Method, system and equipment for detecting working state of wind driven generator and storage medium
The vibration waveform characteristics of wind turbines are extracted through convolutional neural networks and graph neural networks, and combined with the classifier to automatically detect working status, it solves the problem of difficulty in timely discovering safety hazards in the existing technology, and achieves the efficiency of automated detection and accident response.
Patent Information
- Application Number
- CN202311501598.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-10
- Publication Date
- 2025-05-13
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Wind turbines may cause abnormal working conditions under long-term operation and erosion of the external environment. The existing maintenance and inspection methods rely on manual regular inspections, making it difficult to detect safety hazards in a timely manner and respond to accidents.
By obtaining multiple vibration waveform diagrams of the wind turbine, input them to the convolutional neural network and graph neural network, extracting the vibration waveform feature vector and topological feature matrix, and combining with the pre-trained classifier, it automatically detects whether the working state of the wind turbine is normal.
It realizes the ability to automatically detect the working status of the wind turbine, improves the efficiency of timely detection of safety hazards and accident response, and ensures the normal operation of the wind turbine.
Smart Images
Figure CN119989069A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of wind power generation, and in particular to a method, system, device and storage medium for detecting the working status of a wind turbine generator. Background Art
[0002] During the operation of wind turbines, due to the long continuous operation time and the erosion of different parts of the wind turbines by the external environment, the working state of the wind turbines may be abnormal. However, the maintenance and inspection of wind turbines usually need to be carried out manually on a regular basis. This method is not only difficult to find the safety hazards in the operation of wind turbines, but also takes a long time to respond to accidents. Summary of the invention
[0003] The present application provides a method, system, device and storage medium for detecting the working state of a wind turbine, which can detect whether the working state of the wind turbine is normal, thereby ensuring the normal operation of the wind turbine.
[0004] In a first aspect, an embodiment of the present application provides a method for detecting the working state of a wind turbine, comprising: obtaining multiple vibration waveform graphs of the wind turbine in a predetermined time period; inputting the vibration waveform graphs into a first convolutional neural network to obtain multiple vibration waveform feature vectors; based on the multiple vibration waveform feature vectors, obtaining a similarity feature matrix; based on the multiple vibration waveform feature vectors, obtaining a global vibration waveform feature matrix; inputting the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix; inputting the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result; wherein the classifier has learned the ability to predict various types of classification results based on the feature information in the first similar topology global vibration waveform feature matrix, the various types including a first type and a second type, the first type indicating that the working state of the wind turbine is normal, and the second type indicating that the working state of the wind turbine is abnormal; based on the classification result, detecting whether the working state of the wind turbine is normal.
[0005] In this technical solution, multiple vibration signals of the wind turbine can be processed to extract characteristic information from the multiple vibration signals, and a classification result indicating whether the working state of the wind turbine is normal can be obtained based on the characteristic information, thereby detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0006] In one implementation, obtaining a similarity feature matrix based on the multiple vibration waveform feature vectors includes: obtaining multiple similarities between the multiple vibration waveform feature vectors; obtaining a similarity input matrix based on the multiple similarities; and inputting the similarity input matrix into a second convolutional neural network to obtain the similarity feature matrix.
[0007] In an optional implementation, the similarity is obtained by:
[0008]
[0009] Among them, V i and V j are any two of the multiple vibration waveform feature vectors, For the V i The eigenvalues at each position of For the V j The characteristic values at each position, d(V i ,V j ) represents the V i and the V j The similarity between .
[0010] In one implementation, obtaining a global vibration waveform feature matrix based on the multiple vibration waveform feature vectors includes: position encoding the multiple vibration waveform feature vectors to obtain multiple position-coded feature vectors; fusing the multiple vibration waveform feature vectors with the multiple position-coded feature vectors to obtain multiple fused vibration waveform feature vectors; and arranging the multiple fused vibration waveform feature vectors in two dimensions to obtain the global vibration waveform feature matrix.
[0011] In an optional implementation, the formula for obtaining the position encoding feature vector is:
[0012] V p =f pos (V b )
[0013] Among them, V p is the position encoding feature vector, V b is the vibration waveform feature vector, f pos (·) is the position encoding function; the formula for obtaining the fusion vibration waveform feature vector is:
[0014] V c =Concat[V p ,V b ]
[0015] Among them, V cis the fusion vibration waveform feature vector, V p is the position encoding feature vector, V b is the vibration waveform feature vector, and Concat[·,·] is the cascade function.
[0016] In one implementation, the similarity feature matrix and the global vibration waveform feature matrix are input into a graph neural network to obtain a first similar topology global vibration waveform feature matrix, including: inputting the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a second similar topology global vibration waveform feature matrix; expanding the second similar topology global vibration waveform feature matrix to obtain multiple first similar topology global vibration waveform feature vectors; obtaining multiple wavelet function family-like energy aggregation factors of the multiple first similar topology global vibration waveform feature vectors; performing weighted correction on the multiple first similar topology global vibration waveform feature vectors based on the multiple wavelet function family-like energy aggregation factors to obtain multiple second similar topology global vibration waveform feature vectors; and obtaining the first similar topology global vibration waveform feature matrix based on the multiple second similar topology global vibration waveform feature vectors.
