A method and device for removing abnormal scattered points of wind turbine power curve

Abstract

The present invention discloses a method for removing abnormal scatter points from a wind turbine power curve, comprising the following steps: Step S1) obtaining SCADA data for a target turbine, the SCADA data including a status code, wind speed, and power; Step S2) performing preliminary screening according to status code rules to obtain wind speed and power data for the target turbine in a normal power generation state; Step S3) using the DBSCAN clustering algorithm to determine a primary cluster of wind speed-power scatter points; Step S4) classifying data that meets the criteria into the primary cluster, determining the primary cluster as normal data, and removing data from other clusters; Step S5) fitting the power curve based on a RANSAC regression model; and Step S6) setting scatter point boundaries for the power curve to remove abnormal scatter points. This solution accurately removes abnormal scatter points, facilitating accurate calculation and reliable application of power curves.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation, and in particular to a method and device for removing abnormal scattered points on a power curve of a wind turbine generator set. Background Art

[0002] The power curve of a wind turbine represents the operational performance of its power generation system and serves as critical input data for subsequent optimization, control, and evaluation. Due to factors such as icing, power curtailment, status code confusion, and program upgrades, wind turbine power curves may contain abnormal scatter points. Status code rules alone cannot completely eliminate these abnormal scatter points. When these abnormal scatter points are applied to turbine performance optimization, fault warning and diagnosis, and assessment and evaluation, they can introduce errors into the model. Therefore, when applying wind turbine power curve data, it is important to consider eliminating these abnormal scatter points to ensure accurate and reliable results.

[0003] At present, the widely used method for removing abnormal scatter points from the power curve of wind turbines is mainly based on the status code that characterizes the status of the unit. Specifically, the wind speed, power and other data of the unit under normal power generation status are obtained through status code rule screening. However, the wind speed and power data during actual operation of the unit are affected by many factors, and there are still outliers after status code rule screening. The popular method for removing abnormal scatter points nowadays is the density-based clustering algorithm, which assumes that the scatter points distributed on the power curve are relatively dense, and removes abnormal scatter points through the distribution differences of the power curve scatter points. Due to the influence of operating conditions, the scatter points of the unit power curve are not evenly distributed. When the outlier scatter points are densely distributed and difficult to separate from the scatter points on the power curve, the effect of using the clustering algorithm to remove abnormal scatter points will be greatly reduced.

[0004] In summary, how to accurately and completely eliminate abnormal scatter points in the power curve is a technical research problem that technicians in the wind power field need to solve. Summary of the Invention

[0005] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a method and device for removing abnormal scattered points in the power curve of a wind turbine generator set, which can accurately remove abnormal scattered points and facilitate accurate calculation and reliable application of the power curve.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for removing abnormal scatter points from a wind turbine power curve comprises the following steps: step S1) acquiring SCADA data of a target turbine, the SCADA data including a status code, wind speed, and power; step S2) performing preliminary screening according to a status code rule to obtain wind speed and power data of the target turbine in a normal power generation state; step S3) obtaining a main cluster of wind speed-power scatter points based on a DBSCAN clustering algorithm; step S4) classifying data that meets a condition into the main cluster, determining the main cluster as normal data, and removing data from other clusters; step S5) fitting a power curve based on a RANSAC regression model; and step S6) setting a scatter point boundary of the power curve to remove abnormal scatter points. The present invention provides a method for removing abnormal scatter points from a wind turbine power curve. The specific process comprises: obtaining status code data of a target turbine and preliminarily removing abnormal data from the turbine power curve using status code rules; screening out a main cluster of the power curve based on a DBSCAN clustering model; assigning data that reaches a certain power above the rated wind speed to the main cluster based on the rated wind speed and rated power, and removing data from other clusters to form a new normal data set; fitting the power curve based on a trained RANSAC regression model, setting a scatter point boundary for the power curve, and removing abnormal scatter points to obtain normal scatter point data for the turbine power curve. The process of training the RANSAC regression model includes obtaining learning samples, which are extracted based on a power stratification method; and the process of setting the scatter point boundary for the power curve includes obtaining the wind speed width of each power segment and setting a boundary threshold based on the average wind speed width.

