A wind turbine blade stall diagnosis method and system based on data distribution

Abstract

The application discloses a wind turbine blade stall diagnosis method and system based on data distribution, and the method comprises the following steps: S1, obtaining a full-pitch power generation data set D of a wind turbine; S2, filtering the full-pitch power generation data set D by using a DBSCAN spatial density clustering algorithm, filtering out a normal data cluster N, and obtaining an initial lower boundary of the normal data cluster N; S3, performing boundary smoothing on the initial lower boundary of the normal data cluster N, and obtaining a smoothed lower boundary; and S4, judging whether a data point of the wind turbine is in a stall working condition based on the smoothed lower boundary. The application has the advantages of low cost, high diagnosis precision, high feasibility and the like.

Description

Technical Field

[0001] This invention mainly relates to the field of wind power technology, specifically to a method and system for diagnosing wind turbine blade stall based on data distribution. Background Technology

[0002] like Figure 1 As shown, blade stall refers to the phenomenon that occurs during wind power generation when wind speed increases to a certain level, causing the blade angle of attack (the angle between the airflow direction and the blade chord) of the wind turbine to exceed a critical value. At this point, the flow conditions on the back of the blade begin to deteriorate, the boundary layer is disrupted, and a vortex region appears at the trailing edge of the blade. At this time, the lift of the blade decreases sharply, while the drag increases significantly, causing the blade to be unable to continue providing sufficient lift to maintain normal rotational speed, thus leading to a decrease in the wind turbine's power generation and efficiency. Summary of the Invention

[0003] To address the technical problems existing in the prior art, this invention provides a low-cost, high-precision, and highly feasible method and system for diagnosing wind turbine blade stall based on data distribution.

[0004] To solve the above-mentioned technical problems, the technical solution proposed by this invention is as follows:

[0005] A method for diagnosing wind turbine blade stall based on data distribution, comprising the following steps:

[0006] S1. Obtain the data set D of wind turbine generators with fully open blades;

[0007] S2. Use the DBSCAN spatial density clustering algorithm to filter the fully open-pitch power generation dataset D, filter out the normal data cluster N, and obtain the initial lower boundary of the normal data cluster N.

[0008] S3. Perform boundary smoothing on the initial lower boundary of the normal data cluster N to obtain the smoothed lower boundary;

[0009] S4. Determine whether the data points of the wind turbine are in stall condition based on the smoothed lower boundary.

[0010] Preferably, the specific process of step S2 is as follows:

[0011] S201. Divide the fully open-pole power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0012] S202, for each wind speed bin i Clustering is performed by sorting by active power and using the minimum number of points in the ε-neighborhood and MinPts as parameters.

[0013] S203. Select the cluster with the largest number of samples as the bin for that wind speed. i Normal data cluster N i ;

[0014] S204, all wind speeds bin i The normal data cluster set {N1, N2, N3...N} is as follows m} is defined as the initial normal data cluster N.

[0015] Preferably, in step S201, the specific process of obtaining the ε-neighborhood is as follows:

[0016] S1.1. Divide the fully open-pitch power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0017] S1.2, For each bin i Calculate the upper and lower limits L of active power up ,L low Filter bin i Active power between L up ,L low From the samples between, we get:

[0018] G i ={g ij |L low <g ij <L up} (1)

[0019] S1.3, G i Sort the data in ascending order of active power, then calculate the difference between each sorted power value to obtain the difference d between two adjacent active power values. ij j = 1, 2, 3..., take d ij Find the maximum value of j = 1, 2, 3... to obtain ε. i ;

[0020] 1.4, Take the collective {ε} i The 80th percentile of |i=1,2,3...,m} is used as the ε neighborhood value.

