CN116292130A

CN116292130A - Wind driven generator state monitoring system and method based on cooperative analysis

Info

Publication number: CN116292130A
Application number: CN202211089902.8A
Authority: CN
Inventors: 黄力哲; 刘庆伏; 陈思; 张雨军; 程施霖; 安达; 赵仕文; 黄泽伟
Original assignee: Huaneng Renewables Corp Ltd; Beijing Huaneng Xinrui Control Technology Co Ltd
Current assignee: Huaneng Renewables Corp Ltd; Beijing Huaneng Xinrui Control Technology Co Ltd
Priority date: 2022-09-07
Filing date: 2022-09-07
Publication date: 2023-06-23

Abstract

The wind driven generator state monitoring system comprises a multiscale neighborhood feature extraction module, a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector, wherein the multiscale neighborhood feature extraction module is used for processing the acquired wind speed, impeller rotating speed, generator rotating speed and generator power at a plurality of preset time points to obtain the multiscale wind speed feature vector, the multiscale impeller rotating speed feature vector, the multiscale generator rotating speed feature vector and the multiscale power feature vector, corresponding response estimates are calculated respectively to obtain a first cooperated transfer matrix, a second cooperated transfer matrix and a third cooperated transfer matrix, the first cooperated transfer matrix, the second cooperated transfer matrix and the third cooperated transfer matrix, the third cooperated transfer matrix are arranged into three-dimensional input tensors, the three-dimensional input tensors are corrected through a convolution neural network and then a corrected classification feature diagram is obtained through a classifier, and a classification result used for representing whether the state of the wind driven generator is normal or not is obtained. Therefore, the relevant characteristics of the external environment elements and the internal factors of the offshore wind turbine are excavated, and the detection judgment accuracy and reliability are improved.

Description

Wind driven generator state monitoring system and method based on cooperative analysis

Technical Field

The invention relates to the field of new energy, in particular to a system and a method for monitoring the state of a wind driven generator based on cooperative analysis.

Background

Wind generating sets are developing towards high power and high efficiency, so that accidents of the wind generating sets consume high maintenance cost to bring huge losses to enterprises, and even threaten lives and properties of people, and therefore early fault identification and treatment are important.

At present, the state monitoring of the wind driven generator is realized by monitoring mechanical parameters, and common monitoring objects include oil liquid monitoring, temperature monitoring, vibration monitoring and the like. The data acquisition and monitoring control (Supervisory Control And Data Acquisition, SCADA) system is continuously introduced into the state monitoring of the wind turbine, and the SCADA system acquires a large number of data types (such as speed, temperature, electric energy, angle and the like), so that the state monitoring of the wind turbine can be realized by analyzing key data. However, analysis and interpretation of huge SCADA data of the wind turbine are very difficult, and the reliability of the data analysis results is insufficient because the wind turbine is affected by the environment and the operation factors thereof.

Therefore, an optimized wind turbine condition monitoring system is desired to comprehensively monitor abnormal conditions of the wind turbine based on the operation conditions of the wind turbine and external factors.

Disclosure of Invention

The present application has been made in order to solve the above technical problems. The embodiment of the application provides a wind driven generator state monitoring system and a method thereof based on cooperative analysis, wherein a multiscale neighborhood feature extraction module is used for processing acquired wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points to obtain multiscale wind speed feature vectors, multiscale impeller rotating speed feature vectors, multiscale generator rotating speed feature vectors and multiscale power feature vectors, corresponding response estimation is calculated respectively to obtain a first cooperative transfer matrix, a second cooperative transfer matrix and a third cooperative transfer matrix, the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix are arranged into three-dimensional input tensors, the three-dimensional input tensors are corrected through feature matrices of a convolution neural network trailing edge channel dimension to obtain corrected classification feature diagrams, and classification results used for representing whether states of wind driven generators are normal can be obtained through a classifier. Therefore, the relevant characteristics of the external environment elements and the internal factors of the offshore wind turbine are excavated, and the detection judgment accuracy and reliability are improved.

According to one aspect of the present application, there is provided a wind turbine condition monitoring system based on collaborative analysis, comprising:

the data acquisition and monitoring control module is used for acquiring wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points;

the vectorization module is used for respectively arranging the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at a plurality of preset time points into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector;

the multiscale sequential feature extraction module is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into the multiscale neighborhood feature extraction module to obtain a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector;

the coordination analysis module is used for respectively calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector so as to obtain a first coordination transfer matrix, a second coordination transfer matrix and a third coordination transfer matrix;

The association feature extraction module is used for arranging the first coordination transfer matrix, the second coordination transfer matrix and the third coordination transfer matrix into three-dimensional input tensors and then obtaining a classification feature map through a convolution neural network of which adjacent layers use convolution kernels which are transposed with each other;

the correction module is used for correcting each feature matrix of the classification feature map along the channel dimension based on a reference feature vector to obtain a corrected classification feature map, wherein the reference feature vector is a channel feature vector obtained by calculating the global average value of each feature matrix of the classification feature map along the channel dimension; and

and the state monitoring result generation module is used for enabling the corrected classification characteristic diagram to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state of the wind driven generator is normal or not.

In the above system for monitoring a wind turbine status based on collaborative analysis, the multi-scale time sequence feature extraction module includes: the first convolution coding unit is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind speed associated feature vector, a first neighborhood scale impeller rotating speed associated feature vector, a first neighborhood scale generator rotating speed associated feature vector and a first neighborhood scale power associated feature vector, wherein the first convolution layer is provided with a first one-dimensional convolution kernel with a first length; the second convolution encoding unit is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind speed associated feature vector, a second neighborhood scale impeller rotating speed associated feature vector, a second neighborhood scale generator rotating speed associated feature vector and a second neighborhood scale power associated feature vector, wherein the second convolution layer is provided with a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and the cascading unit is used for cascading the first neighborhood scale wind speed association feature vector and the second neighborhood scale wind speed association feature vector to obtain the multi-scale wind speed feature vector, cascading the first neighborhood scale impeller rotating speed association feature vector and the second neighborhood scale impeller rotating speed association feature vector to obtain the multi-scale impeller rotating speed feature vector, cascading the first neighborhood scale generator rotating speed association feature vector and the second neighborhood scale generator rotating speed association feature vector to obtain the multi-scale generator rotating speed feature vector, and cascading the first neighborhood scale power association feature vector and the second neighborhood scale power association feature vector to obtain the multi-scale power feature vector.

