Wind turbine generator fault diagnosis method and system based on SCADA (Supervisory Control And Data Acquisition) data and electronic equipment

The SCADA-based wind turbine fault diagnosis method simplifies model establishment and enhances reliability by using support vector machines and feature fusion, addressing the complexity and cost issues of existing vibration signal-based methods.

CN119062525AActive Publication Date: 2024-12-03HUBEI ENERGY GROUP RENEWABLE TECHNOLOGY CO LTD

Patent Information

Application Number
CN202411105965.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-12-03
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

Existing wind turbine fault detection methods based on vibration signals require complex model establishment and large amounts of experimental data, necessitating professional knowledge and high fault detection costs.

Method used

A wind turbine fault diagnosis method using SCADA data, involving training a fault detection model with support vector machines, generating new samples based on historical operation data, and performing feature fusion to enhance model reliability and reduce complexity.

Benefits of technology

The method simplifies model establishment, increases sample data, and ensures reliable fault detection by generating new samples around classification boundaries, reducing costs and improving detection efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119062525A_ABST
    Figure CN119062525A_ABST
Patent Text Reader

Abstract

The invention provides a wind turbine generator fault diagnosis method and system based on SCADA (Supervisory Control and Data Acquisition) data, namely electronic equipment. The method comprises the following steps: acquiring a plurality of sample operation data; training at least part of the sample operation data in the multiple pieces of sample operation data through a support vector machine algorithm, and determining support sample operation data which is boundary sample operation data for distinguishing positive sample operation data and negative sample operation data; determining a generation parameter according to the distance between the support sample operation data and the K nearest neighbor sample operation data; generating a plurality of new sample operation data according to the generation parameters to obtain training data of a fan fault diagnosis model; performing feature fusion on the training data to obtain fused training data; and training the fusion training data through a support vector machine algorithm to obtain a fan fault diagnosis model. The method solves the problem that an existing detection method is complex in detection model establishment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular to a wind turbine fault diagnosis method, system and electronic equipment based on SCADA data. Background Art

[0002] With the increase of wind power installed capacity, the normal operation of wind turbines is crucial to the reliability of wind power generation.

[0003] In the prior art, some wind farms have deployed condition monitoring systems that provide predictive maintenance. These condition monitoring systems rely on high-frequency sensing equipment to detect faults in wind turbines, resulting in high fault detection costs. In order to reduce operation and maintenance costs, fault detection methods based on equipment operating status are widely used in wind farm wind turbine fault detection. This detection method mainly detects wind turbine faults based on vibration signals collected from the main transmission system of the wind turbine. However, the analysis of vibration signals requires a large amount of experimental data and professional knowledge support. The establishment of the detection model needs to be combined with the structure, working principle and relationship between the various components of the wind turbine set, which makes the model establishment of this detection method complex. Summary of the invention

[0004] In view of this, the present invention proposes a wind turbine fault diagnosis method, system and electronic equipment based on SCADA data. The wind turbine fault detection model trained by this scheme can be obtained by training based on historical operation data, which solves the problem of complex detection model establishment in existing detection methods.

[0005] According to one aspect of the present invention, an embodiment of the present invention provides a wind turbine fault diagnosis method based on SCADA data, the method comprising:

[0006] Acquire multiple sample operation data, wherein the sample operation data refers to operation data corresponding to a fan fault state that has been calibrated, and the multiple sample operation data include multiple positive sample operation data and multiple negative sample operation data, wherein the fan fault state corresponding to the positive sample operation data is normal, and the fan fault state corresponding to the negative sample operation data is faulty; train at least part of the multiple sample operation data through a support vector machine algorithm to determine support sample operation data, wherein the support sample operation data is boundary sample operation data for distinguishing positive sample operation data from negative sample operation data; determine generation parameters according to the distance between the support sample operation data and K nearest neighbor sample operation data, where K is an integer greater than 1; generate multiple new sample operation data according to the generation parameters to obtain training data for a fan fault diagnosis model, wherein the training data include the multiple sample operation data and the multiple new sample operation data; perform feature fusion on the training data to obtain fused training data; train the fused training data through a support vector machine algorithm to obtain a fan fault diagnosis model.

[0007] Optionally, when the number of positive sample running data in the multiple sample running data is less than the number of negative sample running data, determining the generation parameter according to the distance between the supporting sample running data and the K nearest neighboring sample running data includes:

[0008] Determine the positive supporting sample running data in the supporting sample running data; calculate the first distance between each positive supporting sample running data and each positive sample running data in the corresponding K nearest neighbor sample running data; and determine the generation parameters corresponding to each positive supporting sample running data according to the first distance.

[0009] Optionally, determining the generation parameters corresponding to each positive supporting sample running data according to the first distance and the second distance includes:

[0010] The mean of the first distances is calculated to obtain the generation parameter.

[0011] Optionally, generating a plurality of new sample operation data according to the generation parameters includes:

[0012] When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is less than the number of negative sample running data, the new sample running data is:

[0013] X new =x SV+ +δ(x nn+ -x SV+ )+αγ 0 ;

[0014] When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is greater than the number of negative sample running data, the new sample running data is:

[0015] X new =x SV+ -δ(x nn+ -x SV+ )+αγ 0 ;

[0016] Among them, x SV+ is any one of the positive support sample running data, δ is a random number in the interval (0,1), x nn+ For x SV+ Any positive sample running data among the corresponding K nearest neighbor sample running data, α is a random number in the interval (0,1), γ 0 Generate parameters for the above.

