Wind power plant inter-field yaw control method taking field group as unit

Abstract

The invention relates to the technical field of wind power plant control, in particular to a wind power plant inter-field yaw control method with a field group as a unit. The method comprises the following steps: collecting real-time operation data of a target wind power plant; performing time synchronization and space normalization processing on the whole-field wind vector time sequence data, and fusing the whole-field wind vector time sequence data with fan operation data; dividing fan groups by adopting a dynamic clustering method; calculating an inter-group dynamic wake flow influence matrix by adopting a Jensen wake flow model subjected to terrain correction; an optimization model with the yaw angle of the draught fan in each group as a decision variable is constructed, solving is conducted through a particle swarm optimization algorithm, and a field group cooperative yaw instruction set is generated; and adjusting parameters of the Jensen wake flow model through power feedback. Fan groups are divided through dynamic clustering, an inter-group dynamic wake flow influence matrix is constructed, cooperative yaw optimization with the maximized total power of the whole field is carried out, and the problem that the wake flow coupling effect and terrain interference are neglected in traditional independent yaw control of a single fan is solved.

Description

Technical Field

[0001] This invention relates to the field of wind farm control technology, and more specifically, to a method for inter-farm yaw control of wind farms on a farm-group basis. Background Technology

[0002] Yaw control in wind farms is one of the core technologies for improving wind energy utilization efficiency. It maximizes wind energy capture by adjusting the yaw angle of the wind turbines to align the rotors with the wind direction. Existing wind farm yaw control methods often make independent decisions for each individual turbine, failing to fully consider the mutual influence of wakes between turbines and the interference of terrain on wind farm distribution. This results in wind energy loss at downstream turbines due to wake obstruction, and the overall power output of the entire farm does not reach its optimal level. Furthermore, traditional control methods lack real-time adaptation to dynamic wind conditions and closed-loop correction of model parameters, making it difficult to maintain stable control accuracy under complex terrain and wind conditions. Therefore, this invention provides a wind farm inter-farm yaw control method based on a wind farm cluster. Summary of the Invention

[0003] The purpose of this invention is to provide a wind farm yaw control method based on a group of wind farms, in order to solve the problem mentioned in the background art that the existing wind farm yaw control methods mostly make independent decisions on a single wind turbine basis, without fully considering the mutual influence of wakes between wind turbines and the interference of terrain on wind farm distribution.

[0004] To achieve the above objectives, the present invention aims to provide a method for inter-farm yaw control of wind farms at the farm cluster level, comprising the following steps:

[0005] S1. Collect real-time operational data of the target wind farm, including time-series wind vector data, wind turbine operation data, and wind condition forecast data.

[0006] S2. Perform time synchronization and spatial normalization processing on the time series data of the entire wind vector to form a wind field feature matrix, and then merge it with the wind turbine operation data to construct a fused dataset;

[0007] S3. Based on the fused dataset, a dynamic clustering method is used to divide the turbines in the wind farm into several mutually influential turbine groups;

[0008] S4. For the divided wind turbine groups, the Jensen wake model with terrain correction is used to calculate the dynamic wake influence matrix between groups;

[0009] S5. With the goal of maximizing the total predicted power of the entire field, an optimization model is constructed with the yaw angle of the wind turbines in each group as the decision variable. The dynamic wake influence matrix between groups and wind condition forecast data are combined and solved by the particle swarm optimization algorithm to generate a set of coordinated yaw instructions for the field groups in the future.

[0010] S6. Execute the field group coordinated yaw command set, collect the actual power feedback in the SCADA system in real time, and adjust the parameters of the Jensen wake model through the power feedback.

[0011] As a further improvement to this technical solution, the specific steps involved in collecting real-time operating data of the target wind farm in step S1 are as follows:

[0012] By using a virtual lidar system installed in the wind farm, real-time simulation and acquisition of wind vector time-series data for the entire field are achieved, including wind speed, wind direction, and turbulence intensity.

[0013] Real-time operating data of each wind turbine is obtained from the SCADA system of the wind farm, including turbine power, speed, pitch angle, yaw angle, nacelle wind direction, operating mode and fault status;

[0014] Access external numerical weather forecast services to obtain wind condition forecast data for future periods, including forecast wind speed, wind direction, and atmospheric stability parameters;

[0015] Outliers and missing data were removed from the time series data of wind vectors, wind turbine operation data and wind condition forecast data of the entire field, and interpolation methods were used to fill in the short-term missing data appropriately;

[0016] The processed full-field wind vector data, wind turbine operation data, and wind condition forecast data are organized according to time series to form structured real-time operation data.

[0017] As a further improvement to this technical solution, the specific steps involved in S2 for time synchronization and spatial normalization of the full-field wind vector time series data are as follows:

[0018] Based on the full-field wind vector time series data, resampling is performed according to a unified time base to generate time-aligned wind vector data;

[0019] The wind farm is divided into a regular grid. Based on the time-aligned wind vector data, the discrete wind measurement point data is interpolated to each node of the regular grid using the inverse distance weighted interpolation method to obtain a continuous wind field distribution covering the entire field.

