Wind turbine cluster wake control method based on radar wind measurement

By combining radar wind measurement and dynamic wind field data processing with three-dimensional safety load boundaries and structural natural frequencies, the problems of accurate prediction of wake interference and structural overload in the wake control of wind turbine groups were solved. This achieved safe control of wake deflection and lubrication protection of pitch bearings, thus improving the operational reliability of wind turbine units.

CN123109946APending Publication Date: 2026-11-17HUANENG ZHAOJUE WIND POWER CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611423271.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing wake control methods for wind turbine clusters are unable to accurately predict wake interference under complex wind conditions, leading to problems such as overload of the unit structure and lubrication failure of the pitch bearing.

Method used

A wake control method based on radar wind measurement is adopted. By acquiring real wind field data after dynamic compensation, the interference state between the wake trajectory and the swept area of ​​the downstream unit is predicted, a wake cooperative control trigger flag is generated, and a comprehensive pitch command is synthesized by combining the three-dimensional safety load envelope boundary and the structure's natural frequency. The cumulative micro-stroke integral value of the pitch action is monitored to achieve grease redistribution.

Benefits of technology

It enables accurate prediction of wake interference under complex wind conditions, avoids overload and mechanical resonance of the unit structure, reduces the wear risk of pitch bearings, and improves the safe operating life of the wind turbine.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123109946A_ABST
    Figure CN123109946A_ABST
Patent Text Reader

Abstract

This invention relates to the field of wind power generation control technology and discloses a wake control method for wind turbine clusters based on radar wind measurement. The method includes: acquiring real wind field data collected and dynamically compensated by radar; predicting the interference state between the wake trajectory and the swept area of ​​the downstream turbine and generating a coordinated control trigger flag; separating high-frequency wind direction change data based on the flag and converting it into a safe lateral thrust command constrained by a three-dimensional safe load envelope boundary; calculating a local additional pitch angle based on the command and synthesizing a comprehensive pitch command by combining it with a safe wake excitation frequency that avoids the inherent frequency of the turbine structure; monitoring the cumulative micro-motion stroke integral value of the pitch action; if it exceeds a preset wear threshold, superimposing a synchronous sweep command on a unified pitch reference angle to perform grease redistribution; and resuming the original command issuance after completion. This invention improves the accuracy of wake interference prediction, avoids structural overload, and extends the life of the pitch bearing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of wind power generation control technology, specifically to a method for controlling the wake of wind turbine clusters based on radar wind measurement. Background Technology

[0002] In wind farms, the wake generated by upstream wind turbines can reduce wind speeds and increase turbulence intensity at downstream turbines. Conventional active wake control strategies primarily involve adjusting the rotor attitude to deflect the wake trajectory, thereby reducing interference with downstream turbines.

[0003] In practical engineering, due to the complex and variable flow field environment, existing control methods relying on static wake models cannot accurately predict the dynamic evolution of the wake, resulting in a lag in the triggering timing of coordinated control. During active wake deflection, the unit typically needs to perform high-frequency independent pitch control to generate lateral thrust, inevitably introducing additional alternating loads to the entire unit. Existing systems often fail to constrain the real-time load boundaries of the unit when issuing deflection commands and do not avoid the unit's overall structural natural frequencies, easily leading to mechanical resonance and structural overload. Because the wake deflection state needs to be maintained, the pitch bearing is under constant high-frequency reciprocating fretting at small angles. The grease in the contact area between the balls and raceways cannot be effectively replenished, and the dry friction caused by localized lubrication failure accelerates bearing wear and degradation, affecting the safe operating life of the unit. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a wind turbine cluster wake control method based on radar wind measurement. This method solves the problem that existing wind turbine cluster wake control methods mostly use static wake models, which are difficult to accurately predict wake interference states under complex wind conditions. In addition, the active wake deflection strategy introduces additional alternating loads to the unit, which can easily induce structural resonance. At the same time, the high-frequency micro-pitch commands required to maintain wake deflection can easily lead to local grease loss in the pitch bearing contact area, causing excessive bearing wear.

[0005] To address the above problems, the present invention provides the following technical solution:

[0006] This invention provides a wake control method for wind turbine clusters based on radar wind measurement, comprising the following steps:

[0007] Acquire real wind field data collected by radar and dynamically compensated, predict the interference state between the wake trajectory and the swept area of ​​the downstream unit, and generate wake cooperative control trigger flag.