[0017] Optionally, when the number of positive sample running data in the multiple sample running data is greater than the number of negative sample running data, the generation parameters are determined based on the distance between the supporting sample running data and the K nearest neighbor sample running data, including: determining the negative supporting sample running data in the supporting sample running data; calculating the third distance between each negative supporting sample running data and each negative sample running data in the corresponding K nearest neighbor sample running data; and determining the generation parameters corresponding to each negative support sample running data based on the third distance.

[0018] The determining, according to the third distance, a generation parameter corresponding to each of the negative support sample running data comprises:

[0019] The mean of the third distance is calculated to obtain the generation parameter.

[0020] Optionally, generating a plurality of new sample operation data according to the generation parameters includes:

[0021] When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is less than the number of positive sample running data, the new sample running data is:

[0022] X new =x SV- +δ(x nn- -x SV- )+βγ 0 ;

[0023] When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is greater than the number of positive sample running data, the new sample running data is:

[0024] Xnew =x SV- -δ(x nn- -x SV- )+βγ 0 ;

[0025] Among them, x SV- Run the data for any of the negative support samples, δ is a random number in the interval (0,1), x nn- For x SV- The corresponding K nearest neighbor sample running data is any negative sample running data, β is a random number in the interval (0,1), γ 0 Generate parameters for the above.

[0026] Optionally, performing feature fusion on the training data to obtain fused training data includes:

[0027] Determine a neighborhood set corresponding to each of the training data, wherein the neighborhood set includes N training data that are closest to the corresponding training data; calculate a linear weight coefficient between each of the training data and the N training data in the corresponding neighborhood set; optimize a preset objective function based on the linear weight coefficient until the preset objective function converges to obtain the fused training data; the preset objective function includes regularization and the sum of errors.

[0028] According to another aspect of the present invention, the present invention provides a wind turbine fault diagnosis system based on SCADA data, the system comprising:

[0029] A data acquisition module, used to acquire fan operation data;

[0030] A feature fusion module performs feature fusion on the wind turbine operation data to obtain fused data;

[0031] A fault diagnosis module inputs the fused data into a wind turbine fault diagnosis model to perform fault diagnosis on the wind turbine; wherein the wind turbine fault diagnosis model is obtained according to the above-mentioned wind turbine fault diagnosis method based on SCADA data.

[0032] According to another aspect of the present invention, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a calculation program, and when the calculation program is executed by the processor, the processor implements the above-mentioned wind turbine fault diagnosis method based on SCADA data.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] (1) The fan fault detection model trained by this method can be obtained by training based on historical operation data, which solves the problem of complex detection model establishment in existing detection methods.

[0035] (2) During the model training process, new sample operation data is generated based on the acquired historical operation data, which increases the number of samples and makes the trained fault detection model more reliable.

[0036] (3) When generating new sample running data, the boundary between the positive sample running data and the negative sample running data is determined by the support vector machine method, and the new sample running data is generated based on the support sample running data at the boundary, thereby ensuring the reliability of the newly generated sample type.

[0037] (4) When generating new sample running data, the generation parameters are determined according to the distance from the supporting sample running data. The newly generated samples are located around the boundary, ensuring the reliability of the classification boundary. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flow chart of a wind turbine fault diagnosis method, system and electronic device fault diagnosis method based on SCADA data according to an embodiment of the present invention;

[0040] Figure 2 A flow chart of a method for determining generation parameters of a wind turbine fault diagnosis method, system and electronic device fault diagnosis method based on SCADA data according to an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of generating new sample operation data in a wind turbine fault diagnosis method, system and electronic device based on SCADA data according to an embodiment of the present invention;

[0042] Figure 4 A flow chart of a method for determining parameters generated in a wind turbine fault diagnosis method, system and electronic device based on SCADA data according to another embodiment of the present invention;

[0043] Figure 5 A flow chart of the wind turbine fault diagnosis method, system and electronic equipment system based on SCADA data according to an embodiment of the present invention;

[0044] Figure 6The figure is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] like Figure 1 As shown, an embodiment of the present invention provides a wind turbine fault diagnosis method based on SCADA data, comprising the following steps:

[0047] S10: Acquire multiple sample operation data.

[0048] In this step, the sample operation data is the historical operation data of the fan. The historical operation data is the recorded data of the fan operation before model training. In the absence of recorded data, the fan monitoring system can be sampled. When the fan is running, the monitoring system collects and stores the operating parameters of the fan at a certain time interval to facilitate the analysis of the fan operation status.

[0049] In the embodiment of the present application, the wind turbine operation data includes the generator set related parameters, operation status parameters and fault status marks of the wind turbine.

[0050] Specifically, the generator set related parameters include the rated power of the wind turbine generator set, the cut-in wind speed, the cut-out wind speed, the blade speed, the number of stages and the total ratio of the transmission, the speed range of the generator, and the rated speed range of the generator.

[0051] For each turbine, two separate sets of data are required: SCADA data and status data. Specifically, SCADA data contains 58 parameters for each turbine at one-minute intervals. Status data provides information about fault conditions. The relevant parameters for each wind turbine in the wind farm include rated power, cut-in speed, cut-out speed, blade speed, number of stages and total ratio of the transmission, generator speed range, and rated speed range of the generator.

