CN114118213A - VaDE-based wind turbine generator bearing fault diagnosis method and system - Google Patents

Abstract

The invention relates to a wind driven generator bearing fault diagnosis method and system based on VaDE, wherein the method comprises the following steps: acquiring state data of a wind power generator bearing at each moment in a prediction time period; inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model, and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period; and diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period. According to the technical scheme provided by the invention, whether the bearing fails or not is diagnosed through the bearing fault diagnosis model of the wind driven generator, so that the accuracy of bearing fault diagnosis of the wind driven generator can be improved.

Description

VaDE-based wind turbine generator bearing fault diagnosis method and system

Technical Field

The invention relates to the field of bearing faults, in particular to a method and a system for diagnosing a bearing fault of a wind driven generator based on VaDE.

Background

At present, the bearing fault of a wind power generator in the wind power industry is a common phenomenon, and the existing bearing fault monitoring system can only monitor the vibration, temperature and other parameters of a bearing, and still needs experienced technical personnel to carry out manual judgment on the analysis and judgment of monitoring data. Therefore, the existing bearing fault diagnosis system has little significance for prediction and real-time diagnosis of the generator bearing fault, and still provides original data for parameter monitoring of the generator bearing and reason analysis after the fault does not occur.

Therefore, the existing wind turbine bearing fault diagnosis obviously still has inconvenience and defects, and needs to be further improved. How to create a convenient and accurate wind turbine bearing fault diagnosis method and system, so that the method and system can accurately diagnose the fault of the wind turbine bearing, improve the safety of the wind turbine, and become an urgent need for improvement in the current industry.

Disclosure of Invention

The application provides a method and a system for diagnosing faults of a wind turbine bearing based on VaDE (dynamic equivalent error rate), which are used for at least solving the technical problem of low accuracy of a method for diagnosing whether the wind turbine bearing is in fault in the related technology.

An embodiment of the first aspect of the present application provides a VaDE-based wind turbine bearing fault diagnosis method, including:

acquiring state data of a wind power generator bearing at each moment in a prediction time period;

inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model, and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period;

and diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

An embodiment of the second aspect of the present application provides a VaDE-based wind turbine bearing fault diagnosis system, which includes:

the first acquisition module is used for acquiring the state data of the wind turbine bearing at each moment in a prediction time interval;

the second acquisition module is used for inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period;

and the diagnosis module is used for diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

The technical scheme provided by the embodiment of the application at least has the following beneficial effects:

according to the method and the system for diagnosing the fault of the wind driven generator bearing based on the VaDE, provided by the invention, the state data of the wind driven generator bearing at each moment in a prediction time interval is firstly obtained, then the state data of the wind driven generator bearing at each moment in the prediction time interval is input into a pre-established wind driven generator bearing fault diagnosis model, the weight of Gaussian distribution corresponding to the wind driven generator bearing in the prediction time interval is obtained, and finally, whether the wind driven generator bearing has the fault in the prediction time interval is diagnosed based on the weight of the Gaussian distribution corresponding to the wind driven generator bearing in the prediction time interval. According to the technical scheme provided by the invention, whether the bearing fails or not is diagnosed through the bearing fault diagnosis model of the wind driven generator, so that the accuracy of bearing fault diagnosis of the wind driven generator can be improved.

Additional aspects and advantages of the present application will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the present application.

Drawings

The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 is a flow chart of a VaDE-based wind turbine bearing fault diagnosis method provided according to an embodiment of the present application;

FIG. 2 is a detailed flow chart of a VaDE-based wind turbine bearing fault diagnosis method provided according to an embodiment of the present application;

FIG. 3 is a block diagram of a VaDE-based wind turbine bearing fault diagnosis system provided in accordance with an embodiment of the present application.

Detailed Description

Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining the present application and should not be construed as limiting the present application.