In the above system for monitoring a wind turbine status based on cooperative analysis, the cooperative analysis module is further configured to: calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector respectively according to the following formula to obtain the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix; wherein, the formula is:

M _i ＝M _j *M ₁

wherein M is _i Respectively representing the multiscale impeller rotating speed characteristic vector, the multiscale generator rotating speed characteristic vector and the multiscale power characteristic vector, M ₁ Representing the multi-scale wind speed feature vector, M representing the first, second and third co-translational matrices, respectively。

In the above system for monitoring a wind turbine status based on collaborative analysis, the correction module includes: the reference feature vector generation unit is used for calculating the global average value of each feature matrix of the classification feature map along the channel dimension to obtain the reference feature vector; an optimization factor calculation unit, configured to calculate an optimization factor corresponding to a feature value of each position in the reference feature vector based on a mean and a variance of a feature value set of all positions in the reference feature vector; and the correction unit is used for respectively weighting the corresponding feature matrixes in the classified feature graphs by taking the optimization factors corresponding to the feature values of each position in the reference feature vector as weights so as to obtain the corrected classified feature graphs.

In the above system for monitoring a state of a wind turbine based on collaborative analysis, the optimization factor calculation unit is further configured to: calculating the optimization factors corresponding to the feature values of all the positions in the reference feature vector according to the following formulas based on the mean and variance of the feature value sets of all the positions in the reference feature vector; wherein, the formula is:

wherein v is _i And expressing the eigenvalue of the ith position of the reference eigenvector, wherein mu and sigma respectively express the mean value and variance of the eigenvalue set of all positions in the reference eigenvector, exp (-) expresses the exponential operation of the eigenvalue, and the exponential operation of the eigenvalue expresses the natural exponential function value with the eigenvalue as power.

According to another aspect of the present application, there is also provided a method for monitoring a state of a wind turbine based on a collaborative analysis, including:

acquiring wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points;

arranging the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at a plurality of preset time points into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector respectively;

Respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a multiscale neighborhood feature extraction module to obtain a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector;

respectively calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector to obtain a first co-overall transfer matrix, a second co-overall transfer matrix and a third co-overall transfer matrix;

arranging the first, second and third co-overall transfer matrices into three-dimensional input tensors, and then using convolution neural networks which are mutually transposed convolution kernels through adjacent layers to obtain a classification characteristic diagram;

correcting each feature matrix of the classification feature map along the channel dimension based on a reference feature vector to obtain a corrected classification feature map, wherein the reference feature vector is a channel feature vector obtained by calculating the global average value of each feature matrix of the classification feature map along the channel dimension; and

And the corrected classification characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state of the wind driven generator is normal or not.

Compared with the prior art, the system and the method for monitoring the state of the wind driven generator based on the cooperative analysis are characterized in that the acquired wind speed, impeller rotating speed value, generator rotating speed and generator power at a plurality of preset time points are processed through the multi-scale neighborhood feature extraction module to obtain multi-scale wind speed feature vectors, multi-scale impeller rotating speed feature vectors, multi-scale generator rotating speed feature vectors and multi-scale power feature vectors, corresponding response estimates are calculated respectively to obtain a first cooperative transfer matrix, a second cooperative transfer matrix and a third cooperative transfer matrix, the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix are arranged into three-dimensional input tensors, the three-dimensional input tensors are corrected through the feature matrices of the back edge channel dimension of the convolutional neural network to obtain corrected classification feature graphs, and classification results for representing whether the state of the wind driven generator is normal can be obtained through the classifier. Therefore, the relevant characteristics of the external environment elements and the internal factors of the offshore wind turbine are excavated, and the detection judgment accuracy and reliability are improved.

Drawings

The foregoing and other objects, features and advantages of the present application will become more apparent from the following more particular description of embodiments of the present application, as illustrated in the accompanying drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and not constitute a limitation to the application. In the drawings, like reference numerals generally refer to like parts or steps.

FIG. 1 illustrates an application scenario diagram of a collaborative analysis-based wind turbine condition monitoring system according to an embodiment of the present application.

FIG. 2 illustrates a block diagram of a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application.

FIG. 3 illustrates a system architecture diagram of a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application.

FIG. 4 illustrates a block diagram of a multi-scale temporal feature extraction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application.

FIG. 5 illustrates a block diagram of an associated feature extraction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application.

FIG. 6 illustrates a block diagram of a correction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application.

FIG. 7 illustrates a flowchart of a method of monitoring a wind turbine status based on a collaborative analysis according to an embodiment of the present application.

Detailed Description

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application and not all of the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

Summary of the application

It should be appreciated that the wind turbine generator may be affected by the environment and its operating factors when monitoring the state of the wind turbine generator, which may result in insufficient reliability of the data analysis result. Therefore, in order to improve the accuracy and reliability of detection and judgment when the state of the wind turbine is monitored, the selection is performed by implicit characteristic information of an external environment element and an internal factor of the offshore wind turbine and the relevance characteristic of the external environment element and the internal factor of the offshore wind turbine. In other words, in the technical scheme of the application, the external environment elements can be represented through wind speed data, the internal factors of the offshore wind turbine are represented through impeller rotation speed values, generator rotation speeds and generator power data, the deep neural network model is utilized to extract the respective multi-scale neighborhood associated feature representation, and then the data coordination of the high-dimensional feature space is carried out through the response estimation among the data features, so that the effectiveness and the reliability of monitoring the abnormal state of the wind driven generator are improved.