[0008] Based on the wake coordination control trigger flag, high-frequency wind direction change data is separated and converted into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary.

[0009] The local additional pitch angle command is calculated based on the safe lateral thrust command, and combined with the safe wake excitation frequency that avoids the natural frequency of the unit structure, the comprehensive pitch command is synthesized.

[0010] During the execution of the integrated pitch control command, the cumulative micro-motion stroke integral value of the pitch control action is monitored. If the cumulative micro-motion stroke integral value is greater than the preset wear threshold, a synchronous sweep command is superimposed on the unified pitch reference angle of the unit's current operating condition to perform grease redistribution. After the execution is completed, the issuance of the integrated pitch control command is restored.

[0011] Furthermore, the radar collects and acquires dynamically compensated real wind field data, including:

[0012] Perform time definite integral on the acceleration and angular velocity data output by the cabin inertial measurement unit to obtain transient translational displacement and transient rotational deviation angle;

[0013] By using transient translational displacement and transient rotational deviation angle to construct an instantaneous homogeneous coordinate transformation matrix, the original spatial wind field point cloud data matrix collected by the forward lidar is remapped to the reference static coordinate system to generate real wind field data.

[0014] The discrete Fourier transform is performed on the acceleration data synchronously, and the point with the maximum vibration amplitude is located in the effective frequency band sequence retained by the bandpass truncation, and the corresponding unit structure natural frequency is extracted.

[0015] Furthermore, the interference state between the wake trajectory and the swept area of ​​the downstream unit is predicted, and a wake cooperative control trigger flag is generated, including:

[0016] Based on real wind field data, three-dimensional wind field spatiotemporal voxels are divided and constructed. The transmission time is calculated by dividing the absolute longitudinal physical distance from the spatial voxel to the wind turbine plane by the corresponding longitudinal wind speed component.

[0017] The area-weighted average of the spatial voxel feature wind speeds after time alignment is used to generate a predicted equivalent wind speed sequence, and the environmental turbulence intensity is calculated and extracted based on the predicted equivalent wind speed sequence.

[0018] When the predicted equivalent wind speed sequence meets the wind speed operating range and the environmental turbulence intensity is lower than the turbulence intensity suppression threshold, and the predicted absolute wind direction corner falls within the wake interference sector estimated based on the environmental turbulence intensity, the wind direction interference condition is determined to be met and a wake cooperative control trigger flag in an active state is generated.

[0019] Furthermore, before converting high-frequency wind direction change data into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary, the following steps are also included:

[0020] The high-frequency wind direction change data corresponding to the wake-coordinated control trigger flag is converted into a frequency domain to generate a wind direction frequency domain amplitude spectrum. The core frequency component with the largest amplitude and the initial phase information of the core frequency component are extracted in the effective low-frequency range of the wind direction frequency domain amplitude spectrum.

[0021] Calculate the angular deviation between the current predicted absolute wind direction angle and the target escape boundary, and generate the target wake deflection angle by combining it with the preset deflection safety margin angle;

[0022] The lateral thrust coefficient is obtained by multiplying the wake sensitivity coefficient, the target wake deflection angle, and the axial thrust coefficient, and the high-frequency lateral thrust amplitude is calculated based on the lateral thrust coefficient.

[0023] Using the amplitude-limited high-frequency lateral thrust amplitude as an amplitude parameter, combined with the core frequency components and initial phase information, the target lateral thrust time-domain command is reconstructed for aerodynamic mapping.

[0024] Furthermore, the high-frequency wind direction change data is converted into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary, including:

[0025] Based on the current operating parameters of the wind turbine generator, calculate the dynamic margins of axial thrust, pitch moment, and yaw moment, and construct an orthogonal three-dimensional safety load envelope boundary.

[0026] Query the offline aeroelastic sensitivity matrix table, and combine it with the target lateral thrust time domain command to predict and generate the axial additional thrust, additional pitch moment amplitude and additional yaw moment at the current moment;

[0027] Divide the dynamic margins of each dimension by the corresponding additional amplitude after adding the zero offset constant to obtain the thrust margin ratio, pitch margin ratio, and yaw margin ratio. Extract the minimum value among the thrust margin ratio, pitch margin ratio, and yaw margin ratio as the global truncation coefficient and multiply it with the target lateral thrust time domain command to generate the safe lateral thrust command.