[0021] Preferably, in step S201, the process of obtaining the minimum number of points MinPts is as follows:

[0022] S2.1. Divide the fully open-pitch power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0023] S2.2, For each bin i Calculate the upper and lower limits L of active power up ,Llow upper and lower limit distance R i =L up -L low Total number of samples Count i ;

[0024] S2.3, Take R i Find the 80th quantile of i = 1, 2, 3...m to obtain R. Let:

[0025]

[0026] S2.4, Take MinPts i The minimum value of i = 1, 2, 3... is used as the minimum number of points parameter in MinPts.

[0027] Preferably, in step S3, the boundary smoothing process is as follows:

[0028] Let the lower boundary after the nth smoothing be:

[0029] L n (v), n=1,2,3... (3)

[0030] The lower boundary of the nth smoothing is calculated from the lower boundary of the (n-1)th smoothing, that is:

[0031]

[0032] Where L n-1 (v) is the lower boundary curve after the (n-1)th smoothing, v i For bin i The average wind speed, i = 1, 2, 3... m.

[0033] Preferably, the cutoff condition for a smooth lower boundary is:

[0034]

[0035] Where n cut To smooth out the cutoff number, p cut It is a constant.

[0036] Preferably, in step S4, if data point i is below the lower boundary and i is not a noise point, then data point i is a stall sample; otherwise, it is another abnormal sample.

[0037] The present invention also discloses a computer program product, comprising a computer program that, when executed by a processor, performs the steps of the method described above.

[0038] The present invention further discloses a computer-readable storage medium having a computer program stored thereon, the computer program executing the steps of the method described above when run by a processor.

[0039] The present invention also discloses a wind turbine blade stall diagnosis based on data distribution, including a memory and a processor connected to each other, wherein the memory stores a computer program, and the computer program executes the steps of the method described above when run by the processor.

[0040] Compared with the prior art, the advantages of the present invention are as follows:

[0041] This invention considers the spatial distribution characteristics of the data. First, it uses the DBSCAN spatial density clustering algorithm to perform preliminary screening on the wind turbine fully open-pitch power generation dataset D, obtaining an initial normal data cluster N. Then, it smooths the boundaries of data cluster N to obtain a smoothed lower boundary, which is then used to determine whether a data point has stalled. The above scheme has the following advantages:

[0042] 1) DBSCAN plus boundary smoothing provides a new perspective.

[0043] The DBSCAN algorithm is a widely used clustering algorithm that divides data points into several categories based on the spatial distribution density of the data. After initially filtering out normal categories, it uses boundary smoothing to determine the lower boundary of normal data, and then identifies whether the turbine has stalled based on the lower boundary. Compared with other methods, this scheme is based on existing wind turbine SCADA data and does not require the addition of new sensors (blade pressure, deformation, torsion, etc.). In addition, the DBSCAN method combined with boundary smoothing fully explores the distribution patterns of wind turbine SCADA data, providing a new perspective for blade stall diagnosis.

[0044] 2) Improved the feasibility and accuracy of blade stall diagnosis.

[0045] Since this solution does not require the addition of new sensors, no new hardware costs are required for either existing or newly built wind farms. This allows for seamless deployment and implementation in both existing and new wind farms, making the solution highly feasible. Furthermore, because this solution is based on the distribution patterns of a large amount of historical data, the analysis process is clear and straightforward, with each step having a clear basis and purpose, which helps improve the accuracy of diagnosis. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of blade stall in existing technology.

[0047] Figure 2 This is a flowchart of an embodiment of the wind turbine blade stall diagnosis method of the present invention.

[0048] Figure 3 This is a diagram showing the diagnostic results of a normal unit in this invention.

[0049] Figure 4 This is a diagram showing the diagnostic results of the blade stall unit in this invention. Detailed Implementation

[0050] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0051] like Figure 2 As shown, the wind turbine blade stall diagnosis method based on data distribution provided in this embodiment of the invention includes the following steps:

[0052] S1. Obtain the wind turbine fully open pitch power generation dataset D; where the fully open pitch power generation dataset D is collected when the wind turbine is in the fully open pitch power generation condition; where the fully open pitch power generation condition refers to the condition in which the wind turbine is in the power generation state and the wind turbine blades are in the optimal pitch angle state set by the main control program to maximize the acquisition of wind energy.

