A wind turbine generator set abnormal operation state alarm method

Abstract

In the wind turbine abnormal operation state alarm method, system and storage medium provided in the application, a plurality of vibration data packets of a wind turbine are acquired, sparse features of each vibration data packet in the plurality of vibration data packets are extracted, and a training data set and a test data set are constructed by using the sparse features. The training data set and the test data set are used to train a preset operation state recognition model to obtain a target operation state recognition model. A vibration data packet to be analyzed of the wind turbine is acquired, sparse features of the vibration data packet to be analyzed are extracted, and the sparse features of the vibration data packet to be analyzed are analyzed by the target operation state recognition model. If the state analysis result is a fault state, an alarm is given. Thus, the features extracted by the application are more comprehensive, so that the state analysis result of the wind turbine is more accurate, and the accuracy of the wind turbine abnormal operation state alarm is improved. Meanwhile, the process of extracting features by the application is more simple and suitable for a wide range.

Description

Technical Field

[0001] This application relates to the technical field of abnormal alarms for wind turbine generator sets, and in particular to a method, system, and storage medium for alarming abnormal operating conditions of wind turbine generator sets. Background Technology

[0002] Energy equipment, represented by wind turbines, is now widely used in the power industry. As the operating time of wind turbines increases, the probability of abnormal operating conditions also gradually increases. Once a wind turbine malfunctions, it not only affects normal production operations but also causes serious economic losses and may even trigger major safety accidents. Therefore, real-time monitoring of the operating status of wind turbines and identification of abnormal conditions are of great significance for ensuring the stable operation of wind turbines.

[0003] In related technologies, the main method for identifying the operating status of wind turbines is based on the operating status of the data itself. Its focus is on how to extract features from monitoring data, construct feature vectors or feature matrices, and build a mapping relationship between features and patterns to provide a basis for comprehensive decision-making on the operating status and pattern identification of wind turbines.

[0004] Common features generally include statistical features and features extracted using artificial intelligence methods. Traditional statistical features, such as mean, variance, margin, and power spectrum, are technically easy to obtain and are widely used to express specific characteristics of data. However, their simplistic nature leads to lower accuracy in state recognition analysis. Features extracted using artificial intelligence methods, on the other hand, abandon statistical indicators with specific physical or mathematical meanings, relying entirely on neural networks to express features. These networks quantify features using structural parameters after training, directly establishing a mapping relationship between features and patterns through extensive training. However, this method requires increasing the number of neural network layers to improve feature representation, leading to numerous uncertain parameters in the neural network structure design process. A good method for reasonably setting these parameters remains elusive. Summary of the Invention

[0005] This application provides a method, system, and storage medium for alarming abnormal operating conditions of wind turbine generator sets, in order to solve the technical problems that have arisen in the aforementioned related technologies.

[0006] The first aspect of this application proposes a method for alarming abnormal operating conditions of a wind turbine generator set, including:

[0007] Acquire multiple vibration data packets from the wind turbine generator;

[0008] Extract the sparse features of each vibration data packet from the plurality of vibration data packets, and use the sparse features to construct a training dataset and a test dataset;

[0009] The preset running state recognition model is trained using the training dataset and the test dataset to obtain the target running state recognition model;

[0010] The vibration data packet to be analyzed from the wind turbine is obtained, the sparse features of the vibration data packet to be analyzed are extracted, and the sparse features of the vibration data packet to be analyzed are analyzed by the target operating state identification model. If the analysis result is a fault state, an alarm is triggered.

[0011] A second aspect of this application provides an alarm system for abnormal operating status of a wind turbine generator set, comprising:

[0012] The acquisition module is used to acquire multiple vibration data packets from the wind turbine.

[0013] A construction module is used to extract sparse features from each vibration data packet in the plurality of vibration data packets, and to construct training datasets and test datasets using the sparse features;

[0014] The training module is used to train the preset running state recognition model using the training dataset and the test dataset to obtain the target running state recognition model;

[0015] The analysis module is used to acquire the vibration data packets of the wind turbine to be analyzed, extract the sparse features of the vibration data packets to be analyzed, and analyze the sparse features of the vibration data packets to be analyzed through the target operating state identification model. If the analysis result is a fault state, an alarm is triggered.

[0016] The computer storage medium proposed in the third aspect of this application stores computer-executable instructions; after being executed by a processor, the computer-executable instructions can implement the method described in the first aspect above.

[0017] The computer device proposed in the fourth aspect of this application includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it is able to implement the method described in the first aspect above.

