A multi-fan head wind speed abnormal constant value detection method and device

Abstract

The present application relates to wind power prediction technical field, especially point to a kind of multi-wind turbine nose wind speed abnormal constant value detection method, device and computer readable storage medium.The method described in the present application includes: obtaining the nose wind speed time series of multiple wind turbines of wind farm, and grouping multiple wind turbines;Select the interval of constant wind speed in the nose wind speed time series of each wind turbine as the suspicious abnormal constant value interval of this wind turbine;Compare the suspicious abnormal constant value interval of all wind turbines in each group, obtain the voting time series corresponding to each group, respectively calculate the distance between the suspicious abnormal constant value interval of each wind turbine in each group and the corresponding interval in voting time series, determine the abnormal constant value interval of each wind turbine.The present application fully considers the correlation between the nose wind speed time series of multiple wind turbines, reduces the influence of local anomaly on constant value detection, and improves the detection precision of wind speed abnormal constant value.

Description

Technical Field

[0001] This invention relates to the field of wind power prediction technology, and in particular to a method, device, and computer-readable storage medium for detecting abnormal constant values ​​of wind speed at the turbine inlets of multiple wind turbines. Background Technology

[0002] Accurate wind power forecasting provides grid operators and wind farm managers with precise information on future wind power production. By predicting wind power output, power system dispatching and operational planning can be improved, helping to optimize electricity market transactions and supply-demand balance, and preventing power shortages or surpluses caused by wind power volatility. Furthermore, accurate forecasting can help power companies and wind farms develop reasonable energy procurement and sales strategies, thereby reducing electricity transaction costs. Wind speed at the wind turbine inlet is one of the key data points required for predicting wind power output.

[0003] The operating environment of wind farms is complex and variable, and wind speed data acquisition may be subject to interference and noise from various factors, leading to various problems in the raw wind speed data. Since "data and features determine the upper limit of the model," effective methods for detecting abnormal turbine wind speed data can help eliminate abnormal raw data in wind power prediction, improve the quality and accuracy of raw data, and provide a reliable data foundation for the training and performance of wind power prediction models.

[0004] However, most existing methods for detecting abnormal wind speed data at the turbine head are based on two-dimensional scatter plots to identify abnormal data. They are difficult to identify complete sequential abnormal data and cannot fully explore the correlation characteristics between data from multiple turbines, resulting in incomplete removal of abnormal data. Summary of the Invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the existing technology is unable to identify complete sequential abnormal data and cannot fully explore the correlation characteristics between multiple wind turbine data, resulting in incomplete removal of abnormal data.

[0006] To solve the above-mentioned technical problems, the present invention provides a method for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines, comprising:

[0007] S1. Obtain the head wind speed time series of multiple wind turbines in the wind farm, and group the multiple wind turbines based on the head wind speed time series.

[0008] S2. Select the interval where the wind speed is constant in the time series of the wind speed at the head of each wind turbine as the suspected abnormal constant value interval of the wind turbine. Mark the wind speed data in the suspected abnormal constant value interval of each wind turbine as 0, and mark the wind speed data in the normal interval as 1 to obtain the binarized time series of each wind turbine.

[0009] S3. Timestamp alignment is performed on the binarized time series of all wind turbines within each cluster;

[0010] Compare the binarized time series of all wind turbines in each cluster, remove the intervals where all elements of all wind turbines are 1 at the same time point, and obtain the remaining suspicious anomaly constant value intervals corresponding to each cluster.

[0011] The mode of the elements in the remaining suspicious anomaly constant value intervals corresponding to each subgroup at the same time point is selected as the voting result at that time point, and the voting time series composed of the voting results of each subgroup is obtained.

[0012] Calculate the distance between the suspected abnormal constant value interval of each wind turbine in each subgroup and the corresponding interval in the voting time series. The suspected abnormal constant value interval with a distance greater than the preset distance threshold is taken as the abnormal constant value interval of the wind turbine, and the head wind speed time series of the abnormal constant value interval of each wind turbine is removed.