[0020] Based on continuous wind field distribution, a terrain height correction coefficient is introduced. The wind speed at the grid nodes is corrected for terrain features to obtain the corrected wind speed.

[0021] Based on the corrected wind speed and continuous wind field distribution, a wind field feature matrix is ​​constructed using a feature matrix organization method. The wind field feature matrix includes the wind speed, wind direction, and turbulence intensity for each grid point.

[0022] As a further improvement to this technical solution, the specific steps involved in S2, which involve fusing the wind turbine operation data to construct a fused dataset, are as follows:

[0023] For each wind turbine Based on its location coordinates, the wind speed, wind direction, and turbulence intensity of each wind turbine's grid point are extracted from the wind field feature matrix to obtain the wind field characteristics at the turbine level.

[0024] Each wind turbine The wind turbine operation data is aligned and merged with the wind field characteristics of the corresponding wind turbine locations to form wind turbine-level time series samples;

[0025] Organize the turbine-level time series samples of all wind turbines at all times in chronological order to construct a fused dataset.

[0026] As a further improvement to this technical solution, in step S3, the specific steps involved in dividing the turbines in the wind farm into several mutually influential turbine groups using a dynamic clustering method are as follows:

[0027] For each wind turbine Extracting its time window from the fused dataset The average values ​​of wind speed, wind direction, turbulence intensity, power, rotational speed, pitch angle, and yaw angle, as well as the standard deviations of wind speed and wind direction, are used to construct a wind turbine feature vector by combining these with the turbine's location coordinates. ;

[0028] For wind turbine feature vectors Mini-max normalization is performed on each dimension to obtain the normalized feature vector. ;

[0029] Based on the normalized feature vector A density-based dynamic clustering method is used to calculate the Euclidean distance between wind turbines to obtain the similarity distance;

[0030] Similarity distance does not exceed the neighborhood radius And the number of wind turbines in the neighborhood is not less than the minimum number of wind turbines in the neighborhood. A group of wind turbines is grouped into the same group, thus resulting in several wind turbine groups. .

[0031] As a further improvement to this technical solution, the specific steps involved in calculating the dynamic wake influence matrix between groups using the terrain-corrected Jensen wake model in step S4 are as follows:

[0032] Based on the wind field feature matrix, extract each wind turbine Wind speed at the location and wind direction ;

[0033] Based on wind direction For any two wind turbines Japanese wind machine Calculate its relative position vector Calculate the fan The angle between the wind direction and the direction of the relative position vector, if the angle does not exceed the direction threshold. Then determine the fan For the fan There is a wake effect; otherwise, there is no wake effect.

[0034] For wind turbine pairs with wake effects, a terrain-corrected Jensen wake model is used, combined with wind speed. Calculate the wind turbine's thrust coefficient, the distance between the two turbines, the turbine rotor diameter, and the terrain height correction factor at the downstream turbine location. Under the wind turbine Actual wind speed after the wake effect;

[0035] wind turbine The actual wind speed is compared with the original wind speed, and the wind speed loss ratio is calculated as the wake influence coefficient of the upstream wind turbine on the downstream wind turbine. ;

[0036] For the whole audience Typhoon generator, constructed based on all wake influence coefficients. Wind turbine-level dynamic wake influence matrix ;

[0037] Based on the dynamic wake influence matrix of the wind turbine level The wake influence coefficients within and between each wind turbine group are averaged and aggregated to obtain the dynamic wake influence matrix between groups. .

[0038] As a further improvement to this technical solution, in step S5, the specific steps involved in constructing an optimization model with the yaw angle of the wind turbines in each group as the decision variable, with the goal of maximizing the total predicted power of the entire field, are as follows:

[0039] Obtain key basic data for a single wind turbine, including air density, turbine rotor swept area, forecast wind speed, and the correspondence between power coefficient and group common yaw angle;

[0040] Based on key basic data, the predicted power of wind energy conversion of a single wind turbine is calculated, and then the predicted power of all wind turbines in each group is summed to obtain the total predicted power of the group.

[0041] The optimization objective is to maximize the total predicted power of all wind turbines in the future forecast period, and the decision variable is the set of common yaw angles of each group at each forecast time. ;

[0042] Setting a yaw angle range constraint is used to limit the yaw angle change of each group from not exceeding the allowable range, and setting a yaw angle change rate constraint is used to limit the rate of yaw angle change between adjacent moments from not exceeding the limit value.

[0043] The optimization model is obtained by combining the optimization objective, decision variables, yaw angle range constraints, and yaw angle change rate constraints.