According to the method and the system for diagnosing the fault of the wind driven generator bearing based on the VaDE, firstly, state data of the wind driven generator bearing at each moment in a prediction time period are obtained, then the state data of the wind driven generator bearing at each moment in the prediction time period are input into a wind driven generator bearing fault diagnosis model which is established in advance, the weight of Gaussian distribution corresponding to the wind driven generator bearing in the prediction time period is obtained, and finally, whether the wind driven generator bearing has the fault in the prediction time period is diagnosed based on the weight of the Gaussian distribution corresponding to the wind driven generator bearing in the prediction time period. According to the technical scheme provided by the invention, whether the bearing fails or not is diagnosed through the bearing fault diagnosis model of the wind driven generator, so that the accuracy of bearing fault diagnosis of the wind driven generator can be improved.

Example 1

Fig. 1 is a flowchart of a VaDE-based wind turbine bearing fault diagnosis method provided in an embodiment of the present disclosure, and as shown in fig. 1, the method includes:

step 1: and acquiring the state data of the wind turbine bearing at each moment in the prediction time period.

The state data of the wind turbine bearing includes:

the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of a bearing of the wind driven generator and the second-level characteristic value of the vibration acceleration of the bearing;

wherein, bearing vibration speed second level eigenvalue includes: the effective value, the maximum value, the minimum value and the average value of the vibration speed of the bearing;

the bearing vibration acceleration second-level characteristic value comprises: effective value, maximum value, minimum value, mean value, peak value, kurtosis index and skewness index of bearing vibration acceleration.

Step 2: and inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model, and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

In an embodiment of the present disclosure, the process of establishing the pre-established wind turbine bearing fault diagnosis model includes:

acquiring state data of a wind power generator bearing at each moment in a historical period;

processing the acquired state data of the wind driven generator bearing at each moment in the historical time period, and acquiring the processed state data;

training an initial wind turbine bearing fault diagnosis model which is constructed in advance based on a VaDE model by using the processed state data to obtain a wind turbine bearing fault diagnosis model;

wherein the initial wind turbine bearing fault diagnosis model comprises: a first VaDE module, a residual calculation layer, and a second VaDE module.

It should be noted that the processing of the acquired state data of the wind turbine bearing at each time in the historical period includes:

step a: splicing the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the second-level characteristic value of the rotating speed of the wind driven generator and the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in the state data of the wind driven generator bearing at each moment in the historical period at the same moment, and splicing the data;

b, setting the spliced data with the time data when the output power value of the wind driven generator is smaller than the output power threshold value as a null value;

step c: normalizing the data processed in the step b;

step d: and carrying out section division on the state data normalized in the historical time period according to the rotating speed of the wind driven generator, and acquiring the maximum value, the minimum value, the average value and the standard deviation of the temperature difference between the driving end and the non-driving end, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in each section division.

It should be noted that, the process of constructing the initial wind turbine bearing fault diagnosis model includes:

constructing a first VaDE module, establishing a residual error calculation layer and constructing a second VaDE module;

the construction process of the first VaDE module includes:

step A, acquiring the number of Gaussian mixture models corresponding to a first VaDE module, inputting state data processed at each moment in a historical period into a first deep neural network of the first VaDE module consisting of four fully-connected layers for dimension reduction and feature extraction, and outputting the mean value mu of each distribution in Gaussian mixture distribution generating latent variables by a fourth fully-connected layer of the first deep neural network₁Sum standard deviation squared log σ₁ ²；

B, average value mu of each distribution₁Sum standard deviation squared log σ₁ ²Carrying out latent variable sampling transformation, wherein the transformed data is represented by Z1;

and step C, inputting the transformed data Z1 into a second deep neural network of a first VaDE module consisting of four fully-connected layers for dimension increasing and outputting predicted state data at each moment in a historical time period to obtain parameters of the first VaDE module, thereby obtaining the constructed first VaDE module.