[0017] In an optional implementation, the formula for obtaining the energy aggregation factor of the wavelet-like function family is:
[0018]
[0019] Wherein, w is the energy aggregation factor of the wavelet-like function family, v i represents the eigenvalue of each position of the i-th one of the multiple first similar topology global vibration waveform feature vectors, and L is the length of the i-th one of the multiple first similar topology global vibration waveform feature vectors.
[0020] In one implementation, the classifier includes at least one fully connected layer and a normalized exponential Softmax classification function, wherein the step of inputting the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result includes: expanding the first similar topology global vibration waveform feature matrix to obtain a classification feature vector; inputting the classification feature vector into the fully connected layer to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the classification function to obtain the classification result.
[0021] In a second aspect, an embodiment of the present application provides a wind turbine working state detection system, comprising: an acquisition module, used to acquire multiple vibration waveform graphs of the wind turbine in a predetermined time period; a first processing module, used to input the vibration waveform graph into a first convolutional neural network to obtain multiple vibration waveform feature vectors; a second processing module, used to acquire a similarity feature matrix based on the multiple vibration waveform feature vectors; a third processing module, used to acquire a global vibration waveform feature matrix based on the multiple vibration waveform feature vectors; a fourth processing module, used to input the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix; a fifth processing module, used to input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result; wherein the classifier has learned to predict various types of classification results based on feature information in the first similar topology global vibration waveform feature matrix, the various types including a first type and a second type, the first type indicating that the wind turbine is in a normal working state, and the second type indicating that the wind turbine is in an abnormal working state; a detection module, used to detect whether the wind turbine is in a normal working state based on the classification result.
[0022] In one implementation, the second processing module is specifically used to: obtain multiple similarities between the multiple vibration waveform feature vectors; obtain a similarity input matrix based on the multiple similarities; input the similarity input matrix into a second convolutional neural network to obtain the similarity feature matrix.
[0023] In an optional implementation, the similarity is obtained by:
[0024]
[0025] Among them, V i and V j are any two of the multiple vibration waveform feature vectors, For the V i The eigenvalues at each position of For the V j The characteristic values at each position, d(V i ,V j ) represents the V i and V j The similarity between .
[0026] In one implementation, the third processing module is specifically used to: perform position encoding on the multiple vibration waveform feature vectors to obtain multiple position-encoded feature vectors; fuse the multiple vibration waveform feature vectors and the multiple position-encoded feature vectors to obtain multiple fused vibration waveform feature vectors; and arrange the multiple fused vibration waveform feature vectors in two dimensions to obtain the global vibration waveform feature matrix.
[0027] In an optional implementation, the formula for obtaining the position encoding feature vector is:
[0028] V p =f pos (V b )
[0029] Among them, V p is the position encoding feature vector, V b is the vibration waveform feature vector, f pos (·) is the position encoding function; the formula for obtaining the fusion vibration waveform feature vector is:
[0030] V c =Concat[V p ,V b ]
[0031] Among them, V c is the fusion vibration waveform feature vector, V p is the position encoding feature vector, V b is the vibration waveform feature vector, and Concat[·,·] is the cascade function.
[0032] In one implementation, the fourth processing module is specifically used to include: inputting the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a second similar topology global vibration waveform feature matrix; expanding the second similar topology global vibration waveform feature matrix to obtain multiple first similar topology global vibration waveform feature vectors; obtaining multiple wavelet function family-like energy aggregation factors of the multiple first similar topology global vibration waveform feature vectors; performing weighted correction on the multiple first similar topology global vibration waveform feature vectors based on the multiple wavelet function family-like energy aggregation factors to obtain multiple second similar topology global vibration waveform feature vectors; based on the multiple second similar topology global vibration waveform feature vectors, obtaining the first similar topology global vibration waveform feature matrix.
[0033] In an optional implementation, the formula for obtaining the energy aggregation factor of the wavelet-like function family is:
[0034]
[0035] Wherein, w is the energy aggregation factor of the wavelet-like function family, v i represents the eigenvalue of each position of the first similar topology global vibration waveform eigenvector, and L is the length of the first similar topology global vibration waveform eigenvector.
[0036] In one implementation, the classifier includes at least one fully connected layer and a normalized exponential Softmax classification function, wherein the fifth processing module is specifically used to: expand the first similar topology global vibration waveform feature matrix to obtain a classification feature vector; input the classification feature vector into the fully connected layer to obtain an encoded classification feature vector; input the encoded classification feature vector into the Softmax classification function to obtain the classification result.
[0037] In a third aspect, an embodiment of the present application provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for detecting the working status of a wind turbine as described in the first aspect.
[0038] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium for storing instructions, which, when executed, enables the method described in the first aspect to be implemented.
[0039] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method for detecting the working status of a wind turbine as described in the first aspect.
[0040] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present application.
[0042] Figure 1 is a schematic diagram of a method for detecting the working state of a wind turbine provided in an embodiment of the present application;
[0043] Figure 2 is a schematic diagram of another method for detecting the working state of a wind turbine provided in an embodiment of the present application;
[0044] Figure 3is a schematic diagram of another method for detecting the working state of a wind turbine provided in an embodiment of the present application;
[0045] Figure 4 is a schematic diagram of another method for detecting the working state of a wind turbine provided in an embodiment of the present application;
[0046] Figure 5 is a schematic diagram of another method for detecting the working state of a wind turbine provided in an embodiment of the present application;
[0047] Figure 6 This is an application scenario diagram of a method for detecting the working status of a wind turbine provided in an embodiment of the present application;
[0048] Figure 7 It is a schematic diagram of the architecture of a wind turbine operating status detection system provided in an embodiment of the present application;
[0049] Figure 8 is a schematic diagram of a wind turbine operating status detection system provided in an embodiment of the present application;
[0050] Fig. 9 is a schematic block diagram of an example electronic device that can be used to implement embodiments of the present application. DETAILED DESCRIPTION
[0051] The following is a description of exemplary embodiments of the present application in conjunction with the accompanying drawings, including various details of the embodiments of the present application to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.
