A method for calculating the wind speed transfer function of a wind turbine nacelle.

Abstract

This invention discloses a method for calculating the wind speed transfer function of a wind turbine nacelle, comprising: S1, collecting wind speed data from the wind turbine and preprocessing it, filtering out wind speed data that meet the conditions and sorting them; S2, extracting wind speed feature data and dividing all wind speed feature data into several clusters using a clustering algorithm; S3, calculating the fitting function of the data in each cluster and weighting all fitting functions to construct the nacelle wind speed transfer function; S4, constructing a parameter optimization model for the wind turbine nacelle wind speed transfer function, optimizing the parameters to obtain the final nacelle wind speed transfer function. This invention uses a clustering algorithm to identify strongly correlated wind turbine units based on the collected wind speed data, avoiding the limitations of manually judging wind speed data that is less affected by other factors; and further utilizes an optimization algorithm to determine the influence of each main influencing factor on the wind speed transfer function, improving the accuracy of the nacelle wind speed transfer function.

Description

Technical Field

[0001] This invention relates to the field of wind power generation technology, and in particular to a method for calculating the wind speed transfer function of a wind turbine nacelle. Background Technology

[0002] The power curve of a wind turbine is directly related to its power generation. Determining a reasonable nacelle wind speed transfer function is crucial for evaluating the turbine's output performance. Traditional wind turbine power curve testing follows the requirements of IEC 61400-12-1. This involves erecting a wind measurement tower, equipped with meteorological equipment such as a wind tower and laser anemometer, at a distance of 2 to 4 times the rotor diameter from the wind turbine in the upwind direction. The measured free-flow wind speed is used to characterize the wind speed at the rotor center height. The later-updated standard IEC 61400-12-2 defines a method using the nacelle transfer function for power curve testing. This involves establishing a function relating the free-flow wind speed at the rotor center height to the nacelle wind speed across multiple wind direction intervals—the nacelle wind speed transfer function. However, the methods in the two current standards assume or default that the wind measurement data is the inflow wind speed. However, in reality, due to the influence of the wake effect of other wind turbines and the environment, this assumption will have a large error, resulting in low accuracy of the final transfer function. It is necessary to filter the wind speed data and find the data in the wind direction range where the wind measurement tower and wind turbine are less affected by other factors for function fitting, which relies on human experience judgment.

[0003] The "A Method for Zoning Fitting the Transfer Function of a Wind Turbine Nacelle" disclosed in Chinese patent literature, with publication number CN104794347B and publication date of 2017-08-29, includes: S1: Determining the target turbine and collecting operational data of the wind measuring tower and the target turbine at the same time; S2: Determining the hub height H of the target turbine and calculating the average wind speed data of the wind measuring tower at that height H; S3: Using IEC standards and based on the average wind speed data calculated in step S2, eliminating abnormal operating data from step S1; S4: Calculating the transfer function of the target turbine nacelle and correcting the transfer function; Step S4 includes: S4-1: Fitting the function to fit ... This technology obtains the free-flow wind speed at the center of the wind turbine rotor by correcting the nacelle wind speed, solving the problem of cumbersome calculations and poor practicality caused by too many partitions in the transfer function of the IEC61400-12-2 standard method. However, it is still an improvement on the standard method, which inevitably assumes or defaults to the inflow wind speed as the wind measurement data. Therefore, in actual situations, due to the influence of wake effects from other wind turbines and environmental factors, this assumption will have a large error, resulting in low accuracy of the final transfer function. Summary of the Invention

[0004] This invention aims to overcome the problem in existing technologies where the wind measurement data is assumed to be the inflow data before determining the transfer function, resulting in low accuracy of the actual transfer function under the influence of wake effects from other wind turbines and environmental factors. It provides a method for calculating the wind speed transfer function of a wind turbine nacelle. This method uses a clustering algorithm to identify strongly correlated turbines based on the collected wind speed data, avoiding the limitations of manually judging wind speed data that is less affected by other factors. Furthermore, it utilizes an optimization algorithm to determine the influence of various main influencing factors on the wind speed transfer function, thereby improving the accuracy of the nacelle wind speed transfer function.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for calculating the wind speed transfer function of a wind turbine nacelle includes:

[0007] S1. Collect wind speed data from the fan and preprocess it, then filter out the wind speed data that meets the criteria and sort them.

