A wind turbine unit yaw control optimization method based on operation data clustering

Abstract

This invention relates to an optimization method for yaw control of wind turbine generators based on operational data clustering. By filtering massive amounts of operational data from wind turbine generators, key information related to yaw deviation, such as wind speed, power generation, and wind angle, is extracted. The data is then divided into intervals based on wind speed and power generation, and cluster optimization is performed on the power generation and wind angle within each wind speed interval to find the optimal cluster center point for each interval. Finally, with the goal of maximizing wind turbine generator power generation, the comprehensive wind deviation angle is calculated and used as the yaw control compensation angle for the wind turbine generator, thereby improving yaw accuracy and increasing generator power generation efficiency.

Description

Technical Field

[0001] This invention relates to the field of wind turbine generator control, and in particular to a wind turbine generator yaw control optimization method based on operational data clustering, which is used to optimize yaw control when the wind turbine generator is not in sync with the wind, thereby improving the accuracy of wind sync. Background Technology

[0002] During the operation of a wind turbine, the wind direction may change. In order to improve the power generation efficiency of the wind turbine, the wind turbine will yaw to adjust the rotor of the wind turbine to be in the windward position.

[0003] Currently, the mainstream yaw control of wind turbine generators uses the wind vane behind the rotor as the basis for judgment. When the wind direction changes, such as when the 10-minute average wind direction or the 60-second average wind direction change exceeds the yaw start threshold, the yaw system of the wind turbine generator will activate and adjust the wind turbine generator to be in a state facing the wind.

[0004] However, in actual operation, problems such as the influence of wind turbine rotation, inaccurate zeroing due to abnormal installation of the wind vane, loosening of the wind vane due to equipment vibration, and increased measurement errors due to aging of the wind vane can all cause inaccurate wind direction measurement by the wind turbine, resulting in abnormal yaw of the wind turbine. The power generation of the wind turbine is proportional to the cube of the cosine of the yaw angle. When the yaw angle is 0 degrees, the power generation is at its maximum, with zero power loss; when the yaw angle is 5 degrees, the power generation is 98.863% of the normal power, with a power loss of 1.137%; when the yaw angle is 10 degrees, the power generation is 95.552% of the normal power, with a power loss of 4.448%. It is evident that inaccurate yaw has a significant impact on the power generation of the wind turbine, resulting in a substantial loss of electricity.

[0005] Currently, some researchers have conducted studies on the optimization of yaw deviation control for wind turbine generators. For example, the patent application No. 2018114069744, titled "An Adjustment Method and Device for a Wind Turbine Yaw System," calculates the overall optimal wind angle based on the numerical intervals of all wind speed range segments and the optimal wind angle corresponding to each segment. However, it does not consider the influence of wind frequency on the overall optimal wind angle across different wind speed range segments. The patent application No. 201911117633X, titled "A Wind Turbine Yaw Calibration Method and System," establishes a wind speed-yaw error calibration value lookup table and uses this table to determine the yaw error calibration value under the current wind direction and speed. However, it does not provide a comprehensive wind deviation. In practical applications, when the wind direction changes rapidly, this method exhibits a lag, potentially causing greater errors in yaw control.

[0006] Therefore, it is necessary to optimize and improve the existing yaw control methods to avoid power loss caused by wind turbine misalignment, which would affect the economics of wind farms. Summary of the Invention

[0007] In view of the above situation and to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a wind turbine yaw control optimization method based on operational data clustering, which is used to optimize yaw control when the wind turbine is not aligned with the wind, improve the accuracy of wind alignment, avoid power generation loss caused by the wind turbine's inaccurate wind alignment, and improve the economic efficiency of wind farm operation.

[0008] The technical solution solved by this invention is:

[0009] An optimization method for yaw control of wind turbine generators based on operational data clustering is proposed. First, operational data of the wind turbine generators from the previous period (0.5-2 years) is read. This data includes wind speed (v), power generation (P), rotational speed (m), windward angle (α), and blade angle (β). Second, the read operational data is initially screened to remove abnormal data. The screened data is then clustered according to standard power curves and categorized by wind speed to obtain the average power generation for each wind speed range. Clustering is performed based on power generation and windward angle to obtain cluster centers. The windward deviation angle for each wind speed range is determined using these cluster centers. The average power for each wind speed range is corrected based on the windward deviation angle. Finally, with the maximum power generation as the objective, the comprehensive windward deviation angle of the wind turbine generators is calculated and used as the yaw control compensation angle, thereby improving yaw accuracy and increasing the power generation of the wind turbine generators.