[0008] Due to reasons such as icing, power restrictions, status code confusion, and program upgrades, there are abnormal scatter points in the power curve of wind turbines, and the abnormal scatter points cannot be completely eliminated by using only the status code rules. Using the density-based DBSCAN clustering algorithm, the abnormal scatter points of the power curve with a relatively scattered distribution can be eliminated, but it is still difficult to eliminate the abnormal scatter points of the power curve with a relatively dense distribution, and the clustering algorithm has certain shortcomings. The RANSAC regression model is used to fit the power curve scatter points before the rated power into a curve, and the power curve scatter point boundary threshold is set in combination with the wind speed width of each power segment, which can eliminate the abnormal scatter points of the power curve with a relatively dense distribution. On the basis of the existing technology, the present invention first uses the status code rules for preliminary screening, and then eliminates the abnormal scatter points with a relatively scattered distribution based on the DBSCAN clustering algorithm, and then uses the RANSAC regression model to fit the power curve and eliminate the abnormal scatter points with a relatively dense distribution. This method can make the fitted power curve more accurate, and make the power curve have a reasonable and smooth shape, which is conducive to the accurate calculation and reliable application of the subsequent power curve.

[0009] Preferably, the specific process of step S3 includes the following steps:

[0010] Step S301: Assume that the wind speed-power scatter point set is {(v1, p1), (v2, p2), ..., (v n,p n )};

[0011] Step S302: For the wind speed-power scatter point set {(v1, p1), (v2, p2), ..., (v n ,p n )} is normalized to obtain the normalized wind speed-power scatter point set Step S303: normalize the wind speed-power scatter point set Input into the DBSCAN clustering model to obtain the main cluster of wind speed-power scatter points.

[0012] As an example, the condition in step S4 is: wind speed - power scatter (v i ,p i ) i ≥v 额定 And p i ≥p 额定 *0.95.

[0013] Preferably, the specific process of step S5 includes the following steps:

[0014] Step S501: obtaining wind speed-power learning samples using a power stratification method, wherein the power in the wind speed-power scatter points of the learning samples is lower than the rated power;

[0015] Step S502: Based on the learning samples, train the RANSAC regression model with the actual power p i As input, predict the corresponding wind speed v_pr i .

[0016] Preferably, the specific process of step S501 includes the following steps:

[0017] Step S5011: Based on T power intervals {(0,δ),(δ,2δ),…,((T-1)*δ,p 额定 )}, δ represents the set unit power length, divides the wind speed-power scatter points into T data sets {dataset 1,dataset 2,…,dataset T}, and counts the number of samples num of power quartiles of all data sets f ;

[0018] Step S5012: Calculate the wind speed width d of each data set i , according to the set wind speed width threshold ∈, obtain the initial learning sample, specifically: when v i ≤∈+v min When the wind speed-power scatter point (v i ,p i ) is the initial learning sample, where v minIndicates the minimum wind speed in the corresponding data set;

[0019] Step S5013: Based on the total number of samples in each data set {num 1, num 2, ..., num T} and the number of power quartile samples num f , get the final learning sample, specifically: when num i≤num f When num i>num f When , the final learning samples are obtained by sampling at equal intervals based on the initial learning samples.

[0020] As a preference, in step S6, the specific process of setting the power curve scatter point boundary is as follows: removing the wind speed width d in each data set i The wind speed width value greater than the wind speed width threshold ∈ is used to calculate the average value of the remaining wind speed widths. The power curve scatter point boundary threshold σ is set to:

[0021]

[0022] Among them, k and h are actual experience parameters.

[0023] As a preference, in step S6, when -σ≤v_pr i -v i ≤σ, then wind speed-power scatter (v i ,p i ) is the normal scatter point of the power curve; when v_pr i -v i >σ or v_pr i -v i <σ, then the wind speed-power scatter (v i ,p i ) are abnormal scatter points on the power curve.