Specifically, in the technical solution of the present application, first, wind speeds, impeller rotation speed values, generator rotation speeds, and generator power at a plurality of predetermined time points are acquired by respective sensors. And then, in order to facilitate the subsequent deep association feature mining, the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at a plurality of preset time points are further arranged into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector respectively for the distribution information of the whole data in the time dimension.

It should be appreciated that convolutional neural networks were originally models applied in the image domain, but the concept of local feature extraction can be equally applied to time series data analysis. For example, a time series convolution structure with a convolution kernel size of 3, for wind speed time series data input, the convolution kernel moves in the form of a sliding window along the time dimension and outputs a weighted sum of the data within each wind speed time series segment. Each convolution unit stacks a plurality of convolution kernels to output a multi-dimensional wind speed signature. Since the wind speed may exhibit different mode states in different time periods, for example, the wind speed may be strong and weak during a burst in different time periods. The feature extraction in the field is to perform deep associated feature mining on data in different time spans, and a large convolution kernel can extract features from a large-scale time sequence neighborhood of wind speeds, wherein the influence of each wind speed value in the neighborhood is smaller, so that fluctuation of wind speed input data is weakened, and the influence of noise points on the output features of the wind speeds is lightened. However, the large-scale convolution kernel weakens the difference of numerical variation, so that the problem of smooth transition is easily caused, and the output characteristics of the wind speed lose the discrimination capability. In contrast, small scale convolution kernels can better preserve information in wind speed input data, but are also more susceptible to interference from noise therein. Therefore, the characteristics of different wind speed time sequence scales are extracted by combining convolution units with different sizes in consideration of the characteristics of convolution with different scales. And then, feature fusion is completed in a feature splicing mode, so that multi-scale wind speed neighborhood features are obtained. It is worth mentioning that, likewise, the extraction of the multi-scale neighborhood correlation features can also be performed based on this method for the impeller speed, the generator speed and the power input data.

Specifically, in the technical scheme of the application, a convolution layer with one-dimensional convolution kernels of different scales of a multi-scale neighborhood feature extraction module is further used for carrying out one-dimensional convolution coding on the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector respectively, and then the obtained feature vectors corresponding to the two one-dimensional convolution kernels of different scales are cascaded to obtain the multi-scale wind speed feature vector, the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the multi-scale power feature vector. In particular, in this way, the multi-scale neighborhood associated feature representations of the dynamic implicit features of the wind speed, the impeller rotation speed, the generator rotation speed and the power are respectively extracted, so that the output features not only comprise the smoothed features, but also preserve the features of the original input, avoid information loss, and improve the accuracy of the subsequent classification. In particular, in other specific examples of the present application, the multi-scale neighborhood feature extraction module may further include a greater number of one-dimensional convolution layers, which use one-dimensional convolution kernels with different lengths to perform intra-neighborhood correlation feature extraction with different scales, which is not limited by the present application.

Because the wind speed data is an external environment factor, the impeller rotating speed, the generator rotating speed and the power are all internal factors of the offshore wind turbine, and feature scales among the multiscale neighborhood associated features of the wind speed, the impeller rotating speed, the generator rotating speed and the power are not the same, and the multiscale neighborhood associated features of the impeller rotating speed, the generator rotating speed and the power can be respectively regarded as response features of the multiscale neighborhood associated features of the wind speed in a high-dimensional feature space. Therefore, in order to better integrate the feature information, the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the multi-scale power feature vector are further calculated respectively to perform data coordination of a high-dimensional feature space relative to the response estimation of the multi-scale wind speed feature vector so as to obtain a first coordination transfer matrix, a second coordination transfer matrix and a third coordination transfer matrix.

And then, after arranging the first, second and third co-migration matrices into three-dimensional input tensors to integrate data co-integration characteristic information of respective data in a high-dimensional characteristic space, processing the three-dimensional input tensors by using a convolution neural network with adjacent layers being convolution kernels transposed with each other to obtain a classification characteristic diagram. In particular, here, adjacent convolution layers of the neural network use convolution kernels that are transposed with each other, so that updating of network parameters and searching of network parameter structures suitable for a specific data structure can be simultaneously updated during training, and further accuracy of subsequent classification is improved.

Particularly, in the technical solution of the present application, when the first co-migration matrix, the second co-migration matrix and the third co-migration matrix are arranged into three-dimensional input tensors to obtain the classification feature map, convolution kernels with mutually transposed adjacent layers can extract a specific structure in a matrix dimension, so that it is expected that the classification feature map can have better expression in a channel dimension.

Based on this, the global mean of each feature matrix along the channel of the classification feature map is first calculated to obtain channel feature vectors, and then the deep recursive press-excitation optimization factor of each channel is calculated based on the reference feature vectors, specifically:

wherein v is _i Represents the eigenvalue of the ith position of the reference eigenvector, and μ and σ represent the mean and variance, respectively, of the eigenvalue set of all positions in the reference eigenvector.

Here, the deep recursive pressing-excitation optimization factor may activate the deep recursion of the feature distribution along the channel based on the statistical characteristics of the feature set along the channel, so as to infer the depth distribution of the feature at the sampling depth of each channel, and acquire the channel attention-enhancing depth confidence value by adopting a channel pressing-excitation mechanism formed by a ReLU-Sigmoid function, and weight the feature matrix of each channel of the classification feature map by taking the depth confidence value as a coefficient, so that the expression confidence of the depth direction of the high-dimensional feature flow pattern can be improved, and the classification accuracy is further improved. Therefore, the abnormal state of the wind driven generator can be monitored, so that accidents of the wind driven generator set are avoided, and the normal operation of the fan is ensured.