[0018] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:

[0019] The wind turbine generator abnormal operation state alarm method, system, and storage medium proposed in this application acquire multiple vibration data packets from the wind turbine generator, extract sparse features from each vibration data packet, and construct training and testing datasets using these sparse features. A preset operation state recognition model is trained using the training and testing datasets to obtain a target operation state recognition model. Vibration data packets of the wind turbine generator to be analyzed are acquired, sparse features of these packets are extracted, and the target operation state recognition model is used to analyze these sparse features. If the state analysis result indicates a fault state, an alarm is triggered. Therefore, this application can extract sparse features corresponding to each vibration data packet based on multiple features within each data packet, resulting in more comprehensive extracted features and thus more accurate wind turbine generator state analysis results, improving the accuracy of wind turbine generator abnormal operation state alarms. Furthermore, this application does not require constructing complex neural networks or configuring parameters for neural networks during the extraction of sparse features corresponding to vibration data packets, making the feature extraction process simpler and more widely applicable.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0022] Figure 1 This is a flowchart illustrating an alarm method for abnormal operating status of a wind turbine generator set according to an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the structure of an alarm system for abnormal operation status of a wind turbine generator set according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] The following describes, with reference to the accompanying drawings, a method and system for alarming abnormal operating conditions of wind turbine generator sets according to embodiments of this application.

[0026] Example 1

[0027] Figure 1This is a flowchart illustrating an alarm method for abnormal operating status of a wind turbine generator set according to an embodiment of this application, as shown below. Figure 1 As shown, it may include:

[0028] Step 101: Obtain multiple vibration data packets from the wind turbine.

[0029] In the embodiments of this disclosure, the wind turbine may include multiple measurement points, and each vibration data packet may be vibration data at multiple measurement times corresponding to a certain measurement point.

[0030] For example, the vibration data packet is D = {x1, x2, ..., x...} i ,…,x n The vibration data packet contains vibration data at n measurement times, where x i It is the vibration data at the i-th measurement time.

[0031] Step 102: Extract the sparse features of each vibration data packet from multiple vibration data packets, and use the sparse features to construct the training dataset and the test dataset.

[0032] In this embodiment of the disclosure, the method for extracting sparse features of each vibration data packet from multiple vibration data packets may include the following steps:

[0033] Step a: Calculate the RMS (Root Mean Square) of each vibration data packet in the multiple vibration data packets.

[0034] In one embodiment of this disclosure, the method for calculating the RMS of each vibration data packet can be:

[0035]

[0036] Where x1, x2, ..., x n It represents the vibration data corresponding to each measurement moment in the original vibration data packet, where n is the total number of measurement moments corresponding to the vibration data packet.

[0037] Furthermore, in the embodiments of this disclosure, the RMS value of each vibration data packet can reflect the magnitude of the vibration energy of the vibration data packet and can be used as a reference for measuring whether the vibration data is abnormal.

[0038] Step b: Construct a sparse vector corresponding to each vibration data packet using the root mean square value of each vibration data packet.

[0039] In embodiments of this disclosure, the method for constructing a sparse vector corresponding to each vibration data packet using the root mean square value of each vibration data packet may include:

[0040]

[0041] By sequentially applying the vibration data from the vibration data packet to the above formula, the sparse vector 'a' corresponding to the vibration data packet is obtained. Where, 'a'... i For vibration data x i The corresponding vector.

[0042] Step c: Extract the sparse features corresponding to each sparse vector using the compression reconstruction method.

[0043] In the embodiments disclosed herein, the sparse vector a is used as the observation vector, the signal is reconstructed using a basis tracking method based on compressed sensing, and the obtained sparse coefficients θ are used as the sparse features corresponding to the sparse vector.

[0044] Specifically, in the embodiments of this disclosure, compressed sensing theory mainly includes signal sparse reconstruction. Utilizing the sparsity of signals, discrete samples of the signal are obtained through random sampling, and the signal is reconstructed using a nonlinear reconstruction algorithm. The principle of compressed sensing theory is as follows:

[0045]

[0046] Where a represents the sparse vector, i.e., the original signal; y is the obtained observation value; φ is the observation matrix; ψ is the sparse matrix; and θ is the sparse coefficient. Therefore, y = φ·ψ·θ, and the product of the observation matrix φ and the sparse matrix ψ is the sensing matrix. Specifically, the observation matrix φ and the sparse matrix ψ are the Gaussian random matrix and the identity matrix, respectively.