[0053] S2. Use the DBSCAN spatial density clustering algorithm to filter the fully open-pitch power generation dataset D, filter out the normal data cluster N, and obtain the initial lower boundary of the normal data cluster N.

[0054] S3. Perform boundary smoothing on the initial lower boundary of the normal data cluster N to obtain the smoothed lower boundary;

[0055] S4. Determine whether the data points of the wind turbine are in stall condition based on the smoothed lower boundary.

[0056] Specifically, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm widely used in data mining and machine learning. It organizes data by dividing points in a dataset into core points, boundary points, and noise points, thus forming clusters.

[0057] The DBSCAN algorithm is based on two main parameters: ε (neighborhood size) and MinPts (minimum number of points). For each point in the dataset, the algorithm checks if its ε-neighborhood contains at least MinPts points; if so, this point is considered a core point and may be expanded into a cluster. The algorithm starts with any unvisited core point and builds clusters by continuously adding core points within its ε-neighborhood until no more core points can be added. Points that are not core points but belong to the ε-neighborhood of a core point are considered boundary points of clusters, while points that do not belong to any cluster are considered noise points.

[0058] Specifically, in step S2, the process of DBSCAN clustering and filtering the initial normal data clusters is as follows:

[0059] S201. Divide the fully open-pole power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0060] S202, for each wind speed bin i Clustering is performed by sorting by active power and using the minimum number of points in the ε-neighborhood and MinPts as parameters.

[0061] S203. Select the cluster with the largest number of samples as the bin for that wind speed. i Normal data cluster N i ;

[0062] S204, all wind speeds (bin) i The normal data cluster set {N1, N2, N3...N} is as follows m} is defined as the initial normal data cluster N.

[0063] Specifically, the process for determining the DBSCAN parameters is as follows:

[0064] (1) Steps for determining the ε-neighborhood

[0065] S1.1. Divide the fully open-pitch power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0066] S1.2, For each bin i Calculate the upper and lower limits L of active power up ,L low Filter bin i Active power between L up ,L low between samples g ij , to obtain G i :

[0067] G i ={g ij |L low <g ij <L up} (1)

[0068] S1.3, G i Sort the data in ascending order of active power, then calculate the difference between each sorted power value to obtain the difference d between two adjacent active power values. ij j = 1, 2, 3..., take d ij Find the maximum value of j = 1, 2, 3... to obtain ε. i ;

[0069] S1.4, Take the set {ε} i The 80th percentile of |i=1,2,3...,m} is used as the ε neighborhood value.

[0070] (2) Steps for determining the minimum number of points in MinPts

[0071] S2.1. Divide the fully open-pitch power generation dataset D into m bins according to wind speed. i i = 1, 2, 3...m;

[0072] S2.2, For each bin i Calculate the upper and lower limits L of active power up ,L low upper and lower limit distance R i =L up -L low Total number of samples Count i ;

[0073] S2.3, Take R i Find the 80th quantile of i = 1, 2, 3...m to obtain R. Let:

[0074]

[0075] S2.4, Take MinPts i The minimum value of i = 1, 2, 3... is used as the minimum number of points parameter in MinPts.

[0076] Specifically, the fully open-pitch power generation dataset D is divided into m bins according to wind speed. i ,i=1,2,3...m; Let each bin i The average wind speed is v i N i The minimum active power is P i_min In the wind speed-power rectangular coordinate system, from point (v i ,P i The curve formed by the initial normal data cluster N is defined as the initial lower boundary L0(v).

[0077] Because the rate of change of the initial lower boundary L0(v) with respect to wind speed v often exhibits abrupt changes, it appears visually as a distinct "sawtooth" pattern. Based on the analysis experience of a large amount of actual wind farm operation data, it can be found that the normal operation data of wind turbine units should have smooth boundaries in spatial distribution. Therefore, it is necessary to smooth the initial lower boundary L0(v).

