A method for identifying wind power anomalies based on adaptive K-means segmentation and boundary modeling

Abstract

This invention provides a method for identifying wind power anomaly data based on adaptive K-means segmentation and boundary modeling. The method includes: identifying and filtering bottom-stacking anomaly data from the wind power dataset using rule-based filtering based on the characteristic that the power value of bottom-stacking anomaly data is less than 0; adaptively dividing the rule-filtered wind power dataset into K non-overlapping clusters using the CH index and k-means algorithm; dividing each cluster obtained in step S3 into L intervals according to power value, and calculating the wind speed range data and mean wind speed data corresponding to each power interval; and identifying discrete anomaly data and power-limited anomaly data based on the wind speed range data and mean wind speed data corresponding to each power interval using peak correction combined with boundary modeling. This invention improves the accuracy of wind power anomaly data identification.

Description

Technical Field

[0001] This invention belongs to the field of wind power technology, and in particular relates to a method for identifying wind power anomaly data based on adaptive K-means segmentation and boundary modeling. Background Technology

[0002] Currently, the main methods for processing wind power anomaly data can be categorized into machine learning-based methods, statistical analysis-based methods, and image analysis-based methods.

[0003] The basic idea of ​​machine learning-based wind power anomaly data processing methods is to learn the normal patterns of wind power data by training a model, and then use the trained model to identify and process anomaly data. Existing machine learning methods attempt to use the nearest neighbor algorithm (KNN) to construct wind speed-power curves for online anomaly detection; some scholars use the K-means algorithm to identify anomaly data and solve the problem of selecting the K value in the K-means algorithm through secondary clustering; in addition, some scholars have also adopted the method of using support vector machines to build a proximity model of multiple units to identify anomaly data by utilizing the similarity relationship between data. A common drawback of machine learning algorithms is that identification may fail when there is a large amount of power-limiting anomaly data to be identified. At the same time, machine learning-based methods often contain parameters that need to be manually adjusted according to different datasets. Different parameter settings often lead to completely different results, making it difficult to find a suitable parameter selection method.

[0004] The basic idea behind statistical analysis-based methods for processing wind power anomaly data is to detect and handle anomalies through statistical analysis. Based on the statistical properties and distribution characteristics of the data, statistical methods such as probability and hypothesis testing are used to determine whether the data is abnormal. Currently, statistical analysis-based methods mainly include the 3-σ criterion, the quartile method, and the Copula function. While these methods consider power-limited anomaly data, they often require data partitioning, resulting in jagged identification results and a large number of normal data points being incorrectly identified as anomalies. Summary of the Invention

[0005] To address the problems existing in the background technology, and to improve the accuracy of wind power anomaly data identification by employing different methods for identification and cleaning based on the unique characteristics of different types of data, this invention provides a wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling, comprising:

[0006] S1: Based on the distribution characteristics of different wind power anomaly data, wind power anomaly data is divided into bottom-accumulated anomaly data, discrete anomaly data, and power-limited anomaly data; among them, bottom-accumulated anomaly data is wind power anomaly data with power values ​​less than 0;

[0007] S2: Identify and filter out bottom-accumulated anomalies in the wind power dataset based on the characteristic that the power value of bottom-accumulated anomalies is less than 0 using rule-based filtering.

[0008] S3: The wind power dataset after rule filtering is adaptively divided into K non-overlapping clusters using the CH index and k-means algorithm;

[0009] S4: Divide each cluster obtained in step S3 into L intervals according to its power value, and calculate the wind speed range data and wind speed mean data corresponding to each power interval.

[0010] S5: Based on the wind speed range data and average wind speed data corresponding to each power range, peak correction combined with boundary modeling is used to identify discrete abnormal data and power-limited abnormal data.

[0011] Preferably, step S3 includes:

[0012] S31: For a given K value, use the K-means algorithm to cluster the wind power dataset and calculate the CH index after clustering;

[0013] S32: By trying different K values, comparing their corresponding CH indices, selecting the K value with the highest CH index as the optimal K value, and obtaining the K clusters corresponding to the optimal K value.

[0014] Preferably, the CH index includes:

[0015]

[0016]

[0017]

[0018] in, This indicates finding the trace of a matrix; Represents the covariance matrix of inter-cluster data; This represents the covariance matrix of the data within a cluster; n represents the total number of data points in all clusters. Indicates the number of clusters; Indicates the first The total number of data in each cluster; Indicates the first Data center points of each cluster, Indicates the first A cluster; This represents the data center point of all clusters; Indicates the first Data points in a cluster.

[0019] Preferably, the calculation of the wind speed range data and average wind speed data corresponding to each power range includes:

[0020] The wind speed range corresponding to each power interval is calculated using the quartile method, thus obtaining the wind speed range data for each power interval. The calculation formula is as follows:

[0021]

[0022]

[0023] in, This indicates the wind speed range corresponding to the power range; This represents the wind speed value corresponding to the first 25% of the wind power data within the power range. This represents the wind speed value corresponding to the last 25% of the wind power data within the power range. Indicates the center point of the power range;

[0024] The average wind speed data corresponding to each power range is as follows: , This represents the average wind speed of the wind power data corresponding to each power range.

[0025] Preferably, step S5 includes:

[0026] S51: Construct the wind speed range data and wind speed mean data corresponding to each power range into a wind speed range data point set and a wind speed mean data point set, respectively;

[0027] S52: Extract the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set, respectively; wherein, the peak data includes: the degree of peak protrusion, the starting point of the peak and the ending point of the peak.

[0028] S53: Traverse the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set. When the peak protrusion is greater than the set threshold, the data points from the start point to the end point of the peak are corrected using the linear interpolation method to obtain the corrected wind speed range data point set and wind speed mean data point set.

[0029] S54: Use an exponential function to fit the corrected wind speed range data point set to obtain the wind speed range curve; use a cubic function to fit the corrected wind speed mean data point set to obtain the wind speed mean curve.

[0030] S55: Set left and right boundary thresholds based on the wind speed range curve and the wind speed mean curve, and identify discrete and power-limited abnormal data in the wind power dataset based on the set left and right boundary thresholds.

[0031] Preferably, the peak data of the extracted wind speed range data point set and the peak data of the wind speed mean data point set include:

[0032] S521: Iterate through each data point in the dataset , Indicates the first Data points, Representing data points The corresponding power, Representing data points The corresponding wind speed range or average wind speed, comparing data points The size of its adjacent data points, if and Then the data points Add to the peak point set;

[0033] S522: Iterate through each peak point in the set of peak points. From the peak point Starting from the peak point in the dataset The expansion is performed simultaneously on both the left and right sides to calculate the peak point. to the straight line The shortest distance, where the straight line From the peak point Points obtained by expanding to the left and along the peak point Points obtained by expanding to the right Connected to form;

[0034] S523: When the peak point to the straight line The shortest distance is greater than the peak point The degree of protrusion corresponding to the peak Then the point and points As the peak point The corresponding start and end points of the peak value; wherein, the peak value point The degree of protrusion corresponding to the peak include:

[0035]

[0036] in, Indicates peak point The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed.

[0037] Preferably, the exponential function includes:

[0038]

[0039] in, Indicates power, Indicates the wind speed range. , and Indicates a parameter.

[0040] Preferably, the cubic function includes:

[0041]

[0042] in, Indicates power, Indicates the wind speed range. , , and Indicates a parameter.

[0043] Preferably, the left boundary threshold is v(x)-0.5w(x), and the right boundary threshold is v(x)+0.5w(x).

[0044] The present invention has at least the following beneficial effects

[0045] This invention utilizes the CH index combined with the k-means clustering algorithm to cluster wind power data, adaptively determining the number of regions to divide the wind speed-power scatter plot into different categories. This improves the similarity of data within the same category and reduces the similarity between different categories, achieving accurate identification of anomalous data across the entire power range in wind power monitoring data. In particular, by employing peak correction combined with boundary modeling, it overcomes the shortcomings of traditional methods in identifying power-limited anomalous data and overcomes the dependence of traditional clustering algorithms on the number of clusters. By dividing the anomalous wind power monitoring data into regions and obtaining wind speed range data and mean wind speed data, the boundary modeling problem is simplified into a curve fitting problem between the two types of data. This solves the problem that traditional identification methods are difficult to achieve globally optimal identification results, improving the accuracy of wind power anomalous data identification. This provides new ideas for researchers conducting related research in the future, and the research method has significant theoretical research value and practical engineering significance. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0047] Figure 2 This is a schematic diagram illustrating the characteristic distribution of different types of wind power anomaly data;

[0048] Figure 3 A schematic diagram showing the partitioning results using the CH index and the k-means algorithm;

[0049] Figure 4 This is a schematic diagram illustrating the principle of the quartile method;

[0050] Figure 5 A schematic diagram showing the range and average wind speed data;

[0051] Figure 6 A schematic diagram of the boundary modeling results;

[0052] Figure 7 Schematic diagram of the CU identification results in the region;

[0053] Figure 8 This diagram illustrates the comparison of anomaly identification results using different methods. Detailed Implementation

[0054] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0055] This embodiment, through a survey and analysis of the data distribution of existing wind power datasets, reveals that the operating power of wind turbines varies widely, and wind power data in different power ranges have their own unique characteristics. Uniform processing of the entire power range often yields poor results; therefore, the dataset needs to be regionalized. Traditional partitioning methods typically specify the number of categories, ignoring the differences between different datasets. A new partitioning method is needed that can adaptively select the number of categories, improve the correlation of data within the same category, and reduce the correlation between data from different categories. Through a survey and analysis of existing anomaly identification methods, it was found that wind power anomalies can be mainly divided into three categories: bottom-stacking anomalies, discrete anomalies, and power-limited anomalies. Traditional anomaly identification methods uniformly identify the actual collected raw power data, ignoring the unique characteristics of different types of anomalies, making accurate identification difficult. A new joint processing framework is needed to employ different methods for identification and cleaning of different types of data. For details, please refer to [link / reference needed]. Figure 1 This invention provides a method for identifying wind power anomaly data based on adaptive K-means segmentation and boundary modeling, comprising:

[0056] S1: Based on the distribution characteristics of different wind power anomaly data, wind power anomaly data is divided into bottom-accumulated anomaly data, discrete anomaly data, and power-limited anomaly data; among them, bottom-accumulated anomaly data is wind power anomaly data with power values ​​less than 0;

[0057] S2: Identify and filter out bottom-accumulated anomalies in the wind power dataset based on the characteristic that the power value of bottom-accumulated anomalies is less than 0 using rule-based filtering.

[0058] S3: The wind power dataset after rule filtering is adaptively divided into K non-overlapping clusters using the CH index and k-means algorithm;

[0059] S4: Divide each cluster obtained in step S3 into L intervals according to its power value, and calculate the wind speed range data and wind speed mean data corresponding to each power interval.

[0060] In this embodiment, dividing each cluster into L intervals according to its power value includes: dividing cluster CK i Divided into ten smaller regions along the power axis {CK} (i,j) |j=0,1,2,…,9}, where each power interval is of equal size.

[0061] S5: Based on the wind speed range data and average wind speed data corresponding to each power range, peak correction combined with boundary modeling is used to identify discrete abnormal data and power-limited abnormal data.

[0062] In this embodiment, based on the distribution characteristics of different wind power anomaly data, wind power anomaly data is divided into three categories: bottom-accumulated anomaly data, discrete anomaly data, and power-limited anomaly data. Bottom-accumulated anomaly data: located at the bottom of the data, characterized by high density, a banded distribution, and power values ​​near zero. The main causes are shutdown for maintenance, wind curtailment, and self-consumption of the internal monitoring and control system when the wind turbine blades are not running. Discrete anomaly data: characterized by random distribution around the standard power curve, decreasing in number with distance from the curve. The main causes are measurement errors, signal noise, and extreme weather. Power-limited anomaly data: located in the middle of the data, characterized by a horizontally dense power band. The main cause is power curtailment, such as… Figure 2 Bottom-accumulated anomaly data is identified and cleaned using rule-based filtering based on its power value being less than zero.

[0063] Preferably, step S3 includes:

[0064] S31: For a given K value, use the K-means algorithm to cluster the wind power dataset and calculate the CH index after clustering;

[0065] S32: By trying different K values, comparing their corresponding CH indices, selecting the K value with the highest CH index as the optimal K value, and obtaining the K clusters corresponding to the optimal K value.

[0066] Preferably, the CH index includes:

[0067]

[0068]

[0069]

[0070] in, This indicates finding the trace of a matrix; Represents the covariance matrix of inter-cluster data; This represents the covariance matrix of the data within a cluster; n represents the total number of data points in all clusters. Indicates the number of clusters; Indicates the first The total number of data in each cluster; Indicates the first Data center points of each cluster, Indicates the first A cluster; This represents the data center point of all clusters; Indicates the first Data points in a cluster.

[0071] In this embodiment, the Calinski-Harabasz (CH) index and k-means algorithm are used to adaptively divide the rule-filtered wind power dataset into K non-overlapping intervals. This maximizes the similarity between data points within the same cluster and minimizes the similarity between different clusters, adaptively determining each classification interval and separating intervals containing power-limiting anomalies from normal data intervals, facilitating subsequent processing. The CH index is used to adaptively determine the number of intervals K. The basic idea of ​​the CH index is to measure intra-class similarity and inter-class similarity, aiming to increase intra-class similarity and decrease inter-class similarity. The wind power dataset is divided into K intervals, denoted as K. The partitioning results are as follows Figure 3 As shown;

[0072] This invention provides a preferred real-time method, which performs the following operations before dividing the wind power dataset into K intervals:

[0073] Iterate through each data point in the wind power dataset, determining whether its wind speed value is greater than 1.2 times the rated wind speed. If so, remove the data point from the wind power dataset and add it to the CU region. After the iteration is complete, the preprocessed wind power dataset and CU region dataset are obtained. Then, the preprocessed wind power dataset is used in step S3 and subsequent abnormal data identification. For the data point d to be processed in the CU region dataset, if its power value p does not meet the requirement... , If d is considered to be discrete outlier data, then the data points in the CU region dataset are sorted in ascending order of power value. This represents the values ​​of the top 25% of the data in the CU region dataset. This represents the values ​​of the last 25% of the data in the CU region dataset, and its recognition result image. Figure 7 As shown.

[0074] Preferably, the calculation of the wind speed range data and average wind speed data corresponding to each power range includes:

[0075] The wind speed range corresponding to each power interval is calculated using the quartile method, thus obtaining the wind speed range data for each power interval. The calculation formula is as follows:

[0076]

[0077]

[0078] in, This indicates the wind speed range corresponding to the power range; This represents the wind speed value corresponding to the first 25% of the wind power data within the power range. This represents the wind speed value corresponding to the last 25% of the wind power data within the power range. Indicates the center point of the power range;

[0079] The average wind speed data corresponding to each power range is as follows: , This represents the average wind speed of the wind power data corresponding to each power range.

[0080] The principle of the quartile method in this embodiment is as follows: Figure 4 As shown, a set of data is sorted in ascending order and divided into four equal parts, Q L It is the lower quartile, representing the value of the data in the top 25% position, Q. U The upper quartile represents the value of the data in the bottom 25% of the dataset. The interquartile range (IQR) reflects the centrality of the dataset. This article uses the quartile method to calculate the CK. (i,j) Discrete data identification threshold Q L-1.5IQR and Q U The difference of +1.5IQR is used as the wind speed range parameter corresponding to the power range, and the corresponding power parameter is the mean value of the power range, i.e., the center point.

[0081] Preferably, step S5 includes:

[0082] S51: Construct the wind speed range data and wind speed mean data corresponding to each power range into a wind speed range data point set and a wind speed mean data point set, respectively;

[0083] In this embodiment, due to the distribution characteristics of power-limited anomaly data, the wind speed range data set and the wind speed mean data set will show uneven spikes, and the power values ​​of these spikes will be in the same power range as the power-limited anomaly data in the original wind power scatter plot. Therefore, it is necessary to correct the spikes in the wind speed range data set and the wind speed mean data set, such as... Figure 5 As shown.

[0084] S52: Extract the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set, respectively; wherein, the peak data includes: the degree of peak protrusion, the starting point of the peak and the ending point of the peak.

[0085] S53: Traverse the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set. When the peak protrusion is greater than the set threshold, the data points from the start point to the end point of the peak are corrected using the linear interpolation method to obtain the corrected wind speed range data point set and wind speed mean data point set.

[0086] S54: Use an exponential function to fit the corrected wind speed range data point set to obtain the wind speed range curve; use a cubic function to fit the corrected wind speed mean data point set to obtain the wind speed mean curve.

[0087] S55: Set left and right boundary thresholds based on the wind speed range curve and the wind speed mean curve, and identify discrete and power-limited abnormal data in the wind power dataset based on the set left and right boundary thresholds.

[0088] Preferably, the peak data of the extracted wind speed range data point set and the peak data of the wind speed mean data point set include:

[0089] S521: Iterate through each data point in the dataset , Indicates the first Data points, Representing data points The corresponding power, Representing data points The corresponding wind speed range or average wind speed, comparing data points The size of its adjacent data points, if and Then the data points Add to the peak point set;

[0090] S522: Iterate through each peak point in the set of peak points. From the peak point Starting from the peak point in the dataset The expansion is performed simultaneously on both the left and right sides to calculate the peak point. to the straight line The shortest distance, where the straight line From the peak point Points obtained by expanding to the left and along the peak point Points obtained by expanding to the right Connected to form;

[0091] In this embodiment, the data is analyzed along the peak point. Simultaneous expansion on both sides includes: assuming along the peak point The corresponding power coordinates are The point obtained after the first leftward expansion The power coordinates are The point obtained after the first rightward expansion The power coordinates are Similarly, after the first The point obtained after the leftward expansion The power coordinates are , No. The point obtained after lateral expansion The power coordinates are .

[0092] S523: When the peak point to the straight line The shortest distance is greater than the peak point The degree of protrusion corresponding to the peak Then the point and points As the peak point The corresponding start and end points of the peak value; wherein, the peak value point The degree of protrusion corresponding to the peak include:

[0093]

[0094] in, Indicates peak point The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed.

[0095] In step S5 of this embodiment, a framework for identifying power-limited abnormal data of wind turbines is proposed, which solves the problem of difficulty in identifying power-limited abnormal data of wind turbines.

[0096] Preferably, the exponential function includes:

[0097]

[0098] in, Indicates power, Indicates the wind speed range. , and Indicates a parameter.

[0099] In this embodiment, after analysis, it is found that due to the wind speed distribution characteristics and the unit's operating characteristics, the wind speed width parameter defined in this paper will undergo abrupt changes in the rated power region. Therefore, the wind speed range data is theoretically approximately an exponential curve. Thus, an exponential function is selected to fit the corrected wind speed range data point set.

[0100] Preferably, the cubic function includes:

[0101]

[0102] in, Indicates power, Indicates the wind speed range. , , and Indicates a parameter.

[0103] In this embodiment, after analysis, due to the wind speed distribution characteristics and unit operating characteristics, the mean wind speed data is theoretically approximately a cubic function. Therefore, a cubic function is selected to fit the corrected mean wind speed data set.

[0104] Preferably, the left boundary threshold is v(x)-0.5w(x), and the right boundary threshold is v(x)+0.5w(x).

[0105] Finally, based on the obtained left boundary threshold of v(x)-0.5w(x) and right boundary threshold of v(x)+0.5w(x), the remaining data points in the wind power dataset are identified, that is, the power values ​​of the data points are substituted into the function. and The system determines whether the wind speed value of a data point is greater than or equal to v(x) - 0.5w(x) and less than or equal to v(x) + 0.5w(x). If so, the data point is considered normal; otherwise, it is considered abnormal. The system identifies discrete abnormal data and power-limited abnormal data. Figure 6 and Figure 8 As shown.

[0106] To verify the effectiveness and universality of the proposed method, actual operating data from wind turbines at a domestic wind farm from March 2014 to March 2015 were selected for verification and analysis. The wind turbines at this wind farm recorded data every minute, and their basic parameters were as follows: cut-in wind speed of 1 m / s, rated wind speed of 12 m / s, cut-out wind speed of 25 m / s, and rated power of 1500 kW. This paper uses turbine No. 1 as an example. First, rule-based filtering was used to identify bottom-accumulated anomalies. Then, the CH index combined with k-means clustering was used to regionalize the wind power data in the SCADA system of turbine No. 1 (see details). Figure 2 The wind speed range data and mean wind speed data were obtained using the quartile method (see details). Figure 3 By combining peak correction with boundary modeling, discrete and power-limited anomaly data can be identified.

[0107] To further verify the effectiveness of the proposed wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling in wind turbine power anomaly data identification, the identification results of the proposed method are compared and analyzed with the anomaly data identification results of the density clustering-based DBSCAN algorithm. The neighborhood radius Eps and the minimum number of points contained (Minpts) are the main parameters. For different datasets, the appropriate Eps and Minpts values ​​vary significantly. For the data in this paper, Eps=0.01 and Minpts=20 were selected. See the comparison for details. Figure 8 As can be seen from the simulation results, the comparison algorithm is not good at identifying power-limited anomaly data, while the method in this paper is not affected by the density of power-limited anomaly data and can accurately identify power-limited anomaly data.