[0024] A device for removing abnormal scattered points from a wind turbine power curve adopts the above-mentioned method for removing abnormal scattered points from a wind turbine power curve, comprising:

[0025] The SCADA data acquisition module is used to obtain the SCADA data of the target unit, including status code, wind speed, and power data; the status code rule screening module is used to preliminarily screen the SCADA data according to the status code rules to obtain the wind speed and power data of the target unit in the normal power generation state;

[0026] A DBSCAN clustering model screening module is used to input the wind speed and power data in the normal power generation state into the DBSCAN clustering model, and obtain a main cluster of wind speed-power scatter points in combination with the rated power of the target unit;

[0027] A RANSAC regression model module is used to fit a power curve using the main cluster of wind speed-power scatter points;

[0028] The fitted power curve screening module is used to set the power curve scatter point boundary of the target unit by using the fitted power curve and the wind speed width of each power segment, remove abnormal scatter points outside the boundary, and obtain the normal scatter points of the power curve of the target unit;

[0029] The RANSAC regression model module includes:

[0030] a learning sample acquisition unit, configured to select data as learning samples based on the power segment using the main cluster of wind speed-power scatter points;

[0031] The model training unit is used to train the RANSAC regression model using the learning samples.

[0032] A method and device for removing abnormal scattered points of a wind turbine power curve, comprising:

[0033] memory for storing computer programs;

[0034] The processor is configured to implement the steps of the above-mentioned method for eliminating abnormal scattered points in the power curve of a wind turbine generator set when executing the computer program in the memory.

[0035] A readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the above-mentioned method for eliminating abnormal scattered points in a wind turbine power curve.

[0036] Therefore, the advantages of the present invention are: first, preliminary screening is performed using the status code rule, and then the scattered abnormal points are eliminated based on the DBSCAN clustering algorithm, and then the RANSAC regression model is used to fit the power curve to eliminate the densely distributed abnormal points, so as to accurately eliminate the abnormal points of the power curve, which is conducive to the accurate calculation and reliable application of the subsequent power curve. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The figure is a flow chart of a method for removing abnormal scattered points from a wind turbine power curve according to an embodiment of the present invention.

[0038] Figure 2 The figure is a schematic structural diagram of a device for removing abnormal scattered points on a wind turbine power curve according to an embodiment of the present invention.

[0039] Figure 3 The figure is a schematic structural diagram of a device for removing abnormal scattered points on a wind turbine power curve according to an embodiment of the present invention.

[0040] Figure 4The figure is a schematic diagram of the specific structure of a device for removing abnormal scattered points of a wind turbine power curve according to an embodiment of the present invention.

[0041] 1. SCADA data acquisition module 2. Status code rule screening module 3. DBSCAN clustering model screening module 4. RANSAC regression model module 41. Learning sample acquisition unit 42. Model training unit 5. Fitting power curve screening module 6. Memory 7. Processor 8. System bus 9. Power supply device 10. Computer program 11. Data 12. Network interface 13. Display device 14. Input device 15. Storage medium 16. Operating system. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0043] Example 1:

[0044] like Figure 1 As shown, a method for removing abnormal scattered points of a wind turbine power curve includes the following steps:

[0045] Step 1: Obtain the SCADA data corresponding to the target unit, where the SCADA data includes a status code representing the unit status, wind speed, and power;

[0046] Step 2: Perform preliminary screening according to the status code rules to obtain the wind speed and power data of the unit in normal power generation status;

[0047] Step 3: Based on the DBSCAN clustering algorithm, obtain the main cluster of wind speed-power scatter points;

[0048] Step 4: Satisfy v i ≥v 额定 , and p i ≥p 额定 *The data with a condition of 0.95 is classified into the main cluster, and the data of other clusters are eliminated to determine that the main cluster is normal data;

[0049] Step 5: Based on the RANSAC regression model, fit the power curve, set the scatter point boundary of the power curve, remove abnormal scatter points, and obtain the normal scatter points of the unit power curve.

[0050] In step 3, the DBSCAN clustering algorithm is used to determine the main cluster of wind speed-power scatter points, including:

[0051] Assume that the wind speed-power scatter point set is {(v1,p1),(v2,p2),…,(v n ,p n )}, normalize the data to obtain the normalized wind speed-power scatter point set The DBSCAN clustering model is input to obtain the main cluster of wind speed-power scatter points.

[0052] The specific steps of the DBSCAN clustering algorithm are as follows:

[0053] Step 3.1: Determine the radius parameter r and the minimum point parameter MinPts, where the radius parameter r is the radius of the neighborhood around the wind speed-power scatter point. The area within the scatter point radius r is called the r neighborhood. The minimum point parameter MinPts is the minimum number of scatter points contained in the r neighborhood.

[0054] Step 3.2: Randomly select a wind speed-power scatter point as the starting point and calculate the number of scatter points in its r neighborhood;

[0055] Step 3.3: In the r neighborhood of the scattered point, when the number of wind speed-power scattered points is greater than or equal to the minimum point parameter MinPts, the object is a core point and a cluster label is assigned. When the number of wind speed-power scatter points is less than the minimum point parameter MinPts, the object is a noise point;

[0056] Step 3.4: Visit all neighboring scattered points in the neighborhood of the scattered point r. If the visited scattered points do not have cluster labels, assign cluster labels When the visited point is a core point, the neighboring points of the visited core point are visited in sequence until there is no core point in the r neighborhood of the visited point.

[0057] Step 3.5: Select an unvisited wind speed-power scatter point as a new starting point and repeat steps 3.2 to 3.4 until all wind speed-power scatter points are assigned cluster labels. Select the cluster containing the most wind speed-power scatter points as the main cluster.

[0058] In step 5, the RANSAC regression model is used to fit the power curve, including:

[0059] Based on the acquired learning sample data, the RANSAC regression model is trained with the actual power p i As input, predict the corresponding wind speed v_pr i .

[0060] The specific steps of fitting the power curve of the RANSAC model are as follows:

[0061] Step 5.1: Randomly select x data from the learning sample and fit a model H(p) that may meet the requirements;

[0062] Step 5.2: Based on the curve fitted by model H(p), count the data set M whose Euclidean distance to the curve is within the error range [-τ,τ];

[0063] Step 5.3: Assume that there are x data points and the data set M contains m data points in total. Assume that the model requires at least n data points. If m>n, refit the model H′(p) that meets the requirements based on the x randomly selected data points and the data set M. If m<n, return to step 5.1.

[0064] Step 5.4: Based on the curve fitted by model H′(p), calculate the tolerance error μ between the selected x data, the data set M and the curve. Let μ be the historical optimal tolerance error. opt , the historical optimal model is H opt (p), when μ<μ opt , then μ opt =μ,H opt (p)=H′(p); when μ>μ opt , then return to step 5.1 until the given number of iterations S is reached.

[0065] In step 5, the RANSAC regression model is used to fit the power curve, which also includes:

[0066] The power stratification method is used to obtain wind speed and power learning samples, and the learning samples only consider wind speed-power scatter points below the rated power.

[0067] The specific steps for obtaining learning samples using the power stratification method are as follows:

[0068] 1) Based on T equal power intervals {[0,δ],(δ,2δ],…,((T-1)*δ,p 额定 )}, δ is the set unit power length, the wind speed-power scatter points are divided into T data sets {dataset 1,dataset 2,…,dataset T}, and the number of samples num of the power quartile points of all data sets is counted f ;

[0069] 2) Calculate the wind speed width d of each data set i , according to the set wind speed width threshold ∈, obtain the initial learning sample, specifically: when v i ≤∈+v min , obtain wind speed-power scatter (v i ,p i ) is the initial learning sample, where v min is the minimum wind speed in the data set;

[0070] 3) Based on the total number of samples in each data set {num 1, num 2, ..., num T} and the number of power quartile samples num f , get the final learning sample, specifically: when num i≤num f , the final learning sample is consistent with the initial learning sample; when num i>num f ,The final learning samples are obtained by sampling at equal intervals based on the initial learning samples.

[0071] The scatter point boundary of the power curve fitted by the RANSAC regression model in step 5 is calculated as follows:

[0072] Eliminate the wind speed width d in each data set i The wind speed width value greater than the wind speed width threshold ∈ is used to calculate the average value of the remaining wind speed widths. The power curve scatter point boundary threshold σ is set to: Among them, k and h are actual experience parameters.

[0073] The process of obtaining the normal scattered point data of the target unit power curve in step 5 is as follows:

[0074] When -σ≤v_pr i -v i ≤σ, then wind speed-power scatter (v i ,p i ) is the normal scattered data of the power curve; when v_pr i -v i >σ or v_pr i -v i <σ, then the wind speed-power scatter (v i ,p i ) is the abnormal scatter data of the power curve.

[0075] Example 2:

[0076] like Figure 2 As shown, a device for removing abnormal scattered points of a wind turbine power curve comprises:

[0077] SCADA data acquisition module 1 is used to obtain the SCADA data of the target unit, including status code, wind speed, and power data; status code rule screening module 2 is used to screen the SCADA data according to the status code rules to obtain the wind speed and power data of the target unit in the normal power generation state;

[0078] DBSCAN clustering model screening module 3 is used to input the wind speed and power data in the normal power generation state into the DBSCAN clustering model, and combine it with the rated power of the target unit to obtain the main cluster of wind speed-power scatter points;

[0079] RANSAC regression model module 4 is used to fit the power curve using the main cluster of wind speed-power scatter points, including:

[0080] A learning sample acquisition unit 41 is configured to select an appropriate amount of data as learning samples based on the power segment using the main cluster of wind speed-power scatter points;

[0081] A model training unit 42 is used to train a RANSAC regression model using learning samples;

[0082] The power curve fitting screening module 5 is used to fit the power curve and the wind speed width of each power segment using the RANSAC regression model, set the power curve scatter point boundary of the target unit, eliminate the abnormal scatter point data outside the boundary, and obtain the normal scatter point data of the power curve of the target unit.

[0083] The device for removing abnormal scattered points from the power curve of a wind turbine provided in this embodiment can execute the above method embodiment. Its implementation principle and technical effects are similar and will not be described in detail here.

[0084] Example 3:

[0085] like Figure 3 As shown, a device for removing abnormal scattered points of a wind turbine power curve comprises:

[0086] Memory 6, for storing computer programs;

[0087] The processor 7 is configured to implement the steps of the above-mentioned method for removing abnormal scattered points of the wind turbine power curve when executing the computer program in the memory 6 .

[0088] Specifically, such as Figure 4 As shown, a device for removing abnormal scatter points in the power curve of a wind turbine generator system includes a power supply device 9, a processor 7, a memory 6, a network interface 12, a display device 13, an input device 14 and a storage medium 15 connected via a system bus 8. Among them, the processor 7 is used to provide computing and control capabilities. The storage medium 15 stores an operating system 16, a computer program 10 and data 11. The memory 6 provides an operating environment for the operating system 16 and the computer program 10 in the storage medium 15. The network interface 12 can be a wireless or wired network interface for network connection communication. The display device 13 can be various types of displays, including LCD displays, CRT displays, LED displays, etc. The input device 14 can be a keyboard, a mouse, a touch layer of a display screen, etc.

[0089] The steps in the method for removing abnormal scattered points in a wind turbine power curve described above can be implemented by the structure of a device for removing abnormal scattered points in a wind turbine power curve.

[0090] Example 4:

[0091] A readable storage medium having computer program instructions stored thereon. When the computer program is executed by a processor 7, the steps of the above-described method for removing abnormal scatter points from a wind turbine power curve are implemented. The readable storage medium may be a readable storage medium containing various program codes, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0092] The embodiment described above is only a preferred solution of the present invention and does not limit the present invention in any form. Other variations and modifications are possible without exceeding the technical solution described in the claims.

Claims

1. A method for removing abnormal scattered points from a wind turbine power curve, characterized in that: The following steps are involved: Step S1: Acquire SCADA data of the target unit, wherein the SCADA data includes status code, wind speed and power; Step S2: Perform preliminary screening according to the status code rules to obtain the wind speed and power data of the target unit in the normal power generation state; Step S3: Based on the DBSCAN clustering algorithm, obtain the main cluster of wind speed-power scatter points according to the wind speed and power data; Step S4: classify the data that meets the conditions into the main cluster, determine that the main cluster is normal data, and eliminate the data in other clusters; Step S5: training a RANSAC regression model based on the main cluster data, and fitting a power curve based on the RANSAC regression model; Step S6: Setting the scatter point boundary of the power curve to remove abnormal scatter points.

2. A method for removing abnormal scattered points from a wind turbine power curve according to claim 1, characterized in that: The specific process of step S3 includes the following steps: Step S301: Assume that the wind speed-power scatter point set is {(v1, p1), (v2, p2), ..., (v n ,p n )}; Step S302: For the wind speed-power scatter point set {(v1, p1), (v2, p2), ..., (v n ,p n )} is normalized to obtain the normalized wind speed-power scatter point set Step S303: normalize the wind speed-power scatter point set Input into the DBSCAN clustering model to obtain the main cluster of wind speed-power scatter points.