[0052] In the description of this application, unless otherwise specified, " / " means or, for example, A / B can mean A or B; "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The various numbers such as first and second involved in this application are only distinguished for the convenience of description, and are not used to limit the scope of the embodiments of this application, nor do they indicate the order of precedence.
[0053] See also Figure 1 , Figure 1 Schematic diagram of a method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 1 As shown, the method may include but is not limited to the following steps:
[0054] Step S101: Acquire a plurality of vibration waveforms of a wind turbine in a predetermined time period.
[0055] In the embodiments of the present application, the wind turbine generator may be an onshore wind turbine generator.
[0056] For example, based on a plurality of vibration sensors pre-deployed at a plurality of different positions of a wind turbine, a plurality of vibration signals of the wind turbine within a preset period of time are acquired, and the plurality of vibration signals are processed to obtain a plurality of corresponding vibration waveform graphs.
[0057] It is understandable that, during normal operation of the wind turbine, a specific regular vibration signal will be generated, and the vibration signals at different positions of the wind turbine have different correlation characteristic distribution information. When the working state of the wind turbine is abnormal, the regular vibration signal characteristics will change. Therefore, the present application evaluates the working state of the wind turbine by monitoring and analyzing the vibration characteristics at various positions of the wind turbine.
[0058] Step S102: input the vibration waveform image into the first convolutional neural network to obtain multiple vibration waveform feature vectors.
[0059] For example, the vibration waveform graph is input into the first convolutional neural network to perform feature mining on the waveform graph of each vibration signal in the multiple vibration signals, and the local implicit feature distribution of the waveform graph of each vibration signal in the multiple vibration signals is extracted respectively, so as to obtain multiple vibration waveform feature vectors.
[0060] Step S103: obtaining a similarity feature matrix based on multiple vibration waveform feature vectors.
[0061] For example, the similarity between every two of the multiple vibration waveform feature vectors is obtained to obtain multiple similarities, and each similarity is used as an element of a matrix, thereby obtaining a similarity feature matrix.
[0062] Step S104: Based on the multiple vibration waveform feature vectors, a global vibration waveform feature matrix is obtained.
[0063] For example, feature extraction is performed on multiple vibration waveform feature vectors to obtain multiple feature information, and each feature information is used as an element of a matrix, thereby obtaining a global vibration waveform feature matrix.
[0064] Step S105: Input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a first similar topology global vibration waveform feature matrix.
[0065] For example, the similarity feature matrix and the global vibration waveform feature matrix are input into the graph neural network, so that the graph neural network encodes the graph structure data of the similarity feature matrix and the global vibration waveform feature matrix through learnable neural network parameters to obtain a similar topology global vibration waveform feature matrix containing irregular similarity logical association topological features and vibration feature information at each position.
[0066] Step S106: input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result.
[0067] Among them, in an embodiment of the present application, the classifier has learned the ability to predict various types of classification results based on the feature information in the first similar topology global vibration waveform feature matrix, and each type includes a first type and a second type. The first type indicates that the wind turbine is working normally, and the second type indicates that the wind turbine is working abnormally.
[0068] As an example, the first similar topology global vibration waveform feature matrix is input into a pre-trained classifier to obtain a first classification result indicating that the working state of the wind turbine is normal.
[0069] As another example, the first similar topology global vibration waveform feature matrix is input into a pre-trained classifier to obtain a second classification result indicating that the working state of the wind turbine is abnormal.
[0070] Step S107: Based on the classification result, detect whether the working state of the wind turbine generator is normal.
[0071] As an example, in response to the classification result being the first classification result, it is determined that the working state of the wind turbine is normal.
[0072] As another example, in response to the classification result being the second classification result, it is determined that the working state of the wind turbine is abnormal.
[0073] By implementing the embodiments of the present application, multiple vibration signals of the wind turbine obtained can be processed to extract characteristic information from the multiple vibration signals, and a classification result indicating whether the working state of the wind turbine is normal can be obtained based on the characteristic information, thereby detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0074] In one implementation, a similarity input matrix can be obtained based on multiple similarities between multiple vibration waveform feature vectors, and the similarity input matrix is input into the second convolutional neural network to obtain a similarity feature matrix. As an example, see Figure 2 , Figure 2FIG. 1 is a schematic diagram of another method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 2 As shown, the method may include but is not limited to the following steps:
[0075] Step S201: Acquire a plurality of vibration waveforms of a wind turbine in a predetermined time period.
[0076] In the embodiments of the present application, step S201 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0077] Step S202: Input the vibration waveform image into the first convolutional neural network to obtain multiple vibration waveform feature vectors.
[0078] For example, a vibration waveform graph is input into a first convolutional neural network to perform feature mining on the waveform graph of each vibration signal in a plurality of vibration signals, and local implicit feature distributions of the waveform graph of each of the plurality of vibration signals are extracted respectively, thereby obtaining a plurality of vibration waveform feature vectors. The first convolutional neural network has previously learned the ability to extract features from the waveform graph of the vibration signal.