[0047] It should be noted that in the embodiments of this disclosure, since the sparse vector a is known, the observed value y can be obtained from the equation y = φ·a; as can be seen from the above theory, y = φ·ψ·θ, and based on the sparse reconstruction theory, the sparse coefficient θ is obtained. This process was originally intended to obtain the reconstructed sparse vector a, but here the sparse coefficient θ is used to construct a sparse feature matrix, which is used in subsequent steps to extract the sparse heat signature corresponding to each sparse vector.

[0048] Here, the sparse matrix ψ is an n x n identity matrix, meaning the diagonal elements are 1s and the remaining positions are 0s. The sparse coefficients θ corresponding to the sparse vector a are obtained by using a = ψ·θ, and these sparse coefficients θ are used as the sparse features corresponding to the sparse vector.

[0049] Furthermore, in the embodiments of this disclosure, the sparse vectors corresponding to each of the above-mentioned multiple vibration data packets are used to extract the sparse features corresponding to each vibration data packet using the above-described method.

[0050] Furthermore, in embodiments of this disclosure, the method for constructing training and testing datasets using sparse features may include the following steps:

[0051] Step 1: Determine the status label for each vibration data packet, where the status label is used to indicate the status of the vibration data packet.

[0052] In embodiments of this disclosure, the state of the vibration data packet may include a normal state and a fault state. Furthermore, in embodiments of this disclosure, the state label can be represented by 0 and 1. For example, in embodiments of this disclosure, 0 indicates that the vibration data packet is in a normal state, and 1 indicates that the vibration data packet is in a fault state.

[0053] Step 2: Match the sparse features of each vibration data packet with the corresponding state label of each vibration data packet.

[0054] Step 3: Construct training and testing datasets using the sparse features and state labels corresponding to each vibration data packet.

[0055] In the embodiments of this disclosure, the sparse features and state labels corresponding to a vibration data packet constitute a set of training or testing data. Furthermore, the sparse features and state labels corresponding to each vibration data packet from the acquired multiple vibration data packets are divided into training and testing datasets according to the ratio of the training dataset to the testing dataset. For example, if the training dataset accounts for 70% and the testing dataset accounts for 30%, then 70% of the data used for model training is randomly selected as the training dataset, and the remaining 30% is used as the testing dataset.

[0056] Step 103: Train the preset running state recognition model using the training dataset and the test dataset to obtain the target running state recognition model.

[0057] In the embodiments of this disclosure, the aforementioned preset operating state recognition model can be a CNN convolutional neural network model. Furthermore, in the embodiments of this disclosure, the aforementioned preset operating state recognition model may include: a convolutional layer, a pooling layer, a fully connected layer, and a classifier, wherein the last layer is a classifier used to output the state analysis result. Additionally, the input to the aforementioned preset operating state recognition model is the sparse features corresponding to the vibration data packet, and the output of the preset operating state recognition model is the state analysis result.

[0058] Furthermore, in the embodiments of this disclosure, the method of training a preset running state recognition model using a training dataset and a test dataset to obtain a target running state recognition model may include: training the preset running state recognition model using a training dataset to obtain a trained running state recognition model; testing the trained running state recognition model using a test dataset until the accuracy of the model analysis results reaches a threshold, at which point the training is complete and the target running state recognition model is obtained.

[0059] Furthermore, in the embodiments of this disclosure, the above-mentioned state analysis results may include normal state and fault state. Specifically, in the embodiments of this disclosure, the target operating state recognition model obtained through the above training process can identify sparse features corresponding to the normal state. Based on this, when there is a difference between the input sparse features to be analyzed and the sparse features corresponding to the normal state, the analysis result of the input sparse features to be analyzed can be determined as a fault state.

[0060] Step 104: Obtain the vibration data packet of the wind turbine to be analyzed, extract the sparse features of the vibration data packet to be analyzed, and analyze the sparse features of the vibration data packet to be analyzed through the target operating state identification model. If the state analysis result is a fault state, an alarm will be triggered.

[0061] In the embodiments of this disclosure, the sparse features of the vibration data packet to be analyzed can be extracted using the above method.

[0062] Furthermore, in the embodiments of this disclosure, the vibration data packets of the wind turbine can be analyzed in real time using the above method to determine in advance whether the vibration data packets of the wind turbine have malfunctioned. When the analysis result of the vibration data packets indicates a malfunction, an alarm is triggered, allowing maintenance personnel to promptly inspect and repair the wind turbine, thus avoiding serious economic losses. In the embodiments of this disclosure, the alarm methods can be varied; for example, alarms can be triggered by horns, lights, etc., and this disclosure does not limit the methods used.