[0044] As a further improvement to this technical solution, in step S5, the specific steps involved are solved by combining the dynamic wake influence matrix between groups with wind forecast data and using a particle swarm optimization algorithm:

[0045] S5.1. Based on the decision variables, the initial size is... A swarm of particles, each particle Represents a set of possible common yaw angles, where ;

[0046] S5.2, For each particle Based on the optimized model, combined with wind forecast data and the dynamic wake influence matrix between groups, the total predicted power for the entire field is calculated. This is the fitness value of the particle;

[0047] S5.3 Record the historical best position of each particle. and the global optimal position of the particle swarm Based on the velocity and position update formulas of the particle swarm optimization algorithm, the state of all particles is iteratively updated to maximize the fitness value of the particles.

[0048] S5.4 Repeat steps S5.2 to S5.3 until the maximum number of iterations is reached. Output the yaw angle decision set corresponding to the globally optimal position. That is, the future The set of coordinated yaw commands for the field group during the time period.

[0049] As a further improvement to this technical solution, in step S6, the specific steps involved in executing the field group coordinated yaw command set and real-time acquisition of actual power feedback from the SCADA system are as follows:

[0050] The wind turbine group coordinated yaw command set is sent to the controller of each wind turbine group, and each group of wind turbines executes a unified common yaw angle command;

[0051] The actual power data of each wind turbine is collected in real time from the wind farm's SCADA system to form a power feedback sequence.

[0052] As a further improvement to this technical solution, the specific steps involved in adjusting the parameters of the Jensen wake model through power feedback in step S6 are as follows:

[0053] Selecting the wake attenuation coefficient Terrain height correction factor As parameters to be adjusted in the Jensen wake model;

[0054] Based on the power feedback sequence, at the end of each control cycle, the average actual power of all wind turbines during the control cycle is calculated. And compared with the average power predicted by the optimized model for the same period. Compare and calculate the power deviation ratio. ;

[0055] If the power deviation ratio is greater than the positive deviation threshold, the wake attenuation coefficient is reduced proportionally. If the power deviation ratio is less than the negative deviation threshold, then the wake attenuation coefficient is increased proportionally. Otherwise, maintain the wake attenuation coefficient. constant;

[0056] If the wind turbine power deviation ratio in a certain area If the value remains large, then adjust the terrain height correction coefficient for the grid points in that area. Make an adaptive upward adjustment;

[0057] Adjusted wake attenuation coefficient Terrain height correction factor Updated to the Jensen wake model.

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] 1. In this wind farm inter-farm yaw control method based on farm groups, wind turbine groups are divided by dynamic clustering, and a dynamic wake influence matrix between groups is constructed by combining the terrain-corrected Jensen wake model. Cooperative yaw optimization is carried out with the goal of maximizing the total predicted power of the entire field. Compared with the traditional independent yaw control of a single wind turbine, this method fully considers the wake coupling effect between wind turbines and the interference of terrain on the wind farm, and solves the problems of neglecting wake loss and not achieving optimal power output of the entire field in the traditional control method.

[0060] 2. In this wind farm inter-farm yaw control method based on a farm cluster, a coordinated yaw command is generated by solving the optimization model through a particle swarm optimization algorithm, and the wake model parameters are dynamically adjusted based on the actual power feedback of the SCADA system to form a closed-loop control mechanism. Compared with the lack of adaptability of traditional fixed parameter control, it can adapt to wind conditions and terrain changes in real time, improve yaw control accuracy, and ensure that the wind farm maintains high wind energy capture efficiency under complex operating conditions. Attached Figure Description

[0061] Figure 1 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example:

[0064] Please see Figure 1 As shown in the figure, this embodiment provides a method for inter-farm yaw control of wind farms on a farm group basis, including the following steps:

[0065] S1. Collect real-time operational data of the target wind farm, including time-series wind vector data, wind turbine operation data, and wind condition forecast data.

[0066] In this embodiment, a virtual lidar system installed in the wind farm is used to simulate and acquire real-time wind vector time-series data of the entire field, including at least wind speed, wind direction, and turbulence intensity.

[0067] Real-time operating data of each wind turbine is obtained from the SCADA system of the wind farm, including at least the turbine power, speed, pitch angle, yaw angle, nacelle wind direction, operating mode and fault status;

[0068] Access external numerical weather forecast services to obtain wind condition forecast data for future periods, including forecast wind speed, wind direction, and atmospheric stability parameters;

[0069] Outliers and missing data were removed from the time series data of wind vectors, wind turbine operation data and wind condition forecast data of the entire field, and interpolation methods were used to fill in short-term missing data to ensure the integrity and reliability of the data.

[0070] The processed full-field wind vector data, wind turbine operation data, and wind condition forecast data are organized according to time series to form structured real-time operation data.