[0028] Furthermore, the local additional pitch angle command is calculated based on the safe lateral thrust command, including:

[0029] Divide the safe lateral thrust command by the lateral thrust pitch angle sensitivity coefficient after the lower limit clamping process to calculate the total independent pitch amplitude.

[0030] The cosine value is obtained by adding the absolute azimuth angle of each blade to the phase compensation angle calculated using the first-order inertial time constant. The cosine value is then used to modulate the total independent pitch amplitude, and the spatial azimuth projection is used to generate the local additional pitch angle command for each blade.

[0031] Furthermore, by combining a safe wake excitation frequency that avoids the inherent frequency of the generator structure, a comprehensive pitch command is synthesized, including:

[0032] The predicted equivalent wind speed and environmental turbulence intensity under the current operating conditions of the unit are obtained. The initial wake excitation frequency is calculated by combining the predicted equivalent wind speed and the rotor diameter with the optimal Strouhal number matched according to the environmental turbulence intensity.

[0033] When the initial wake excitation frequency falls into the interference forbidden zone constructed by the rotor speed frequency multiplier, an adaptive frequency shift is performed on the boundary value of the low-frequency side or the side with the closer absolute distance to obtain the candidate frequency; after the candidate frequency is verified to not fall into the mechanical resonant frequency forbidden zone corresponding to the unit structure's natural frequency, the candidate frequency is subjected to time-domain smoothing processing with first-order low-pass filtering to confirm it as a safe wake excitation frequency.

[0034] A high-frequency micro-motion disturbance wave is generated using a safe wake excitation frequency, and then superimposed onto the local additional pitch angle command using an algebraic addition method to output a comprehensive pitch command.

[0035] Furthermore, the cumulative integral value of the pitch control stroke is monitored, including:

[0036] Within a set sliding time window, the single-step pitch increment of the real-time pitch angle feedback signal is extracted and discretely accumulated to obtain the sliding cumulative pitch stroke.

[0037] The theoretical maximum pitch stroke is calculated by dividing the cumulative slip pitch stroke by the slip time window and combining it with the rated maximum pitch rate. The pitch micro-motion duty cycle is then calculated.

[0038] When the pitch micro-motion duty cycle of any blade reaches the upper limit threshold, a fatigue degradation flag is sent to the command synthesis stage. When the cumulative activation time of the flag reaches the sweep trigger time threshold, it is determined that the cumulative micro-motion stroke integral value is greater than the preset wear threshold, thereby suspending the integrated pitch command.

[0039] Furthermore, a synchronous sweep command is superimposed on the unified pitch reference angle of the unit's current operating condition to perform grease redistribution, including: extracting the unified pitch reference angle of the unit's current operating condition, and calculating the positive and negative stroke margins respectively in combination with the preset blade mechanical limit angle range.

[0040] Select a motion direction with a stroke margin greater than the target amplitude, and construct a sweep pulse command in half-sine wave shape based on the motion direction;

[0041] After verifying the maximum command change rate to ensure that the speed limit is not exceeded, the sweep pulse command is applied to the three blades in the same direction and with the same amplitude as the synchronous sweep command, forcing the internal grease to re-cover the contact area.

[0042] Furthermore, after execution is complete, the issuance of integrated pitch commands will resume, including:

[0043] When the execution time of the synchronous sweep command reaches the execution cycle and the actual pitch angle of each blade falls back to the tolerance range, a recovery wake coordination control trigger flag is generated.

[0044] Based on the recovery wake coordination control trigger flag, the historical state accumulation items are cleared to zero, and the linear climb logic is started within the preset recovery time window to generate a transition weight that increases over time.

[0045] The local additional pitch angle command is multiplied by the transition weight and then superimposed on the unified pitch reference angle to perform a smooth cut-in. After the transition weight climbs to the peak value, the accumulated micro-motion stroke integral value is cleared and the full-function closed loop is restored.