[0079] In an optional implementation, after the vibration waveform image is input into the first convolutional neural network, each layer of the first convolutional neural network model performs the following operations on the vibration waveform image in the forward transfer of the layer: convolution processing is performed on the vibration waveform image to obtain a convolution feature image; the convolution feature image is pooled based on the feature matrix to obtain a pooled feature image; the pooled feature image is nonlinearly activated to obtain an activation feature image; wherein the output of the last layer of the first convolutional neural network is a plurality of vibration waveform feature vectors.
[0080] Step S203: obtaining multiple similarities between multiple vibration waveform feature vectors.
[0081] For example, the similarity obtaining formula is used to calculate the similarity between every two vibration waveform feature vectors in the multiple vibration waveform feature vectors, thereby obtaining multiple similarities between the multiple vibration waveform feature vectors.
[0082] In an optional implementation, the formula for obtaining the similarity is:
[0083]
[0084] Among them, V i and V j are any two of the multiple vibration waveform feature vectors, V i The eigenvalues at each position of V jThe characteristic values at each position, d(V i ,V j ) indicates V i and V j The similarity between them, i and j are positive integers.
[0085] Step S204: obtaining a similarity input matrix based on multiple similarities.
[0086] For example, each similarity is taken as an element of a matrix, thereby obtaining a similarity input matrix based on multiple similarities.
[0087] Step S205: Input the similarity input matrix into the second convolutional neural network to obtain a similarity feature matrix.
[0088] For example, the similarity input matrix is input into the pre-trained second convolutional neural network model for correlation feature mining to extract similarity logical association topological features between vibration mode features at various positions of the wind turbine, thereby obtaining a similarity feature matrix. The second convolutional neural network model has been pre-learned to obtain the ability to extract similarity logical association topological features between vibration mode features at various positions of the wind turbine.
[0089] Step S206: Based on the multiple vibration waveform feature vectors, a global vibration waveform feature matrix is obtained.
[0090] In the embodiments of the present application, step S206 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0091] Step S207: Input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a first similar topology global vibration waveform feature matrix.
[0092] In the embodiments of the present application, step S207 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0093] Step S208: inputting the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result.
[0094] In the embodiments of the present application, step S208 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0095] Step S209: Based on the classification result, detect whether the working state of the wind turbine generator is normal.
[0096] In the embodiments of the present application, step S209 can be implemented in any of the embodiments of the present application. The embodiments of the present application do not limit this and will not be described in detail.
[0097] By implementing the embodiments of the present application, a similarity input matrix can be obtained based on multiple similarities between multiple vibration waveform feature vectors, and the similarity input matrix can be input into a second convolutional neural network to obtain a similarity feature matrix. The similarity feature matrix is subsequently processed to obtain a classification result indicating whether the working state of the wind turbine is normal, thereby detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0098] In one implementation, multiple vibration waveform feature vectors may be position-encoded to obtain multiple position-encoded feature vectors, thereby obtaining a global vibration waveform feature matrix based on the multiple vibration waveform feature vectors and the multiple position-encoded feature vectors. As an example, see Figure 3 , Figure 3 is a schematic diagram of another method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 3 As shown, the method may include but is not limited to the following steps:
[0099] Step S301: Acquire a plurality of vibration waveforms of a wind turbine in a predetermined time period.
[0100] In the embodiments of the present application, step S301 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0101] Step S302: input the vibration waveform image into the first convolutional neural network to obtain multiple vibration waveform feature vectors.
[0102] In the embodiments of the present application, step S302 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0103] Step S303: obtaining a similarity feature matrix based on multiple vibration waveform feature vectors.
[0104] In the embodiments of the present application, step S303 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0105] Step S304: Position-encode the multiple vibration waveform feature vectors to obtain multiple position-encoded feature vectors.
[0106] For example, a position encoding function is used to perform position encoding on each of the multiple vibration waveform feature vectors to obtain a corresponding position encoding feature vector, thereby obtaining multiple position encoding feature vectors.
[0107] Among them, in the embodiment of the present application, the position encoding function can learn the relative position correlation feature information of the vibration implicit features at each position of each vibration waveform feature vector in a plurality of vibration waveform feature vectors.
[0108] In an optional implementation, the formula for obtaining the position encoding feature vector is:
[0109] V p =f pos (V b )
[0110] Among them, V p is the position encoding feature vector, V b is the vibration waveform characteristic vector, f pos (·) is the position encoding function;
[0111] Step S305: Fusing the multiple vibration waveform feature vectors and the multiple position coding feature vectors to obtain multiple fused vibration waveform feature vectors.
[0112] For example, each vibration waveform feature vector is fused with the corresponding position encoding feature vector to obtain a corresponding fused vibration waveform feature vector, thereby obtaining multiple fused vibration waveform feature vectors.
[0113] In an optional implementation, the formula for obtaining the fused vibration waveform feature vector is:
[0114] V c =Concat[V p ,V b ]
[0115] Among them, V c is the fusion vibration waveform feature vector, V p is the position encoding feature vector, V b is the vibration waveform feature vector, and Concat[·,·] is the cascade function.
[0116] Step S306: Arrange the multiple fused vibration waveform feature vectors in two dimensions to obtain a global vibration waveform feature matrix.
[0117] In the embodiments of the present application, step S306 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0118] Step S307: Input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a first similar topology global vibration waveform feature matrix.