[0071] S2. Perform time synchronization and spatial normalization processing on the time series data of the entire wind vector to form a wind field feature matrix, and then merge it with the wind turbine operation data to construct a fused dataset;

[0072] In this embodiment, based on the full-field wind vector time series data, resampling is performed according to a unified time reference (such as the SCADA system timestamp) to ensure that all data points have the same time resolution, thereby generating time-aligned wind vector data;

[0073]

[0074] in, for Wind vector at any given moment; Indicates the first wind measurement point Wind speed at any moment Indicates the first wind measurement point The wind direction at any moment Indicates the first wind measurement point Turbulence intensity at any given moment;

[0075] The wind farm is divided into regular grids (e.g., 100m×100m). Based on the time-aligned wind vector data, the discrete wind measurement point data is interpolated to each node of the regular grid using the inverse distance weighted interpolation method to obtain a continuous wind field distribution covering the entire field.

[0076]

[0077] in, For grid points exist The normalized wind speed at any given time reflects the continuous wind field distribution across the entire field; For grid points to the th The distance between each wind measurement point This is the attenuation coefficient (usually taken as 2). The number of wind measurement points participating in the interpolation; For the first The weighting factor of each wind measurement point is used in the inverse distance weighted interpolation method to reflect the degree of influence of the wind measurement point on the current grid point.

[0078] Based on continuous wind field distribution, a terrain height correction coefficient is introduced. The wind speed at the grid nodes is corrected for terrain features to obtain the corrected wind speed.

[0079]

[0080] in, For grid points The terrain height correction factor is obtained based on preset terrain undulation data; For grid points exist Wind speed after real-time correction;

[0081] Based on the corrected wind speed and continuous wind field distribution, a wind field feature matrix is ​​constructed using a feature matrix organization method. The wind field feature matrix includes the wind speed, wind direction, and turbulence intensity for each grid point.

[0082] in, The wind field characteristic matrix, For grid height, For grid width, The number of characteristic channels (wind speed, wind direction, turbulence intensity).

[0083] In this embodiment, for each wind turbine According to its position coordinates Extract the wind speed of each wind turbine grid point from the wind field feature matrix. ,wind direction turbulence intensity The characteristics of the wind field at the wind turbine level were obtained;

[0084] Each wind turbine The wind turbine operation data is aligned and merged with the wind field characteristics of the corresponding wind turbine locations to form wind turbine-level time series samples;

[0085]

[0086] in, For wind turbine exist Time series samples of wind turbines at any given moment. For wind turbine exist Power at any moment For wind turbine exist Rotation speed at any given moment For wind turbine exist The pitch angle at that moment, For wind turbine exist Yaw angle at any moment For wind turbine exist The cabin wind direction at any given moment. For wind turbine exist Wind speed at any moment For wind turbine exist The wind direction at any moment For wind turbine exist Turbulence intensity at any given moment;

[0087] Organize the turbine-level time series samples of all wind turbines at all times in chronological order to construct a fused dataset;

[0088]

[0089] in, To merge datasets, This represents the total number of wind turbines. This represents the number of time steps.

[0090] S3. Based on the fused dataset, a dynamic clustering method is used to divide the turbines in the wind farm into several mutually influential turbine groups;

[0091] In this embodiment, for each wind turbine Extracting its time window from the fused dataset The average values ​​of wind speed, wind direction, turbulence intensity, power, rotational speed, pitch angle, and yaw angle, as well as the standard deviations of wind speed and wind direction, are used to construct a wind turbine feature vector by combining these with the turbine's location coordinates. :

[0092]

[0093] in, For wind turbine The characteristic vector of the wind turbine, The wind speed is the average value over the time window. The average wind direction over the time window. The mean of turbulence intensity over the time window. The power is the average value over the time window. The average rotational speed over the time window. The pitch angle is the average value over the time window. This represents the average yaw angle over the time window; The standard deviation of wind speed, The standard deviation of wind direction reflects the volatility of wind conditions; The coordinates of the wind turbine's location in the wind farm (unit: meters);

[0094] For wind turbine feature vectors Mini-max normalization is performed on each dimension to obtain the normalized feature vector. ;

[0095] in, For wind turbine The normalized eigenvectors;

[0096] Based on the normalized feature vector A density-based dynamic clustering method is used to calculate the Euclidean distance between wind turbines to obtain the similarity distance;

[0097]

[0098] in, For wind turbine Japanese wind machine The similarity distance between them; Indicates Euclidean distance;

[0099] Similarity distance does not exceed the neighborhood radius And the number of wind turbines in the neighborhood is not less than the minimum number of wind turbines in the neighborhood. A group of wind turbines is grouped into the same group, thus resulting in several wind turbine groups. ;

[0100] If And the number of wind turbines in the neighborhood Then the fan and Grouped into the same group;

[0101] in, The neighborhood radius (e.g., 0.3) is set according to the scale of the wind farm and the distribution of wind turbines; The minimum number of neighboring wind turbines (e.g., 3);

[0102]

[0103] in, The output of the clustering results A group of wind turbines; For the first A group of wind turbines , The total number of groups is represented by the number of wind turbine groups, each containing several wind turbines that significantly influence each other.