Based on this, the application proposes a wind turbine condition monitoring system based on collaborative analysis, which includes: the data acquisition and monitoring control module is used for acquiring wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points; the vectorization module is used for respectively arranging the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at a plurality of preset time points into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector; the multiscale sequential feature extraction module is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into the multiscale neighborhood feature extraction module to obtain a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector; the coordination analysis module is used for respectively calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector so as to obtain a first coordination transfer matrix, a second coordination transfer matrix and a third coordination transfer matrix; the association feature extraction module is used for arranging the first coordination transfer matrix, the second coordination transfer matrix and the third coordination transfer matrix into three-dimensional input tensors and then obtaining a classification feature map through a convolution neural network of which adjacent layers use convolution kernels which are transposed with each other; the correction module is used for correcting each feature matrix of the classification feature map along the channel dimension based on a reference feature vector to obtain a corrected classification feature map, wherein the reference feature vector is a channel feature vector obtained by calculating the global average value of each feature matrix of the classification feature map along the channel dimension; and the state monitoring result generation module is used for enabling the corrected classification characteristic diagram to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state of the wind driven generator is normal or not.

FIG. 1 illustrates an application scenario diagram of a collaborative analysis-based wind turbine condition monitoring system according to an embodiment of the present application. As shown in fig. 1, in this application scenario, first, wind speeds, impeller rotation speed values, generator rotation speed, and generator power at a plurality of predetermined time points are acquired by respective sensors (e.g., se1, se2, se3, se4 as illustrated in fig. 1). Then, the wind speed, the impeller rotation speed value, the generator rotation speed, and the generator power at the plurality of predetermined time points are input to a server (e.g., S as illustrated in fig. 1) in which a collaborative analysis-based wind turbine state monitoring algorithm is deployed, wherein the server is capable of processing the input information with the collaborative analysis-based wind turbine state monitoring algorithm to obtain a classification result for indicating whether the state of the wind turbine is normal.

Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.

Exemplary System

FIG. 2 illustrates a block diagram of a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application. As shown in fig. 2, a system 100 for monitoring a state of a wind turbine based on a cooperative analysis according to an embodiment of the present application includes: the data acquisition and monitoring control module 110 is used for acquiring wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points; a vectorization module 120, configured to arrange the wind speed, the impeller rotation speed value, the generator rotation speed, and the generator power at the plurality of predetermined time points into a wind speed input vector, an impeller rotation speed input vector, a generator rotation speed input vector, and a power input vector, respectively; the multiscale sequential feature extraction module 130 is configured to input the wind speed input vector, the impeller rotation speed input vector, the generator rotation speed input vector, and the power input vector into a multiscale neighborhood feature extraction module respectively to obtain a multiscale wind speed feature vector, a multiscale impeller rotation speed feature vector, a multiscale generator rotation speed feature vector, and a multiscale power feature vector; the cooperative analysis module 140 is configured to calculate a response estimate of the multi-scale impeller rotational speed feature vector, the multi-scale generator rotational speed feature vector, and the multi-scale power feature vector relative to the multi-scale wind speed feature vector, to obtain a first cooperative transfer matrix, a second cooperative transfer matrix, and a third cooperative transfer matrix; the correlation feature extraction module 150 is configured to arrange the first cooperative transfer matrix, the second cooperative transfer matrix, and the third cooperative transfer matrix into three-dimensional input tensors, and then obtain a classification feature map by using a convolutional neural network of mutually transposed convolution kernels through adjacent layers; a correction module 160, configured to correct each feature matrix of the classification feature map along a channel dimension based on a reference feature vector to obtain a corrected classification feature map, where the reference feature vector is a channel feature vector obtained by calculating a global average value of each feature matrix of the classification feature map along the channel dimension; and a state monitoring result generating module 170, configured to pass the corrected classification feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the state of the wind turbine is normal.

FIG. 3 illustrates a system architecture diagram of a wind turbine condition monitoring system 100 based on collaborative analysis according to an embodiment of the present application. As shown in fig. 3, in the system architecture of the system 100 for monitoring the state of the wind driven generator based on the cooperative analysis, wind speed, impeller rotation speed value, generator rotation speed and generator power at a plurality of predetermined time points are first obtained and respectively arranged as a wind speed input vector, an impeller rotation speed input vector, a generator rotation speed input vector and a power input vector. And then, a multiscale neighborhood feature extraction module is utilized to obtain a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector, and the response estimation of the multiscale impeller rotating speed feature vector, the multiscale generator rotating speed feature vector and the multiscale power feature vector relative to the multiscale wind speed feature vector is respectively calculated to obtain a first co-migration matrix, a second co-migration matrix and a third co-migration matrix. And then, arranging the first, second and third co-overall transfer matrixes into three-dimensional input tensors, using convolution neural networks which are mutually transposed convolution kernels through adjacent layers to obtain a classification characteristic diagram, and correcting each characteristic matrix of the classification characteristic diagram along the channel dimension to obtain a corrected classification characteristic diagram. And then, the corrected classification characteristic diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state of the wind driven generator is normal or not.

In the above-mentioned wind turbine status monitoring system 100 based on cooperative analysis, the data collection and monitoring control module 110 is configured to obtain wind speeds, impeller rotational speed values, generator rotational speeds and generator power at a plurality of predetermined time points. It should be appreciated that the wind turbine generator may be affected by the environment and its operating factors when monitoring the state of the wind turbine generator, which may result in insufficient reliability of the data analysis result. Therefore, in order to improve the accuracy and reliability of detection and judgment when the state of the wind turbine is monitored, the selection is performed by implicit characteristic information of an external environment element and an internal factor of the offshore wind turbine and the relevance characteristic of the external environment element and the internal factor of the offshore wind turbine. That is, in the technical solution of the present application, the external environmental element may be represented by wind speed data, and the internal factor of the offshore wind turbine may be represented by an impeller rotation speed value, a generator rotation speed and generator power data.