Wherein the expression of the transformed data Z1 is

ε₁The data is randomly sampled from a standard normal distribution corresponding to the first VaDE module;

the second VaDE module construction process, comprising:

step I1: obtaining a residual error norm of the state data output by the first VaDE module and the state data input by the first VaDE module, which are obtained by a residual error calculation layer, determining the number of Gaussian mixture models based on preset wind turbine generator bearings by frequent classification, inputting the residual error norm into a first deep neural network of a second VaDE module consisting of two fully-connected layers, and outputting a mean value mu of each distribution in Gaussian mixture distribution of latent variables generated by the second fully-connected layer of the first deep neural network₂Logarithm of sum standard deviation squared

Step I2 mean value μ for each distribution₂Sum standard deviation squared log σ₂ ²Carrying out latent variable sampling transformation, wherein the transformed data is represented by Z2;

step I3, inputting the transformed data Z2 into a second deep neural network of a second VaDE module formed by two layers of full connection layers to obtain parameters of the second VaDE module, and further obtaining the constructed second VaDE module;

step I4: recording the Gaussian distribution corresponding to the maximum mean value in the Gaussian mixture distribution of the step I1 as alpha distribution, and recording the distribution corresponding to the maximum standard deviation as beta distribution;

wherein the expression of the transformed data Z2 is

ε₂Is randomly sampled from a corresponding standard normal distribution of the second VaDE module.

The residual calculation layer is configured to calculate a sum of absolute values of differences between the state data output by the first VaDE module and the state data input by the first VaDE module, and record the sum of the absolute values as a residual-norm.

And step 3: and diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

In an embodiment of the present disclosure, the step 3 specifically includes:

selecting Gaussian distribution with the highest weight corresponding to a wind power generator bearing;

and judging whether the selected Gaussian distribution is alpha distribution or beta distribution, if so, diagnosing that the bearing of the wind driven generator is abnormal, and otherwise, judging that the bearing of the wind driven generator is normal.

The specific method of the present application is exemplified by combining the wind turbine generator bearing fault diagnosis method:

as shown in fig. 2, a specific flowchart of a wind turbine bearing fault diagnosis method is provided, where the method includes:

f1, acquiring the state data of the wind driven generator bearing at each moment in the historical period;

step F2, processing the acquired state data of the wind driven generator bearing at each moment in the historical period, and acquiring the processed state data;

f3, constructing an initial wind turbine bearing fault diagnosis model;

step F4, training the initial wind turbine bearing fault diagnosis model by using the processed state data to obtain a wind turbine bearing fault diagnosis model;

step F5: inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model, obtaining the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period, selecting the Gaussian distribution with the highest weight corresponding to the wind turbine bearing, then judging whether the selected Gaussian distribution is alpha distribution or beta distribution, if so, diagnosing that the wind turbine bearing is abnormal, otherwise, diagnosing that the wind turbine bearing is normal.

In summary, in the wind turbine bearing fault diagnosis method based on VaDE provided in the embodiment of the present disclosure, whether the bearing has a fault is diagnosed through the wind turbine bearing fault diagnosis model, so that the accuracy of the wind turbine bearing fault diagnosis can be improved.

Example 2

FIG. 3 is a block diagram of a VaDE-based wind turbine bearing fault diagnosis system according to an embodiment of the present disclosure, as shown in FIG. 3, the system includes:

the first acquisition module is used for acquiring the state data of the wind turbine bearing at each moment in a prediction time interval;

the second acquisition module is used for inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period;

and the diagnosis module is used for diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

In an embodiment of the disclosure, the state data of the wind turbine bearing comprises:

the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of a bearing of the wind driven generator and the second-level characteristic value of the vibration acceleration of the bearing;

wherein, bearing vibration speed second level eigenvalue includes: the effective value, the maximum value, the minimum value and the average value of the vibration speed of the bearing;

the bearing vibration acceleration second-level characteristic value comprises: effective value, maximum value, minimum value, mean value, peak value, kurtosis index and skewness index of bearing vibration acceleration.

Further, the process of establishing the pre-established wind turbine bearing fault diagnosis model comprises the following steps:

acquiring state data of a wind power generator bearing at each moment in a historical period;

processing the acquired state data of the wind driven generator bearing at each moment in the historical time period, and acquiring the processed state data;

training an initial wind turbine bearing fault diagnosis model which is constructed in advance based on a VaDE model by using the processed state data to obtain a wind turbine bearing fault diagnosis model;

wherein the initial wind turbine bearing fault diagnosis model comprises: a first VaDE module, a residual calculation layer, and a second VaDE module.