[0104] S4. For the divided wind turbine groups, the Jensen wake model with terrain correction is used to calculate the dynamic wake influence matrix between groups;

[0105] In this embodiment, each wind turbine is extracted based on the wind field feature matrix. Wind speed at the location and wind direction ;

[0106] Based on wind direction For any two wind turbines (Upstream) and wind turbines (Downstream), calculate its relative position vector Calculate the fan The angle between the wind direction and the direction of the relative position vector, if the angle does not exceed the direction threshold. Then determine the fan For the fan There is a wake effect; otherwise, there is no wake effect.

[0107]

[0108]

[0109] in, For wind turbine Japanese wind machine The relative position vector; For wind turbine exist The angle between the wind direction at any given moment and the direction of the relative position vector; For example, the direction threshold. ;

[0110] For wind turbine pairs with wake effects, a terrain-corrected Jensen wake model is used, combined with wind speed. Calculate the downstream wind turbine's thrust coefficient, the distance between the two turbines, the turbine rotor diameter, and the terrain height correction factor at the downstream turbine location. Under the wind turbine Actual wind speed after wake effect ;

[0111]

[0112] in, For wind turbine exist Constantly affected by wind turbines Actual wind speed after wake effect For wind turbine exist The wind speed at any given time is derived from the wind field characteristic matrix; For wind turbine The thrust coefficient is determined by its pitch angle. Calculation of tip speed ratio; This is the wake attenuation coefficient (generally taken as 0.04 to 0.08). For wind turbine arrive The straight-line distance; The diameter of the fan rotor; For wind turbine The terrain height correction factor is used to account for the impact of terrain on wake diffusion;

[0113] downstream wind turbines actual wind speed The wind speed loss ratio is calculated by comparing it with the original wind speed, and this ratio is used as the wake influence coefficient of the upstream wind turbine on the downstream wind turbine. ;

[0114]

[0115] If there is no wake effect, then ;

[0116] in, For wind turbine For the fan exist The wake effect coefficient at time [time]. ;

[0117] For the whole audience Typhoon generator, constructed based on all wake influence coefficients. Wind turbine-level dynamic wake influence matrix It is used to reflect the wake influence relationship between any two wind turbines;

[0118]

[0119] in, for The dynamic wake influence matrix of wind turbines at any given time is a sparse matrix, and the non-zero elements reflect the dynamic wake interaction relationship between wind turbines.

[0120] Based on the dynamic wake influence matrix of the wind turbine level The wake influence coefficients within and between each wind turbine group are averaged and aggregated to obtain the dynamic wake influence matrix between groups. ;

[0121] For each wind turbine group and Calculate the average wake influence coefficient among groups:

[0122]

[0123] in, for Time Wind Turbine Group For wind turbine groups The average wake influence coefficient characterizes the overall wake influence among groups; For the first The number of wind turbines contained in a wind turbine group. For the first The number of wind turbines contained in a wind turbine group;

[0124] Based on the average wake influence coefficient among all groups This forms a dynamic wake influence matrix among groups. This is used for subsequent field-group collaborative optimization.

[0125] S5. With the goal of maximizing the total predicted power of the entire field, an optimization model is constructed with the yaw angle of the wind turbines in each group as the decision variable. The dynamic wake influence matrix between groups and wind condition forecast data are combined and solved by the particle swarm optimization algorithm to generate a set of coordinated yaw instructions for the field groups in the future.

[0126] In this embodiment, key basic data of a single wind turbine are obtained, including air density, wind turbine rotor swept area, forecast wind speed (from wind condition forecast data), and the correspondence between power coefficient and group common yaw angle.

[0127] Based on key basic data, the predicted power of wind energy conversion of a single wind turbine is calculated, and then the predicted power of all wind turbines in each group is summed to obtain the total predicted power of the group.

[0128]

[0129]

[0130] in, Let the air density be denoted as . ; The swept area of ​​the wind turbine rotor. The diameter of the fan rotor; for Time Fan Forecast wind speed at the location, For groups At the predicted time The common yaw angle (all turbines in the group use the same yaw command); is the power factor of the wind turbine, which characterizes the efficiency of the wind turbine in converting wind energy into mechanical energy, and its value ranges from 0 to 16 / 27 (Bates limit). For wind turbine exist The predicted power should always take into account the effects of wake turbulence and yaw angle. Indicates group exist Predicted total power at any given time;

[0131] The optimization objective is to maximize the total predicted power of all wind turbines in the future forecast period, and the decision variable is the set of common yaw angles of each group at each forecast time. ;

[0132] Optimization goal:

[0133]

[0134] in, Indicates the prediction time domain length, for example, taking minute;

[0135]

[0136] in, For a common set of yaw angles;

[0137] Setting a yaw angle range constraint is used to limit the yaw angle change of each group from not exceeding the allowable range, and setting a yaw angle change rate constraint is used to limit the rate of yaw angle change between adjacent moments from not exceeding the limit value.