[0119] In the embodiments of the present application, step S307 can be implemented in any of the embodiments of the present application. The embodiments of the present application do not limit this and will not be described in detail.
[0120] Step S308: input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result.
[0121] In the embodiments of the present application, step S308 can be implemented in any of the embodiments of the present application. The embodiments of the present application do not limit this and will not be described in detail.
[0122] Step S309: Based on the classification result, detect whether the working state of the wind turbine generator is normal.
[0123] In the embodiments of the present application, step S309 can be implemented in any of the ways in the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0124] By implementing the embodiments of the present application, multiple vibration waveform feature vectors can be position-encoded to obtain multiple position-encoded feature vectors, thereby obtaining a global vibration waveform feature matrix based on the multiple vibration waveform feature vectors and the multiple position-encoded feature vectors, and then subsequently processing the global vibration waveform feature matrix to obtain a classification result indicating whether the working state of the wind turbine is normal, thereby detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0125] In one implementation, multiple first similar topology global vibration waveform feature vectors can be weighted corrected based on multiple wavelet-like function family energy aggregation factors of the first similar topology global vibration waveform feature vectors to obtain multiple second similar topology global vibration waveform feature vectors, so that the second similar topology global vibration waveform feature matrix can be input into a pre-trained classifier to obtain a classification result. As an example, see Figure 4 , Figure 4 is a schematic diagram of another method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 4 As shown, the method may include but is not limited to the following steps:
[0126] Step S401: Acquire a plurality of vibration waveforms of a wind turbine in a predetermined time period.
[0127] In the embodiments of the present application, step S401 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0128] Step S402: Input the vibration waveform image into the first convolutional neural network to obtain multiple vibration waveform feature vectors.
[0129] In the embodiments of the present application, step S402 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0130] Step S403: obtaining a similarity feature matrix based on multiple vibration waveform feature vectors.
[0131] In the embodiments of the present application, step S403 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0132] Step S404: Based on the multiple vibration waveform feature vectors, a global vibration waveform feature matrix is obtained.
[0133] In the embodiments of the present application, step S404 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0134] Step S405: Input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a second similar topology global vibration waveform feature matrix.
[0135] In the embodiments of the present application, step S405 can be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0136] Step S406: Expand the second similar topology global vibration waveform feature matrix to obtain a plurality of first similar topology global vibration waveform feature vectors.
[0137] As an example, the second similar topology global vibration waveform feature matrix is expanded according to row vectors to obtain multiple first similar topology global vibration waveform feature vectors.
[0138] As another example, the second similar topology global vibration waveform feature matrix is expanded as a column vector to obtain a plurality of first similar topology global vibration waveform feature vectors.
[0139] Step S407: Obtain multiple wavelet-like function family energy aggregation factors of multiple first similar topology global vibration waveform feature vectors.
[0140] It can be understood that since each first similar topology global vibration waveform feature vector of the second similar topology global vibration waveform feature matrix is a similarity topological expression of the waveform signal characteristics of each vibration sensor, and the waveform signal characteristics acquired by each vibration sensor have a certain independence in terms of the vibration waveform, this makes the information expression consistency between the first similar topology global vibration waveform feature vectors poor. If the similar topology global vibration waveform feature matrix is directly classified by a classifier, it may cause inductive deviations between the similar topology global vibration waveform feature vectors. Therefore, the present application performs weighted correction on multiple first similar topology global vibration waveform feature vectors through multiple wavelet function family energy aggregation factors of multiple first similar topology global vibration waveform feature vectors to reduce the above-mentioned inductive deviation.
[0141] For example, using the wavelet function family energy aggregation factor acquisition formula, the wavelet function family energy aggregation factor corresponding to each first similar topology global vibration waveform feature vector is calculated respectively to obtain multiple wavelet function family energy aggregation factors of multiple first similar topology global vibration waveform feature vectors.
[0142] In an optional implementation, the formula for obtaining the energy aggregation factor of the wavelet-like function family is:
[0143]
[0144] Among them, w is the energy aggregation factor of the wavelet-like function family, v i Represents the eigenvalue of each position of the i-th one of multiple first similar topology global vibration waveform feature vectors, L is the length of the i-th one of multiple first similar topology global vibration waveform feature vectors, and i is a positive integer.
[0145] Step S408: performing weighted correction on a plurality of first similar topology global vibration waveform feature vectors based on a plurality of wavelet function family-like energy aggregation factors to obtain a plurality of second similar topology global vibration waveform feature vectors.
[0146] It is understandable that for high-dimensional manifolds, information representation tends to be concentrated on high-frequency components, so information tends to be distributed on the edge of the manifold. The wavelet-like function family, as a separable transformation for separating the edges in the feature distribution dimension, can convert the hidden state of high-dimensional features into frequency components and express the amount of information in a wavelet-like energy manner. Therefore, after weighting the first similar topology global vibration waveform feature vector using the wavelet-like function family as a weighting coefficient, multiple second similar topology global vibration waveform feature vectors are obtained, which can improve the information aggregation degree of each similar topology global vibration waveform feature vector of the second similar topology global vibration waveform feature matrix in its feature expression space, so as to improve the consistency of information expression between each similar topology global vibration waveform feature vector, thereby avoiding inductive bias and improving the accuracy of subsequent classification.
[0147] Step S409: Based on a plurality of second similar topology global vibration waveform feature vectors, a first similar topology global vibration waveform feature matrix is obtained.