In the above-mentioned wind turbine condition monitoring system 100 based on cooperative analysis, the vectorizing module 120 is configured to arrange the wind speed, the impeller rotation speed value, the generator rotation speed and the generator power at the plurality of predetermined time points into a wind speed input vector, an impeller rotation speed input vector, a generator rotation speed input vector and a power input vector, respectively. And in order to facilitate the subsequent deep association feature mining, the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at the preset time points are further respectively arranged into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector.

In the above-mentioned wind turbine state monitoring system 100 based on cooperative analysis, the multi-scale time sequence feature extraction module 130 is configured to input the wind speed input vector, the impeller rotational speed input vector, the generator rotational speed input vector, and the power input vector into the multi-scale neighborhood feature extraction module respectively to obtain a multi-scale wind speed feature vector, a multi-scale impeller rotational speed feature vector, a multi-scale generator rotational speed feature vector, and a multi-scale power feature vector. It should be appreciated that convolutional neural networks were originally models applied in the image domain, but the concept of local feature extraction can be equally applied to time series data analysis. For example, a time series convolution structure with a convolution kernel size of 3, for wind speed time series data input, the convolution kernel moves in the form of a sliding window along the time dimension and outputs a weighted sum of the data within each wind speed time series segment. Each convolution unit stacks a plurality of convolution kernels to output a multi-dimensional wind speed signature. Since the wind speed may exhibit different mode states in different time periods, for example, the wind speed may be strong and weak during a burst in different time periods. And the feature extraction of the neighborhood is to perform deep associated feature mining on data in different time spans, and a large convolution kernel can extract features from a large-scale time sequence neighborhood of wind speeds, wherein the influence of each wind speed value in the neighborhood is smaller, so that fluctuation of wind speed input data is weakened, and the influence of noise points on the output features of the wind speeds is lightened. However, the large-scale convolution kernel weakens the difference of numerical variation, so that the problem of smooth transition is easily caused, and the output characteristics of the wind speed lose the discrimination capability. In contrast, small scale convolution kernels can better preserve information in wind speed input data, but are also more susceptible to interference from noise therein. Therefore, the characteristics of different wind speed time sequence scales are extracted by combining convolution units with different sizes in consideration of the characteristics of convolution with different scales. And then, feature fusion is completed in a feature splicing mode, so that multi-scale wind speed neighborhood features are obtained. It is worth mentioning that, likewise, the extraction of the multi-scale neighborhood correlation features can also be performed based on this method for the impeller speed, the generator speed and the power input data.

FIG. 4 illustrates a block diagram of a multi-scale temporal feature extraction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application. As shown in fig. 4, the multi-scale timing feature extraction module 130 includes: a first convolution encoding unit 131, configured to input the wind speed input vector, the impeller rotation speed input vector, the generator rotation speed input vector, and the power input vector to a first convolution layer of the multi-scale neighborhood feature extraction module, respectively, to obtain a first neighborhood scale wind speed associated feature vector, a first neighborhood scale impeller rotation speed associated feature vector, a first neighborhood scale generator rotation speed associated feature vector, and a first neighborhood scale power associated feature vector, where the first convolution layer has a first one-dimensional convolution kernel with a first length; a second convolution encoding unit 132, configured to input the wind speed input vector, the impeller rotation speed input vector, the generator rotation speed input vector, and the power input vector to a second convolution layer of the multi-scale neighborhood feature extraction module, respectively, to obtain a second neighborhood scale wind speed associated feature vector, a second neighborhood scale impeller rotation speed associated feature vector, a second neighborhood scale generator rotation speed associated feature vector, and a second neighborhood scale power associated feature vector, where the second convolution layer has a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and a cascading unit 133, configured to concatenate the first neighborhood scale wind speed correlation feature vector and the second neighborhood scale wind speed correlation feature vector to obtain the multi-scale wind speed feature vector, concatenate the first neighborhood scale impeller rotational speed correlation feature vector and the second neighborhood scale impeller rotational speed correlation feature vector to obtain the multi-scale impeller rotational speed feature vector, concatenate the first neighborhood scale generator rotational speed correlation feature vector and the second neighborhood scale generator rotational speed correlation feature vector to obtain the multi-scale generator rotational speed feature vector, and concatenate the first neighborhood scale power correlation feature vector and the second neighborhood scale power correlation feature vector to obtain the multi-scale power feature vector.

In one example, in the above-mentioned wind turbine condition monitoring system 100 based on the cooperative analysis, the first convolution encoding unit 131 is further configured to: using a first convolution layer of the multiscale neighborhood feature extraction module to perform one-dimensional convolution coding on the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector respectively according to the following formula to obtain a first neighborhood scale wind speed correlation feature vector, a first neighborhood scale impeller rotating speed correlation feature vector, a first neighborhood scale generator rotating speed correlation feature vector and a first neighborhood scale power correlation feature vector;

wherein, the formula is:

wherein a is the width of the first convolution kernel in the X direction, F (a) is a first convolution kernel parameter vector, G (X-a) is a local vector matrix calculated by a convolution kernel function, w is the size of the first convolution kernel, and X represents the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector.

In one example, in the above-mentioned wind turbine condition monitoring system 100 based on the cooperative analysis, the second convolution encoding unit 132 is further configured to: using a second convolution layer of the multiscale neighborhood feature extraction module to perform one-dimensional convolution coding on the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector respectively according to the following formula to obtain a second neighborhood scale wind speed correlation feature vector, a second neighborhood scale impeller rotating speed correlation feature vector, a second neighborhood scale generator rotating speed correlation feature vector and a second neighborhood scale power correlation feature vector;

Wherein, the formula is:

wherein b is the width of the second convolution kernel in the X direction, F (b) is a second convolution kernel parameter vector, G (X-b) is a local vector matrix calculated by a convolution kernel function, m is the size of the second convolution kernel, and X represents the wind speed input vector, the impeller rotation speed input vector, the generator rotation speed input vector and the power input vector.