[0008] S2. Extract wind speed feature data and divide all wind speed feature data into several clusters using a clustering algorithm;

[0009] S3. Calculate the fitting function for the data in each cluster, and weight all the fitting functions to construct the cabin wind speed transfer function.

[0010] S4. Construct a parameter optimization model for the nacelle wind speed transfer function of the wind turbine, and optimize the parameters to obtain the final nacelle wind speed transfer function.

[0011] The wind speed data collected in this invention includes wind speed data measured by anemometer towers and wind speed data measured in the nacelle. After preprocessing the two types of data, corresponding wind speed feature data are extracted in pairs, and then the nacelle wind speed transfer function is calculated. Compared with the existing standard methods for determining the transfer function, this method does not require data judgment or assumption that the anemometer data is the inflow data. It only requires filtering the acquired data for wind speed data that meets the conditions, extracting features, and then using a clustering algorithm to determine the units that are strongly correlated with the anemometer tower. This avoids the limitations of manually judging the data of the anemometer tower and the wind turbine in wind direction ranges that are less affected by other factors. At the same time, the transfer function parameter optimization model is used to optimize and determine the influence of each influencing factor on the overall transfer function, and the accuracy of the transfer function is improved through continuous iterative optimization.

[0012] Preferably, the process of extracting wind speed feature data is as follows:

[0013] x i =V p,i -V nacelle,i

[0014] Where V p,i V represents the i-th wind speed data measured by the anemometer tower. nacelle,i This represents the i-th wind speed data measured in the cabin.

[0015] In this invention, the wind speed data measured by the meteorological tower and the wind speed data measured in the nacelle are aligned by timestamps. Therefore, there will be a set of wind speed data from the meteorological tower and wind speed data measured in the nacelle at the same time point. The wind speed feature data at that time point is extracted using this pair of data to obtain the feature set according to the time series for subsequent transfer function calculation. This eliminates the need for data judgment and avoids the errors caused by wind speed zoning.

[0016] As a preferred approach, for a set containing N wind speed feature data, the clustering process using the K-means algorithm includes:

[0017] S21. Let m = 1, and randomly select K cluster centers from all wind speed characteristic data;

[0018] S22. Calculate each wind speed characteristic data x respectively. i Calculate the distance to the cluster center and reclassify based on the set distance threshold;

[0019] S23. Calculate the cluster center of each cluster again and let m = m + 1;

[0020] S24. Construct an objective function related to wind speed characteristic data and cluster centers, and determine the difference between the objective function of the m-th iteration and the objective function of the (m+1)-th iteration;

[0021] S25. If the absolute value of the difference is less than the convergence threshold, the clustering is completed; otherwise, return to S22.

[0022] In this invention, clustering is a process that, based on the principle of similarity, divides data objects with high similarity into the same cluster and data objects with high dissimilarity into different clusters, thereby making data points in the same cluster more similar to data points in other clusters. In this invention, cluster analysis is used to classify all wind speed characteristic data, grouping highly correlated data together to determine strongly correlated units, and dividing all units into several clusters of strongly correlated units. Each cluster corresponds to a different fitting function for the unit, making the subsequently obtained cabin wind speed transfer function more accurate.

[0023] Preferably, the objective function for constructing the wind speed characteristic data and cluster centers is:

[0024]

[0025] Where J(m) represents the value of the objective function in the m-th iteration. Represents the i-th data in the j-th cluster, n j This represents the number of data points in the j-th cluster.

[0026] In this invention, the change in the objective function after iterative clustering calculation is less than the convergence threshold or remains unchanged as the indicator that clustering is complete. 2 The square of the norm is used to represent the square of the Euclidean norm, which is a function of distance. It is used to represent the sum of squares of the error distances after clustering iterations.