[0148] For example, each of a plurality of second similar topology global vibration waveform feature vectors is used as a vector in a matrix, thereby obtaining a first similar topology global vibration waveform feature matrix.
[0149] Step S410: input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result.
[0150] In the embodiments of the present application, step S410 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0151] Step S411: Based on the classification result, detect whether the working state of the wind turbine generator is normal.
[0152] In the embodiments of the present application, step S411 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0153] By implementing the embodiments of the present application, the second similar topology global vibration waveform feature matrix can be weighted corrected to obtain the first similar topology global vibration waveform feature matrix, thereby improving the accuracy of the classification results obtained subsequently, thereby more accurately detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0154] In one implementation, the classifier includes multiple fully connected layers and a normalized exponential Softmax classification function, so that the first similar topology global vibration waveform feature matrix can be sequentially input into the multiple fully connected layers and the Softmax classification function to obtain a classification result. As an example, see Figure 5 , Figure 5 is a schematic diagram of another method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 5 As shown, the method may include but is not limited to the following steps:
[0155] Step S501: Acquire a plurality of vibration waveforms of a wind turbine in a predetermined time period.
[0156] In the embodiments of the present application, step S501 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0157] Step S502: Input the vibration waveform image into the first convolutional neural network to obtain multiple vibration waveform feature vectors.
[0158] In the embodiments of the present application, step S502 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0159] Step S503: obtaining a similarity feature matrix based on multiple vibration waveform feature vectors.
[0160] In the embodiments of the present application, step S503 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0161] Step S504: Based on the multiple vibration waveform feature vectors, a global vibration waveform feature matrix is obtained.
[0162] In the embodiments of the present application, step S504 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0163] Step S505: Input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a first similar topology global vibration waveform feature matrix.
[0164] In the embodiments of the present application, step S505 can be implemented in any of the ways in the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0165] Step S506: Expand the first similar topology global vibration waveform feature matrix to obtain a classification feature vector.
[0166] As an example, the first similar topology global vibration waveform feature matrix is expanded according to row vectors to obtain a classification feature vector.
[0167] As another example, the first similar topology global vibration waveform feature matrix is expanded as a column vector to obtain a classification feature vector.
[0168] Step S507: Input the classification feature vector into the fully connected layer to obtain the encoded classification feature vector.
[0169] For example, the classification feature vector is input into the fully connected layer for full connection encoding to obtain the encoded classification feature vector.
[0170] Step S508: Input the encoded classification feature vector into the Softmax classification function to obtain the classification result.
[0171] For example, the encoded classification feature vector is processed by the Softmax classification function to obtain the classification result.
[0172] Step S509: Based on the classification result, detect whether the working state of the wind turbine generator is normal.
[0173] In the embodiments of the present application, step S509 may be implemented in any of the embodiments of the present application, and the embodiments of the present application do not limit this and will not be described in detail.
[0174] By implementing the embodiments of the present application, the obtained first similar topology global vibration waveform feature matrix can be input into a classifier including multiple fully connected layers and a Softmax classification function to obtain a classification result to detect whether the working state of the wind turbine is normal and ensure the normal operation of the wind turbine.
[0175] In some embodiments of the present application, the above-mentioned method for detecting the working state of the wind turbine may further include: in response to determining that the working state of the wind turbine is abnormal, issuing a warning message, so that maintenance personnel can timely inspect and repair the wind turbine with abnormal working state based on the warning message to ensure the normal operation of the wind turbine.
[0176] See also Figure 6 , Figure 6 : is an application scenario diagram of a method for detecting the working status of a wind turbine provided in an embodiment of the present application. Figure 6 As shown, in this application scenario, multiple vibration sensors (for example, Figure 6 V1-Vn) in the figure obtains a plurality of vibration signals (for example, Figure 6 Then, the above signal is input to a server (for example, Figure 6In S) shown in , the server can process the above-mentioned input signal using the wind turbine working status detection method provided in any embodiment of the present application to generate a classification result indicating whether the working status of the wind turbine is normal.
[0177] See also Figure 7 , Figure 7 Schematic diagram of the structure of a wind turbine working status detection system provided in an embodiment of the present application. Figure 7 As shown, in the architecture of the system, first, a plurality of vibration signals of a predetermined time period collected by a plurality of vibration sensors deployed on a wind turbine are obtained; then, the waveform graphs of each vibration signal in the plurality of vibration signals obtained are respectively passed through a first convolutional neural network model as a filter to obtain a plurality of vibration waveform feature vectors; a position encoding function is used to positionally encode each vibration waveform feature vector in the plurality of vibration waveform feature vectors to obtain a plurality of position encoding feature vectors; then, each group of position encoding feature vectors and vibration waveform feature vectors are fused to obtain a plurality of fused vibration waveform feature vectors; the similarity between each two vibration waveform feature vectors in the plurality of vibration waveform feature vectors is calculated to obtain similarity input matrix; then, the similarity input matrix is passed through a second convolutional neural network as a feature extractor to obtain a similarity feature matrix; multiple fused vibration waveform feature vectors are arranged in two dimensions to obtain a global vibration waveform feature matrix; the obtained global vibration waveform feature matrix and the similarity feature matrix are passed through a graph neural network to obtain a similar topology global vibration waveform feature matrix; the high-dimensional feature manifold of the similar topology global vibration waveform feature matrix is corrected to obtain a corrected similar topology global vibration waveform feature matrix; further, the corrected similar topology global vibration waveform feature matrix is passed through a classifier to obtain a classification result, which is used to indicate whether the working state of the wind turbine is normal.