In the above-mentioned wind turbine state monitoring system 100 based on cooperative analysis, the cooperative analysis module 140 is configured to calculate the multi-scale impeller rotational speed feature vector, the multi-scale generator rotational speed feature vector, and the responsiveness estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector, respectively, so as to obtain a first cooperative transfer matrix, a second cooperative transfer matrix, and a third cooperative transfer matrix. Because the wind speed data is an external environment factor, the impeller rotating speed, the generator rotating speed and the power are all internal factors of the offshore wind turbine, and feature scales among the multiscale neighborhood associated features of the wind speed, the impeller rotating speed, the generator rotating speed and the power are not the same, and the multiscale neighborhood associated features of the impeller rotating speed, the generator rotating speed and the power can be respectively regarded as response features of the multiscale neighborhood associated features of the wind speed in a high-dimensional feature space. Therefore, in order to better integrate the feature information, the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the multi-scale power feature vector are further calculated respectively to perform data coordination of a high-dimensional feature space relative to the response estimation of the multi-scale wind speed feature vector so as to obtain a first coordination transfer matrix, a second coordination transfer matrix and a third coordination transfer matrix.

In one example, in the above-mentioned wind turbine condition monitoring system 100 based on the coordination analysis, the coordination analysis module 140 is further configured to: calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector respectively according to the following formula to obtain the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix;

wherein, the formula is:

M _i ＝M _j *M ₁

wherein M is _i Respectively representing the multiscale impeller rotating speed characteristic vector, the multiscale generator rotating speed characteristic vector and the multiscale power characteristic vector, M ₁ And representing the multi-scale wind speed characteristic vector, wherein M represents the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix respectively.

In the above-mentioned wind turbine condition monitoring system 100 based on cooperative analysis, the associated feature extraction module 150 is configured to arrange the first cooperative transfer matrix, the second cooperative transfer matrix, and the third cooperative transfer matrix into three-dimensional input tensors, and then obtain the classification feature map by using a convolutional neural network of mutually transposed convolution kernels through adjacent layers. That is, after the first, second, and third co-migration matrices are arranged into three-dimensional input tensors to integrate data co-ordinated feature information of respective data in a high-dimensional feature space, the three-dimensional input tensors are processed by using a convolutional neural network in which adjacent layers are mutually transposed convolution kernels to obtain a classification feature map. In particular, here, adjacent convolution layers of the neural network use convolution kernels that are transposed with each other, so that updating of network parameters and searching of network parameter structures suitable for a specific data structure can be simultaneously updated during training, and further accuracy of subsequent classification is improved.

FIG. 5 illustrates a block diagram of an associated feature extraction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application. As shown in fig. 5, the associated feature extraction module 150 includes: a shallow feature map extracting unit 151 for extracting a shallow feature map from an mth layer of a convolutional neural network in which the adjacent layers use mutually transposed convolution kernels, M being an even number; a deep feature map extracting unit 152, configured to extract a deep feature map from an nth layer of a convolutional neural network in which adjacent layers use mutually transposed convolutional kernels, where N is an even number and N is greater than 2 times of M; and a feature map fusion unit 153 for fusing the shallow feature map and the deep feature map to generate the classification feature map.

In the above-mentioned wind turbine condition monitoring system 100 based on cooperative analysis, the correction module 160 is configured to correct each feature matrix of the classification feature map along the channel dimension based on a reference feature vector to obtain a corrected classification feature map, where the reference feature vector is a channel feature vector obtained by calculating a global average value of each feature matrix of the classification feature map along the channel dimension. Particularly, in the technical solution of the present application, when the first co-migration matrix, the second co-migration matrix and the third co-migration matrix are arranged into three-dimensional input tensors to obtain the classification feature map, convolution kernels with mutually transposed adjacent layers can extract a specific structure in a matrix dimension, so that it is expected that the classification feature map can have better expression in a channel dimension.

Based on this, a global mean of each feature matrix along a channel of the classification feature map is first calculated to obtain channel feature vectors, and then a deep recursive press-excitation optimization factor for each channel is calculated based on the reference feature vectors. Here, the deep recursive pressing-excitation optimization factor may activate the deep recursion of the feature distribution along the channel based on the statistical characteristics of the feature set along the channel, so as to infer the depth distribution of the feature at the sampling depth of each channel, and acquire the channel attention-enhancing depth confidence value by adopting a channel pressing-excitation mechanism formed by a ReLU-Sigmoid function, and weight the feature matrix of each channel of the classification feature map by taking the depth confidence value as a coefficient, so that the expression confidence of the depth direction of the high-dimensional feature flow pattern can be improved, and the classification accuracy is further improved. Therefore, the abnormal state of the wind driven generator can be monitored, so that accidents of the wind driven generator set are avoided, and the normal operation of the fan is ensured.

FIG. 6 illustrates a block diagram of a correction module in a wind turbine condition monitoring system based on collaborative analysis according to an embodiment of the present application. As shown in fig. 6, the correction module 160 includes: a reference feature vector generation unit 161, configured to calculate a global average value of feature matrices along a channel dimension of the classification feature map to obtain the reference feature vector; an optimization factor calculating unit 162, configured to calculate an optimization factor corresponding to a feature value of each position in the reference feature vector based on the mean and variance of the feature value sets of all positions in the reference feature vector; and a correction unit 163, configured to weight the feature matrices corresponding to the classification feature graphs with the optimization factors corresponding to the feature values of the respective positions in the reference feature vector as weights, so as to obtain the corrected classification feature graphs.