Specifically, the processing of the acquired state data of the wind turbine bearing at each time in the historical period includes:

step Q1: splicing the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the second-level characteristic value of the rotating speed of the wind driven generator and the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in the state data of the wind driven generator bearing at each moment in the historical period at the same moment, and splicing the data;

step Q2, setting the spliced data with the time data when the output power value of the wind driven generator is smaller than the output power threshold value as a null value;

step Q3: normalizing the data processed in the step Q2;

step Q4: and carrying out section division on the state data normalized in the historical time period according to the rotating speed of the wind driven generator, and acquiring the maximum value, the minimum value, the average value and the standard deviation of the temperature difference between the driving end and the non-driving end, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in each section division.

Further, the process of constructing the initial wind turbine bearing fault diagnosis model includes:

constructing a first VaDE module, establishing a residual error calculation layer and constructing a second VaDE module;

the construction process of the first VaDE module includes:

y1, obtaining the number of Gaussian mixture models corresponding to the first VaDE module, and inputting the state data processed at each moment in the historical period into the first VaDE module consisting of four full-connection layersThe fourth full-connection layer of the first deep neural network outputs the mean value mu of each distribution in the Gaussian mixture distribution of the generated latent variable₁Sum standard deviation squared log σ₁ ²；

Y2 mean value of each distribution₁Logarithm of sum standard deviation squared log σ 1²Carrying out latent variable sampling transformation, wherein the transformed data is represented by Z1;

and Y3, inputting the transformed data Z1 into a second deep neural network of a first VaDE module consisting of four fully-connected layers for dimension increasing and outputting predicted state data at each moment in a historical time period to obtain parameters of the first VaDE module, thereby obtaining the constructed first VaDE module.

Wherein the expression of the transformed data Z1 is

ε₁The data is randomly sampled from a standard normal distribution corresponding to the first VaDE module;

the second VaDE module construction process, comprising:

step R1: obtaining a residual error norm of the state data output by the first VaDE module and the state data input by the first VaDE module, which are obtained by a residual error calculation layer, determining the number of Gaussian mixture models based on preset wind turbine generator bearings by frequent classification, inputting the residual error norm into a first deep neural network of a second VaDE module consisting of two fully-connected layers, and outputting a mean value mu of each distribution in Gaussian mixture distribution of latent variables generated by the second fully-connected layer of the first deep neural network₂Logarithm of sum standard deviation squared

Step R2 mean value μ for each distribution₂Logarithm of sum standard deviation squared

To perform a latent transformationQuantity sampling transformation, and the transformed data is represented by Z2;

step R3, inputting the transformed data Z2 into a second deep neural network of a second VaDE module formed by two layers of full connection layers to obtain parameters of the second VaDE module, and further obtaining the constructed second VaDE module;

step R4: recording the Gaussian distribution corresponding to the maximum mean value in the Gaussian mixture distribution of the step I1 as alpha distribution, and recording the distribution corresponding to the maximum standard deviation as beta distribution;

wherein the expression of the transformed data Z2 is

ε₂Is randomly sampled from a corresponding standard normal distribution of the second VaDE module.

In an embodiment of the present disclosure, the diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the gaussian distribution corresponding to the wind turbine bearing in the prediction time period includes:

selecting Gaussian distribution with the highest weight corresponding to a wind power generator bearing;

and judging whether the selected Gaussian distribution is alpha distribution or beta distribution, if so, diagnosing that the bearing of the wind driven generator is abnormal, and otherwise, judging that the bearing of the wind driven generator is normal.

In summary, in the wind turbine bearing fault diagnosis system based on VaDE provided in the embodiment of the present disclosure, whether the bearing is faulty or not is diagnosed through the wind turbine bearing fault diagnosis model, so that the accuracy of diagnosing the wind turbine bearing fault can be improved.