[0178] See also Figure 8 , Figure 8 Schematic diagram of a wind turbine operating status detection system provided in an embodiment of the present application. Figure 8As shown, the system 800 includes: an acquisition module 801, which is used to acquire multiple vibration waveform graphs of the wind turbine in a predetermined time period; a first processing module 802, which is used to input the vibration waveform graph into a first convolutional neural network to obtain multiple vibration waveform feature vectors; a second processing module 803, which is used to acquire a similarity feature matrix based on multiple vibration waveform feature vectors; a third processing module 804, which is used to acquire a global vibration waveform feature matrix based on multiple vibration waveform feature vectors; a fourth processing module 805, which is used to input the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix; a fifth processing module 806, which is used to input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result; a detection module 807, which is used to detect whether the working state of the wind turbine is normal based on the classification result.
[0179] In one implementation, the second processing module 803 is specifically used to: obtain multiple similarities between multiple vibration waveform feature vectors; obtain a similarity input matrix based on the multiple similarities; input the similarity input matrix into the second convolutional neural network to obtain a similarity feature matrix.
[0180] In an optional implementation, the formula for obtaining the similarity is:
[0181]
[0182] Among them, V i and V j are any two of the multiple vibration waveform feature vectors, V i The eigenvalues at each position of V j The characteristic values at each position, d(V i ,V j ) indicates V i and V j The similarity between .
[0183] In one implementation, the third processing module 804 is specifically used to: perform position encoding on multiple vibration waveform feature vectors to obtain multiple position-encoded feature vectors; fuse multiple vibration waveform feature vectors and multiple position-encoded feature vectors to obtain multiple fused vibration waveform feature vectors; and arrange multiple fused vibration waveform feature vectors in two dimensions to obtain a global vibration waveform feature matrix.
[0184] In an optional implementation, the formula for obtaining the position encoding feature vector is:
[0185] V p =f pos (Vb )
[0186] Among them, V p is the position encoding feature vector, V b is the vibration waveform characteristic vector, f pos (·) is the position encoding function; the formula for obtaining the fusion vibration waveform feature vector is:
[0187] V c =Concat[V p ,V b ]
[0188] Among them, V c is the fusion vibration waveform feature vector, V p is the position encoding feature vector, V b is the vibration waveform feature vector, and Concat[·,·] is the cascade function.
[0189] In one implementation, the fourth processing module 805 is specifically used to: input the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a second similar topology global vibration waveform feature matrix; expand the second similar topology global vibration waveform feature matrix to obtain multiple first similar topology global vibration waveform feature vectors; obtain multiple wavelet function family-like energy aggregation factors of the multiple first similar topology global vibration waveform feature vectors; perform weighted correction on the multiple first similar topology global vibration waveform feature vectors based on the multiple wavelet function family-like energy aggregation factors to obtain multiple second similar topology global vibration waveform feature vectors; based on the multiple second similar topology global vibration waveform feature vectors, obtain the first similar topology global vibration waveform feature matrix.
[0190] In an optional implementation, the formula for obtaining the energy aggregation factor of the wavelet-like function family is:
[0191]
[0192] Among them, w is the energy aggregation factor of the wavelet-like function family, v i Represents the eigenvalue of each position of the first similar topology global vibration waveform feature vector, and L is the length of the first similar topology global vibration waveform feature vector.
[0193] In one implementation, the classifier includes at least one fully connected layer and a normalized exponential Softmax classification function, wherein the fifth processing module 806 is specifically used to: expand the first similar topology global vibration waveform feature matrix to obtain a classification feature vector; input the classification feature vector into the fully connected layer to obtain an encoded classification feature vector; input the encoded classification feature vector into the Softmax classification function to obtain a classification result.
[0194] Through the system of the embodiment of the present application, multiple vibration signals of the wind turbine can be processed to extract characteristic information from the multiple vibration signals, and a classification result indicating whether the working state of the wind turbine is normal can be obtained based on the characteristic information, thereby detecting whether the working state of the wind turbine is normal to ensure the normal operation of the wind turbine.
[0195] Based on the embodiments of the present application, the present application further provides a computer-readable storage medium, wherein computer instructions are used to enable a computer to execute a method for detecting the working status of a wind turbine generator according to any of the aforementioned embodiments provided in the embodiments of the present application.
[0196] See also Fig. 9 ,like Fig. 9 , is a schematic block diagram of an example electronic device that can be used to implement an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.
[0197] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 to a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.
[0198] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0199] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSP), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as a method for detecting the working state of a wind turbine. For example, in some embodiments, the method for detecting the working state of a wind turbine may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the method for detecting the working state of a wind turbine described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the method for detecting the working status of the wind turbine generator in any other appropriate manner (for example, by means of firmware).
[0200] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system on chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0201] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0202] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0203] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) or LCD (Liquid Crystal Display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0204] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN), the Internet, and blockchain networks.
[0205] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS (Virtual Private Server) services. The server may also be a server of a distributed system, or a server combined with a blockchain.
[0206] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in the present application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of the present application can be achieved, and this document does not limit this.
[0207] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.