In one example, in the above-described wind turbine condition monitoring system 100 based on the collaborative analysis, the optimization factor calculating unit 162 is further configured to: calculating the optimization factors corresponding to the feature values of all the positions in the reference feature vector according to the following formulas based on the mean and variance of the feature value sets of all the positions in the reference feature vector;

wherein, the formula is:

In the above-mentioned wind turbine condition monitoring system 100 based on cooperative analysis, the condition monitoring result generating module 170 is configured to pass the corrected classification feature map through a classifier to obtain a classification result, where the classification result is used to indicate whether the condition of the wind turbine is normal.

In one example, in the above-mentioned wind turbine status monitoring system 100 based on the collaborative analysis, the status monitoring result generating module 170 is further configured to: the classifier processes the corrected classification feature map to generate a classification result with the following formula: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) Project (F), where Project (F) represents projecting the corrected classification feature map as a vector, W ₁ To W _n Weight matrix for all the connection layers of each layer, B ₁ To B _n Representing the bias matrix for each fully connected layer.

In summary, the system 100 for monitoring the state of a wind driven generator based on cooperative analysis according to the embodiment of the present application is illustrated, where the obtained wind speed, the impeller rotational speed value, the generator rotational speed and the generator power at a plurality of predetermined time points are processed by a multi-scale neighborhood feature extraction module to obtain a multi-scale wind speed feature vector, a multi-scale impeller rotational speed feature vector, a multi-scale generator rotational speed feature vector and a multi-scale power feature vector, corresponding response estimations are calculated respectively to obtain a first cooperative transfer matrix, a second cooperative transfer matrix and a third cooperative transfer matrix, and then the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix are arranged into a three-dimensional input tensor, and then corrected by a convolutional neural network along each feature matrix of a channel dimension to obtain a corrected classification feature map, and then a classifier is used to obtain a classification result for indicating whether the state of the wind driven generator is normal. Therefore, the relevant characteristics of the external environment elements and the internal factors of the offshore wind turbine are excavated, and the detection judgment accuracy and reliability are improved.

As described above, the wind turbine state monitoring system 100 based on the collaborative analysis according to the embodiment of the present application may be implemented in various terminal devices, for example, a server or the like having wind turbine state monitoring based on the collaborative analysis. In one example, the collaborative analysis-based wind turbine condition monitoring system 100 according to embodiments of the present application may be integrated into the terminal device as a software module and/or hardware module. For example, the collaborative analysis-based wind turbine condition monitoring system 100 may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the wind turbine condition monitoring system 100 based on collaborative analysis may also be one of a plurality of hardware modules of the terminal device.

Alternatively, in another example, the collaborative analysis-based wind turbine status monitoring system 100 and the terminal device may be separate devices, and the collaborative analysis-based wind turbine status monitoring system 100 may be connected to the terminal device through a wired and/or wireless network and transmit interaction information according to a agreed data format.

Exemplary method

According to another aspect of the application, a method for monitoring the state of the wind driven generator based on the coordination analysis is also provided. As shown in fig. 7, the method for monitoring the state of the wind driven generator based on the cooperative analysis according to the embodiment of the application includes the following steps: s110, acquiring wind speeds, impeller rotating speed values, generator rotating speeds and generator power at a plurality of preset time points; s120, arranging the wind speed, the impeller rotating speed value, the generator rotating speed and the generator power at a plurality of preset time points into a wind speed input vector, an impeller rotating speed input vector, a generator rotating speed input vector and a power input vector respectively; s130, respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a multiscale neighborhood feature extraction module to obtain a multiscale wind speed feature vector, a multiscale impeller rotating speed feature vector, a multiscale generator rotating speed feature vector and a multiscale power feature vector; s140, respectively calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector to obtain a first co-overall transfer matrix, a second co-overall transfer matrix and a third co-overall transfer matrix; s150, arranging the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix into three-dimensional input tensors, and then using convolution neural networks which are mutually transposed convolution kernels through adjacent layers to obtain a classification characteristic diagram; s160, correcting each feature matrix of the classification feature map along the channel dimension based on a reference feature vector to obtain a corrected classification feature map, wherein the reference feature vector is a channel feature vector obtained by calculating the global average value of each feature matrix of the classification feature map along the channel dimension; and S170, passing the corrected classification characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the state of the wind driven generator is normal or not.

In summary, the method for monitoring the state of the wind driven generator based on the collaborative analysis according to the embodiment of the application is illustrated, the obtained wind speed, impeller rotation speed value, generator rotation speed and generator power at a plurality of preset time points are processed through a multi-scale neighborhood feature extraction module to obtain a multi-scale wind speed feature vector, a multi-scale impeller rotation speed feature vector, a multi-scale generator rotation speed feature vector and a multi-scale power feature vector, corresponding response estimates are calculated respectively to obtain a first collaborative transfer matrix, a second collaborative transfer matrix and a third collaborative transfer matrix, the first collaborative transfer matrix, the second collaborative transfer matrix and the third collaborative transfer matrix are arranged into three-dimensional input tensors, the three-dimensional input tensors are corrected through each feature matrix of a convolution neural network trailing edge channel dimension to obtain a corrected classification feature map, and classification results for representing whether the state of the wind driven generator is normal or not can be obtained through a classifier. Therefore, the relevant characteristics of the external environment elements and the internal factors of the offshore wind turbine are excavated, and the detection judgment accuracy and reliability are improved.