[0027] Preferably, the nacelle wind speed transfer function includes:

[0028] y = α1f1(x) + a2f2(x) + ... + α j f j (x)+…+α K f K (x)

[0029] a1+α2++αj++α K =1

[0030] Where y is the fitted free inflow velocity of the fan; α j f is the weighting coefficient; j (x) represents the fitting function for the wind speed data in the j-th cluster.

[0031] In this invention, each cluster corresponds to a class of strongly correlated wind turbines. Therefore, the wind speed data of each cluster has its own more accurate fitting function. By weighted calculation, the fitting functions of different clusters are integrated into the nacelle wind speed transfer function. Taking into account the impact of most wind turbines on the overall performance, the final transfer function is more accurate and conforms to the actual situation.

[0032] Preferably, S4 includes the following steps:

[0033] S41, using the number of clusters K and the weight coefficient α j To optimize parameters, particle position and particle velocity are initialized based on normalized wind speed characteristic data;

[0034] S42. Based on the generated optimized parameter values, cluster the wind speed feature data and construct the cabin wind speed transfer function, and calculate the fitness of all particles.

[0035] S43. Compare the fitness of each particle to obtain the globally optimal particle and the locally optimal particle;

[0036] S44. Update particle velocity and particle position, and recalculate the fitness of all particles;

[0037] S45. If the fitness of the new particle is better than the fitness of the original particle, then update the historical best solution of the particle; if the fitness of the new particle is better than the global best particle, then update the global best particle.

[0038] S46. Repeat S42 to S45 until the maximum number of iterations is reached to complete parameter optimization.

[0039] This invention utilizes the particle swarm optimization algorithm, a heuristic algorithm that simulates the foraging behavior of bird flocks, to find the optimal solution by updating particle velocity and particle position, which can make the final cabin wind speed transfer function more accurate. In this invention, the particle position represents the optimization parameters (number of clusters K and weight coefficients), i.e., a solution, and the particle velocity represents the step size of the change of this solution. The optimization of the parameters in the cabin wind speed transfer function is completed by iterative calculation.

[0040] As a preferred method, the result after normalization is obtained by dividing the wind speed characteristic data by the minimum value among all wind speed characteristic data.

[0041] The fitness function is:

[0042]

[0043] Where N is the number of wind speed characteristic data, V p,i y represents the i-th wind speed data measured by the wind tower, and y represents the fitted free inflow wind speed value of the wind turbine.

[0044] In this invention, normalization is performed on all data uniformly, but the original wind speed feature data is still used for clustering. Furthermore, in the parameter optimization algorithm, if the wind speed feature data values ​​are too small, it will affect the normalization effect. Therefore, the wind speed feature data is divided by the minimum value among all wind speed feature data as the result of normalization, ensuring that all normalized results are greater than or equal to 1 to facilitate subsequent search for the optimal solution. The fitness function can be calculated by the sum of squared errors between the wind speed data measured by the anemometer and the wind speed data fitted by the transfer function; a smaller fitness value indicates better fitness.

[0045] Preferably, the updated particle velocity and particle position can be expressed as:

[0046]

[0047]

[0048] in Let be the velocity of the i-th particle at time t; This represents the current particle's historical best position. The historical optimal position of all particles; r1 and r2 are random numbers between [0,1], wi l 1,i l 2,i These are inertia weight, local learning factor, and global learning factor, respectively.

[0049] In this invention, the inertia weight w maintains the particle's inertia, giving it a tendency to expand the search space and enabling it to explore new regions; local learning factor and global learning factor l 1,i l 2,i This represents the weight of the statistical acceleration term that pushes each particle toward the local and global optimum; low values ​​allow particles to linger outside the target region before being pulled back, while high values ​​cause particles to suddenly rush toward or cross the target region.

[0050] The present invention has the following beneficial effects: It uses a clustering algorithm to identify strongly correlated wind speed units based on the collected wind speed data, avoiding the limitations of manual judgment of wind speed data that is less affected by other factors; furthermore, it utilizes an optimization algorithm to determine the influence of each main influencing factor on the wind speed transfer function, improving the accuracy of the nacelle wind speed transfer function; it eliminates the need for manual judgment of wind speed data, avoiding errors caused by wind speed zoning in existing technologies; and considering the combined influence of interfering and non-interfering wind turbines further improves the correlation between the wind speed generated by the nacelle wind speed transfer function and the actual wind speed. Attached Figure Description

[0051] Figure 1 This is a flowchart of the calculation method for the cabin wind speed transfer function in this invention. Detailed Implementation

[0052] The present invention will now be further described with reference to the accompanying drawings and specific embodiments.