[0052] SCADA data can be collected once per second. For each turbine, 58 parameters are generally available in the SCADA data, which are divided into four categories.

[0053] 1) Vibration parameters: including vibration data such as spindle radial vibration, spindle axial vibration and spindle torsional vibration, which can reflect a series of problems such as mechanical looseness, bearing or gear failure, etc.

[0054] 2) Health parameters: including main bearing temperature, low-speed shaft temperature, high-speed shaft temperature and gearbox oil temperature, which help analyze the health of the wind turbine.

[0055] 3) Performance parameters: including rotor speed, impeller speed, generator speed and active power, etc., which are used to measure the operating performance of wind turbines.

[0056] 4) Control parameters: refers to the status of the programmable logic controller code (PLC). There are 17 PLC status codes, including system initialization, system interruption, system maintenance, system power generation, etc.

[0057] The status data provides information about the fault status and mainly includes five parameters: wind turbine number, fault cause, maintenance activity of the fault, start date of maintenance and end date of maintenance.

[0058] The fault status mark is identification data of the fan fault status. A set of fan operation data corresponds to one status mark. The status mark of the normal operation of the fan is different from the status mark of the fan fault.

[0059] The obtained multiple sample operation data can be divided into positive sample operation data and negative sample operation data according to the fault status identification. The fan fault status corresponding to the positive sample operation data is normal, and the fan fault status corresponding to the negative sample operation data is fault.

[0060] In some embodiments, the sample operation data obtained is obtained by processing the historical monitoring data obtained. Specifically, the original operation data obtained directly may cause data errors due to the failure of the data acquisition system. Data cleaning is performed on the original operation data obtained to ensure the reliability of the sample operation data obtained. Specifically, errors are identified based on the attribute fields and record types in the original operation data obtained. In the scope of attributes, there are two problems: missing values ​​and wrong field values. Missing values ​​are generally values ​​that are unavailable during data input, and wrong field values ​​are generally inappropriate data input locations. In the scope of record types, there are two problems: duplicate records and contradictory records. Duplicate records are generally because the same record is represented twice; contradictory records are generally caused by records with the same key being described by different values. Delete the original operation data with errors to clean the acquired original operation data.

[0061] In some embodiments, the amount of data after cleaning is reduced. In order to expand the amount of data, the fan monitoring system is resampled to obtain sample operation data. For the resampled data, the data integrity is checked by the following formula:

[0062]

[0063] Where P di is data integrity; N r is the number of instances measured in a sampling period; N cc It is data that meets the consistency, that is, data that has passed the internal consistency check. The internal consistency check is to see whether there is logical consistency between the various attributes in the data. Here, the focus is on checking whether the data values ​​are within a reasonable range. When the data integrity of the collected data reaches the preset value, the collected data is valid, otherwise it is resampled. For example, the data is resampled at 10-minute intervals. If the proportion of data that meets the consistency reaches 90%, the collected data is valid. The resampled SCADA data is defined as the data set of this method.

[0064] Most wind turbines are fault-free most of the time. Therefore, the number of normal instances in the dataset is significantly greater than the number of faulty instances, which leads to an imbalanced dataset, making it impossible for the algorithm or model to accurately identify the complete signal features, resulting in overfitting of the learning process and low reliability and accuracy of the results. Therefore, it is necessary to generate a balanced dataset.

[0065] S20: Training at least part of the sample operation data among the multiple sample operation data by using a support vector machine algorithm to determine supporting sample operation data.

[0066] The support vector machine (SVM) algorithm is a supervised learning algorithm. Its core idea is to find an optimal hyperplane in the feature space for classification and maximize the interval between two categories.

[0067] In this step, at least part of the sample operation data used to train the support vector machine includes positive sample operation data and negative sample operation data. The support sample operation data is the boundary sample operation data that distinguishes the positive sample operation data from the negative sample operation data. The distance from each support sample operation data to the optimal hyperplane is the shortest. When training the support vector machine, any kernel function can be selected for training, such as a linear kernel function, a Gaussian kernel function, etc., but the embodiment of the present invention is not limited thereto.

[0068] S30: Determine generation parameters according to the distance between the supporting sample running data and the K nearest neighbor sample running data.

[0069] In this step, the generation parameter is a calculation coefficient when generating new sample running data. The generation parameter represents the distance between the new sample and the supporting sample running data. The larger the generation parameter, the closer the distance between the new sample running data and the supporting sample running data. When generating multiple new sample running data, determine the K nearest neighbor sample running data closest to each supporting sample running data. Among them, K is an integer greater than 1, and the value of K can be selected according to the number of new sample running data to be generated. The more the number of new sample running data to be generated, the larger the value of K. Conversely, the fewer the number of new sample running data to be generated, the smaller the value of K.

[0070] For different supporting sample operation data, the number of selected nearest neighbor sample operation data can be the same or different. For any supporting sample operation data, the distance between the surrounding sample operation data and the supporting sample operation data is calculated with the supporting sample operation data as the center, and the first K sample operation data are selected according to the order of distance from small to large. For any sample operation data among the K sample operation data, the generation parameter is determined according to the distance between it and the supporting sample operation data. The larger the distance, the larger the generation parameter. The embodiment of the present invention does not limit the specific determination method of the generation parameter. For example, in one embodiment, K generation parameters of different sizes can be pre-set. When determining the generation parameter, the generation parameter of the corresponding size is selected according to the distance between the K sample operation data and the same supporting sample operation data. Assuming that K is 3, 3 generation parameters of different sizes can be pre-set, such as 0.2, 0.5, and 0.8. Among the 3 sample operation data, the generation parameter corresponding to the sample operation data closest to the supporting sample operation data is 0.8, and the generation parameter corresponding to the sample operation data closest to the supporting sample operation data is 0.2.

[0071] S40: Generate a plurality of new sample operation data according to the generated parameters to obtain training data for the fan fault diagnosis model.

[0072] In this step, the position of the new sample operation data is between the supporting sample operation data and the nearest neighbor sample operation data. The distance difference between each nearest neighbor sample operation data and the supporting sample operation data is calculated, and the product of the difference and the generation parameter and the sum of the supporting sample operation data is the new sample operation data. The new sample operation data is combined with the sample operation data in step S10 to obtain the training data of the fan fault diagnosis model.

[0073] In some embodiments, multiple new sample operation data may be generated between a supporting sample operation data and a nearest neighbor sample operation data. That is, after determining the new sample operation data according to the generation parameters, new sample operation data may be generated between the new sample operation data and the supporting sample operation data, and new sample operation data may be generated between the new sample operation data and the nearest neighbor sample data. The specific generation method is the same as steps S30 and S40, and is not described here for brevity.

[0074] S50: Perform feature fusion on the training data to obtain fused training data.

[0075] Since not all features in the SCADA data are related to generator failure, we first determine the subset of features collected by the SCADA system that reflect the operating status of the generator. More specifically, the features for analyzing wind turbine failures belong to the vibration parameters, health parameters, and performance parameters mentioned above. Table 1 summarizes the selected features and their potential relevance for the purpose of generator fault prediction and diagnosis.

[0076] Table 1 Inference rules for fault diagnosis

[0077]

[0078] The features described in Table 1 contain a certain degree of redundancy. The training data includes feature data of multiple dimensions, and the feature data of each dimension corresponds to a parameter of the fan, and the features of each dimension are correlated. For example, the main shaft radial vibration data and the main bearing temperature are healthy data. The larger the vibration amplitude of the fan, the larger the main shaft radial vibration parameter, and the higher the main bearing temperature. The temperature of the low-speed shaft is closely related to the temperature of the high-speed shaft, and their correlation coefficient is greater than 0.98. The active power and reactive power of the generator are also closely related, and their correlation coefficient is greater than 0.95. Feature fusion is performed on the feature data of multiple dimensions, that is, multiple correlated feature data are represented by a fusion feature to obtain fused training data. The specific fusion method can be a principal component analysis method, a linear discriminant analysis method, etc. The embodiment of the present invention is not limited to a specific fusion method. The dimension of the fused training data is significantly smaller than the dimension of the training data, thereby reducing the complexity of model training.

[0079] S60: Train the fused training data using a support vector machine algorithm to obtain a fan fault diagnosis model.

[0080] To use the SCADA system to monitor the operating status of the generator, it is necessary to map from the feature space to the state space and determine whether the generator is in a normal working state through the features. This application achieves the ultimate goal of fault diagnosis of wind turbines through an improved support vector machine algorithm SVM.

[0081] In this step, the improved support vector machine algorithm SVM uses the radial basis function RBF as the kernel function to map the fused training data to the high-dimensional space so that the fused training data is linearly separable in the high-dimensional space. In some embodiments, the kernel function may also use other types of kernel functions, such as Gaussian kernel function, etc., and the embodiments of the present invention are not limited thereto. In high-dimensional space, the model is prone to overfitting the training data, resulting in poor generalization ability on new data, so further adding regularization to the objective function can control the complexity of the model, prevent overfitting, and improve the generalization ability of the model.

[0082] The core of the improved support vector machine algorithm SVM is to find a hyperplane that maximizes the classification interval, and the standard Hinge loss can effectively punish misclassification and samples close to the classification interval. By minimizing the Hinge loss function, it helps to find a better classification hyperplane and better distinguish different categories of data.

[0083] The wind turbine fault diagnosis method based on SCADA data provided by the embodiment of the present invention can obtain a detection model by training historical operation data, thereby solving the problem of complex detection model establishment of existing detection methods; in the model training process, new sample operation data is generated based on the acquired sample operation data, thereby increasing the number of samples and making the fault detection model obtained by training more reliable; when generating new sample operation data, the boundary between the positive sample operation data and the negative sample operation data is determined by the support vector machine method, and new sample operation data is generated based on the support sample operation data at the boundary, thereby ensuring the reliability of the newly generated sample type. When generating new sample operation data, the generation parameters are determined according to the distance from the support sample operation data, and the newly generated sample is located around the boundary, thereby ensuring the reliability of the classification boundary.

[0084] In some embodiments, the number of positive sample running data in the plurality of sample running data is less than the number of negative sample running data. In this case, new positive sample running data is generated to achieve a balance between the positive sample running data and the negative sample running data. Figure 2 , Figure 2 The method for determining the generation parameters in this embodiment is shown, which specifically includes the following steps:

[0085] S301: Determine positive supporting sample operation data among the supporting sample operation data.

[0086] In this step, the supporting sample operation data whose corresponding fan fault status is calibrated as normal is the positive supporting sample operation data.

[0087] S302: Calculate a first distance between each positive supporting sample running data and each positive sample running data in the corresponding K nearest neighbor sample running data.

[0088] In this step, the K nearest neighbor sample running data of a positive support sample running data include positive sample running data and negative sample running data. The distance between a positive support sample running data and any positive sample running data of its corresponding K nearest neighbor sample running data is the first distance. The first distance can be the absolute value of the difference between the positive support sample running data and the nearest neighbor positive sample running data, or the Euclidean distance between the positive support sample running data and the nearest neighbor positive sample running data.

[0089] S303: Determine generation parameters corresponding to each positive supporting sample running data according to the first distance.

[0090] In this step, the generation parameter reflects the distance between the newly generated sample and the positive support sample running data. For the same positive support sample running data, the corresponding generation parameters are the same, and the generation parameters corresponding to different positive support vectors are different. When multiple positive sample running data in the K nearest neighbor sample running data are close to the positive support sample running data, the generation parameter is small, and the newly generated sample running data is close to the positive sample running data of the nearest neighbor. When multiple positive sample running data in the K nearest neighbor sample running data are far away from the positive support sample running data, the generation parameter is large, and the newly generated sample running data is close to the positive support sample running data of the nearest neighbor. By generating parameters, the decision boundary can be balanced while ensuring the category of the newly generated samples. In one embodiment, the first distance mean of each positive sample running data in a positive support sample running data and its corresponding K nearest neighbor sample running data is used as a generation parameter. In other embodiments, other methods can also be used, for example, the generation parameters are determined by using a pre-set distance and generation parameter correspondence table.

[0091] After the generation parameters are determined, new sample operation data is generated according to the method described in step S40 to obtain training data.

[0092] In some embodiments, when the number of positive sample operation data among the acquired multiple sample operation data is less than the number of negative sample operation data, the generated new sample operation data is the positive sample operation data. The specific generation method is:

[0093] When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is less than the number of negative sample running data, the new sample running data is:

[0094] X new =x SV+ +δ(x nn+ -x SV+ )+αγ 0 ;

[0095] When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is greater than the number of negative sample running data, the new sample running data is:

[0096] X new =x SV+ -δ(x nn+ -x SV+ )+αγ 0 ;

[0097] Among them, x SV+ is any positive support sample running data, δ is a random number in the interval (0,1), x nn+ For x SV+ Any positive sample running data among the corresponding K nearest neighbor sample running data, α is a random number in the interval (0,1), γ 0 is the generation parameter. Both δ and α are random numbers in the interval (0,1), and they may be the same or different.

[0098] Here the generating parameter γ 0 The first distance mean of the samples is expressed as where k + is the number of nearest neighbors of the positive class; x SV is the minority class support vector; Represents x SV The jth positive nearest neighbor of .

[0099] In order to more clearly illustrate the new sample generation method in the embodiment of the present invention, Figure 3 A schematic diagram of a new sample generation method is shown. Figure 3 As shown, the square is the positive sample running data, and the circle is the negative sample running data. The number of positive sample running data is less than the number of negative sample running data. In order to achieve sample balance, positive sample running data needs to be generated. Assume x i Run the data for any positive support sample in the support vector, x i The number of positive support sample running data in the corresponding K nearest neighbor sample running data is greater than the number of negative support sample running data, where any positive sample running data is x k , then according to the above calculation method, the new sample running data generated is shown as the triangle in the figure. j Run the data for any positive support sample in the support vector, x j The number of negative support sample running data in the corresponding K nearest neighbor sample running data is greater than the number of positive support sample running data, and any positive sample running data is x l , then according to the above calculation method, the new sample running data generated is shown as the star in the figure.

[0100] In this embodiment, when the number of positive sample operation data among the acquired multiple sample operation data is less than the number of negative sample operation data, the generated new sample operation data is positive sample operation data, which ensures the balance of the samples and makes the trained fault diagnosis model more reliable.

[0101] Furthermore, if the positive sample running data accounts for the majority among the nearest neighbor sample running data corresponding to the positive support sample running data, it can be known that the boundary sample is in an area where the two types of data are far apart, so the positive support sample running data is interpolated; when the negative sample running data accounts for the majority among the nearest neighbor sample running data, it can be known that the positive support sample running data sample is in the overlapping area of ​​the positive sample running data and the negative sample running data in the kernel space, so the positive support sample running data is interpolated; using this method, the classification surface is made more ideal.

[0102] In addition, when generating new sample running data, the generation parameters are determined based on the distance from the supporting sample running data. Therefore, adding generation parameter operations to generate new sample running data can ensure that the generated new sample running data is located around the boundary, thereby ensuring the reliability of the classification boundary.

[0103] In some embodiments, the number of negative sample running data in the plurality of sample running data is less than the number of positive sample running data. In this case, new negative sample running data is generated to achieve a balance between the positive sample running data and the negative sample running data. Figure 4 , Figure 4 The method for determining the generation parameters in this embodiment is shown, which specifically includes the following steps:

[0104] S311: Determine negative supporting sample running data in the supporting sample running data.

[0105] In this step, the supporting sample operation data whose corresponding wind turbine fault status is marked as fault is negative supporting sample operation data.

[0106] S312: Calculate the third distance between each negative support sample running data and each negative sample running data in the corresponding K nearest neighbor sample running data.

[0107] In this step, the K nearest neighbor sample running data of a negative support sample running data include positive sample running data and negative sample running data. The distance between a negative support sample running data and any positive sample running data of its corresponding K nearest neighbor sample running data is the third distance. The third distance can be the absolute value of the difference between the negative support sample running data and the nearest neighbor negative sample running data, or the Euclidean distance between the negative support sample running data and the nearest neighbor negative sample running data.

[0108] S313: Determine generation parameters corresponding to each negative support sample running data according to the third distance.

[0109] In this step, the generation parameter reflects the distance between the newly generated sample and the negative support sample running data. For the same negative support sample running data, the corresponding generation parameters are the same, and the generation parameters corresponding to different negative support vectors are different. When the multiple negative sample running data in the K nearest neighbor sample running data are close to the negative support sample running data, the generation parameter is small, and the newly generated sample running data is close to the negative sample running data of the nearest neighbor. When the multiple negative sample running data in the K nearest neighbor sample running data are far away from the negative support sample running data, the generation parameter is large, and the newly generated sample running data is close to the negative support sample running data of the nearest neighbor. By generating parameters, the decision boundary can be balanced while ensuring the category of the newly generated samples. In one embodiment, the third distance mean of each negative sample running data in a negative support sample running data and its corresponding K nearest neighbor sample running data is used as a generation parameter. In other embodiments, other methods can also be used, for example, the generation parameters are determined by using a pre-set distance and generation parameter correspondence table.

[0110] After the generation parameters are determined, new sample operation data is generated according to the method described in step S40 to obtain training data.

[0111] In some embodiments, when the number of negative sample running data among the acquired multiple sample running data is less than the number of positive sample running data, new sample running data is generated by the following method:

[0112] When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is less than the number of positive sample running data, the new sample running data is:

[0113] X new =x SV- +δ(x nn- -x SV- )+βγ 0 ;

[0114] When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is greater than the number of positive sample running data, the new sample running data is:

[0115] X new =x SV- -δ(x nn- -x SV- )+βγ 0 ;

[0116] Among them, x SV-Run the data for any negative support sample, δ is a random number in the interval (0,1), x nn -x SV- Any negative sample running data among the corresponding K nearest neighbor sample running data, β is a random number in the interval (0, 1), γ 0 is the generation parameter. Both δ and β are random numbers in the interval (0, 1), and they may be the same or different.

[0117] Here the generating parameter γ 0 Select the third distance mean, its expression is in Represents x SV The jth negative nearest neighbor, k - is the number of negative nearest neighbors.

[0118] It should be understood that δ in this embodiment and δ in the embodiment of generating positive sample running data are both random numbers, and the two may be the same or different.

[0119] In this embodiment, when the number of negative sample operation data among the acquired multiple sample operation data is less than the number of positive sample operation data, the generated new sample operation data is negative sample operation data, which ensures the balance of the samples and makes the trained fault diagnosis model more reliable.

[0120] Furthermore, if the negative sample running data is in the majority among the nearest neighbor sample running data corresponding to the negative support sample running data, it can be known that the negative sample running data is in the area where the two types of data are far apart, so the negative support sample running data is interpolated; if the positive sample running data is in the majority among the nearest neighbor sample running data corresponding to the negative support sample running data, it can be known that the boundary sample is in the overlapping area of ​​the positive sample running data and the negative sample running data in the kernel space, so the negative support sample running data is interpolated; using this method, the classification surface is more ideal. In addition, when generating new sample running data, the generation parameters are determined according to the distance from the support sample running data. Therefore, adding the generation parameter operation to generate new sample running data can ensure that the generated new sample running data is located around the boundary, ensuring the reliability of the classification boundary.

[0121] For step S50, in some embodiments, an objective function is constructed by local linear embedding and principal component analysis to perform feature fusion on the training data. Specifically, a neighborhood set corresponding to each training data is determined, and the neighborhood set includes N training data that are closest to the corresponding training data; the linear weight coefficients of each training data and the N training data in the corresponding neighborhood set are calculated; the preset objective function is optimized based on the linear weight coefficients until the preset objective function converges to obtain fused training data; wherein the preset objective function includes regularization and the sum of errors. The error includes the error corresponding to the local linear embedding algorithm and the error corresponding to the principal component analysis method.

[0122] Specifically, for a high-dimensional dataset X∈R n×d For each training data x i , n represents the number of samples in the high-dimensional dataset, d represents the number of features, and calculates its difference from all other training data x j The Euclidean distance of where x i and x j Respectively represent the i-th and j-th training data of the high-dimensional data set, x ij represents the kth feature of the i-th training data, x jk Represents the kth feature of the jth training data; for each training data x i Select the N training data with the closest Euclidean distance as the neighborhood set N(x i ).

[0123] For each training data x i , which corresponds to any training data x in the N training data in the neighborhood set j The linear weight coefficient is w ij : in, Keeping the weight composition unchanged, solve the low-dimensional representation Y: This part ensures a low-dimensional representation of each data point y in Y i It can be reconstructed by a linear combination of its domain, thus preserving the local structure. Then the objective function J(Y) is optimized using gradient descent.

[0124] The preset objective function is optimized using the gradient descent method based on the linear weight coefficient as J(Y):

[0125]

[0126] Among them, y i and j The training data x i and x jThe corresponding fault status mark. The linear weight coefficient regularization term prevents overfitting and ensures the sparsity of the weight. Trace is to find the trace of the two-dimensional matrix. It is iteratively updated through the gradient descent method until the objective function converges to a stable value, and finally outputs the fusion training data after dimensionality reduction. Regularization term Prevent overfitting and ensure the sparsity of weights. λ is the regularization parameter. According to the gradient calculation formula Gradient Where I is the identity matrix, W i is the weight matrix, Represents the local structure, due to the regularization term on y i The derivative of does not directly affect y i , so the gradient is 0. 2αY represents the global structure, and αj is a weight parameter used to balance the importance of local structure and global structure. In this way, by updating Y in the direction of the gradient, we get the updated low-dimensional representation: Where η is the learning rate. Iteratively update the low-dimensional representation Y until the objective function J(Y) converges to a stable value. Finally, the reduced-dimensional data Y is output.

[0127] The local structure of the original data is retained during the dimensionality reduction process, so that the low-dimensional representation after dimensionality reduction can still reflect the local neighborhood relationship of the original data. This method of retaining local structure makes the results after dimensionality reduction more interpretable, can more clearly reflect the local characteristics and mutual relationships of the original data, and is suitable for subsequent classification tasks.

[0128] For step S60, when training the fan fault diagnosis model, the improved support vector machine algorithm SVM selects the RBF kernel function and uses the standard Hinge loss plus regularization as the objective function optimization parameter to better solve the nonlinear feature mapping and high-dimensional data problems. The specific objective function is expressed as:

[0129]

[0130] where y i and j The training data x i and x j The corresponding fault state label; w is the model weight vector; b is the bias constant; is the regularization term; C is the regularization constant, which controls the trade-off between Hinge loss and regularization term; γ is the hyperparameter of the RBF kernel function, which is a preset constant. The objective function is used to train the model, and w and b are optimized during the training process. α j is the support vector.

[0131] After obtaining the optimized w and b through training, the parameter w is not used directly in the decision function. The kernel function is used to calculate the dot product without explicitly calculating the high-dimensional mapping, and the optimized parameters are directly used to quickly classify the new sample x. The decision function is expressed as:

[0132] α i The improved support vector machine algorithm SVM can effectively process high-dimensional data and perform nonlinear classification without explicitly calculating high-dimensional mapping, and can obtain the label of whether the wind turbine fails. Combined with Table 1, the location and cause of the wind turbine failure can be determined.

[0133] Figure 5 A fan fault diagnosis method according to an embodiment of the present invention is shown, which specifically includes the following steps:

[0134] S100: Obtain fan operation data.

[0135] S200: Perform feature fusion on the wind turbine operation data to obtain fused data.

[0136] S300: Inputting the fused data into a fan fault diagnosis model to perform fault diagnosis on the fan.

[0137] In this embodiment, the dimension of the obtained wind turbine operation data is the same as that of the sample operation data obtained when training the wind turbine fault diagnosis model, and the features represented by each dimension also correspond one to one. The method for performing feature fusion is the same as the method for performing feature fusion on the training data in step S50. Please refer to the training feature fusion method in step S50, which will not be repeated here. The wind turbine fault diagnosis model used in this embodiment is the model obtained by the wind turbine fault diagnosis method based on SCADA data in any of the above embodiments.

[0138] Through this embodiment, fan faults can be diagnosed only by inputting the acquired fan operation data into the fan fault diagnosis model, which greatly reduces the cost and complexity of fan fault diagnosis and improves the efficiency of fan fault diagnosis.

[0139] Figure 6 A schematic diagram of an electronic device according to an embodiment of the present invention is shown. Figure 6 As shown, the electronic device of the embodiment of the present invention includes a memory 61 and a processor 62, the memory 61 stores a calculation program, and the calculation program can be executed by the processor 62. When the calculation program is executed by the processor 62, the processor 62 can implement the steps corresponding to the wind turbine fault diagnosis method based on SCADA data in any of the above-mentioned embodiments, and / or the processor 62 can implement the steps corresponding to the wind turbine fault diagnosis method in any of the above-mentioned embodiments.

[0140] It should be understood that in the embodiment of the present invention, the memory 61 can be any computer-readable memory, such as RAM, Flash, etc., and the embodiment of the present invention is not limited to the specific type of the memory 61. There can be one or more processors 62, and the processor 62 can execute the logic corresponding to the calculation program and output the calculation result.

[0141] An embodiment of the present invention also provides a computer-readable storage medium, which stores a calculation program. When the calculation program is executed, it can implement the steps corresponding to the wind turbine fault diagnosis method based on SCADA data in any of the above-mentioned embodiments, and / or implement the steps corresponding to the wind turbine fault diagnosis method in any of the above-mentioned embodiments.

[0142] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A wind turbine fault diagnosis method based on SCADA data, Features: The method comprises: Acquire a plurality of sample operation data, wherein the sample operation data refers to operation data corresponding to a fan fault state that has been calibrated, and the plurality of sample operation data includes a plurality of positive sample operation data and a plurality of negative sample operation data, wherein the fan fault state corresponding to the positive sample operation data is normal, and the fan fault state corresponding to the negative sample operation data is fault; Training at least part of the sample operation data among the plurality of sample operation data by a support vector machine algorithm to determine support sample operation data, wherein the support sample operation data is boundary sample operation data for distinguishing positive sample operation data from negative sample operation data; Determine the generation parameter according to the distance between the supporting sample running data and the K nearest neighbor sample running data, where K is an integer greater than 1; Generate a plurality of new sample operation data according to the generation parameters to obtain training data of a fan fault diagnosis model, wherein the training data includes the plurality of sample operation data and the plurality of new sample operation data; Performing feature fusion on the training data to obtain fused training data; The fused training data is trained by a support vector machine algorithm to obtain a fan fault diagnosis model.

2. The wind turbine fault diagnosis method based on SCADA data as claimed in claim 1, Features: When the number of positive sample running data in the plurality of sample running data is less than the number of negative sample running data, determining the generation parameter according to the distance between the supporting sample running data and the K nearest neighboring sample running data includes: determining positive supporting sample operation data among the supporting sample operation data; Calculate the first distance between each positive support sample running data and each positive sample running data in the corresponding K nearest neighbor sample running data; Determine generation parameters corresponding to each of the positive supporting sample running data according to the first distance.

3. The wind turbine fault diagnosis method based on SCADA data as claimed in claim 2, Features: The determining, according to the first distance, a generation parameter corresponding to each positive supporting sample running data includes: The mean of the first distances is calculated to obtain the generation parameter.

4. The wind turbine fault diagnosis method based on SCADA data according to claim 2 or 3, Features: The step of generating a plurality of new sample operation data according to the generation parameters comprises: When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is less than the number of negative sample running data, the new sample running data is: X new =x SV+ +δ(x nn+ -x SV+ )+ag 0 ; When the number of positive sample running data in the K nearest neighbor sample running data corresponding to the positive support sample running data is greater than the number of negative sample running data, the new sample running data is: X new =x SV+ -δ(x nn+ -x SV+ )+ag 0 ; Among them, x SV+ is any one of the positive support sample running data, δ is a random number in the interval (0, 1), x nn+ For x SV+ Any positive sample running data among the corresponding K nearest neighbor sample running data, α is a random number in the interval (0, 1), γ 0 Generate parameters for the above.

5. The wind turbine fault diagnosis method based on SCADA data as claimed in claim 1, Features: When the number of positive sample running data in the plurality of sample running data is greater than the number of negative sample running data, determining the generation parameter according to the distance between the supporting sample running data and the K nearest neighboring sample running data includes: determining negative supporting sample running data in the supporting sample running data; Calculate the third distance between each negative support sample running data and each negative sample running data in the corresponding K nearest neighbor sample running data; A generation parameter corresponding to each of the negative support sample running data is determined according to the third distance.

6. The wind turbine fault diagnosis method based on SCADA data as claimed in claim 5, Features: The determining, according to the third distance, a generation parameter corresponding to each of the negative support sample running data comprises: The mean of the third distance is calculated to obtain the generation parameter.

7. The wind turbine fault diagnosis method based on SCADA data according to claim 5 or 6, Features: The step of generating a plurality of new sample operation data according to the generation parameters comprises: When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is less than the number of positive sample running data, the new sample running data is: X new =x SV- +δ(x nn- -x SV- )+βγ 0 ; When the number of negative sample running data in the K nearest neighbor sample running data corresponding to the negative support sample running data is greater than the number of positive sample running data, the new sample running data is: X new =x SV- -δ(x nn- -x SV- )+βγ 0 ; Among them, x SV- Run the data for any of the negative support samples, δ is a random number in the interval (0,1), x nn- For x SV- The corresponding K nearest neighbor sample running data is any negative sample running data, β is a random number in the interval (0,1), γ 0 Generate parameters for the above.

8. The wind turbine fault diagnosis method based on SCADA data according to claim 1, Features: The step of performing feature fusion on the training data to obtain fused training data includes: Determine a neighborhood set corresponding to each of the training data, wherein the neighborhood set includes N training data that are closest to the corresponding training data; Calculate the linear weight coefficient of each training data and N training data in the corresponding neighborhood set; The preset objective function is optimized based on the linear weight coefficient until the preset objective function converges to obtain the fusion training data; the preset objective function includes regularization and the sum of errors.

9. A wind turbine fault diagnosis system based on SCADA data. Features: The system comprises: A data acquisition module, used to acquire fan operation data; A feature fusion module performs feature fusion on the wind turbine operation data to obtain fused data; A fault diagnosis module inputs the fused data into a wind turbine fault diagnosis model to perform fault diagnosis on the wind turbine; wherein the wind turbine fault diagnosis model is obtained according to the wind turbine fault diagnosis method based on SCADA data as described in any one of claims 1 to 8.

10. An electronic device, Features: The electronic device includes a memory and a processor, the memory stores a calculation program, and when the calculation program is executed by the processor, the processor implements the wind turbine fault diagnosis method based on SCADA data as described in any one of claims 1-8.

Citation Information

Patent Citations

  • Method for diagnosing surface icing faults of wind power generator blade of power system based on machine visual images

    CN103982378A

  • Integrated learning based mountain wind generation set behavior predicating model

    CN110685857A

  • Fan foundation and tower structure detection method

    CN111946559A

  • Wind turbine generator fault diagnosis method and system based on WOA-KELM algorithm

    CN116738308A

  • Analogue circuit fault diagnosis method based on generalized multiple kernel learning-support vector machine

    WO2017128455A1

Cited By

  • Power distribution method and system for integrated energy electrical system

    CN119253641A

  • Fan blade inspection local dynamic path planning method and system based on unmanned aerial vehicle

    CN121364743A