Claims

1. Wind driven generator state monitoring system based on cooperation analysis, characterized by comprising:

2. The system for monitoring the state of a wind turbine based on collaborative analysis according to claim 1, wherein the multi-scale temporal feature extraction module includes:

the first convolution coding unit is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a first convolution layer of the multi-scale neighborhood feature extraction module to obtain a first neighborhood scale wind speed associated feature vector, a first neighborhood scale impeller rotating speed associated feature vector, a first neighborhood scale generator rotating speed associated feature vector and a first neighborhood scale power associated feature vector, wherein the first convolution layer is provided with a first one-dimensional convolution kernel with a first length;

The second convolution encoding unit is used for respectively inputting the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector into a second convolution layer of the multi-scale neighborhood feature extraction module to obtain a second neighborhood scale wind speed associated feature vector, a second neighborhood scale impeller rotating speed associated feature vector, a second neighborhood scale generator rotating speed associated feature vector and a second neighborhood scale power associated feature vector, wherein the second convolution layer is provided with a second one-dimensional convolution kernel with a second length, and the first length is different from the second length; and

the cascade unit is used for cascading the first neighborhood scale wind speed association feature vector and the second neighborhood scale wind speed association feature vector to obtain a multi-scale wind speed feature vector, cascading the first neighborhood scale impeller rotating speed association feature vector and the second neighborhood scale impeller rotating speed association feature vector to obtain a multi-scale impeller rotating speed feature vector, cascading the first neighborhood scale generator rotating speed association feature vector and the second neighborhood scale generator rotating speed association feature vector to obtain a multi-scale generator rotating speed feature vector, and cascading the first neighborhood scale power association feature vector and the second neighborhood scale power association feature vector to obtain the multi-scale power feature vector.

3. The system for monitoring the state of a wind turbine based on collaborative analysis according to claim 2, wherein the first convolution encoding unit is further configured to: using a first convolution layer of the multiscale neighborhood feature extraction module to perform one-dimensional convolution coding on the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector respectively according to the following formula to obtain a first neighborhood scale wind speed correlation feature vector, a first neighborhood scale impeller rotating speed correlation feature vector, a first neighborhood scale generator rotating speed correlation feature vector and a first neighborhood scale power correlation feature vector;

wherein, the formula is:

4. A system for monitoring the condition of a wind turbine based on a collaborative analysis according to claim 3, wherein the second convolution encoding unit is further configured to: using a second convolution layer of the multiscale neighborhood feature extraction module to perform one-dimensional convolution coding on the wind speed input vector, the impeller rotating speed input vector, the generator rotating speed input vector and the power input vector respectively according to the following formula to obtain a second neighborhood scale wind speed correlation feature vector, a second neighborhood scale impeller rotating speed correlation feature vector, a second neighborhood scale generator rotating speed correlation feature vector and a second neighborhood scale power correlation feature vector;

Wherein, the formula is:

5. The system for monitoring the condition of a wind turbine based on a collaborative analysis according to claim 4, wherein the collaborative analysis module is further configured to: calculating the multi-scale impeller rotating speed feature vector, the multi-scale generator rotating speed feature vector and the response estimation of the multi-scale power feature vector relative to the multi-scale wind speed feature vector respectively according to the following formula to obtain the first cooperative transfer matrix, the second cooperative transfer matrix and the third cooperative transfer matrix;

wherein, the formula is:

M _i ＝M _j *M ₁

wherein M is _i Respectively represent theMultiscale impeller rotating speed feature vector, multiscale generator rotating speed feature vector and multiscale power feature vector, M ₁ Representing the multiscale wind speed feature vector, M _j Representing the first, second and third co-overall transfer matrices, respectively.

6. The collaborative analysis based wind turbine condition monitoring system of claim 5, wherein the associated feature extraction module comprises:

a shallow feature map extracting unit, configured to extract a shallow feature map from an mth layer of a convolutional neural network in which adjacent layers use mutually transposed convolutional kernels, where M is an even number;

a deep feature map extracting unit, configured to extract a deep feature map from an nth layer of a convolutional neural network in which adjacent layers use mutually transposed convolutional kernels, where N is an even number and N is greater than 2 times of M; and

and the feature map fusion unit is used for fusing the shallow feature map and the deep feature map to generate the classification feature map.

7. The collaborative analysis based wind turbine condition monitoring system of claim 6, wherein the correction module includes:

the reference feature vector generation unit is used for calculating the global average value of each feature matrix of the classification feature map along the channel dimension to obtain the reference feature vector;

an optimization factor calculation unit, configured to calculate an optimization factor corresponding to a feature value of each position in the reference feature vector based on a mean and a variance of a feature value set of all positions in the reference feature vector; and

And the correction unit is used for respectively weighting the feature matrixes corresponding to the classification feature graphs by taking the optimization factors corresponding to the feature values of each position in the reference feature vector as weights so as to obtain the corrected classification feature graphs.

8. The system for monitoring the state of a wind turbine based on collaborative analysis according to claim 7, wherein the optimization factor calculation unit is further configured to: calculating the optimization factors corresponding to the feature values of all the positions in the reference feature vector according to the following formulas based on the mean and variance of the feature value sets of all the positions in the reference feature vector;

wherein, the formula is:

wherein v is _i Represents the eigenvalue of the ith position of the reference eigenvector, and μ and σ represent the mean and variance of the eigenvalue set of all positions in the reference eigenvector, respectively, exp (- σ) represents calculating the natural exponential function value raised to the power of the negative value of the variance.

9. The system for monitoring the state of a wind turbine based on collaborative analysis according to claim 8, wherein the state monitoring result generation module is further configured to: the classifier processes the corrected classification feature map to generate a classification result with the following formula: softmax { (W) _n ,B _n ):…:(W ₁ ,B ₁ ) Project (F), where Project (F) represents projecting the corrected classification feature map as a vector, W ₁ To W _n Weight matrix for all the connection layers of each layer, B ₁ To B _n Representing the bias matrix for each fully connected layer.

10. A wind driven generator state monitoring method based on cooperative analysis is characterized by comprising the following steps: