;

Using the sample operating condition parameters and the input parameters of the target sample gearbox as input, and the output parameters of the sample gearbox as output, the machine learning model to be trained is trained to obtain the machine learning model.
The fault diagnosis method according to claim 7, further comprising:

calculating a correlation coefficient between a sample gearbox input parameter and said sample gearbox output parameter;

A first preset number of the sample gearbox input parameters with relatively large correlation coefficients are determined as the target sample gearbox input parameters.
The fault diagnosis method according to claim 7, further comprising:

A second preset number of sample gearbox input parameters with greater importance is determined as the target sample gearbox input parameters.
The fault diagnosis method according to any one of claims 1 to 9, wherein the generating a machine learning diagnosis result according to the predicted value of the output parameter of the gearbox and the output parameter of the gearbox includes:

calculating a residual value between the predicted value of the gearbox output parameter and the gearbox output parameter;

The machine learning diagnosis result is generated according to the residual value and the residual value threshold.
The fault diagnosis method according to any one of claims 1 to 10, further comprising:

Calculate the mean and standard deviation of the residual values in the machine learning model training process;

The residual value threshold is calculated according to the mean value and the standard deviation.
A fault diagnosis device for a wind turbine gearbox, characterized in that it includes:

The data collection module is used to collect the operating parameters of the wind turbine, and the operating parameters include working condition parameters and gearbox parameters;

A determination module, configured to sequentially use one of the gearbox parameters as a gearbox output parameter, and use other parameters of the gearbox parameters except the one parameter as a gearbox input parameter;

A prediction module, configured to input the working condition parameters and the input parameters of the gearbox into a machine learning model corresponding to the output parameters of the gearbox to obtain a predicted value of the output parameters of the gearbox;

A diagnosis module, configured to generate a machine learning diagnosis result according to the predicted value of the gearbox output parameter and the gearbox output parameter;

A rule diagnosis module, configured to input the working condition parameters and the gearbox input parameters into a rule model corresponding to the gearbox output parameters, to obtain a rule diagnosis result;

A comprehensive diagnosis module, configured to perform comprehensive diagnosis according to the machine learning diagnosis result, the rule diagnosis result and the rule set, to obtain a target diagnosis result.
A wind turbine, characterized by comprising: the fault diagnosis device for a gearbox of a wind turbine according to claim 12.
An electronic device, characterized in that it comprises: a memory, a processor, and a computer program stored in the memory and operable on the processor, when the processor executes the program, it realizes any one of claims 1-11. The fault diagnosis method of the wind turbine gearbox described in the item.
A computer-readable storage medium, on which a computer program is stored, characterized in that, when the program is executed by a processor, the fault diagnosis method for a gearbox of a wind turbine set according to any one of claims 1-11 is implemented.
A computer program product, characterized in that the computer program product includes computer program code, and when the computer program code is run on a computer, the wind turbine generator according to any one of claims 1-11 can be executed Gearbox fault diagnosis method.
A computer program, characterized in that the computer program includes computer program code, when the computer program code is run on the computer, so that the computer executes the wind turbine gear according to any one of claims 1-11 Box fault diagnosis method.