3. The method for removing abnormal scattered points from a wind turbine power curve according to claim 1, characterized in that: The conditions in step S4 are: wind speed - power scatter (v i ,p i ) i ≥v 额定 And p i ≥p 额定 *0.

95.

4. A method for removing abnormal scattered points from a wind turbine power curve according to claim 1, characterized in that: The specific process of step S5 includes the following steps: Step S501: obtaining wind speed-power learning samples using a power stratification method, wherein the power in the wind speed-power scatter points of the learning samples is lower than the rated power; Step S502: Based on the learning samples, train the RANSAC regression model with the actual power p i As input, predict the corresponding wind speed v_pr i .

5. A method for removing abnormal scattered points from a wind turbine power curve according to claim 4, characterized in that: The specific process of step S501 includes the following steps: Step S5011: Based on T power intervals {(0,δ),(δ,2δ),…,((T-1)*δ,p 额定 )}, δ represents the set unit power length, divides the wind speed-power scatter points into T data sets {dataset 1,dataset 2,…,dataset T}, and counts the number of samples num of power quartiles of all data sets f ; Step S5012: Calculate the wind speed width d of each data set i , according to the set wind speed width threshold ∈, obtain the initial learning sample, specifically: when v i ≤∈+v min When the wind speed-power scatter point (v i ,p i ) is the initial learning sample, where v min Indicates the minimum wind speed in the corresponding data set; Step S5013: Based on the total number of samples in each data set {num 1, num 2, ..., num T} and the number of power quartile samples num f , get the final learning sample, specifically: when num i≤num f When num i>num f When , the final learning samples are obtained by sampling at equal intervals based on the initial learning samples.

6. A method for removing abnormal scattered points from a wind turbine power curve according to claim 1, characterized in that: In step S6, the specific process of setting the power curve scatter point boundary is as follows: eliminating the wind speed width d in each data set i The wind speed width value greater than the wind speed width threshold ∈ is used to calculate the average value of the remaining wind speed widths. The power curve scatter point boundary threshold σ is set to: Among them, k and h are actual experience parameters.

7. A method for removing abnormal scattered points from a wind turbine power curve according to claim 1, characterized in that: In step S6, v_pr i is the wind speed predicted by the RANSAC regression model, when -σ≤v_pr i -v i ≤σ, then wind speed-power scatter (v i ,p i ) is the normal scatter point of the power curve; when v_pr i -v i >σ or v_pr i -v i <σ, then the wind speed-power scatter (v i ,p i ) are abnormal scatter points on the power curve.

8. A device for removing abnormal scatter points from a wind turbine power curve, using a method for removing abnormal scatter points from a wind turbine power curve according to any one of claims 1 to 7, characterized in that: include: The SCADA data acquisition module is used to obtain the SCADA data of the target unit, including status code, wind speed, and power data; the status code rule screening module is used to preliminarily screen the SCADA data according to the status code rules to obtain the wind speed and power data of the target unit in the normal power generation state; A DBSCAN clustering model screening module is used to input the wind speed and power data in the normal power generation state into the DBSCAN clustering model, and obtain a main cluster of wind speed-power scatter points in combination with the rated power of the target unit; A RANSAC regression model module is used to fit a power curve using the main cluster of wind speed-power scatter points; The fitted power curve screening module is used to set the power curve scatter point boundary of the target unit by using the fitted power curve and the wind speed width of each power segment, remove abnormal scatter points outside the boundary, and obtain the normal scatter points of the power curve of the target unit; The RANSAC regression model module includes: a learning sample acquisition unit, configured to select data as learning samples based on the power segment using the main cluster of wind speed-power scatter points; The model training unit is used to train the RANSAC regression model using the learning samples.

9. A device for removing abnormal scattered points from a wind turbine power curve, characterized in that: include: memory for storing computer programs; The processor is configured to implement the steps of the method for removing abnormal scattered points of a wind turbine power curve as described in any one of claims 1 to 7 when executing the computer program on the memory.

10. A readable storage medium, characterized in that: The readable storage medium stores computer program instructions, and when the computer program instructions are executed by the processor, the steps of the method for removing abnormal scattered points of a wind turbine power curve according to any one of claims 1 to 7 are implemented.

Images