[0053] Traditional wind turbine power curve testing follows the requirements of IEC 61400-12-1. This involves erecting a wind measurement tower, equipped with meteorological equipment such as a wind tower and laser anemometer, at a distance of 2 to 4 times the rotor diameter from the wind turbine in the upwind direction. The measured free-flow wind speed is used to characterize the wind speed at the rotor center height. The updated standard IEC 61400-12-2 defines a method for power curve testing using the nacelle transfer function. This involves establishing a function relating the free-flow wind speed at the rotor center height to the nacelle wind speed across multiple wind direction intervals—the nacelle wind speed transfer function.

[0054] In the existing process of constructing the transfer function according to the standard IEC 61400-12-2, the wind measurement data is filtered by sector. Simultaneous wind measurement data in front of the nacelle and wind speed data are filtered by sector, removing sectors affected by obstacles and invalid data. Then, the data is further processed by the Bin method, and the wind speed data at the filtered points is used as the data for fitting the transfer function. The formula for calculating the transfer function is:

[0055]

[0056] Where V free,i+1 and V free,i V represents the interval average of the wind speed measured in front of the nacelle in wind speed intervals i and i+1. nacelle,i+1 and V nacelle,i V represents the interval average of the nacelle wind speed in wind speed intervals i and i+1. nacelle To measure the nacelle wind speed, V free The free-flow wind speed is estimated using measured nacelle wind speed and wind speed measured in front of the nacelle, and corrected for airflow distortion caused by terrain. For each zone, it can be divided into 0.5 m / s zones, or into low-speed and high-speed zones to reduce the number of zones. Then, the data for each zone is fitted using a polynomial.

[0057] However, these existing algorithms have the following problems: The IEC 61400-12-1 method is impractical because it's impossible to install a wind measurement device in front of every wind turbine, resulting in long testing cycles, high costs, and complex processes; the IEC 61400-12-2 method is affected by blade disturbances and other factors, making it difficult to accurately determine the complex relationship between the measured wind speed and the incoming wind speed; both methods in the current standard assume or default that the measured wind speed is the incoming wind speed, but this assumption can lead to significant errors due to the wake effects of other wind turbines and environmental influences; the wind data needs to be filtered in the early stages to find data in wind direction ranges less affected by other factors for function fitting, relying on human experience; the transfer function in the IEC 61400-12-2 standard uses multi-segment linear function interpolation, resulting in large errors; and it only fits the nacelle wind speed of some units strongly correlated with the wind speed of the wind measurement tower, failing to quantitatively analyze the impact of other non-strongly correlated units on overall performance. To address these shortcomings, this invention proposes a new method for calculating the nacelle wind speed transfer function.

[0058] like Figure 1 As shown, a method for calculating the wind speed transfer function of a wind turbine nacelle includes:

[0059] S1. Collect wind speed data from the fan and preprocess it, then filter out the wind speed data that meets the criteria and sort them.

[0060] Collect wind speed data measured by the anemometer tower and wind speed data measured in the nacelle, and align the timestamps; then filter the wind speed data according to the characteristics of the unit, and select wind speed data between the cut-in wind speed and the cut-out wind speed. Under normal circumstances, select a cut-in wind speed of 3 m / s or 4 m / s and a cut-out wind speed of 25 m / s, and select wind speed data within this range.

[0061] S2. Extract wind speed feature data and divide all wind speed feature data into several clusters using a clustering algorithm.

[0062] The process of extracting wind speed feature data is as follows:

[0063] x i =V p，i -V nacelle，i

[0064] Where V p,i V represents the i-th wind speed data measured by the anemometer tower. nacelle,i This represents the i-th wind speed data measured in the cabin.