[0046] This invention provides a wake control method for wind turbine clusters based on radar wind measurement. It has the following beneficial effects:

[0047] 1. This invention acquires radar wind measurement data and nacelle inertial measurement data, uses coordinate transformation to correct spatial sampling distortion caused by nacelle sway, and combines this with three-dimensional wind field spatiotemporal voxel modeling to predict the wake trajectory. This design enables the unit to accurately determine the interference state of the wake on the downstream swept area before the wake actually arrives, thereby activating cooperative control in advance and solving the problems of prediction lag and low accuracy of conventional static wake models under complex wind conditions.

[0048] 2. When generating integrated pitch commands, this invention constructs a three-dimensional safety load envelope boundary based on the current operating conditions, performs dynamic margin truncation on the lateral thrust command, and performs adaptive frequency shifting by combining the dual forbidden zones of the rotational speed doubling frequency and the structural natural frequency. This control mechanism, while achieving wake deflection and structural disturbance, strictly limits the upper limit of the amplitude of independent pitch, avoiding structural overload and mechanical resonance of the unit.

[0049] 3. This invention addresses the problem of lubrication loss in pitch bearings caused by high-frequency micro-motion pitch control by introducing a micro-motion duty cycle monitoring mechanism. When the cumulative micro-motion stroke integral value exceeds the limit, the system superimposes a half-sine synchronous sweep pulse on the current pitch reference angle. The large-stroke sweep action redistributes the grease inside the bearing and covers the micro-motion contact area, then smoothly restores the original wake control, reducing the risk of wear caused by local dry friction in the bearing. Attached Figure Description

[0050] Figure 1 This is a block diagram of the wake control system based on radar and asymmetric pitch of the present invention.

[0051] Figure 2 This is a flowchart of the wake control method based on radar and asymmetric pitch of the present invention.

[0052] Figure 3 This is a flowchart of the steps for wind direction frequency decoupling and yaw thrust command generation in this invention;

[0053] Figure 4 This is a flowchart of the independent pitch command synthesis and adaptive frequency shift steps of the present invention;

[0054] Figure 5 The wind direction time-frequency response and feature extraction curves based on frequency decoupling of the present invention are shown in the figure; wherein, (a) is a three-dimensional wind field voxelization and interferometric prediction profile of forward radar; (b) is a wake excitation frequency distribution and frequency shift avoidance spectrum.

[0055] Figure 6 This is a schematic diagram of the load prediction and command saturation truncation based on dynamic safety envelope of the present invention;

[0056] Figure 7 This is a timing diagram of the pitch micro-motion wear suppression and sweep pulse smooth transition control of the present invention;

[0057] Figure 8 This is a comparison diagram of the overall wind farm performance and fatigue load under different wake control strategies of the present invention. Detailed Implementation

[0058] The technical solutions in 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.

[0059] See attached document Figure 1 This invention provides a wake control method for wind turbine clusters based on radar wind measurement. To implement the above control method, this invention also provides a wake control system, which may include: a forward lidar, a nacelle inertial measurement unit, a blade root load sensor, a yaw servo actuator, an independent pitch actuator, and a main control computing unit.

[0060] Forward-facing lidar is installed on the top of the nacelle of the wind turbine to collect data on wind speed and direction in the space ahead.

[0061] The nacelle inertial measurement unit is deployed on the load-bearing structure inside the nacelle and is used to collect acceleration and angular velocity data of the wind turbine generator.

[0062] The blade root load sensor is installed at the root of each blade of the wind turbine generator to collect the flapping bending moment data and the swaying bending moment data of each blade.

[0063] The yaw servo actuator is connected to the main control computing unit and is used to adjust the yaw angle of the cabin according to the received yaw control commands.

[0064] The independent pitch control system is connected to the main control computing unit for adjusting the pitch angle for a single blade.

[0065] The main control computing unit contains a coordinate compensation module, an interferometry prediction module, a frequency decoupling module, a load constraint module, an instruction synthesis module, and a wear suppression module.

[0066] The coordinate compensation module receives spatial wind speed and direction matrix data collected by the forward-facing lidar, and acceleration and angular velocity data collected by the nacelle inertial measurement unit. The module dynamically compensates the spatial coordinates of the spatial wind speed and direction matrix data using displacement data processed by time integration, generating compensated, realistic wind field data. Simultaneously, the coordinate compensation module performs a fast Fourier transform on the acceleration data to extract the structural natural frequencies of the wind turbine generator.