[0010] The principle for initially screening the read runtime data is as follows:

[0011] A) Data with speeds lower than the grid-connected speed is excluded;

[0012] B. Discard data where the power does not reach the rated power, but the blade angle is greater than the specified value (can be set to 3°);

[0013] C. Remove data where the angle deviation of the three blades exceeds the specified value (which can be set to 1°-2°);

[0014] D. Exclude data with rated wind speed and above.

[0015] The specific method for clustering the filtered operational data according to the standard power curve and classifying it according to wind speed to obtain the average power generation corresponding to each wind speed range is as follows:

[0016] The filtered operational data is arranged according to wind speed-power, and the standard power curve of the wind turbine generator is used as the cluster center. Using the standard power curve as the center, the power curve is shifted left and right according to a set threshold (0.5m / s-1m / s) (left side is the upper limit of the power curve, right side is the lower limit of the power curve) to cluster the operational data. The clustered operational data is then classified according to wind speed range. The read wind speed data is sorted by size and divided into n equal parts, resulting in [v1-Δv,v1+Δv), [v1-Δv,v1+Δv), ..., [v1-Δv,v1+Δv). i -Δv,v i +Δv), ..., [v] n -Δv,v n +Δv), in the wind speed range data, v1 is the cut-in wind speed, v n Rated wind speed, v i Let n be the average wind speed of the i-th wind speed interval, n be the number of intervals, and Δv be half the length of the wind speed interval.

[0017]

[0018] Count the number of data points corresponding to each wind speed interval, i.e., the wind frequency (T1, T2, L, T). n ), where T i For [v i -Δv,v i Wind frequency in the wind speed range of +Δv,)

[0019] Calculate the average power for each wind speed range. Where [v] i -Δv,v i The average power generation in the wind speed range of +Δv,) is

[0020]

[0021] Where P j For [v i -Δv,v i The power generation corresponding to the j-th data point in the wind speed range of +Δv,);

[0022] The specific method for clustering operational data based on power generation and wind angle to obtain cluster centroids is as follows:

[0023] 1) Arrange the wind speed, power generation, and wind angle data for each wind speed range, and randomly select (P, α) as the initial cluster center for the data in each wind speed range; using [v i -Δv,v iTaking the wind speed range of +Δv as an example, its wind speed, power generation, and wind angle data are (v1(j), P1(j), α1(j)), ... (v i (j),P i (j),α i (j))...、 Randomly select the kth data point, i.e. (v i (k),P i (k),α i (k)) represents the initial cluster centers, where 1 ≤ k ≤ T. i ;

[0024] 2) Calculate the distance from each data point in each interval sample to the cluster center. And the point with the smallest distance from other points in the sample is taken as the new cluster center; with [v i -Δv,v i Taking the wind speed range of +Δv as an example, calculate the distance from each data point in the sample of each range to the cluster center. The point with the smallest distance from other points in the sample is taken as the new cluster center of the sample.

[0025] 3) Repeat steps 1) and 2) to iteratively optimize the cluster centers until the cluster centers remain unchanged. The clustering process ends, and the cluster center points (v1, P1, α1), ..., (v...) for each wind speed range are obtained. n ,P n ,α n );

[0026] The windward angle at the cluster center point within each wind speed range is the windward deviation angle for each wind speed range: (α1, α2, L, α n );

[0027] The specific method for correcting the average power of each wind speed range based on the wind deviation angle is as follows:

[0028] Based on the wind deviation angle, the average power for each wind speed range Make corrections;

[0029] Based on the power generation function relationship of the wind turbine generator set, the power generation P can be obtained as follows:

[0030]

[0031] Where ρ is the air density, A is the swept area of ​​the wind turbine, and C p The wind energy utilization coefficient;

[0032] Therefore, based on the wind deviation angle (α1, α2, L, α) n ), for the original average power After correction, the power generation under no wind deviation is calculated, and the corrected power generation is obtained. for:

[0033]

[0034] Then, with the maximum power generation as the target, the comprehensive wind deviation angle α of the wind turbine generator is calculated. ref The specific method is as follows:

[0035] For the wind deviation angle (α1, α2, L, α) in each wind speed range n By overlaying wind frequency data and performing weighted calculations, the comprehensive wind deviation angle α is obtained. ref :

[0036]

[0037] To further improve the accuracy of the comprehensive wind deviation angle, the target value of the comprehensive wind deviation angle is optimized by left and right search using the following formula. The objective function of the optimization is to maximize the power generation S of the wind turbine generator, with a step size of kα. ref It can search left and right n times. Taking three searches on each side as an example, the step size is -3kα. ref -2kα ref -kα ref 、kα ref 2kα ref 3kα ref ;

[0038]

[0039]

[0040] By comparing the wind deviation angle corresponding to the maximum power generation S in n searches, the comprehensive wind deviation angle is used as the yaw control compensation angle, thereby improving the yaw wind accuracy and increasing the power generation of the wind turbine.

[0041] This invention utilizes a method that filters massive amounts of operational data from wind turbine generators to extract key information related to yaw alignment deviation, such as wind speed, power generation, and alignment angle. The data is then divided into intervals based on wind speed and power generation, and clustered for optimal power generation and alignment angle within each wind speed interval to find the best cluster center. Finally, with the goal of maximizing wind turbine generator power generation, the comprehensive alignment deviation angle is calculated and used as the yaw control compensation angle for the wind turbine generator, thereby improving yaw alignment accuracy and increasing generator power generation efficiency. Compared with existing technologies, this invention has the following advantages:

[0042] (1) Using the unit's operating data as the basis for yaw control optimization, it can be continuously updated and the accuracy of the unit's yaw control can be continuously optimized.

[0043] (2) Data clustering can effectively remove abnormal running data and reduce the interference of abnormal data on optimization results;

[0044] (3) Find the wind deviation angle in different wind speed ranges to ensure the accuracy of the wind deviation angle;

[0045] (4) Calculate the comprehensive wind deviation angle to avoid the disadvantage of the lag in compensation for multiple wind speed intervals, which facilitates the application and implementation of yaw control parameter compensation and ensures the effect of yaw control optimization. Attached Figure Description

[0046] Figure 1 Flowchart of the wind turbine yaw control optimization method based on operational data clustering in this invention;

[0047] Figure 2 The embodiment of this invention uses a standard power curve as a basis to filter the effect diagram of the operating data;

[0048] Figure 3 Schematic diagram of wind deviation angle and overall wind deviation angle for each wind speed range in this embodiment of the invention;

[0049] Figure 4 A comparison of power curves before and after yaw control compensation in an embodiment of the present invention. Detailed Implementation

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

[0051] This invention discloses a wind turbine yaw control optimization method based on operational data clustering, comprising the following steps:

[0052] (1) First, read the operating data of the wind turbine generator set for the past year. The operating data includes wind speed v, power generation P, rotational speed m, windward angle α, and blade angle β. Perform preliminary screening on the read operating data. Screening principles: First, remove data with rotational speed lower than the grid-connected speed; second, remove data with power not reaching the rated power but blade angle greater than the specified value (3° in this embodiment); third, remove data with the deviation of the three blade angles exceeding the specified value (1° in this embodiment); fourth, remove data with rated wind speed and above.

[0053] (2) The filtered operating data were arranged according to wind speed-power, and the standard power curve of the wind turbine generator was used as the cluster center. Using the standard power curve as the center, the power curve was shifted left and right according to a set threshold of 1 m / s (the left side is the upper limit of the power curve, and the right side is the lower limit of the power curve) to cluster the data. The data distribution after clustering is as follows: Figure 2 As shown. The clustered operational data is then categorized by wind speed range. The read wind speed data is sorted by size and divided into n equal parts (8 equal parts in this embodiment), resulting in [v1-Δv, v1+Δv), [v1-Δv, v1+Δv), ..., [v1-Δv, v1+Δv). i -Δv,v i +Δv), ..., [v] n -Δv,v n +Δv) refers to the wind speed range [2.5, 3.5), [3.5, 4.5), ..., [10.5, 11.5), where v1 is the cut-in wind speed of 3 m / s. n For a rated wind speed of 11 m / s, v i The cut-in wind speed is Δv, which is half the length of the wind speed range, 0.5 m / s.