In the description herein, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.

Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing steps of a custom logic function or process, and alternate implementations are included within the scope of the preferred embodiment of the present application in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present application.

Although embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application, and that variations, modifications, substitutions and alterations may be made to the above embodiments by those of ordinary skill in the art within the scope of the present application.

Claims (10)

1. A VaDE-based wind turbine bearing fault diagnosis method is characterized by comprising the following steps:

acquiring state data of a wind power generator bearing at each moment in a prediction time period;

inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model, and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period;

and diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

2. The method of claim 1, wherein the wind turbine bearing condition data comprises:

the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of a bearing of the wind driven generator and the second-level characteristic value of the vibration acceleration of the bearing;

wherein, bearing vibration speed second level eigenvalue includes: the effective value, the maximum value, the minimum value and the average value of the vibration speed of the bearing;

the bearing vibration acceleration second-level characteristic value comprises: effective value, maximum value, minimum value, mean value, peak value, kurtosis index and skewness index of bearing vibration acceleration.

3. The method of claim 2, wherein the pre-established wind turbine bearing fault diagnosis model is established by:

acquiring state data of a wind power generator bearing at each moment in a historical period;

processing the acquired state data of the wind driven generator bearing at each moment in the historical time period, and acquiring the processed state data;

training an initial wind turbine bearing fault diagnosis model which is constructed in advance based on a VaDE model by using the processed state data to obtain a wind turbine bearing fault diagnosis model;

wherein the initial wind turbine bearing fault diagnosis model comprises: a first VaDE module, a residual calculation layer, and a second VaDE module.

4. The method according to claim 3, wherein the processing of the acquired state data of the wind turbine bearing at each time within the historical period comprises:

step a: splicing the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the second-level characteristic value of the rotating speed of the wind driven generator and the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in the state data of the wind driven generator bearing at each moment in the historical period at the same moment, and splicing the data;

b, setting the spliced data with the time data when the output power value of the wind driven generator is smaller than the output power threshold value as a null value;

step c: normalizing the data processed in the step b;

step d: and carrying out section division on the state data normalized in the historical time period according to the rotating speed of the wind driven generator, and acquiring the maximum value, the minimum value, the average value and the standard deviation of the temperature difference between the driving end and the non-driving end, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in each section division.

5. The method of claim 3, wherein the initial wind turbine bearing fault diagnosis model is constructed by a process comprising:

constructing a first VaDE module, establishing a residual error calculation layer and constructing a second VaDE module;

the construction process of the first VaDE module includes:

step A, acquiring the number of Gaussian mixture models corresponding to a first VaDE module, inputting state data processed at each moment in a historical period into a first deep neural network of the first VaDE module consisting of four fully-connected layers for dimension reduction and feature extraction, and outputting the mean value mu of each distribution in Gaussian mixture distribution generating latent variables by a fourth fully-connected layer of the first deep neural network₁Logarithm of sum standard deviation squared

B, average value mu of each distribution₁Logarithm of sum standard deviation squared

Carrying out latent variable sampling transformation, wherein the transformed data is represented by Z1;

and step C, inputting the transformed data Z1 into a second deep neural network of a first VaDE module consisting of four fully-connected layers for dimension increasing and outputting predicted state data at each moment in a historical time period to obtain parameters of the first VaDE module, thereby obtaining the constructed first VaDE module.