[0108] In summary, this invention utilizes the CH index combined with the k-means clustering algorithm to cluster wind power data, adaptively determining the number of regions to divide the wind speed-power scatter plot into different categories. This improves the similarity of data within the same category and reduces the similarity between different categories, achieving accurate identification of anomalous data across the entire power range in wind power monitoring data. In particular, by employing peak correction combined with boundary modeling, it overcomes the shortcomings of traditional methods in identifying power-limited anomalous data and overcomes the dependence of traditional clustering algorithms on the number of clusters. By dividing the anomalous wind power monitoring data into regions and obtaining wind speed range data and mean wind speed data, the boundary modeling problem is simplified into a curve fitting problem between the two types of data. This solves the problem that traditional identification methods are difficult to achieve globally optimal identification results, improving the accuracy of wind power anomalous data identification. This provides new ideas for researchers conducting related research in the future, and the research method has significant theoretical research value and practical engineering significance.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying wind power anomaly data based on adaptive K-means segmentation and boundary modeling, characterized in that, include: S1: Based on the distribution characteristics of different wind power anomaly data, wind power anomaly data is divided into bottom-accumulated anomaly data, discrete anomaly data, and power-limited anomaly data; among them, bottom-accumulated anomaly data is wind power anomaly data with power values ​​less than 0; S2: Identify and filter out bottom-accumulated anomalies in the wind power dataset based on the characteristic that the power value of bottom-accumulated anomalies is less than 0 using rule-based filtering. S3: The wind power dataset after rule filtering is adaptively divided into K non-overlapping clusters using the CH index and k-means algorithm; Step S3 includes: S31: For a given K value, use the K-means algorithm to cluster the wind power dataset and calculate the CH index after clustering; S32: By trying different K values, comparing their corresponding CH indices, selecting the K value with the highest CH index as the optimal K value, and obtaining the K clusters corresponding to the optimal K value; S4: Divide each cluster obtained in step S3 into L intervals according to its power value, and calculate the wind speed range data and wind speed mean data corresponding to each power interval. The calculation of wind speed range data and average wind speed data corresponding to each power range includes: The wind speed range corresponding to each power interval is calculated using the quartile method, thus obtaining the wind speed range data for each power interval. The calculation formula is as follows: in, This indicates the wind speed range corresponding to the power range; This represents the wind speed value corresponding to the first 25% of the wind power data within the power range. This represents the wind speed value corresponding to the last 25% of the wind power data within the power range. Indicates the center point of the power range; The average wind speed data corresponding to each power range is as follows: , This represents the average wind speed of the wind power data corresponding to each power range. S5: Based on the wind speed range data and average wind speed data corresponding to each power range, peak correction combined with boundary modeling is used to identify discrete and power-limited anomalies. Step S5 includes: S51: Construct the wind speed range data and wind speed mean data corresponding to each power range into a wind speed range data point set and a wind speed mean data point set, respectively; S52: Extract the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set, respectively; wherein, the peak data includes: the degree of peak protrusion, the starting point of the peak and the ending point of the peak. S53: Traverse the peak data of the wind speed range data point set and the peak data of the wind speed mean data point set. When the peak protrusion is greater than the set threshold, the data points from the start point to the end point of the peak are corrected using the linear interpolation method to obtain the corrected wind speed range data point set and wind speed mean data point set. S54: Use an exponential function to fit the corrected wind speed range data point set to obtain the wind speed range curve; use a cubic function to fit the corrected wind speed mean data point set to obtain the wind speed mean curve. S55: Set left and right boundary thresholds based on the wind speed range curve and the wind speed mean curve, and identify discrete and power-limited abnormal data in the wind power dataset based on the set left and right boundary thresholds.

2. The wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling according to claim 1, characterized in that, The CH index includes: in, This indicates finding the trace of a matrix; Represents the covariance matrix of inter-cluster data; This represents the covariance matrix of the data within a cluster; n represents the total number of data points in all clusters. Indicates the number of clusters; Indicates the first The total number of data in each cluster; Indicates the first Data center points of each cluster, Indicates the first A cluster; This represents the data center point of all clusters; Indicates the first Data points in a cluster.

3. The wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling according to claim 1, characterized in that, The peak data of the extracted wind speed range data point set and the peak data of the wind speed mean data point set include: S521: Iterate through each data point in the dataset , Indicates the first Data points, Representing data points The corresponding power, Representing data points The corresponding wind speed range or average wind speed, comparing data points The size of its adjacent data points, if and Then the data points Add to the peak point set; S522: Iterate through each peak point in the set of peak points From the peak point Starting from the peak point in the dataset The expansion is performed simultaneously on both the left and right sides to calculate the peak point. to the straight line The shortest distance, where the straight line From the peak point Points obtained by expanding to the left and along the peak point Points obtained by expanding to the right Connected to form; S523: When the peak point to the straight line The shortest distance is greater than the peak point The degree of protrusion corresponding to the peak Then the point and points As the peak point The corresponding start and end points of the peak value; wherein, the peak value point The degree of protrusion corresponding to the peak include: in, Indicates peak point The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed; Representing data points The corresponding wind speed range or average wind speed.

4. The wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling according to claim 1, characterized in that, The exponential function includes: in, Indicates power, Indicates the wind speed range. , and Indicates a parameter.

5. The wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling according to claim 4, characterized in that, The cubic function includes: in, Indicates power, Indicates the wind speed range. , , and Indicates a parameter.

6. The wind power anomaly data identification method based on adaptive K-means segmentation and boundary modeling according to claim 5, characterized in that, The left boundary threshold is v(x)-0.5w(x), and the right boundary threshold is v(x)+0.5w(x).

Images