[0054]

[0055] Count the number of data points corresponding to each wind speed interval, i.e., the wind frequency (T1, T2, L, T8), where T i For [v i -Δv,v i +Δv,) Wind frequency in the wind speed range.

[0056] Calculate the average power for each wind speed range. Where [v] i -Δv,v i The average power generation in the wind speed range of +Δv,) is

[0057]

[0058] Where P j For [v i -Δv,v i The power generation corresponding to the j-th data point in the wind speed range of +Δv,)

[0059] (3) Cluster the operation data according to power generation and wind angle.

[0060] 1) Arrange the wind speed, power generation, and wind angle data for each interval, and randomly select (P, α) as the initial cluster center for the data in each wind speed interval, with [v i -Δv,v iTaking the wind speed range of +Δv as an example, its wind speed, power generation, and wind angle data are (v1(j),P1(j),α1(j)), ... (v i (j),P i (j),α i (j))...、 Randomly select the kth data point, i.e. (v i (k),P i (k),α i (k)) represents the initial cluster centers, where 1 ≤ k ≤ T. i ;

[0061] 2) Calculate the distance from each data point in each interval sample to the cluster center. And the point with the smallest distance from other points in the sample is taken as the new cluster center; with [v i -Δv,v i Taking the wind speed range of +Δv as an example, calculate the distance from each data point in the sample of each range to the cluster center. The point with the smallest distance from other points in the sample is taken as the new cluster center of the sample.

[0062] 3) Repeat steps 1) and 2) to iteratively optimize the cluster centers until the cluster centers remain unchanged. The clustering process ends, and the cluster center points (v1, P1, α1), ..., (v...) for each wind speed range are obtained. n ,P n ,α n );

[0063] (4) Determine the wind deviation angle for each wind speed range. The cluster centers obtained through (3) are (v1, P1, α1), ..., (v n ,P n ,α n The windward angle at the cluster center point in each wind speed range is the windward deviation angle in each wind speed range, which is (α1, α2, L, α). n );

[0064] (5) According to the wind deviation angle, the average power of each wind speed segment Make corrections.

[0065] Based on the power generation function relationship of the wind turbine generator set, the power generation P can be obtained as follows:

[0066]

[0067] Where ρ is the air density, A is the swept area of ​​the wind turbine, and C p This represents the wind energy utilization coefficient.

[0068] Therefore, based on the wind deviation angle (α1, α2, L, α) n ), for the original average power After correction, the power generation under no wind deviation is calculated, and the corrected power generation is obtained.

[0069]

[0070] (6) Calculate the comprehensive windward deviation angle α of the wind turbine generator set. ref For the wind deviation angles (α1, α2, L, α) in each wind speed range... n By overlaying wind frequency data and performing weighted calculations, the comprehensive wind deviation angle α is obtained. ref .

[0071]

[0072] To further improve the accuracy of the overall wind alignment deviation angle, a left-right search optimization can be performed using the target value of the overall wind alignment deviation angle. The objective function for optimization is to maximize the power generation S of the wind turbine generator, with a step size of kα. ref It can search left and right n times. In this embodiment, it searches left and right three times each, that is, the movement step size is -3kα respectively. ref -2kα ref -kα ref 、kα ref 2kα ref 3kα ref

[0073]

[0074]

[0075] By comparing the wind deviation angle corresponding to the maximum power generation S in n searches, the comprehensive wind deviation angle is used as the yaw control compensation angle, thereby improving the yaw wind accuracy and increasing the power generation of the wind turbine.

[0076] like Figure 3 As shown, through the calculations in the above examples, this embodiment obtains the wind deviation angles (6.2°, 6.2°, 6.1°, 8.6°, 8.9°, 8.8°, 8.4°, 8.2°, 8.5°) for each wind speed range, with a total wind deviation angle of 7.6°.

[0077] like Figure 4 As shown, the power curve of the wind turbine generator is significantly improved after yaw control compensation angle is applied according to the comprehensive wind deviation.