[0138] Yaw angle range constraints (wind turbine mechanical limitations):

[0139]

[0140] Yaw angle change rate constraint (to avoid mechanical shock):

[0141]

[0142] The optimization model is obtained by combining the optimization objective, decision variables, yaw angle range constraints, and yaw angle change rate constraints.

[0143] In this embodiment, S5.1, based on the decision variables, the initial size is... A swarm of particles, each particle Represents a set of possible common yaw angles, where ;

[0144] The initial position of each particle is randomly generated within the constraints:

[0145]

[0146] in, It is a uniform distribution function. This is the minimum yaw angle. This represents the maximum value of the yaw angle; For the first Of the particles, the first one individual wind turbine clusters at the predicted time The yaw angle value;

[0147] S5.2, For each particle Based on the optimized model, combined with wind forecast data and the dynamic wake influence matrix between groups, the total predicted power for the entire field is calculated. This is the fitness value of the particle;

[0148]

[0149] in, For the first The fitness value of a particle is physically represented as the total prediction power of the entire field in the prediction time domain under the yaw angle decision scheme corresponding to that particle. For the first The set of yaw angle decision variables corresponding to each particle includes the yaw angle values ​​of all groups represented by that particle in the prediction time domain.

[0150] S5.3 Record the historical best position of each particle. and the global optimal position of the particle swarm Based on the velocity and position update formulas of the particle swarm optimization algorithm, the state of all particles is iteratively updated to maximize the fitness value of the particles.

[0151] Update particle velocity:

[0152]

[0153] Update particle positions:

[0154]

[0155] Perform constraint checks and corrections on the updated position to ensure that it does not violate any constraints;

[0156] in, For the first The optimal position of a particle is the set of yaw angle decision variables corresponding to the particle when its fitness value is maximized during the iteration process. The global optimal position of the particle swarm is the set of yaw angle decision variables corresponding to the maximum fitness value of all particles during the iteration process. for The first random number within, for The second random number within; For the first The velocity of each particle, with the initial velocity of the particles set to 0; For individual learning factors, For group learning factors; Inertial weights;

[0157] S5.4 Repeat steps S5.2 to S5.3 until the maximum number of iterations is reached. (like Output the yaw angle decision set corresponding to the globally optimal position. That is, the future The field group coordinated yaw command set within the time period;

[0158]

[0159] in, For field group coordinated yaw command set; The position of the global optimal position A group in The optimal yaw angle command value at any given time.

[0160] S6. Execute the field group coordinated yaw command set, collect the actual power feedback in the SCADA system in real time, and adjust the parameters of the Jensen wake model through the power feedback;

[0161] In this embodiment, the wind turbine group coordinated yaw command set is sent to the controller of each wind turbine group, and each group of wind turbines executes a unified common yaw angle command;

[0162] Real-time power data of each wind turbine is collected from the wind farm's SCADA system, with a sampling interval of 1 second, to form a power feedback sequence;

[0163] Furthermore, the wake attenuation coefficient is selected. (Initially set to 0.05), terrain height correction factor (Initially set to 1.0, indicating no terrain influence) as the parameter to be adjusted for the Jensen wake model;

[0164] Based on the power feedback sequence, at the end of each control cycle (e.g., every 10 minutes), the average actual power of all wind turbines within the control cycle is calculated. And compared with the average power predicted by the optimized model for the same period. Compare and calculate the power deviation ratio. ;

[0165]

[0166] in, This is the power deviation ratio. This represents the average actual power. This is the average value of the power prediction;

[0167] If the power deviation ratio is greater than the positive deviation threshold (value...) If so, the wake attenuation coefficient will be reduced proportionally. If the power deviation ratio is less than the negative deviation threshold, then the wake attenuation coefficient is increased proportionally. Otherwise, maintain the wake attenuation coefficient. constant;

[0168] Specifically, if The fact that the actual power was higher than predicted indicates that the wake effect was overestimated and should be reduced. The amplitude is:

[0169]

[0170] like (Actual power is lower than predicted) This indicates that the wake effect has been underestimated, and the power output should be increased. :

[0171]

[0172] like If so, no adjustment will be made for the time being;

[0173] If the wind turbine power deviation ratio in a certain area (such as a valley or ridge) is... If the deviation remains large (e.g., a negative power deviation exceeding 8% occurs for three consecutive control cycles), then the terrain height correction coefficient for the grid points in that area should be adjusted. Make an adaptive upward adjustment;

[0174]

[0175] in, This represents the proportion of the average power deviation in that region. This is the terrain height correction factor before the update. This is the updated terrain height correction factor;

[0176] Adjusted wake attenuation coefficient Terrain height correction factor Update to the Jensen wake model for wake calculation and yaw optimization in the next cycle;

[0177] To prevent excessive parameter adjustment, set upper and lower limits for the parameters:

[0178] .