[0067] The interferometric prediction module constructs a three-dimensional spatiotemporal voxel of the wind field based on real wind field data. Combining the Taylor frozen turbulence assumption, the interferometric prediction module calculates the propagation time of the wind direction deviation voxel to the rotor plane and the downstream wind turbine plane. When the interferometric prediction module determines that the overlap rate between the wake center trajectory and the swept area of ​​the downstream wind turbine exceeds the interferometric threshold, it sends a wake cooperative control trigger flag to the frequency decoupling module.

[0068] Upon receiving the wake coordination control trigger flag, the frequency decoupling module performs frequency domain conversion on the wind direction deviation sequence. The module then separates the low-frequency wind direction change data using a digital frequency division network and converts it into yaw control commands, which are subsequently sent to the yaw servo execution system. Simultaneously, the module separates the high-frequency wind direction change data and converts it into lateral thrust demand data.

[0069] The load constraint module acquires flapping and yaw bending moment data from the blade root load sensor. Based on the acquired bending moment data, the load constraint module establishes a three-dimensional safe load envelope boundary. The load constraint module then uses an aeroelastic model to calculate the predicted additional load after superimposing the lateral thrust demand data. If the predicted additional load exceeds the three-dimensional safe load envelope boundary, the load constraint module performs a saturation truncation operation to generate a safe lateral thrust command.

[0070] The command synthesis module receives the safe lateral thrust command and calculates the additional pitch angle command for each blade at the current azimuth. Based on the free-flow wind speed and rotor diameter, the module calculates the wake vortex excitation frequency and compares it with the structural natural frequencies extracted by the coordinate compensation module. If frequency bands overlap, the module executes a frequency shift algorithm to generate a comprehensive pitch command that includes the safe excitation frequency.

[0071] The wear suppression module monitors the additional pitch angle command portion of the integrated pitch command and integrates its time derivative to obtain the cumulative micro-motion stroke integral value. When the cumulative micro-motion stroke integral value exceeds a preset wear threshold, the wear suppression module suspends the issuance of integrated pitch commands and superimposes a synchronous sweep command onto the reference pitch angle command before issuing it to the independent pitch execution system. After the synchronous sweep command is executed, the wear suppression module resets the cumulative micro-motion stroke integral value to zero and resumes the issuance of integrated pitch commands.

[0072] The yaw servo actuator adjusts the nacelle position based on received yaw control commands. The independent pitch actuator adjusts the blade angle based on integrated pitch commands or synchronized sweep commands. During the operation of the yaw servo actuator and the independent pitch actuator, the nacelle inertial measurement unit continuously collects acceleration and angular velocity data and sends the data to the coordinate compensation module, forming a closed-loop control loop.

[0073] See attached document Figure 2 This invention provides a wake control method for wind turbine clusters based on radar wind measurement, comprising the following steps:

[0074] S10: Acquire spatial wind speed and wind direction matrix data collected by forward-looking lidar, as well as acceleration and angular velocity data collected by the nacelle inertial measurement unit. Use the displacement data after integration processing to dynamically compensate the spatial coordinates of the spatial wind speed and wind direction matrix data to generate real wind field data. Perform spectrum analysis on the acceleration data to extract the structural natural frequency of the wind turbine generator.

[0075] S20: Construct a three-dimensional wind field spatiotemporal voxel based on real wind field data, calculate the transmission time of the wind direction deviation voxel to the wind turbine plane and the downstream wind turbine plane, and generate a wake cooperative control trigger flag when the overlap rate between the wake center trajectory and the swept area of ​​the downstream wind turbine exceeds the interference threshold.

[0076] S30 performs frequency domain conversion on the wind direction deviation sequence based on the wake coordination control trigger flag, and separates the low-frequency wind direction change data into yaw control commands through the frequency division network and sends them to the yaw servo execution system. At the same time, it separates the high-frequency wind direction change data into lateral thrust demand data.

[0077] S40: Acquire the flapping bending moment data and oscillation bending moment data collected by the blade root load sensor and establish a three-dimensional safe load envelope boundary. Calculate the predicted additional load after superimposing the lateral thrust demand data. When the predicted additional load exceeds the three-dimensional safe load envelope boundary, perform saturation truncation operation to generate a safe lateral thrust command.

[0078] S50 calculates the additional pitch angle command for each blade at the current azimuth angle based on the safe lateral thrust command, calculates the wake vortex excitation frequency and compares it with the structure's natural frequency, and executes a frequency shift algorithm to generate a comprehensive pitch command that includes the safe excitation frequency when the frequency bands overlap.

[0079] S60: Integrate the time derivative of the additional pitch angle command in the integrated pitch command to obtain the cumulative micro-motion stroke integral value. When the cumulative micro-motion stroke integral value is greater than the preset wear threshold, suspend the issuance of the integrated pitch command, superimpose the synchronous sweep command on the reference pitch angle command and issue it to the independent pitch execution system. After the synchronous sweep command is executed, clear the cumulative micro-motion stroke integral value to zero and resume the issuance of the integrated pitch command.

[0080] To enable those skilled in the art to more clearly understand the technical solution of the present invention, the following will describe in detail each processing step in the above system and method in conjunction with specific calculation processes and control logic.

[0081] In the radar-based wind turbine wake control method provided by this invention, when the asymmetric independent pitch system performs dynamic actions, the unbalanced aerodynamic load generated on the rotor plane is transmitted to the nacelle, causing vibration. A forward-facing lidar is installed in the nacelle, and vibration causes a periodic deflection of the radar's detection beam. To obtain an accurate incoming wind field structure, it is necessary to calculate the transient attitude of the nacelle. The specific steps for calculating the nacelle motion attitude data are as follows:

[0082] S101, the coordinate compensation module in the main control computing unit acquires the data sequence collected in real time by the cabin inertial measurement unit. The cabin inertial measurement unit is physically rigidly connected to the main load-bearing structure inside the cabin and is used to sense the cabin's spatial dynamic response. The cabin spatial coordinate system is defined, with the cabin's horizontal coordinates set as... The axis, the vertical upward direction of the nacelle is set as The axis, the longitudinal setting of the nacelle is The axis coordinate compensation module receives the high-frequency lateral acceleration output from the cabin inertial measurement unit. Vertical acceleration and around pitch angular velocity of the shaft and around Yaw rate of the shaft Simultaneously, the relative spatial offset vector between the geometric center of the cabin inertial measurement unit and the mounting reference point of the forward-facing lidar is acquired.

[0083] S102, the coordinate compensation module performs bandpass filtering on the received acceleration and angular velocity data. For bandpass filtering of inertial sensor signals, those skilled in the art design corresponding digital filters based on the operating fundamental frequency of the wind turbine generator and known environmental noise frequency bands. The signal preprocessing and denoising methods are well-known technologies in the field and will not be elaborated upon here. The high-pass cutoff frequency of this bandpass filtering is set to filter out the static component of gravitational acceleration and low-frequency zero-bias drift, while the low-pass cutoff frequency is set to filter out the frequency of high-frequency electromagnetic interference from the electrical equipment inside the nacelle, extracting the effective dynamic frequency band signals related to the nacelle structural vibration.

[0084] S103, the coordinate compensation module performs time integration on the filtered acceleration and angular velocity data. To reduce integration drift caused by inertial sensor noise, the coordinate compensation module uses a definite integral calculation method within a sliding time window. Let the current calculation time be... The sliding time window length is Sliding time window length The value is set according to the natural frequency period of the first-order tower of the wind turbine generator, and is taken as the duration of 3 to 5 complete vibration cycles.

[0085] The coordinate compensation module in the sliding time window Internal, lateral acceleration With vertical acceleration Perform two definite-integral operations over time to calculate the transient translational displacement of the cabin along the two coordinate axes, and obtain the lateral translational displacement. and vertical translation displacement .

[0086] Coordinate compensation module in the same sliding time window Inside, regarding pitch angular velocity With yaw rate Perform a time definite integral operation once to obtain the cabin rotation. Transient pitch and rotation deviation angle calculated by axis and around Transient yaw rotation deviation angle calculated by shaft The specific formula for calculating the integral of angular velocity is as follows:

[0087] ;

[0088] ;

[0089] In the formula, Indicates the current calculation time; Indicates the length of the sliding time window; This represents the time variable in the integral calculation process; Indicates the transient pitch and rotation deviation angle; This represents the transient yaw and rotational deviation angle. During the numerical discrete calculation, since the acceleration data has already filtered out static low-frequency components, to obtain the pure dynamic relative translation, the coordinate compensation module performs continuous-time cumulative integration on the discrete data. After integration, a digital high-pass filter, consistent with the aforementioned low-frequency cutoff point of the bandpass filter, is connected in series to eliminate cumulative drift, replacing the traditional method of forcibly setting the initial window state to zero. This ensures that the extracted transient translational displacement and transient rotational deviation angle remain continuous and smooth in the time series. The four transient displacement and rotational deviation parameters obtained from the aforementioned calculation, serving as quantitative indicators characterizing the nacelle spatial attitude changes of the wind turbine generator under asymmetric aerodynamic loads, are stored in the memory register of the main control computing unit for subsequent spatial coordinate remapping calculations.

[0090] The specific execution steps of dynamic coordinate remapping of spatial wind field point cloud data in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0091] S104, the coordinate compensation module in the main control computing unit receives the raw spatial wind field point cloud data matrix output by the forward-facing lidar. The forward-facing lidar acquires aerosol motion information within the detection space. The photoelectric signal processing and wind speed vector inversion within the forward-facing lidar are well-known technologies in this field and will not be elaborated upon here. The coordinate compensation module defines the acquired point cloud analysis results as the raw spatial wind field point cloud data matrix. This matrix contains the relative three-dimensional coordinates of discrete points in space, as well as the original wind speed and direction vectors corresponding to each point.

[0092] S105, the coordinate compensation module constructs an instantaneous homogeneous coordinate transformation matrix based on the transient translational displacement and transient rotational deviation angle stored in the memory register. The structural vibration of the cabin causes the measurement reference coordinate system of the forward-facing lidar to deviate from the static absolute coordinate system. The coordinate compensation module uses the transient pitch and rotational deviation angle... Build around Pitch and rotation matrix of the axis Based on the transient yaw rotation deviation angle Build around Yaw rotation matrix of the axis The two are then multiplied to generate a combined rotation matrix. Simultaneously, the coordinate compensation module extracts the lateral translation displacement. Vertical translational displacement Construct spatial translation column vectors The coordinate compensation module sets the longitudinal translation displacement and roll angle deviation of the cabin to zero. The coordinate compensation module integrates the rotation matrix and the spatial translation column vector. Combination, generating dimension is The instantaneous homogeneous coordinate transformation matrix. The expression for the instantaneous homogeneous coordinate transformation matrix is:

[0093] ;

[0094] In the formula, Represents the instantaneous homogeneous coordinate transformation matrix; Indicates circling The yaw rotation matrix of the axis; Indicates circling The pitch and rotation matrix of the axis; Represents a column vector of spatial translation; The dimension is The zero row vector; Indicates the current calculation time.

[0095] S106, the coordinate compensation module performs dynamic coordinate remapping and velocity vector compensation calculations. The coordinate compensation module obtains the lateral and vertical translational velocities and pitch angular velocity through a single integration of the acceleration. and yaw rate Vector synthesis is performed. During this process, the coordinate compensation module, in conjunction with the relative spatial offset vector obtained in step S101, calculates the linear velocity compensation term of the nacelle top boom arm caused by the angular velocity, and superimposes it with the translational velocity vector to calculate the line-of-sight projection velocity components of the forward-facing lidar in each detection beam direction. The coordinate compensation module obtains the linear velocity compensation term from the original spatial wind field point cloud data matrix. The absolute wind speed vector data is obtained by subtracting the line-of-sight projected velocity component from the original wind speed and wind direction vectors.

[0096] The coordinate compensation module extracts the original spatial wind field point cloud data matrix. The column vector containing the coordinates of the probe points is expanded into homogeneous coordinate form to generate the original spatial position coordinate matrix. The coordinate compensation module utilizes the instantaneous homogeneous coordinate transformation matrix. Left multiply by the original spatial position coordinate matrix The position data in the vibration coordinate system is transformed to the reference static coordinate system to reduce the spatial field-of-view error introduced by the nacelle structural vibration. The remapping calculation formula is as follows:

[0097] ;

[0098] In the formula, This represents the actual spatial coordinate matrix generated after coordinate remapping; Represents the instantaneous homogeneous coordinate transformation matrix; This represents the original spatial position coordinate matrix expanded into homogeneous form.