[0013] In one embodiment of the present invention, S1 involves using an agglomerative hierarchical clustering algorithm to group multiple wind turbines, including:

[0014] Each wind turbine is treated as an initial cluster;

[0015] The inter-group distance between all groups is calculated using the longest distance method. If there is an inter-group distance less than or equal to a preset inter-group distance threshold, the two groups with the smallest inter-group distance are selected for merging until the inter-group distance of all groups is greater than the preset inter-group distance threshold.

[0016] In one embodiment of the present invention, the step of calculating the inter-group distances between all pairs of clusters according to the longest distance method specifically includes:

[0017] Choose any two clusters, calculate the distance between the head wind speed time series X of any fan in one cluster and the head wind speed time series Y of any fan in the other cluster, until the distance between the head wind speed time series of any two fans that do not belong to the same cluster is obtained, and select the maximum distance value as the inter-cluster distance between the two clusters.

[0018] Repeat the above steps until the inter-cluster distances between all pairs of clusters are obtained.

[0019] In one embodiment of the present invention, the distance between the head wind speed time series X of any wind turbine in one cluster and the head wind speed time series Y of any wind turbine in another cluster is 1-C. pearson C pearsonLet X be the Pearson coefficient between the time series X and Y of the nose wind speed, and let Y be the distance between them, which is [-1, 1].

[0020] The Pearson coefficient C between the time series X and Y of the nose wind speed pearson The expression is:

[0021]

[0022] Where X = [x1, x2, ..., x n ], Y = [y1, y2, ..., y n ], where n is the number of elements in the wind speed time series of the wind turbine head. and These are the average values ​​of the elements in X and Y, respectively.

[0023] In one embodiment of the present invention, S2, selecting the interval of constant wind speed in the time series of the turbine head wind speed of each wind turbine as the interval of suspected abnormal constant value for that wind turbine, includes:

[0024] For each wind turbine, the head wind speed time series W hub Perform moving average filtering;

[0025] The time series W of the inlet wind speed for each wind turbine after applying moving average filtering. filter The elements in the array are subjected to a difference operation, and their absolute values ​​are taken. If the absolute value is greater than or equal to a preset difference threshold, it is recorded as 1; if the absolute value is less than the preset difference threshold, it is recorded as 0. This yields the time series W for each wind turbine. note =[w note_1 ,w note_2 ,…,w note_n ];

[0026] Calculate the time series W for each wind turbine note The accumulated value at each time point is used to obtain the accumulated time series W for each wind turbine. cum ;

[0027] The cumulative time series W for each wind turbine cum Continuous identical values ​​are grouped into a set of intervals, and the length of each set of intervals is calculated. Intervals whose length is greater than a preset interval length threshold are identified as suspicious abnormal constant value intervals for the wind turbine.

[0028] In one embodiment of the present invention, the cumsum function in Python is used to calculate the time series W for each wind turbine. note The accumulated value at each time point is used to obtain the accumulated time series W. cum =[w note_1 ,wnote_1 +w note_2 ,…,w note_1 +w note_2 +…+w note_n ] = [w cum_1 ,w cum_2 ,…,w cum_n ].

[0029] In one embodiment of the present invention, the groupby function in Python is used to calculate the interval length of each group of intervals.

[0030] In one embodiment of the present invention, in S3, the Euclidean distance between the interval of the suspected abnormal constant value of each wind turbine in each cluster and the corresponding interval in the voting time series is calculated, and the expression is:

[0031]

[0032] Where [t] a1 ,t a2 ,…,t an ] represents the remaining range of constant values ​​for suspected anomalies, [t] v1 ,t v2 ,…,t vn [ ] represents the interval in the voting time series that corresponds to the interval of the remaining suspicious anomaly constant value.

[0033] The present invention also provides a device for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines, comprising:

[0034] The wind turbine grouping module is used to obtain the head wind speed time series of multiple wind turbines in a wind farm, and to group the multiple wind turbines based on the head wind speed time series.

[0035] The single-fan anomaly detection module is used to select the interval of constant wind speed in the head wind speed time series of each fan as the suspicious abnormal constant value interval of the fan. The wind speed data in the suspicious abnormal constant value interval of each fan is marked as 0, and the wind speed data in the normal interval is marked as 1, so as to obtain the binarized time series of each fan.

[0036] The multi-fan joint anomaly detection module is used to align the timestamps of the binarized time series of all fans in each subgroup; compare the binarized time series of all fans in each subgroup, remove the intervals where all elements are 1 at the same time point for all fans, and obtain the remaining suspicious anomaly constant value intervals corresponding to each subgroup; select the mode of the elements at the same time point of the remaining suspicious anomaly constant value intervals corresponding to each subgroup as the voting result at that time point, and obtain the voting time series composed of the voting results for each subgroup; calculate the distance between the suspicious anomaly constant value interval of each fan in each subgroup and the corresponding interval in the voting time series, and take the suspicious anomaly constant value interval with a distance greater than a preset distance threshold as the anomaly constant value interval of that fan, and remove the head wind speed time series of the anomaly constant value interval of each fan.

[0037] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines.

[0038] The technical solution of the present invention has the following advantages compared with the prior art:

[0039] The present invention discloses a method for detecting abnormal constant values ​​of wind speed at the turbine heads. First, the collected wind speed time series of multiple turbine heads are grouped. Then, abnormal constant values ​​of wind speed are detected for the wind speed time series of individual turbine heads within each group. Finally, correlation anomaly detection is performed on the wind speed time series of all turbine heads within the same group. This method fully considers the correlation between the wind speed time series of multiple turbine heads, obtaining more comprehensive data characteristics of the wind speed time series, reducing the impact of local anomalies on constant value detection, improving the detection accuracy of abnormal constant values ​​of wind speed, and thus improving the accuracy of wind power prediction. Attached Figure Description

[0040] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein...

[0041] Figure 1 This is a flowchart of a method for detecting abnormal constant wind speed at the head of multiple fans according to the present invention;

[0042] Figure 2 This is a schematic diagram illustrating the calculation of the remaining suspicious anomaly constant value interval in an embodiment of the present invention;

[0043] Figure 3 This is a schematic diagram illustrating the calculation of the voting time series in an embodiment of the present invention;

[0044] Figure 4This is a schematic diagram illustrating the distance between the interval of suspected abnormal constant values ​​of wind turbines and the corresponding interval in the voting time series in an embodiment of the present invention;

[0045] Figure 5 This is a comparison chart of the wind speed curves in this embodiment of the invention without removing the abnormal constant value intervals and the wind speed curves after removing the abnormal constant value intervals using the method of this invention. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0047] Example 1

[0048] Reference Figure 1 As shown, this embodiment provides a method for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines, including:

[0049] S1. Obtain the head wind speed time series of multiple wind turbines in the wind farm, and group the multiple wind turbines based on the head wind speed time series.

[0050] This embodiment uses an agglomerative hierarchical clustering algorithm to group multiple wind turbines, including:

[0051] S11. Treat each wind turbine as an initial cluster;

[0052] S12. Calculate the inter-group distances between all pairs of clusters using the longest distance method, and construct a distance matrix, specifically including:

[0053] Choose any two clusters and calculate the distance 1-C between the time series X of the wind turbine's head velocity in one cluster and the time series Y of the wind turbine's head velocity in the other cluster. pearson C pearson Let X be the Pearson coefficient between the time series X and Y of the nose wind speed, and let Y be the distance between them, which is [-1, 1].

[0054] The Pearson coefficient C between the time series X and the time series Y of the nose wind speed is... pearson The expression is:

[0055]

[0056] Where X = [x1, x2, ..., x n ], Y = [y1, y2, ..., y n ], where n is the number of elements in the wind speed time series of the wind turbine head. and These are the average values ​​of the elements in X and Y, respectively.

[0057] The process continues until the distance between the head wind speed time series of any two wind turbines that do not belong to the same cluster is obtained, and the maximum distance value is selected as the inter-cluster distance between the two clusters.

[0058] Repeat the above steps until the inter-cluster distances between all pairs of clusters are obtained.

[0059] S13. If there is a group distance less than or equal to the preset group distance threshold, then select the two groups with the smallest group distance to merge until the group distance of all groups is greater than the preset group distance threshold.

[0060] S2. Select the intervals in the time series of the wind speed at the turbine head of each wind turbine where the wind speed remains constant as the intervals of suspected abnormal constant values ​​for that wind turbine, specifically including:

[0061] S211. Use the `convolve` function in Python to process the time series of the wind speed at the turbine head for each wind turbine. hub Perform moving average filtering.

[0062] S212. The time series of the wind speed at the turbine head for each wind turbine after the moving average filtering is W. filter The elements in the array are subjected to a difference operation, and their absolute values ​​are taken. If the absolute value is greater than or equal to a preset difference threshold, it is recorded as 1; if the absolute value is less than the preset difference threshold, it is recorded as 0. This yields the time series W for each wind turbine, consisting of 0s and 1s. note =[w note_1 ,w note_2 ,…,w note_n ].

[0063] S213. Use the cumsum function in Python to calculate the time series w for each wind turbine. note The accumulated value at each time point is used to obtain the accumulated time series W for each wind turbine. cum =[w note_1 ,w note_1 +w note_2 ,…,w note_1 +w note_2 +…+w note_n ] = [w cum_1 ,w cum_2 ,…,w cum_n ].

[0064] S214, The cumulative time series W of each wind turbine. cumContinuous identical values ​​are grouped into a single interval. The groupby function in Python is used to calculate the length of each interval. Intervals with lengths greater than a preset threshold are identified as suspicious abnormal constant value intervals for the wind turbine.

[0065] For example, time series W cum Chinese w cum_1 w cum_2 w cum_3 If they have the same value, then w will be... cum_1 w cum_2 w cum_3 Grouped together, the corresponding interval length is 3; w cum_7 w cum_8 w cum_9 w cum_10 w cum_11 w cum_12 If they have the same value, then [w] cum_7 ,…,w cum12 They are grouped together, with a corresponding interval length of 6. In this embodiment, the interval length threshold is set to 5, and the interval [w] is grouped together. cum_7 ,…,w cum12 If the length of ] is 6, then the interval is a suspicious constant value abnormal interval.

[0066] The wind speed data in the suspicious abnormal constant value range of each wind turbine is marked as 0, and the wind speed data in the normal range is marked as 1, thus obtaining the binarized time series of each wind turbine.

[0067] S3. Compare the suspected abnormal constant value intervals of all wind turbines within each cluster, determine the constant value intervals for each wind turbine, and eliminate them. Specifically, this includes:

[0068] S31. Timestamp alignment is performed on the binarized time series of all wind turbines within each cluster.

[0069] S32. Compare the binarized time series of all wind turbines in each cluster, remove the intervals where all elements of all wind turbines are 1 at the same time point, and obtain the remaining suspicious constant value intervals corresponding to each cluster.

[0070] In this embodiment, refer to Figure 2 As shown, the elements of the binarized time series corresponding to the timestamps are multiplied together. The intervals with a multiplication result of 1 are classified as normal intervals corresponding to the group, and the intervals with a multiplication result of 0 are classified as the remaining suspicious anomaly constant value intervals corresponding to the group. No further operations are performed on the normal intervals corresponding to the group.

[0071] S33. Select the mode of the elements in the remaining suspicious anomaly constant value intervals corresponding to each subgroup at the same time point as the voting result at that time point, and obtain the voting time series composed of the voting results for each subgroup.

[0072] In this embodiment, refer to Figure 3 As shown, there are n wind turbines in a certain cluster. The mode of the binarized time series after timestamp alignment is 1 at the first time point. Therefore, the element of the voting time series of this cluster at the first time point is recorded as 1. And so on, the voting time series of this cluster is obtained.

[0073] S34. Calculate the Euclidean distance between the interval of suspicious anomaly constant values ​​for each wind turbine within each cluster and the corresponding interval in the voting time series:

[0074]

[0075] Where [t] a1 ,t a2 ,…,t an ] represents the remaining range of constant values ​​for suspected anomalies, [t] v1 ,t v2 ,…,t vn [ ] represents the interval in the voting time series that corresponds to the interval of the remaining suspicious anomaly constant value.

[0076] In this embodiment, refer to Figure 4 As shown, the elements corresponding to the first three time points in the binarized time series of wind turbine k are 0. Therefore, the interval corresponding to the first three time points is the interval of suspicious anomaly constant value for wind turbine k. The elements corresponding to the first three time points in the voting time series of the cluster to which wind turbine k belongs are 1, 1, and 0, respectively. Calculate the Euclidean distance between the interval of suspicious anomaly constant value for wind turbine k and the corresponding interval in the voting time series:

[0077] The interval of suspicious abnormal constant values ​​with a distance greater than a preset distance threshold is taken as the abnormal constant value interval of the wind turbine, and the head wind speed time series of each wind turbine's abnormal constant value interval is removed.

[0078] The multi-fan head wind speed anomaly constant value detection method described in this embodiment first divides the collected head wind speed time series of multiple fans into clusters, performs anomaly constant value detection on the head wind speed time series of individual fans within each cluster, and then performs correlation anomaly detection on the head wind speed time series of all fans within the same cluster. This fully considers the correlation between the head wind speed time series of multiple fans, obtains more comprehensive head wind speed time series data features, reduces the impact of local anomalies on constant value detection, and improves the detection accuracy of anomaly constant value wind speed.

[0079] Example 2

[0080] To verify the beneficial effects of the present invention, this embodiment is scientifically demonstrated through economic benefit calculations and simulation experiments.

[0081] In this embodiment, the wind speed time series of the turbine heads of multiple wind turbines at a wind farm in North China was collected as sample data. The installed capacity of the wind farm is 80.0MW. The measured power data for 16 days from June 10 to June 25, 2023, and the average value of the short-term predicted power data for the corresponding time period (15 minutes) were collected as sample data.

[0082] The formula for the short-term forecast accuracy of the power grid to which this wind farm belongs is as follows:

[0083]

[0084] Where Acc is the day-ahead forecast accuracy, Cap is the maximum operating capacity of the new energy power plant on the assessment day, and P is the maximum operating capacity of the new energy power plant on the assessment day. j Let P' be the actual power output of the new energy power station at time j. j Let j be the predicted power output of the new energy power station, and n be the number of prediction points assessed on that day.

[0085] To verify the impact of removing abnormal constant values ​​from multiple wind turbines on the accuracy of the short-term power prediction model, this embodiment uses the extreme gradient boosting tree model to construct the short-term power prediction model.

[0086] During the training phase, the average wind speed time series of multiple wind turbines before removing outlier data and the average wind speed time series of multiple wind turbines after removing outlier constant value intervals using the method of this invention are used as inputs, and the sum of the measured power of multiple wind turbines is used as the output. Figure 5 This is a comparison chart of wind speed curves without removing abnormal constant value intervals and wind speed curves after removing abnormal constant value intervals using the method of this invention.

[0087] During the forecasting phase, the forecast wind speed from the numerical weather prediction provided by the meteorological supplier is used as input, and the output is the predicted power.

[0088] The accuracy rates before and after using the method of this invention to remove abnormal constant value intervals are shown in Table 1.

[0089] 2023-06-10 75.22％ 74.50％ 2023-06-11 68.79％ 65.49％ 2023-06-12 76.72％ 77.21％ 2023-06-13 88.61％ 88.44％ 2023-06-14 80.67％ 81.65％ 2023-06-15 73.20％ 73.15％ 2023-06-16 86.05％ 89.89％ 2023-06-17 70.72％ 73.33％ 2023-06-18 86.76％ 91.65％ 2023-06-19 26.53％ 25.56％ 2023-06-20 84.84％ 85.56％ 2023-06-21 84.97％ 87.28％ 2023-06-22 77.79％ 80.94％ 2023-06-23 66.03％ 71.07％ 2023-06-24 81.13％ 81.69％ 2023-06-25 67.18％ 68.37％

[0090] Table 1. Comparison of daily accuracy before and after removing constant-value outliers.

[0091] Table 1 shows the daily accuracy comparison before and after removing constant value outliers in the 16-day instance data. Table 2 shows the average accuracy comparison before and after removing constant value outliers.

[0092] average value 74.7％ 75.98％

[0093] Table 2. Comparison of average accuracy before and after removing constant-value outliers.