[0179] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for inter-farm yaw control in wind farms based on a cluster of farms, characterized in that: Includes the following steps: S1. Collect real-time operational data of the target wind farm, including time-series wind vector data, wind turbine operation data, and wind condition forecast data. S2. Perform time synchronization and spatial normalization processing on the time series data of the entire wind vector to form a wind field feature matrix, and then merge it with the wind turbine operation data to construct a fused dataset; S3. Based on the fused dataset, a dynamic clustering method is used to divide the turbines in the wind farm into several mutually influential turbine groups; S4. For the divided wind turbine groups, the Jensen wake model with terrain correction is used to calculate the dynamic wake influence matrix between groups; S5. With the goal of maximizing the total predicted power of the entire field, an optimization model is constructed with the yaw angle of the wind turbines in each group as the decision variable. The dynamic wake influence matrix between groups and wind condition forecast data are combined and solved by the particle swarm optimization algorithm to generate a set of coordinated yaw instructions for the field groups in the future. S6. Execute the field group coordinated yaw command set, collect the actual power feedback in the SCADA system in real time, and adjust the parameters of the Jensen wake model through the power feedback.

2. The wind farm inter-farm yaw control method based on a wind farm cluster as described in claim 1, characterized in that: In step S1, the specific steps involved in collecting real-time operational data of the target wind farm are as follows: By using a virtual lidar system installed in the wind farm, real-time simulation and acquisition of wind vector time-series data for the entire field are achieved, including wind speed, wind direction, and turbulence intensity. Real-time operating data of each wind turbine is obtained from the SCADA system of the wind farm, including turbine power, speed, pitch angle, yaw angle, nacelle wind direction, operating mode and fault status; Access external numerical weather forecast services to obtain wind condition forecast data for future periods, including forecast wind speed, wind direction, and atmospheric stability parameters; Outliers and missing data were removed from the time series data of wind vectors, wind turbine operation data and wind condition forecast data of the entire field, and interpolation methods were used to fill in the short-term missing data appropriately; The processed full-field wind vector data, wind turbine operation data, and wind condition forecast data are organized according to time series to form structured real-time operation data.

3. The wind farm inter-farm yaw control method based on a wind farm cluster as described in claim 1, characterized in that: In step S2, the specific steps involved in performing time synchronization and spatial normalization processing on the full-field wind vector time series data are as follows: Based on the full-field wind vector time series data, resampling is performed according to a unified time base to generate time-aligned wind vector data; The wind farm is divided into a regular grid. Based on the time-aligned wind vector data, the discrete wind measurement point data is interpolated to each node of the regular grid using the inverse distance weighted interpolation method to obtain a continuous wind field distribution covering the entire field. Based on continuous wind field distribution, a terrain height correction coefficient is introduced. The wind speed at the grid nodes is corrected for terrain features to obtain the corrected wind speed. Based on the corrected wind speed and continuous wind field distribution, a wind field feature matrix is ​​constructed using a feature matrix organization method. The wind field feature matrix includes the wind speed, wind direction, and turbulence intensity for each grid point.

4. The wind farm inter-farm yaw control method based on a wind farm cluster as described in claim 3, characterized in that: In step S2, the specific steps involved in fusing the wind turbine operation data to construct the fused dataset are as follows: For each wind turbine Based on its location coordinates, the wind speed, wind direction, and turbulence intensity of each wind turbine's grid point are extracted from the wind field feature matrix to obtain the wind field characteristics at the turbine level. Each wind turbine The wind turbine operation data is aligned and merged with the wind field characteristics of the corresponding wind turbine locations to form wind turbine-level time series samples; Organize the turbine-level time series samples of all wind turbines at all times in chronological order to construct a fused dataset.

5. The wind farm inter-farm yaw control method according to claim 1, characterized in that: In step S3, the specific steps involved in dividing the turbines in the wind farm into several mutually influential turbine groups using a dynamic clustering method are as follows: For each wind turbine Extracting its time window from the fused dataset The average values ​​of wind speed, wind direction, turbulence intensity, power, rotational speed, pitch angle, and yaw angle, as well as the standard deviations of wind speed and wind direction, are used to construct a wind turbine feature vector by combining these with the turbine's location coordinates. ; For wind turbine feature vectors Mini-max normalization is performed on each dimension to obtain the normalized feature vector. ; Based on the normalized feature vector A density-based dynamic clustering method is used to calculate the Euclidean distance between wind turbines to obtain the similarity distance; Similarity distance not exceeding the neighborhood radius And the number of wind turbines in the neighborhood is not less than the minimum number of wind turbines in the neighborhood. A group of wind turbines is grouped into the same group, thus resulting in several wind turbine groups. .