Wherein the expression of the transformed data Z1 is

ε₁The data is randomly sampled from a standard normal distribution corresponding to the first VaDE module;

the second VaDE module construction process, comprising:

step I1: obtaining a residual error norm of the state data output by the first VaDE module and the state data input by the first VaDE module, which are obtained by a residual error calculation layer, determining the number of Gaussian mixture models based on preset wind turbine generator bearings by frequent classification, inputting the residual error norm into a first deep neural network of a second VaDE module consisting of two fully-connected layers, and outputting a mean value mu of each distribution in Gaussian mixture distribution of latent variables generated by the second fully-connected layer of the first deep neural network₂Logarithm of sum standard deviation squared

Step I2 mean value μ for each distribution₂Logarithm of sum standard deviation squared

Carrying out latent variable sampling transformation, wherein the transformed data is represented by Z2;

step I3, inputting the transformed data Z2 into a second deep neural network of a second VaDE module formed by two layers of full connection layers to obtain parameters of the second VaDE module, and further obtaining the constructed second VaDE module;

step I4: recording the Gaussian distribution corresponding to the maximum mean value in the Gaussian mixture distribution of the step I1 as alpha distribution, and recording the distribution corresponding to the maximum standard deviation as beta distribution;

wherein the expression of the transformed data Z2 is

ε₂Is randomly sampled from a corresponding standard normal distribution of the second VaDE module.

6. The method of claim 5, wherein diagnosing whether the wind turbine bearing has failed during the prediction period based on the corresponding Gaussian distributed weights of the wind turbine bearing during the prediction period comprises:

selecting Gaussian distribution with the highest weight corresponding to a wind power generator bearing;

and judging whether the selected Gaussian distribution is alpha distribution or beta distribution, if so, diagnosing that the bearing of the wind driven generator is abnormal, and otherwise, judging that the bearing of the wind driven generator is normal.

7. A VaDE-based wind turbine bearing fault diagnosis system, characterized in that the system comprises:

the first acquisition module is used for acquiring the state data of the wind turbine bearing at each moment in a prediction time interval;

the second acquisition module is used for inputting the state data of the wind turbine bearing at each moment in the prediction time period into a pre-established wind turbine bearing fault diagnosis model and acquiring the weight of Gaussian distribution corresponding to the wind turbine bearing in the prediction time period;

and the diagnosis module is used for diagnosing whether the wind turbine bearing has a fault in the prediction time period based on the weight of the Gaussian distribution corresponding to the wind turbine bearing in the prediction time period.

8. The system of claim 7, wherein the wind turbine bearing status data comprises:

the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of a bearing of the wind driven generator and the second-level characteristic value of the vibration acceleration of the bearing;

wherein, bearing vibration speed second level eigenvalue includes: the effective value, the maximum value, the minimum value and the average value of the vibration speed of the bearing;

the bearing vibration acceleration second-level characteristic value comprises: effective value, maximum value, minimum value, mean value, peak value, kurtosis index and skewness index of bearing vibration acceleration.

9. The system of claim 8, wherein the pre-established wind turbine bearing fault diagnosis model is established by:

acquiring state data of a wind power generator bearing at each moment in a historical period;

processing the acquired state data of the wind driven generator bearing at each moment in the historical time period, and acquiring the processed state data;

training an initial wind turbine bearing fault diagnosis model which is constructed in advance based on a VaDE model by using the processed state data to obtain a wind turbine bearing fault diagnosis model;

wherein the initial wind turbine bearing fault diagnosis model comprises: a first VaDE module, a residual calculation layer, and a second VaDE module.

10. The system of claim 9, wherein the processing of the acquired state data of the wind turbine bearing at each time during the historical period comprises:

step Q1: splicing the temperature difference between the driving end and the non-driving end of the wind driven generator, the output power of the wind driven generator, the second-level characteristic value of the rotating speed of the wind driven generator and the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in the state data of the wind driven generator bearing at each moment in the historical period at the same moment, and splicing the data;

step Q2, setting the spliced data with the time data when the output power value of the wind driven generator is smaller than the output power threshold value as a null value;

step Q3: normalizing the data processed in the step Q2;

step Q4: and carrying out section division on the state data normalized in the historical time period according to the rotating speed of the wind driven generator, and acquiring the maximum value, the minimum value, the average value and the standard deviation of the temperature difference between the driving end and the non-driving end, the output power of the wind driven generator, the rotating speed of the wind driven generator, the second-level characteristic value of the vibration speed of the bearing of the wind driven generator and the second-level characteristic value data of the vibration acceleration of the bearing in each section division.

Images