[0065] For a set containing N wind speed feature data, the process of clustering using the K-means algorithm includes:

[0066] S21. Let m = 1, and randomly select K cluster centers from all wind speed characteristic data;

[0067] S22. Calculate each wind speed characteristic data x respectively. i The distance to the cluster center is used to determine the distance, and the data is reclassified according to the set distance threshold. Wind speed feature data whose distance to a certain cluster center is less than the distance threshold are all assigned to the corresponding cluster.

[0068] S23. Calculate the cluster center of each cluster again and let m = m + 1; the calculation method for the cluster center of each cluster is the conventional calculation method, such as calculating the centroid of all wind speed characteristic data in the cluster as the new cluster center, etc.

[0069] S24. Construct an objective function related to wind speed characteristic data and cluster centers, and determine the difference between the objective function of the m-th iteration and the objective function of the (m+1)-th iteration;

[0070] The objective function related to wind speed characteristic data and cluster centers is constructed as follows:

[0071]

[0072] Where J(m) represents the value of the objective function in the m-th iteration. Represents the i-th data in the j-th cluster, n j Represents the number of data points in the j-th cluster, O j(m) represents the cluster center of the j-th cluster in the m-th iteration;

[0073] S25. If the absolute value of the difference is less than the convergence threshold θ, then clustering is complete; otherwise, return to S22; determine whether the following conditions are met:

[0074] |J(m+1)-J(m)|<θ

[0075] If the conditions are met, the clustering is complete. Select the number of clusters K after the (m+1)th iteration and the wind speed characteristic data corresponding to each cluster. If the conditions are not met, return to S22 and perform iterative calculations until the conditions are met.

[0076] S3. Calculate the fitting function for the data in each cluster, and weight all the fitting functions to construct the cabin wind speed transfer function.

[0077] The data for each cluster is specifically fitted using a quadratic function: y j =f j (x)=ax 2 +bx+c, and then the cabin wind speed transfer function constructed based on the fitted results includes:

[0078] y = α1f1(x) + α2f2(x) + ... + α j f j (x)+…+α K f K (x)

[0079] α1+α2+…+α j +…+α K =1

[0080] Where y is the fitted free inflow velocity of the fan; α j f is the weighting coefficient; j (x) represents the fitting function for the wind speed data in the j-th cluster.

[0081] S4. Construct a parameter optimization model for the nacelle wind speed transfer function of the wind turbine, and optimize the parameters to obtain the final nacelle wind speed transfer function. S4 includes the following steps:

[0082] S41, using the number of clusters K and the weight coefficient α j To optimize parameters, particle position and particle velocity are initialized based on normalized wind speed characteristic data;

[0083] The result after normalization is obtained by dividing the wind speed feature data by the minimum value among all wind speed feature data.

[0084]

[0085] The population size of particles is set, and the range of values ​​for particle position and particle velocity are defined. Particle position is the optimization parameter; in this invention, the optimization parameters are the number of clusters K and the weight coefficient α. j Based on the normalized data, initialize the particle positions and randomly generate particle velocities.

[0086] S42. Based on the generated optimized parameter values, return to S2 and S3 to perform clustering of wind speed feature data and construction of the nacelle wind speed transfer function, and calculate the fitness of all particles in population i.

[0087] The fitness function is:

[0088]

[0089] Where N is the number of wind speed characteristic data, V p,i y represents the i-th wind speed data measured by the wind tower, and y is the fitted free inflow wind speed value of the wind turbine.

[0090] S43. Compare the fitness of each particle to obtain the globally optimal particle Q. g and the local optimal particle Q i ;

[0091] S44. Update particle velocity and particle position, and recalculate the fitness of all particles;

[0092] Updating particle velocity and particle position can be expressed as:

[0093]

[0094]

[0095] in Let be the velocity of the i-th particle at time t; This represents the current particle's historical best position. The historical optimal position of all particles; r1 and r2 are random numbers between [0,1], w i l 1,i l 2,i These are inertia weight, local learning factor, and global learning factor, respectively.

[0096] S45. If the fitness of the new particle is better than the fitness of the original particle, then update the historical best solution of the particle; if the fitness of the new particle is better than the global best particle, then update the global best particle.