6. The wind farm inter-farm yaw control method according to claim 1, characterized in that: In step S4, the specific steps involved in calculating the dynamic wake influence matrix between groups using the terrain-corrected Jensen wake model are as follows: Based on the wind field feature matrix, extract each wind turbine Wind speed at the location and wind direction ; Based on wind direction For any two wind turbines Japanese-style fan Calculate its relative position vector Calculate the fan The angle between the wind direction and the direction of the relative position vector, if the angle does not exceed the direction threshold. Then determine the fan For the fan There is a wake effect; otherwise, there is no wake effect. For wind turbine pairs with wake effects, a terrain-corrected Jensen wake model is used, combined with wind speed. Calculate the wind turbine's thrust coefficient, the distance between the two turbines, the turbine rotor diameter, and the terrain height correction factor at the downstream turbine location. Under the wind turbine Actual wind speed after the wake effect; wind turbine The actual wind speed is compared with the original wind speed, and the wind speed loss ratio is calculated as the wake influence coefficient of the upstream wind turbine on the downstream wind turbine. ; For the whole audience Typhoon generator, constructed based on all wake influence coefficients. Wind turbine-level dynamic wake influence matrix ; Based on the dynamic wake influence matrix of the wind turbine level The wake influence coefficients within and between each wind turbine group are averaged and aggregated to obtain the dynamic wake influence matrix between groups. .

7. The wind farm inter-farm yaw control method according to claim 1, characterized in that: In step S5, the specific steps involved in constructing an optimization model with the yaw angle of wind turbines in each group as the decision variable, with the goal of maximizing the total predicted power of the entire field, are as follows: Obtain key basic data for a single wind turbine, including air density, turbine rotor swept area, forecast wind speed, and the correspondence between power coefficient and group common yaw angle; Based on key basic data, the predicted power of wind energy conversion of a single wind turbine is calculated, and then the predicted power of all wind turbines in each group is summed to obtain the total predicted power of the group. The optimization objective is to maximize the total predicted power of all wind turbines in the future forecast period, and the decision variable is the set of common yaw angles of each group at each forecast time. ; Setting a yaw angle range constraint is used to limit the yaw angle change of each group from not exceeding the allowable range, and setting a yaw angle change rate constraint is used to limit the rate of yaw angle change between adjacent moments from not exceeding the limit value. The optimization model is obtained by combining the optimization objective, decision variables, yaw angle range constraints, and yaw angle change rate constraints.

8. The wind farm inter-farm yaw control method based on a wind farm cluster as described in claim 7, characterized in that: In step S5, the specific steps involved are solved by combining the dynamic wake influence matrix between groups and wind forecast data using the particle swarm optimization algorithm: S5.

1. Based on the decision variables, the initial size is... A swarm of particles, each particle Represents a set of possible common yaw angles, where ; S5.2, For each particle Based on the optimized model, combined with wind forecast data and the dynamic wake influence matrix between groups, the total predicted power for the entire field is calculated. This is the fitness value of the particle; S5.3 Record the historical best position of each particle. and the global optimal position of the particle swarm Based on the velocity and position update formulas of the particle swarm optimization algorithm, the state of all particles is iteratively updated to maximize the fitness value of the particles. S5.4 Repeat steps S5.2 to S5.3 until the maximum number of iterations is reached. Output the yaw angle decision set corresponding to the globally optimal position. That is, the future The set of coordinated yaw commands for the field group during the time period.

9. The wind farm inter-farm yaw control method according to claim 1, characterized in that: In step S6, the specific steps involved in executing the field group coordinated yaw command set and real-time acquisition of actual power feedback from the SCADA system are as follows: The wind turbine group coordinated yaw command set is sent to the controller of each wind turbine group, and each group of wind turbines executes a unified common yaw angle command; The actual power data of each wind turbine is collected in real time from the wind farm's SCADA system to form a power feedback sequence.

10. A method for inter-farm yaw control of wind farms based on a cluster as described in claim 9, characterized in that: In step S6, the specific steps involved in adjusting the parameters of the Jensen wake model through power feedback are as follows: Selecting the wake attenuation coefficient Terrain height correction factor As parameters to be adjusted in the Jensen wake model; Based on the power feedback sequence, at the end of each control cycle, the average actual power of all wind turbines during the control cycle is calculated. And compared with the average power predicted by the optimized model for the same period. Compare and calculate the power deviation ratio. ; If the power deviation ratio is greater than the positive deviation threshold, the wake attenuation coefficient is reduced proportionally. ; If the power deviation ratio is less than the negative deviation threshold, then the wake attenuation coefficient is increased proportionally. Otherwise, maintain the wake attenuation coefficient. constant; If the wind turbine power deviation ratio in a certain area If the value remains large, then adjust the terrain height correction coefficient for the grid points in that area. Make an adaptive upward adjustment; Adjusted wake attenuation coefficient and terrain height correction factor Updated to the Jensen wake model.

Images