Claims

1. A method for optimizing yaw control of wind turbine generators based on operational data clustering, characterized in that, First, read the operating data of the wind turbine generator set in the previous stage. The operating data includes wind speed v, power generation P, rotational speed m, windward angle α, and blade angle β. Second, perform preliminary screening on the read operating data to remove abnormal data. The filtered operational data are clustered according to the standard power curve and classified according to wind speed to obtain the average power generation corresponding to each wind speed range; the operational data are clustered according to power generation and wind angle to obtain the cluster center point; Using the cluster center point, determine the wind deviation angle for each wind speed range; and correct the average power for each wind speed range according to the wind deviation angle. With the goal of maximizing power generation, the comprehensive wind deviation angle of the wind turbine is calculated and used as the yaw control compensation angle, thereby improving the accuracy of yaw wind alignment and increasing the power generation of the wind turbine. The specific method for clustering the filtered operating data according to the standard power curve and classifying it according to wind speed to obtain the average power generation corresponding to each wind speed range is as follows: The filtered operational data is arranged according to wind speed-power, and the standard power curve of the wind turbine generator is used as the cluster center. Using the standard power curve as the center, the power curve is shifted left and right according to a set threshold to cluster the operational data. The clustered operational data is then classified according to wind speed ranges. The read wind speed data is sorted by size and divided into n equal parts, resulting in the following categories: In the wind speed range data, v1 is the cut-in wind speed, v n Rated wind speed, v i Let n be the average wind speed of the i-th wind speed interval, n be the number of intervals, and Δv be half the length of the wind speed interval. The number of data points corresponding to each wind speed range is counted, i.e., the wind frequency is... T i for Wind frequency within a wind speed range; Calculate the average power for each wind speed range. ,in The average power generation in the wind speed range is : in for The power generation corresponding to the j-th data point in the wind speed range; The specific method for clustering operational data based on power generation and wind angle to obtain cluster centroids is as follows: 1) Arrange the wind speed, power generation, and wind angle data for each wind speed range, and randomly select data for each wind speed range. As the initial cluster centers; 2) Calculate the distance from each data point in each interval sample to the cluster center. And the point with the smallest distance from other points in the sample is taken as the new cluster center; 3) Repeat steps 1) and 2) to iteratively optimize the cluster centers until the cluster centers remain unchanged. The clustering process ends, and the cluster center points for each wind speed range are obtained. ; The windward angle at the cluster center point in each wind speed range is the windward deviation angle in each wind speed range: ; The specific method for correcting the average power of each wind speed range based on the wind deviation angle is as follows: Based on the wind deviation angle, the average power for each wind speed range Make corrections; Based on the power generation function relationship of the wind turbine generator set, the power generation P can be obtained as follows: Where ρ is the air density, A is the swept area of ​​the wind turbine, and C p The wind energy utilization coefficient; Therefore, based on the wind deviation angle For the original average power After correction, the power generation under no wind deviation is calculated, and the corrected power generation is obtained. for: Then, with the maximum power generation as the target, the comprehensive wind deviation angle of the wind turbine generator is calculated. The specific method is as follows: Wind deviation angle in various wind speed ranges By overlaying wind frequency data and performing weighted calculations, the comprehensive wind deviation angle is obtained. : To further improve the accuracy of the overall wind deviation angle, the target value of the overall wind deviation angle is optimized by left and right search using the following formula. The objective function of the optimization is to maximize the power generation S of the wind turbine generator, with a step size of [missing value]. It can search left and right n times: By comparing the wind deviation angle corresponding to the maximum power generation S in n searches, the comprehensive wind deviation angle is used as the yaw control compensation angle, thereby improving the yaw wind accuracy and increasing the power generation of the wind turbine.

2. The wind turbine yaw control optimization method based on operational data clustering according to claim 1, characterized in that, The time interval for reading the previous stage of wind turbine operation data is 0.5-2 years.

3. The wind turbine yaw control optimization method based on operational data clustering according to claim 1, characterized in that, The principle for initially screening the read runtime data is as follows: A) Data with speeds lower than the grid-connected speed is excluded; B. Discard data where the power output does not reach the rated power, but the blade angle is greater than the specified value; C. Remove data where the angle deviation of the three blades exceeds the specified value; D. Exclude data with rated wind speed and above.

Images