[0097] S46. Repeat S42 to S45 until the maximum number of iterations is reached to complete parameter optimization.

[0098] Based on the optimized parameters, the number of clusters K and the weight coefficient α jThe wind speed characteristic data were re-clustered and the cabin wind speed transfer function was reconstructed to obtain the final cabin wind speed transfer function.

[0099] The wind speed data collected in this invention includes wind speed data measured by anemometer towers and wind speed data measured in the nacelle. After preprocessing the two types of data, corresponding wind speed feature data are extracted in pairs, and then the nacelle wind speed transfer function is calculated. Compared with the existing standard methods for determining the transfer function, this method does not require data judgment or assumption that the anemometer data is the inflow data. It only requires filtering the acquired data for wind speed data that meets the conditions, extracting features, and then using a clustering algorithm to determine the units that are strongly correlated with the anemometer tower. This avoids the limitations of manually judging the data of the anemometer tower and the wind turbine in wind direction ranges that are less affected by other factors. At the same time, the transfer function parameter optimization model is used to optimize and determine the influence of each influencing factor on the overall transfer function, and the accuracy of the transfer function is improved through continuous iterative optimization.

[0100] In this invention, the wind speed data measured by the meteorological tower and the wind speed data measured in the nacelle are aligned by timestamps. Therefore, there will be a set of wind speed data from the meteorological tower and wind speed data measured in the nacelle at the same time point. The wind speed feature data at that time point is extracted using this pair of data to obtain the feature set according to the time series for subsequent transfer function calculation. This eliminates the need for data judgment and avoids the errors caused by wind speed zoning.

[0101] In this invention, clustering is a process that, based on the principle of similarity, divides data objects with high similarity into the same cluster and data objects with high dissimilarity into different clusters, thereby making data points in the same cluster more similar to data points in other clusters. In this invention, cluster analysis is used to classify all wind speed characteristic data, grouping highly correlated data together to determine strongly correlated units, and dividing all units into several clusters of strongly correlated units. Each cluster corresponds to a different fitting function for the unit, making the subsequently obtained cabin wind speed transfer function more accurate.

[0102] In this invention, the change in the objective function after iterative clustering calculation is less than the convergence threshold or remains unchanged as the indicator that clustering is complete. 2 The square of the norm is used to represent the square of the Euclidean norm, which is a function of distance. It is used to represent the sum of squares of the error distances after clustering iterations.

[0103] In this invention, each cluster corresponds to a class of strongly correlated wind turbines. Therefore, the wind speed data of each cluster has its own more accurate fitting function. By weighted calculation, the fitting functions of different clusters are integrated into the nacelle wind speed transfer function. Taking into account the impact of most wind turbines on the overall performance, the final transfer function is more accurate and conforms to the actual situation.

[0104] This invention utilizes the particle swarm optimization algorithm, a heuristic algorithm that simulates the foraging behavior of bird flocks, to find the optimal solution by updating particle velocity and particle position, which can make the final cabin wind speed transfer function more accurate. In this invention, the particle position represents the optimization parameters (number of clusters K and weight coefficients), i.e., a solution, and the particle velocity represents the step size of the change of this solution. The optimization of the parameters in the cabin wind speed transfer function is completed by iterative calculation.

[0105] In this invention, normalization is performed on all data uniformly, but the original wind speed feature data is still used for clustering. Furthermore, in the parameter optimization algorithm, if the wind speed feature data values ​​are too small, it will affect the normalization effect. Therefore, the wind speed feature data is divided by the minimum value among all wind speed feature data as the result of normalization, ensuring that all normalized results are greater than or equal to 1 to facilitate subsequent search for the optimal solution. The fitness function can be calculated by the sum of squared errors between the wind speed data measured by the anemometer and the wind speed data fitted by the transfer function; a smaller fitness value indicates better fitness.

[0106] In this invention, the inertia weight w maintains the particle's inertia, giving it a tendency to expand the search space and enabling it to explore new regions; local learning factor and global learning factor l 1,i l 2,i This represents the weight of the statistical acceleration term that pushes each particle toward the local and global optimum; low values ​​allow particles to linger outside the target region before being pulled back, while high values ​​cause particles to suddenly rush toward or cross the target region.

