Abnormal detection and correction method and device for fan operation data

Abstract

The invention provides a fan operation data anomaly detection correction method and device, and belongs to the field of data anomaly value detection. The method comprises the following steps: carrying out principal component analysis on collected fan operation data to obtain principal component data; determining the number of target clustering centers according to the principal component data; clustering the principal component data according to the number of target clustering centers, and performing multi-target optimization by taking intra-cluster difference minimization and inter-cluster difference maximization as target functions to obtain cluster labels and abnormal data corresponding to each cluster; and correcting the abnormal data according to the cluster label corresponding to each clustering cluster to obtain corrected data. According to the method, the clustering algorithm is adopted to identify abnormal data, the abnormal data can be accurately identified, abnormal data correction is carried out based on the clustering cluster, the true value of the data can be well simulated, and the correction effect is good.

Description

Technical Field

[0001] The present invention relates to the technical field of data outlier detection, and in particular to a method and device for detecting and correcting anomaly of fan operation data. Background Art

[0002] The rapid development of the global economy has led to an increasing demand for energy. Traditional energy resources are limited and have a great impact on the environment. At present, the world is committed to the development of renewable energy, which has the characteristics of wide resource distribution, sustainability, and high economic benefits. Wind power has become one of the most widely used renewable energy sources due to its rich resources and low construction costs.

[0003] Due to geographical conditions and environmental factors, wind farms are mostly located in remote plains, mountains or seas. Therefore, it is necessary to introduce data acquisition and monitoring control to wind farms, collect wind turbine operation data, and centrally monitor, dispatch and manage the daily operation of wind farms. However, due to factors such as equipment, environment, and operating status, the collected wind turbine operation data will contain a large number of abnormal values ​​and missing values. The existence of these "dirty data" seriously affects the subsequent wind turbine status analysis, fault diagnosis and other functions. Therefore, it is necessary to clean and fill the collected wind turbine operation data.

[0004] In the prior art, statistical methods, machine learning methods and other methods are usually used to detect outliers, but the accuracy is not high and there is a lack of data correction, which seriously affects the subsequent status analysis and fault diagnosis work. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for detecting and correcting abnormality of wind turbine operation data, so as to solve the problem that the statistical method and machine learning method in the prior art have low accuracy in detecting abnormal values ​​and lack of legal data correction, which affects the processing of procedural data.

[0006] In a first aspect, an embodiment of the present invention provides a method for detecting and correcting abnormality of wind turbine operation data, comprising:

[0007] S101: performing principal component analysis on the collected fan operation data to obtain principal component data;

[0008] S102: Determine the number of target cluster centers according to the principal component data;

[0009] S103: clustering the principal component data according to the number of target cluster centers, performing multi-objective optimization with minimizing the intra-cluster difference and maximizing the inter-cluster difference as the objective function, and obtaining cluster labels and abnormal data corresponding to each cluster;

[0010] S104: Correcting the abnormal data according to the cluster labels corresponding to the clusters to obtain corrected data.

[0011] In a second aspect, an embodiment of the present invention provides a device for detecting and correcting abnormality of wind turbine operation data, comprising:

[0012] A dimension reduction module is used to perform principal component analysis on the collected fan operation data to obtain principal component data;

[0013] A cluster center determination module is used to determine the number of target cluster centers based on principal component data;

[0014] The abnormal data identification module is used to cluster the principal component data according to the number of target cluster centers, and to perform multi-objective optimization with minimization of intra-cluster differences and maximization of inter-cluster differences as objective functions to obtain cluster labels and abnormal data corresponding to each cluster;

[0015] The data correction module is used to correct the abnormal data according to the cluster labels corresponding to each cluster cluster to obtain corrected data.

[0016] The embodiment of the present invention provides a method and device for detecting and correcting abnormalities in fan operation data. The above method includes: performing principal component analysis on the collected fan operation data to obtain principal component data; determining the number of target cluster centers according to the principal component data; clustering the principal component data according to the number of target cluster centers, performing multi-objective optimization with minimizing the intra-cluster difference and maximizing the inter-cluster difference as the objective function, and obtaining cluster labels and abnormal data corresponding to each cluster cluster; correcting the abnormal data according to the cluster labels corresponding to each cluster cluster to obtain corrected data. Since there will be certain differences between abnormal data and most normal data, based on this, the embodiment of the present invention adopts a clustering algorithm to identify abnormal data, which can accurately identify abnormal data in fan operation data and improve the accuracy of identification. Based on the clustering algorithm in the above process, the embodiment of the present invention effectively uses the cluster clusters generated in the above clustering process to correct the abnormal data, which can well simulate the true value of the data, has a good correction effect, and reduces the difficulty of data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0018] Figure 1 It is a flow chart of an implementation method of abnormality detection and correction of wind turbine operation data provided by an embodiment of the present invention;

[0019] Figure 2It is a comparison diagram of data before and after correction using the abnormality detection and correction method for fan operation data provided by an embodiment of the present invention;

[0020] Figure 3 It is a structural schematic diagram of an abnormality detection and correction device for fan operation data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0022] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0023] refer to Figure 1 , see Figure 1 , which shows a flow chart of an implementation method of abnormal detection and correction of wind turbine operation data provided by an embodiment of the present invention, which is described in detail as follows:

[0024] The abnormal detection and correction method of fan operation data includes:

[0025] S101: performing principal component analysis on the collected fan operation data to obtain principal component data;

[0026] The wind turbine operation data may include: wind speed, wind direction, wind rotor speed, wind turbine power and other data, including multiple types of operation data, that is, the wind turbine operation data is multi-dimensional data.

[0027] Since the fan operation data is multidimensional data, in order to reduce the difficulty of data processing, the embodiment of the present invention uses principal component analysis to reduce the dimension of complex data, remove noise and redundancy, select the most important "elements" in the fan operation data, find out the simple features behind the complex data, and use subsequent data analysis.

[0028] S102: Determine the number of target cluster centers according to the principal component data;

[0029] S103: clustering the principal component data according to the number of target cluster centers, performing multi-objective optimization with minimizing the intra-cluster difference and maximizing the inter-cluster difference as the objective function, and obtaining cluster labels and abnormal data corresponding to each cluster;

[0030] The fan operation data has certain regularity. The abnormal data is a small amount of data, while most of the fan operation data is normal data. There is a certain difference between the abnormal data and most of the normal data.

[0031] Cluster analysis is a commonly used data mining technology. Its main goal is to divide the data set into several non-overlapping subsets based on certain features in the data, so that data of the same category are put together and data of different categories are separated.

[0032] Abnormal data refers to data points in a data set that do not conform to the norm. After clustering, they are far away from the cluster center relative to most normal data. Based on this, the embodiment of the present invention adopts a clustering algorithm to identify abnormal data, with high recognition accuracy, and can achieve accurate recognition of abnormal data.

[0033] S104: Correcting the abnormal data according to the cluster labels corresponding to the clusters to obtain corrected data.

[0034] During the clustering process, multiple clusters will be obtained, and the cluster centers and cluster labels of the clusters will be obtained at the same time. In the embodiment of the present invention, the data generated in the clustering process is used to correct the abnormal data. Compared with using other algorithms for data correction, the difficulty of data processing is reduced due to the use of the data generated in the clustering process, while maintaining the consistency of the algorithm, further improving the accuracy of data correction.

[0035] Based on the above, it can be seen that clustering is used in the embodiment of the present invention to accurately identify abnormal data, and cluster labels generated in the clustering process are used to correct data, which has high accuracy, small data volume and simple processing process.

[0036] In a possible implementation, before S101, the method may further include:

[0037] S105: Screening the collected fan operation data to obtain screened data;

[0038] S101 may include: performing principal component analysis on the screened data to obtain principal component data.

[0039] In a possible implementation manner, the wind turbine operation data includes multiple types of operation data; S105 may include:

[0040] S1051: for any type of operating data, calculating the correlation coefficient between the type of operating data and the target operating data, and obtaining the correlation coefficient value corresponding to the type of operating data; if the correlation coefficient value corresponding to the type of operating data is less than a preset value, deleting the type of operating data in the fan operating data;

[0041] The wind turbine operation data includes target operation data.

[0042] Since the wind turbine operation data includes multiple types of operation data, such as wind speed, wind direction, wind rotor speed, wind turbine power, etc., the target operation data (such as wind turbine power) is the most concerned data type. In the embodiment of the present invention, the correlation between various types of operation data and the target operation data can be calculated, and only variables with a large correlation with the target operation data are retained, which reduces the complexity of the data and improves the subsequent data processing speed.

[0043] The correlation is calculated as follows:

[0044]

[0045] Among them, E(·) represents mathematical expectation; cov(·) represents covariance.

[0046] For example, only the operation data with the greatest correlation with the target operation data may be retained. For example, after correlation calculation, the wind speed at 10 meters in the latitudinal direction has the greatest correlation with the wind power, and the wind speed at 10 meters in the latitudinal direction is retained.

[0047] In a possible implementation, S102 may include:

[0048] S1021: According to the principal component data, the COA raccoon optimization algorithm is used to optimize the cluster centers to obtain the target number of cluster centers.

[0049] The rationality of the number of cluster centers selected plays a crucial role in the effect of clustering. In the embodiment of the present invention, the COA raccoon optimization algorithm is used to optimize the number of cluster centers, which makes the number of cluster centers more objective and reasonable, and improves the accuracy and effectiveness of cluster analysis.

[0050] The specific optimization process is as follows:

[0051] (1) Randomly initialize spatial position and boundary check:

[0052] X i :x ij =lb j +r·(ub j -lb j ),i=1,2,…,N,j=1,2,…,m

[0053] Among them, X i represents the position of the i-th raccoon in space; lb j Indicates the lower bound of optimization; ub j represents the upper bound of optimization; r represents a random number between [0,1]; N represents the number of raccoons; m represents the number of decision variables.

[0054] Check and correct the population positions that exceed the boundaries to ensure that the population individuals do not exceed the specified boundaries:

[0055]

[0056] (2) Fitness calculation and ranking:

[0057] In the embodiment of the present invention, a silhouette coefficient function may be used as the fitness of each individual in the population.

[0058] Specifically, the calculation formula of fitness in the COA raccoon optimization algorithm may include:

[0059]

[0060] Among them, S(i) is the fitness of sample i, a(i) is the average distance from sample i to other samples in the same cluster, and b(i) is the average distance from sample i to the nearest cluster.

[0061] Sort the fitness values ​​and sort the positions according to the fitness sort results.

[0062] (3) Attack strategy (exploration phase):

[0063] (3.1) Half of the raccoons climbed up the tree to scare the iguana, and the other half waited for the iguana to land under the tree. The process of climbing the tree can be expressed as:

[0064]

[0065] in, represents the new position of the i-th raccoon in the j-th dimension; j represents the position of the iguana in the jth dimension.

[0066] (3.2) The position of the iguana after landing is random, and the other half of the raccoons move towards the iguana to prey on it:

[0067]

[0068] Among them, Iguana G represents the position of the iguana after landing in the jth dimension.

[0069]

[0070] in, represents the objective function value after the iguana lands on the jth dimension; F i represents the objective function value of the i-th raccoon in the j-th dimension.

[0071] (3.3) Update individual optimality:

[0072]

[0073] Among them, F i P1 represents the objective function value of the i-th raccoon in its new position; F i represents the objective function value at the previous position of the i-th raccoon.

[0074] (4) Escape from predators strategy (exploitation stage):

[0075] (4.1) Generate a random location near each raccoon's location:

[0076]

[0077] in, represents a local lower bound; represents the local upper bound; t represents the number of iterations; T represents the maximum number of iterations, represents the new position of the i-th raccoon in the j-th dimension.

[0078] (4.2) Update optimality:

[0079]

[0080] Thus, the optimal number of cluster centers is obtained, which is also the target number of cluster centers.

[0081] In a possible implementation, S103 may include:

[0082] S1031: clustering the principal component data according to the number of target cluster centers to obtain clusters with the number of target cluster centers, and determining the cluster center of each cluster and the cluster label corresponding to each cluster;

[0083] S1032: For any cluster, determine the distance between each data in the cluster and the cluster center of the cluster; determine the data with the largest distance to the cluster center of the cluster as abnormal data in the cluster.

[0084] The data is clustered based on the optimized target cluster center, and a joint objective function is defined. Clustering is performed when the difference within the cluster is the smallest and the difference between clusters is the largest. The objective function is as follows:

[0085]

[0086] Among them, C k is the Kth cluster, c k is the center of the Kth cluster, x i is the ith data point, α and β are weight parameters.

[0087] Calculate the distance from each data point to the cluster center, update the cluster center, calculate the new intra-cluster difference and inter-cluster difference, update the objective function value, and repeat the above steps until the cluster center no longer changes significantly and clustering is completed.

[0088] Abnormal points are usually far away from the cluster center, so the point with the largest distance from the cluster center in each cluster is taken as the abnormal value, and the other points are normal values, so as to realize the detection of abnormal data.

[0089] For example, each data point can be labeled, and the abnormal value can be labeled as 1, and the normal value can be labeled as 0.

[0090] In a possible implementation, an optimized K-means clustering algorithm may be used for clustering.

[0091] After the abnormal data is identified, it is necessary to correct the abnormal data to ensure the integrity of the fan operation data.

[0092] Based on this, in a possible implementation, S104 may include:

[0093] S1041: normalizing the normal data in each cluster, and the normalized data forms a training set;

[0094] S1042: construct an initial WGAN model, and input the cluster labels corresponding to each cluster into the initial WGAN model as conditions;

[0095] To facilitate data processing, normal data are first normalized to the interval [-1,1]. At the same time, missing values ​​(abnormal data) are introduced into the data in a random manner according to the specified missing rate to form a training set.

[0096] The samples in the training set can be divided into training and testing in a ratio of 8:2. The specific ratio is not limited.

[0097] The WGAN model consists of two parts: the generator and the discriminator. The generator takes a random variable Z as input and generates an output vector with the same dimension as the real data X. Normal data is concatenated with the corresponding cluster label, and the cluster label is input as a condition into the discriminator to distinguish the generated data from the real data. In the WGAN model, the loss functions of the generator and the discriminator are as follows:

[0098]

[0099] Among them, P r is the real data distribution; P g To generate data distribution; f ω is the discriminator network, g θ is the generator network;

[0100] During the training process, L D The larger the value, the smaller the Wasserstein distance between the true distribution and the generated distribution, and the better the GAN training effect.

[0101] In the embodiment of the present invention, the hyper parameters of training are set, including epochs, learning rate, weight clipping value, etc., and the generator and discriminator are optimized respectively using the RMSprop optimizer. First, the training data filled with random noise and missing value masks are input into the generator to generate data after filling the missing values, and the score difference of the discriminator on the real data and the generated data is calculated to optimize the discriminator. Then, the generator is optimized so that the data it generates can deceive the discriminator, and the weight of the discriminator is clipped to ensure the stability of the model.

[0102] S1043: training the initial WGAN model using the training set to obtain a trained WGAN model;

[0103] S1044: Correct each abnormal data using the trained WGAN model to obtain corrected data.

[0104] The trained WGAN model is used to fill in the abnormal data, and then the data is inversely normalized to the original data range and used as the corrected data to update the wind turbine operation data, so as to obtain accurate wind turbine operation data, which provides a more scientific data basis for subsequent data processing and can be used for more accurate data analysis and prediction model construction, optimize power generation scheduling and operation strategies, and improve power generation rate and utilization rate.

[0105] In a possible implementation, when training the initial WGAN model, root mean square error, visualized PCA and t-SNE may be used as training evaluation indicators.

[0106] The above method is described in detail below with reference to specific embodiments.

[0107] The wind turbine operation data are selected from wind power data of 10 wind farms in Australia. The types of wind turbine operation data include: wind speed and wind turbine power at 10m and 100m in the latitudinal wind component and the longitudinal wind component.

[0108] The above-mentioned fan operation data is analyzed using the abnormality detection and correction method for fan operation data provided by an embodiment of the present invention.

[0109] Table 1 Correlation coefficient between each fan operation data and fan power

[0110] Variable Name U10 V10 U100 V100 power 0.378 -0.192 0.366 -0.202

[0111] The WGAN model is used to fill in the abnormal data, and the RMSE is used to evaluate the test set. The RMSE can reach 0.15, which is very close to 0. Figure 2 PCA and t-SNE of the data before and after correction are shown, by Figure 2 It can be seen that the overlap between the original data and the corrected data is very high, and the correction effect is very good. In summary, the clustering algorithm used in the present invention can accurately detect abnormal data, and at the same time, the WGAN model can accurately fill in the abnormal data, which can be used in practical engineering applications.

[0112] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0113] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0114] Figure 3 The structure diagram of the abnormality detection and correction device for fan operation data provided by an embodiment of the present invention is shown. For the convenience of description, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0115] like Figure 3 As shown, the abnormality detection and correction device for fan operation data includes:

[0116] A dimension reduction module 21 is used to perform principal component analysis on the collected wind turbine operation data to obtain principal component data;

[0117] A cluster center determination module 22, used to determine the number of target cluster centers according to the principal component data;

[0118] The abnormal data identification module 23 is used to cluster the principal component data according to the number of target cluster centers, and perform multi-objective optimization with minimizing the intra-cluster difference and maximizing the inter-cluster difference as the objective function to obtain the cluster labels and abnormal data corresponding to each cluster;

[0119] The data correction module 24 is used to correct the abnormal data according to the cluster labels corresponding to each cluster to obtain corrected data.

[0120] In a possible implementation, the cluster center determination module 22 may be specifically used to: optimize the cluster centers using the COA raccoon optimization algorithm according to the principal component data to obtain the target number of cluster centers.

[0121] In a possible implementation, the calculation formula for fitness in the COA raccoon optimization algorithm may include:

[0122]

[0123] Among them, S(i) is the fitness of sample i, a(i) is the average distance from sample i to other samples in the same cluster, and b(i) is the average distance from sample i to the nearest cluster.

[0124] In a possible implementation, the abnormal data identification module 23 may include:

[0125] A clustering unit is used to cluster the principal component data according to the number of target cluster centers, obtain clusters with the number of target cluster centers, and determine the cluster centers of each cluster and the cluster labels corresponding to each cluster;

[0126] The abnormality identification unit is used to determine the distance between each data in any cluster and the cluster center of the cluster; and determine the data with the largest distance from the cluster center of the cluster as the abnormal data in the cluster.

[0127] In a possible implementation, the data correction module 24 may include:

[0128] A training set forming unit is used to normalize the normal data in each cluster, and the normalized data forms a training set;

[0129] A model building unit, used to build an initial WGAN model and input the cluster labels corresponding to each cluster as conditions into the initial WGAN model;

[0130] A model training unit is used to train the initial WGAN model using the training set to obtain a trained WGAN model;

[0131] The data correction unit is used to correct each abnormal data using the trained WGAN model to obtain corrected data.

[0132] In a possible implementation, when training the initial WGAN model, root mean square error, visualized PCA and t-SNE may be used as training evaluation indicators.

[0133] In a possible implementation manner, the above device may further include:

[0134] The data screening module 25 is used to screen the collected fan operation data to obtain screened data;

[0135] The dimension reduction module 21 can be specifically used to perform principal component analysis on the filtered data to obtain principal component data.

[0136] In a possible implementation manner, the wind turbine operation data includes multiple types of operation data; the data screening module 25 may be specifically used for:

[0137] For any type of operating data, the correlation coefficient between the type of operating data and the target operating data is calculated to obtain the correlation coefficient value corresponding to the type of operating data; if the correlation coefficient value corresponding to the type of operating data is less than a preset value, the type of operating data in the fan operating data is deleted;

[0138] The wind turbine operation data includes target operation data.

[0139] In a possible implementation manner, the target operating data may be wind power.

[0140] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0141] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0142] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned abnormal detection and correction method embodiments of each wind turbine operation data can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device that can carry computer program code, recording media, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution media.

[0143] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for detecting and correcting abnormality of wind turbine operation data, characterized in that: include: Performing principal component analysis on the collected fan operation data to obtain principal component data; Determining the number of target cluster centers according to the principal component data; Clustering the principal component data according to the number of target cluster centers, performing multi-objective optimization with minimizing intra-cluster differences and maximizing inter-cluster differences as objective functions, and obtaining cluster labels and abnormal data corresponding to each cluster; The abnormal data is corrected according to the cluster labels corresponding to each cluster to obtain corrected data.

2. The abnormality detection and correction method for wind turbine operation data according to claim 1, characterized in that: Determining the number of target cluster centers according to the principal component data includes: According to the principal component data, the COA raccoon optimization algorithm is used to optimize the cluster centers to obtain the target number of cluster centers.

3. The abnormality detection and correction method for wind turbine operation data according to claim 2, characterized in that: The calculation formula of fitness in the COA raccoon optimization algorithm includes: Among them, S(i) is the fitness of sample i, a(i) is the average distance from sample i to other samples in the same cluster, and b(i) is the average distance from sample i to the nearest cluster.

4. The abnormality detection and correction method for wind turbine operation data according to any one of claims 1 to 3, characterized in that: The principal component data is clustered according to the number of target cluster centers, and multi-objective optimization is performed with minimization of intra-cluster differences and maximization of inter-cluster differences as objective functions to obtain cluster labels and abnormal data corresponding to each cluster, including: Clustering the principal component data according to the number of target cluster centers to obtain clusters with the number of target cluster centers, and determining the cluster center of each cluster and the cluster label corresponding to each cluster; For any cluster, the distance between each data in the cluster and the cluster center of the cluster is determined; the data with the largest distance from the cluster center of the cluster is determined as the abnormal data in the cluster.

5. The abnormality detection and correction method for wind turbine operation data according to claim 4, characterized in that: The step of correcting the abnormal data according to the cluster labels corresponding to the clusters to obtain corrected data includes: Normalize the normal data in each cluster, and the normalized data forms a training set; Constructing an initial WGAN model, and inputting the cluster labels corresponding to each cluster cluster as conditions into the initial WGAN model; Using the training set to train the initial WGAN model to obtain a trained WGAN model; The trained WGAN model is used to correct each abnormal data to obtain corrected data.

6. The abnormality detection and correction method for wind turbine operation data according to claim 5, characterized in that: When training the initial WGAN model, root mean square error, visualized PCA and t-SNE are used as evaluation indicators for training.

7. The abnormality detection and correction method for wind turbine operation data according to any one of claims 1 to 3, characterized in that: Before performing principal component analysis on the collected wind turbine operation data to obtain principal component data, the method further includes: Screening the collected fan operation data to obtain screened data; The principal component analysis is performed on the collected fan operation data to obtain the principal component data, including: The screened data is subjected to principal component analysis to obtain the principal component data.

8. The abnormality detection and correction method for wind turbine operation data according to claim 7, characterized in that: The fan operation data includes multiple types of operation data; the data filtering of the collected fan operation data to obtain the filtered data includes: For any type of operating data, the correlation coefficient between the type of operating data and the target operating data is calculated to obtain the correlation coefficient value corresponding to the type of operating data; if the correlation coefficient value corresponding to the type of operating data is less than a preset value, the type of operating data in the fan operating data is deleted; Wherein, the wind turbine operation data includes the target operation data.

9. The abnormality detection and correction method for wind turbine operation data according to claim 8, characterized in that: The target operating data is wind power.

10. A device for detecting and correcting abnormality of fan operation data, characterized in that: include: A dimension reduction module is used to perform principal component analysis on the collected fan operation data to obtain principal component data; A cluster center determination module, used to determine the number of target cluster centers according to the principal component data; An abnormal data identification module is used to cluster the principal component data according to the number of target cluster centers, perform multi-objective optimization with minimization of intra-cluster differences and maximization of inter-cluster differences as objective functions, and obtain cluster labels and abnormal data corresponding to each cluster cluster; The data correction module is used to correct the abnormal data according to the cluster labels corresponding to each clustering cluster to obtain corrected data.

Images