Claims
1. A method for detecting the working state of a wind turbine generator, characterized in that: include: Acquire multiple vibration waveforms of the wind turbine in a predetermined time period; Inputting the vibration waveform graph into a first convolutional neural network to obtain a plurality of vibration waveform feature vectors; Based on the multiple vibration waveform feature vectors, obtaining a similarity feature matrix; Based on the multiple vibration waveform feature vectors, obtaining a global vibration waveform feature matrix; Inputting the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix; Inputting the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result; wherein the classifier has learned to obtain the ability to predict various types of classification results based on the feature information in the first similar topology global vibration waveform feature matrix, and the various types include a first type and a second type, the first type indicates that the wind turbine is in a normal working state, and the second type indicates that the wind turbine is in an abnormal working state; Based on the classification result, it is detected whether the working state of the wind turbine is normal.
2. The method according to claim 1, characterized in that The obtaining of a similarity feature matrix based on the plurality of vibration waveform feature vectors comprises: Acquire multiple similarities between the multiple vibration waveform feature vectors; Acquire a similarity input matrix based on the multiple similarities; The similarity input matrix is input into a second convolutional neural network to obtain the similarity feature matrix.
3. The method according to claim 2, characterized in that The formula for obtaining the similarity is: Among them, V i and V j are any two of the multiple vibration waveform feature vectors, For the V i The eigenvalues at each position of For the V j The eigenvalues of each position, d(V i ,V j ) represents the V i and the V j The similarity between .
4. The method according to claim 1, characterized in that The step of obtaining a global vibration waveform feature matrix based on the multiple vibration waveform feature vectors includes: Position-encoding the plurality of vibration waveform feature vectors to obtain a plurality of position-encoded feature vectors; Fusing the multiple vibration waveform feature vectors and the multiple position coding feature vectors to obtain multiple fused vibration waveform feature vectors; The multiple fused vibration waveform feature vectors are arranged in two dimensions to obtain the global vibration waveform feature matrix.
5. The method according to claim 4, characterized in that: The formula for obtaining the position encoding feature vector is: V p =f pos (V b ) Among them, V p is the position encoding feature vector, V b is the vibration waveform feature vector, f pos (·) is the position encoding function; The formula for obtaining the fusion vibration waveform feature vector is: In c =Concat[V p ,V b ] Among them, V c is the fusion vibration waveform feature vector, V p is the position encoding feature vector, V b is the vibration waveform feature vector, and Concat[·,·] is the cascade function.
6. The method according to claim 1, characterized in that The step of inputting the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix comprises: Inputting the similarity feature matrix and the global vibration waveform feature matrix into the graph neural network to obtain a second similar topology global vibration waveform feature matrix; Expanding the second similar topology global vibration waveform feature matrix to obtain a plurality of first similar topology global vibration waveform feature vectors; Obtaining a plurality of wavelet-like function family energy aggregation factors of the plurality of first similar topology global vibration waveform feature vectors; Performing weighted correction on the plurality of first similar topology global vibration waveform feature vectors based on the plurality of wavelet-like function family energy aggregation factors to obtain a plurality of second similar topology global vibration waveform feature vectors; Based on the multiple second similar topology global vibration waveform feature vectors, the first similar topology global vibration waveform feature matrix is obtained.
7. The method according to claim 6, characterized in that The formula for obtaining the energy aggregation factor of the wavelet-like function family is: Wherein, w is the energy aggregation factor of the wavelet-like function family, v i represents the eigenvalue of each position of the i-th one of the multiple first similar topology global vibration waveform feature vectors, and L is the length of the i-th one of the multiple first similar topology global vibration waveform feature vectors.
8. The method according to claim 1, characterized in that The classifier includes at least one fully connected layer and a normalized exponential Softmax classification function, wherein the first similar topology global vibration waveform feature matrix is input into a pre-trained classifier to obtain a classification result, including: Expanding the first similar topology global vibration waveform feature matrix to obtain a classification feature vector; Inputting the classification feature vector into the fully connected layer to obtain an encoded classification feature vector; The encoded classification feature vector is input into the Softmax classification function to obtain the classification result.
9. A wind turbine operating status detection system, characterized in that: include: An acquisition module, used for acquiring a plurality of vibration waveforms of the wind turbine in a predetermined time period; A first processing module, used for inputting the vibration waveform image into a first convolutional neural network to obtain a plurality of vibration waveform feature vectors; A second processing module, configured to obtain a similarity feature matrix based on the plurality of vibration waveform feature vectors; A third processing module, configured to obtain a global vibration waveform feature matrix based on the plurality of vibration waveform feature vectors; A fourth processing module, used for inputting the similarity feature matrix and the global vibration waveform feature matrix into a graph neural network to obtain a first similar topology global vibration waveform feature matrix; a fifth processing module, configured to input the first similar topology global vibration waveform feature matrix into a pre-trained classifier to obtain a classification result; wherein the classifier has learned to obtain the ability to predict various types of classification results based on the feature information in the first similar topology global vibration waveform feature matrix, wherein the various types include a first type and a second type, wherein the first type indicates that the wind turbine is in a normal working state, and the second type indicates that the wind turbine is in an abnormal working state; The detection module is used to detect whether the working state of the wind turbine generator is normal based on the classification result.
10. An electronic device, characterized in that: include: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for detecting the working status of a wind turbine according to any one of claims 1 to 8.
11. A computer-readable storage medium for storing instructions, characterized in that: When the instructions are executed, the method according to any one of claims 1 to 8 is implemented.