[0107] The above embodiments are further elaborations and descriptions of the present invention to facilitate understanding, and are not intended to limit the present invention in any way. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calculating the wind speed transfer function of a wind turbine nacelle, characterized in that, include: S1. Collect wind speed data from the fan and preprocess it, then filter out the wind speed data that meets the criteria and sort them. S2. Extract wind speed feature data and divide all wind speed feature data into several clusters using a clustering algorithm; S3. Calculate the fitting function for the data in each cluster, and weight all the fitting functions to construct the cabin wind speed transfer function. S4. Construct a parameter optimization model for the nacelle wind speed transfer function of the wind turbine, and optimize the parameters to obtain the final nacelle wind speed transfer function.

2. The method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 1, characterized in that, The process of extracting wind speed feature data is as follows: x i =V p,i -V nacelle,i Where V p,i V represents the i-th wind speed data measured by the anemometer tower. nacelle,i This represents the i-th wind speed data measured in the cabin.

3. A method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 1 or 2, characterized in that, For a set containing N wind speed feature data, the process of clustering using the K-means algorithm includes: S21. Let m = 1, and randomly select K cluster centers from all wind speed characteristic data; S22. Calculate each wind speed characteristic data x respectively. i Calculate the distance to the cluster center and reclassify based on the set distance threshold; S23. Calculate the cluster center of each cluster again and let m = m + 1; S24. Construct an objective function related to wind speed characteristic data and cluster centers, and determine the difference between the objective function of the m-th iteration and the objective function of the (m+1)-th iteration; S25. If the absolute value of the difference is less than the convergence threshold, the clustering is completed; otherwise, return to S22.

4. The method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 3, characterized in that, The objective function for constructing the data related to wind speed characteristics and cluster centers is as follows: Where J(m) represents the value of the objective function in the m-th iteration. Represents the i-th data in the j-th cluster, n j Represents the number of data points in the j-th cluster, O j (m) represents the cluster center of the j-th cluster in the m-th iteration.

5. A method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 1, 2, or 4, characterized in that, The constructed nacelle wind speed transfer function includes: y=α1f1(x)+α2f2(x)+…+α j f j (x)+…+a K f K (x) α1+α2+…+α j +…+α K ＝1 Where y is the fitted free inflow velocity of the fan; α j f is the weighting coefficient; j (x) represents the fitting function for the wind speed data in the j-th cluster.

6. A method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 1, 2, or 4, characterized in that, S4 includes the following steps: S41, using the number of clusters K and the weight coefficient α j To optimize parameters, particle position and particle velocity are initialized based on normalized wind speed characteristic data; S42. Based on the generated optimized parameter values, cluster the wind speed feature data and construct the cabin wind speed transfer function, and calculate the fitness of all particles. S43. Compare the fitness of each particle to obtain the globally optimal particle and the locally optimal particle; S44. Update particle velocity and particle position, and recalculate the fitness of all particles; S45. If the fitness of the new particle is better than the fitness of the original particle, then update the historical best solution of the particle; if the fitness of the new particle is better than the global best particle, then update the global best particle. S46. Repeat S42 to S45 until the maximum number of iterations is reached to complete parameter optimization.

7. The method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 6, characterized in that, The result after normalization is obtained by dividing the wind speed characteristic data by the minimum value among all wind speed characteristic data. The fitness function is: Where N is the number of wind speed characteristic data, V p,i is the i-th wind speed data measured by the wind tower, and u is the fitted free inflow wind speed value of the wind turbine.

8. The method for calculating the wind speed transfer function of a wind turbine nacelle according to claim 6, characterized in that, The updated particle velocity and particle position can be expressed as: in Let be the velocity of the i-th particle at time t; This represents the current particle's historical best position. The historical optimal position of all particles; r1 and r2 are random numbers between [0,1], w i l 1,i l 2,i These are inertia weight, local learning factor, and global learning factor, respectively. Let be the position of the i-th particle at time t.

Images