[0063] In summary, the wind turbine generator abnormal operation status alarm method proposed in this application involves acquiring multiple vibration data packets from the wind turbine generator, extracting sparse features from each vibration data packet, constructing training and testing datasets using these sparse features, training a preset operation status recognition model using the training and testing datasets to obtain a target operation status recognition model, acquiring the vibration data packets to be analyzed from the wind turbine generator, extracting sparse features from these vibration data packets, and analyzing these sparse features using the target operation status recognition model. If the status analysis result indicates a fault state, an alarm is triggered. Therefore, this application can extract sparse features corresponding to each vibration data packet based on multiple features of the data in each vibration data packet, resulting in more comprehensive extracted features and thus more accurate wind turbine generator status analysis results, improving the accuracy of wind turbine generator abnormal operation status alarms. Furthermore, this application does not require constructing complex neural networks or configuring parameters for neural networks during the extraction of sparse features corresponding to vibration data packets, making the feature extraction process simpler and more widely applicable.

[0064] Example 2

[0065] Figure 2 This is a schematic diagram of the structure of a wind turbine generator abnormal operation alarm system according to an embodiment of this application, as shown below. Figure 2 As shown, the system may include:

[0066] The acquisition module 201 is used to acquire multiple vibration data packets of the wind turbine.

[0067] Module 202 is used to extract sparse features from each vibration data packet in multiple vibration data packets, and to use the sparse features to construct training and testing datasets.

[0068] Training module 203 is used to train a preset running state recognition model using training dataset and test dataset to obtain a target running state recognition model;

[0069] The analysis module 204 is used to acquire the vibration data packets of the wind turbine to be analyzed, extract the sparse features of the vibration data packets to be analyzed, and analyze the sparse features of the vibration data packets to be analyzed through the target operating state identification model. If the state analysis result is a fault state, an alarm will be triggered.

[0070] Optionally, module 202 is used specifically for:

[0071] Calculate the root mean square value of each vibration data packet in multiple vibration data packets;

[0072] Construct a sparse vector corresponding to each vibration data packet using the root mean square value of each vibration data packet;

[0073] The sparse features corresponding to each sparse vector are extracted using a compression reconstruction method.

[0074] Based on each sparse feature, a training dataset and a test dataset are constructed.

[0075] Optionally, module 202 is further specifically used for:

[0076] Determine the status label for each vibration data packet, where the status label includes normal status and fault status;

[0077] The sparse features of each vibration data packet are mapped one-to-one with the state label corresponding to each vibration data packet;

[0078] Training and testing datasets are constructed using the sparse features and state labels corresponding to each vibration data packet.

[0079] In summary, the wind turbine generator abnormal operation status alarm system proposed in this application acquires multiple vibration data packets from the wind turbine generator, extracts sparse features from each vibration data packet, and uses these sparse features to construct training and testing datasets. The system then trains a preset operation status recognition model using these datasets to obtain a target operation status recognition model. Finally, it acquires the vibration data packets to be analyzed from the wind turbine generator, extracts sparse features from these packets, and analyzes these features using the target operation status recognition model. If the analysis results indicate a fault state, an alarm is triggered. Therefore, this application can extract sparse features corresponding to each vibration data packet based on multiple features within each packet, resulting in more comprehensive features and thus more accurate wind turbine generator status analysis results, improving the accuracy of wind turbine generator abnormal operation status alarms. Furthermore, this application does not require the construction of complex neural networks or the configuration of parameters for these networks during the extraction of sparse features from vibration data packets, making the feature extraction process simpler and more widely applicable.

[0080] To implement the above embodiments, this disclosure also proposes a computer storage medium.

[0081] The computer storage medium provided in this embodiment stores an executable program; after the executable program is executed by a processor, it can achieve the following: Figure 1 The method shown.

[0082] To implement the above embodiments, this disclosure also proposes a computer device.

[0083] The computer device provided in this disclosure includes a memory, a processor, and a computer program stored in the memory and executable on the processor; when the processor executes the program, it can achieve the following: Figure 1 Any of the methods shown.