[0078] The method for boundary smoothing is as follows:

[0079] Let L be the lower boundary after the nth smoothing.n (v) is:

[0080] L n (v), n=1,2,3... (3)

[0081] The lower boundary of the nth smoothing is calculated from the lower boundary of the (n-1)th smoothing, that is:

[0082]

[0083] Where L n-1 (v) is the lower boundary curve after the (n-1)th smoothing, v i For bin i The average wind speed, i = 1, 2, 3... m.

[0084] The number of times boundary smoothing is performed depends on the data itself. Generally speaking, after a certain number of boundary smoothing operations, the degree of change in the lower boundary will become very small when smoothing is continued.

[0085] Therefore, the cutoff condition for lower boundary smoothing is set as follows:

[0086]

[0087] Where n cut To smooth out the cutoff number, p cut It is a constant.

[0088] Specifically, in step S4, if data point i is below the lower boundary and i is not a noise point, then data point i is a stall sample; otherwise, it is another abnormal sample.

[0089] In practical applications, a wind field in a mountainous area in southern China was selected as the sample set. Some information about this wind field is shown in Table 1.

[0090] Table 1 Basic Information of the Case Wind Field

[0091]

[0092] The wind field's comprehensive and balanced composition is key to its success. Specifically, it covers both positive samples (i.e., blade stall samples) and negative samples (i.e., blade non-stall samples), a characteristic crucial for building accurate and reliable analytical models. The coexistence of positive and negative samples not only ensures that multiple dimensions of the problem are covered during the analysis but also effectively enhances the model's ability to identify and generalize real-world data.

[0093] Normal unit diagnostic results are as follows Figure 3 As shown, the diagnostic results for the blade stall unit are as follows: Figure 4 As shown in the figure. The red dashed line represents the lower boundary of the normal data scatter points, and the scatter points represent the data points of fully loaded propeller power generation.

[0094] Figure 3 All data points for the fully open-blade power generation of the unit are above the lower boundary, which clearly shows that the unit has not experienced blade stall.

[0095] Figure 4 When the data points for fully open-pitch power generation are close to the rated power of the unit, the power characteristics of the wind turbine unit decrease significantly, and a large number of data points are below the lower boundary. According to the algorithm diagnosis, the data below the lower boundary are determined to be stall sample data.

[0096] This invention considers the spatial distribution characteristics of the data. First, it uses the DBSCAN spatial density clustering algorithm to perform preliminary screening on the wind turbine fully open-pitch power generation dataset D, obtaining an initial normal data cluster N. Then, it smooths the boundaries of data cluster N to obtain a smoothed lower boundary, which is then used to determine whether a data point has stalled. The above scheme has the following advantages:

[0097] 1) DBSCAN plus boundary smoothing provides a new perspective.

[0098] The DBSCAN algorithm is a widely used clustering algorithm that divides data points into several categories based on the spatial distribution density of the data. After initially filtering out normal categories, it uses boundary smoothing to determine the lower boundary of normal data, and then identifies whether the turbine has stalled based on the lower boundary. Compared with other methods, this scheme is based on existing wind turbine SCADA data and does not require the addition of new sensors (blade pressure, deformation, torsion, etc.). In addition, the DBSCAN method combined with boundary smoothing fully explores the distribution patterns of wind turbine SCADA data, providing a new perspective for blade stall diagnosis.

[0099] 2) Improved the feasibility and accuracy of blade stall diagnosis.

[0100] Since this solution does not require the addition of new sensors, no new hardware costs are required for either existing or newly built wind farms. This allows for seamless deployment and implementation in both existing and new wind farms, making the solution highly feasible. Furthermore, because this solution is based on the distribution patterns of a large amount of historical data, the analysis process is clear and straightforward, with each step having a clear basis and purpose, which helps improve the accuracy of diagnosis.