[0094] As can be seen from Tables 1 and 2, in the 16 days of example data, after removing constant value anomalies, the accuracy was higher than the accuracy before adjustment for 11 days. The average accuracy after adjustment was 0.28% higher than the average accuracy before adjustment, indicating that the multi-fan head wind speed anomaly constant value detection method described in this invention can effectively improve the accuracy of short-term prediction models.

[0095] Example 3

[0096] This embodiment provides a device for detecting abnormal constant wind speed at the head of multiple wind turbines, including:

[0097] The wind turbine grouping module is used to obtain the head wind speed time series of multiple wind turbines in a wind farm, and to group the multiple wind turbines based on the head wind speed time series.

[0098] The single-fan anomaly detection module is used to select the interval of constant wind speed in the head wind speed time series of each fan as the suspicious abnormal constant value interval of the fan. The wind speed data in the suspicious abnormal constant value interval of each fan is marked as 0, and the wind speed data in the normal interval is marked as 1, so as to obtain the binarized time series of each fan.

[0099] The multi-fan joint anomaly detection module is used to align the timestamps of the binarized time series of all fans in each subgroup; compare the binarized time series of all fans in each subgroup, remove the intervals where all elements are 1 at the same time point for all fans, and obtain the remaining suspicious anomaly constant value intervals corresponding to each subgroup; select the mode of the elements at the same time point of the remaining suspicious anomaly constant value intervals corresponding to each subgroup as the voting result at that time point, and obtain the voting time series composed of the voting results for each subgroup; calculate the distance between the suspicious anomaly constant value interval of each fan in each subgroup and the corresponding interval in the voting time series, and take the suspicious anomaly constant value interval with a distance greater than a preset distance threshold as the anomaly constant value interval of that fan, and remove the head wind speed time series of the anomaly constant value interval of each fan.

[0100] Example 4

[0101] This embodiment provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the above-described method for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines.

[0102] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0103] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0104] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0105] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0106] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for detecting abnormal constant values ​​of wind speed at the head of multiple fans, characterized in that, include: S1. Obtain the head wind speed time series of multiple wind turbines in the wind farm, and group the multiple wind turbines based on the head wind speed time series. S2. Select the interval where the wind speed is constant in the time series of the wind speed at the head of each wind turbine as the suspected abnormal constant value interval of the wind turbine. Mark the wind speed data in the suspected abnormal constant value interval of each wind turbine as 0, and mark the wind speed data in the normal interval as 1 to obtain the binarized time series of each wind turbine. S3. Timestamp alignment is performed on the binarized time series of all wind turbines within each cluster; Compare the binarized time series of all wind turbines in each cluster, remove the intervals where all elements of all wind turbines are 1 at the same time point, and obtain the remaining suspicious anomaly constant value intervals corresponding to each cluster. The mode of the elements in the remaining suspicious anomaly constant value intervals corresponding to each subgroup at the same time point is selected as the voting result at that time point, and the voting time series composed of the voting results of each subgroup is obtained. Calculate the distance between the suspected abnormal constant value interval of each wind turbine in each subgroup and the corresponding interval in the voting time series. The suspected abnormal constant value interval with a distance greater than the preset distance threshold is taken as the abnormal constant value interval of the wind turbine, and the head wind speed time series of the abnormal constant value interval of each wind turbine is removed.

2. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 1, characterized in that, In S1, an agglomerative hierarchical clustering algorithm is used to group multiple wind turbines, including: Each wind turbine is treated as an initial cluster; The inter-group distance between all groups is calculated using the longest distance method. If there is an inter-group distance less than or equal to a preset inter-group distance threshold, the two groups with the smallest inter-group distance are selected for merging until the inter-group distance of all groups is greater than the preset inter-group distance threshold.

3. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 2, characterized in that, The calculation of inter-group distances between all pairs of clusters using the longest distance method specifically includes: Choose any two clusters, calculate the distance between the head wind speed time series X of any fan in one cluster and the head wind speed time series Y of any fan in the other cluster, until the distance between the head wind speed time series of any two fans that do not belong to the same cluster is obtained, and select the maximum distance value as the inter-cluster distance between the two clusters. Repeat the above steps until the inter-cluster distances between all pairs of clusters are obtained.

4. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 3, characterized in that, The distance between the head wind speed time series X of any wind turbine in one cluster and the head wind speed time series Y of any wind turbine in another cluster is 1-C. pearson C pearson Let X be the Pearson coefficient between the time series X and Y of the nose wind speed, and let Y be the distance between them, which is [-1, 1]. The Pearson coefficient C between the time series X and Y of the nose wind speed pearson The expression is: Where X = [x1, x2, ..., x] n ], Y = [y1, y2, ..., y n ], where n is the number of elements in the wind speed time series of the wind turbine head. and These are the average values ​​of the elements in X and Y, respectively.

5. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 1, characterized in that, In S2, the step of selecting the interval of constant wind speed in the time series of the turbine head wind speed of each wind turbine as the interval of suspected abnormal constant value for that wind turbine includes: For each wind turbine, the head wind speed time series W hub Perform moving average filtering; The time series W of the inlet wind speed for each wind turbine after applying moving average filtering. filter The elements in the array are subjected to a difference operation, and their absolute values ​​are taken. If the absolute value is greater than or equal to a preset difference threshold, it is recorded as 1; if the absolute value is less than the preset difference threshold, it is recorded as 0. This yields the time series W for each wind turbine. note =[w note_1 w note_2 , ..., w note_n ]; Calculate the time series W for each wind turbine note The accumulated value at each time point is used to obtain the accumulated time series W for each wind turbine. cum ; The cumulative time series W of each wind turbine cum Continuous identical values ​​are grouped into a set of intervals, and the length of each set of intervals is calculated. Intervals whose length is greater than a preset interval length threshold are identified as suspicious abnormal constant value intervals for the wind turbine.

6. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 5, characterized in that, The time series W for each wind turbine is calculated using the cumsum function in Python. note The accumulated value at each time point is used to obtain the accumulated time series W. cum =[w note_1 w note_1 +w note_2 , ..., w note_1 +w note_2 +…+w note_n ] = [w cum_1 w cum_2 , ..., w cum_n ].

7. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 5, characterized in that, Use the groupby function in Python to calculate the length of each interval.

8. The method for detecting abnormal constant wind speed at the head of multiple fans according to claim 1, characterized in that, In S3, the Euclidean distance between the interval of suspicious anomaly constant values ​​for each wind turbine within each cluster and the corresponding interval in the voting time series is calculated, expressed as: Where [t] a1 , t a2 , ..., t an ] represents the remaining range of constant values ​​for suspected anomalies, [t] v1 , t v2 , ..., t vn [ ] represents the interval in the voting time series that corresponds to the interval of the remaining suspicious anomaly constant value.

9. A device for detecting abnormal constant wind speed at the head of multiple fans, characterized in that, include: The wind turbine grouping module is used to obtain the head wind speed time series of multiple wind turbines in a wind farm, and to group the multiple wind turbines based on the head wind speed time series. The single-fan anomaly detection module selects the intervals where the wind speed at the fan head remains constant in the time series of each fan as the suspected abnormal constant value intervals for that fan. The wind speed data in the suspected abnormal constant value intervals for each fan are marked as 0, while the wind speed data in the normal intervals are marked as 1. Obtain the binarized time series for each wind turbine; The multi-fan joint anomaly detection module is used to align the timestamps of the binarized time series of all fans in each subgroup; compare the binarized time series of all fans in each subgroup, remove the intervals where all elements are 1 at the same time point for all fans, and obtain the remaining suspicious anomaly constant value intervals corresponding to each subgroup; select the mode of the elements at the same time point of the remaining suspicious anomaly constant value intervals corresponding to each subgroup as the voting result at that time point, and obtain the voting time series composed of the voting results for each subgroup; calculate the distance between the suspicious anomaly constant value interval of each fan in each subgroup and the corresponding interval in the voting time series, and take the suspicious anomaly constant value interval with a distance greater than a preset distance threshold as the anomaly constant value interval of that fan, and remove the head wind speed time series of the anomaly constant value interval of each fan.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for detecting abnormal constant values ​​of wind speed at the head of multiple wind turbines as described in any one of claims 1 to 8.

Images