[0084] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0085] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0086] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for alarming abnormal operating status of a wind turbine generator set, characterized in that, The method includes: Acquire multiple vibration data packets from the wind turbine generator; Extract the sparse features of each vibration data packet from the plurality of vibration data packets, and use the sparse features to construct a training dataset and a test dataset; The preset running state recognition model is trained using the training dataset and the test dataset to obtain the target running state recognition model; The vibration data packet to be analyzed from the wind turbine is obtained, the sparse features of the vibration data packet to be analyzed are extracted, and the sparse features of the vibration data packet to be analyzed are analyzed by the target operating state identification model. If the state analysis result is a fault state, an alarm is triggered. The extraction of sparse features from each of the plurality of vibration data packets includes: Calculate the root mean square value of each vibration data packet in the plurality of vibration data packets: in, It represents the vibration data corresponding to each measurement moment in the original vibration data packet, where n is the total number of measurement moments corresponding to the vibration data packet. Construct a sparse vector corresponding to each vibration data packet using the root mean square value of each vibration data packet: By sequentially applying the vibration data from the vibration data packet to the above formula, the sparse vector corresponding to the vibration data packet can be obtained. α ,in, For vibration data The corresponding vector; The sparse features corresponding to each sparse vector are extracted using a compression reconstruction method. Extracting sparse features corresponding to each sparse vector using the compression reconstruction method includes: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] a As the observation vector, the signal is reconstructed using a basis tracking method based on compressed sensing, and the obtained sparse coefficients are... Sparse features are the sparse features corresponding to sparse vectors.

2. The method according to claim 1, characterized in that, The construction of the training and test datasets using the sparse features includes: Determine the status label for each vibration data packet, wherein the status label is used to indicate the status of the vibration data packet; The sparse features of each vibration data packet are mapped one-to-one with the state label corresponding to each vibration data packet; Training and testing datasets are constructed using the sparse features and state labels corresponding to each vibration data packet.

3. The method according to claim 1, characterized in that, The preset operating state recognition model includes: a convolutional layer, a pooling layer, a fully connected layer, and a classifier. The classifier is used to output the state analysis results. The input of the preset operating state recognition model is the sparse features corresponding to the vibration data packet, and the output of the preset operating state recognition model is the state analysis results.

4. The method according to claim 3, characterized in that, The step of training a preset running state recognition model using the training dataset and the test dataset to obtain a target running state recognition model includes: The preset running state recognition model is trained using the training dataset to obtain the trained running state recognition model; The trained running state recognition model is tested using the test dataset until the accuracy of the model analysis results reaches a threshold. Then the training is complete, and the target running state recognition model is obtained.

5. An alarm system for abnormal operating status of a wind turbine generator set, characterized in that, The system includes: The acquisition module is used to acquire multiple vibration data packets from the wind turbine. A construction module is used to extract sparse features from each vibration data packet in the plurality of vibration data packets, and to construct training datasets and test datasets using the sparse features; The training module is used to train the preset running state recognition model using the training dataset and the test dataset to obtain the target running state recognition model. The analysis module is used to acquire the vibration data packets of the wind turbine to be analyzed, extract the sparse features of the vibration data packets to be analyzed, and analyze the sparse features of the vibration data packets to be analyzed through the target operating state identification model. If the state analysis result is a fault state, an alarm is triggered. The building module is specifically used for: Calculate the root mean square value of each vibration data packet in the plurality of vibration data packets: in, It represents the vibration data corresponding to each measurement moment in the original vibration data packet, where n is the total number of measurement moments corresponding to the vibration data packet. Construct a sparse vector corresponding to each vibration data packet using the root mean square value of each vibration data packet: By sequentially applying the vibration data from the vibration data packet to the above formula, the sparse vector corresponding to the vibration data packet can be obtained. α ,in, For vibration data The corresponding vector; The sparse features corresponding to each sparse vector are extracted using a compression reconstruction method. Based on each of the sparse features, a training dataset and a test dataset are constructed; Extracting sparse features corresponding to each sparse vector using the compression reconstruction method includes: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] a As the observation vector, the signal is reconstructed using a basis tracking method based on compressed sensing, and the obtained sparse coefficients are... Sparse features are the sparse features corresponding to sparse vectors.

6. The system according to claim 5, characterized in that, The building module is further specifically used for: Determine the status label for each vibration data packet, wherein the status label includes normal status and fault status; The sparse features of each vibration data packet are mapped one-to-one with the state label corresponding to each vibration data packet; Training and testing datasets are constructed using the sparse features and state labels corresponding to each vibration data packet.

7. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it is able to implement the method as described in any one of claims 1-4.

8. A computer storage medium, wherein, The computer storage medium stores computer-executable instructions; when executed by a processor, the computer-executable instructions can implement the method described in any one of claims 1-4.

Images