[0101] This invention also discloses a computer program product, including a computer program that, when run by a processor, performs the steps of the method described above. This invention further discloses a computer-readable storage medium storing a computer program that, when run by a processor, performs the steps of the method described above. This invention also discloses a wind turbine blade stall diagnosis based on data distribution, including an interconnected memory and a processor, wherein the memory stores a computer program that, when run by a processor, performs the steps of the method described above. The products, media, and systems of this invention, corresponding to the methods described above, also possess the advantages described above.

[0102] The present invention can implement all or part of the processes in the methods of the above embodiments, or it can be implemented by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium includes: any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. The memory is used to store computer programs and / or modules. The processor implements various functions by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may include high-speed random access memory, as well as non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart media cards (SMC), secure digital (SD) cards, flash cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0103] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for diagnosing stall in wind turbine blades based on data distribution, characterized in that, Including the following steps: S1. Obtain the data set D of wind turbine generators with fully open blades; S2. Use the DBSCAN spatial density clustering algorithm to filter the fully open-pitch power generation dataset D, filter out the normal data cluster N, and obtain the initial lower boundary of the normal data cluster N. S3. Perform boundary smoothing on the initial lower boundary of the normal data cluster N to obtain the smoothed lower boundary; S4. Determine whether the data points of the wind turbine are in stall condition based on the smoothed lower boundary. The specific process of step S2 is as follows: S201, The fully open-pole power generation dataset D is divided according to wind speed. indivual ; S202, for various wind speeds Sort by active power, with Neighborhood and Clustering calculations are performed using the minimum number of points as a parameter. S203. Select the cluster with the largest number of samples as the wind speed. Normal data clusters ; S204, All wind speeds Normal data cluster set Defined as the initial normal data cluster .

2. The wind turbine blade stall diagnosis method based on data distribution according to claim 1, characterized in that, In step S201, we obtain The specific process of neighborhood is as follows: S1.1, The fully open-pitch power generation dataset D is divided according to wind speed into... indivual ; S1.2, for each Calculate the upper and lower limits of active power ,filter Active power between From the samples between, we get: (1) S1.3, will Sort the data in ascending order of active power, then calculate the difference between each sorted power value to obtain the difference between two adjacent active power data. ,Pick The maximum value is obtained. ; S1.4, Take the collective The 80th percentile as Neighborhood value.

3. The wind turbine blade stall diagnosis method based on data distribution according to claim 1, characterized in that, In step S201, we obtain The process for minimizing the number of points is as follows: S2.1, The fully open-pitch power generation dataset D is divided according to wind speed into... indivual ; S2.2, for each Calculate the upper and lower limits of active power upper and lower limit distance Total number of samples ; S2.3, Take The 80th percentile is obtained. ,make: (2) S2.4, Take The minimum value, as Minimum number of points parameter.

4. The wind turbine blade stall diagnosis method based on data distribution according to any one of claims 1-3, characterized in that, In step S3, the boundary smoothing process is as follows: Let the first The lower boundary after smoothing is: (3) No. The lower boundary of the second smoothing is determined by the first... The lower boundary of the sub-smoothing method is obtained through calculation, i.e.: (4) in For the first The lower boundary curve after secondary smoothing. for The average wind speed.

5. The wind turbine blade stall diagnosis method based on data distribution according to claim 4, characterized in that, The cutoff condition for smoothing the lower boundary is: (5) in To smooth out the number of cutoffs, It is a constant.

6. The wind turbine blade stall diagnosis method based on data distribution according to any one of claims 1-3, characterized in that, In step S4, if data point i is below the lower boundary and i is not a noise point, then data point i is a stall sample; otherwise, it is another abnormal sample.

7. A computer program product, comprising a computer program, characterized in that, The computer program is executed by a processor to perform the steps of the method as described in any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-6.

9. A wind turbine blade stall diagnosis system based on data distribution, comprising an interconnected memory and a processor, wherein the memory stores a computer program, characterized in that, The computer program, when run by a processor, performs the steps of the method as described in any one of claims 1-6.

Images