[0099] The coordinate compensation module is based on the real spatial position coordinate matrix. Update data nodes with absolute wind speed vector data to generate a real wind field data matrix. The coordinate compensation module will use the actual wind field data matrix. The data is sent to the interferometric prediction module within the main control computing unit to provide a feedforward data source for spatiotemporal voxel reconstruction.

[0100] The specific execution steps for real-time spectrum extraction of the structure's natural frequency in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0101] S107, the coordinate compensation module in the main control computing unit extracts the lateral acceleration sequence data collected by the nacelle inertial measurement unit. During operation, the natural frequencies of the tower and drive train drift due to ambient temperature and aerodynamic loads. To avoid structural resonance caused by wake control commands, the coordinate compensation module sets a data buffer time window and slices the lateral acceleration sequence data according to the data buffer time window. The length of the data buffer time window is set according to the reciprocal of the preset frequency resolution. The coordinate compensation module multiplies the truncated lateral acceleration sequence data by a Hanning window function. For the selection of the window function and data smoothing, those skilled in the art can use the Hanning window to suppress spectral leakage; the data windowing method is a well-known technique in the field and will not be described in detail here.

[0102] S108, the coordinate compensation module performs a discrete Fourier transform operation on the windowed lateral acceleration sequence data, converting the time-domain signal into a frequency-domain signal, calculating the vibration amplitude of the signal at each discrete frequency point, and generating an acceleration frequency-domain amplitude spectrum.

[0103] S109, the coordinate compensation module reads the preset first-order lateral a priori frequency range of the tower. This a priori frequency range is determined based on the theoretical modal analysis parameters of the wind turbine generator set and is located in the frequency band of 0.1Hz to 0.5Hz. The vibration signal of the wind turbine generator set nacelle is a mixture of the fundamental frequency of the rotor rotation, the passing frequency of the blades, and high-frequency vibration noise from the transmission chain. The coordinate compensation module performs bandpass truncation on the acceleration frequency domain amplitude spectrum based on this a priori frequency range, retaining the effective frequency band sequence falling within the a priori frequency range and eliminating out-of-band interference frequency components.

[0104] S110, the coordinate compensation module performs a local maximum search in the effective frequency band sequence, comparing the vibration amplitudes of adjacent discrete frequency points to locate the frequency index corresponding to the point with the maximum vibration amplitude. The coordinate compensation module uses this frequency index to calculate the current structural natural frequency of the wind turbine generator. The calculation formula is as follows:

[0105] ;

[0106] In the formula, Indicates the extracted real-time structural intrinsic frequency; This indicates the frequency index corresponding to the point of maximum vibration amplitude within the effective frequency band; This indicates the sampling frequency of the lateral acceleration sequence data; This indicates the total number of sampling points within the data buffer time window. The coordinate compensation module will calculate the structure's natural frequency. The instruction synthesis module within the main control computing unit stores and sends the instruction to provide reference parameters for the interference comparison of the safety excitation frequency in subsequent pitch commands.

[0107] The specific execution steps of the three-dimensional wind field data spatial discretization processing in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0108] S201, the wind point cloud acquired by the forward-looking lidar exhibits an unstructured scanning distribution in space. The interferometric prediction module in the main control computing unit receives the actual wind field data matrix sent by the coordinate compensation module. The interferometric prediction module reads the maximum detection range and scanning field of view of the forward-facing lidar, and combines this with the current physical location of the wind turbine generator to define the spatial calculation boundary of the three-dimensional flow field. The spatial calculation boundary includes the longitudinal depth range, lateral width range, and vertical height range in front of the nacelle.

[0109] In S202, the interferometric prediction module performs spatial mesh generation on the defined 3D flow field, establishing a 3D voxel set. The module sets the spatial mesh resolution parameters based on the wind turbine rotor diameter, setting the longitudinal, lateral, and vertical resolutions to 5%–10% of the rotor diameter. Based on the spatial calculation boundary and the spatial mesh resolution parameters, the module divides the continuous 3D flow field into discrete cuboid meshes, defining each cuboid mesh as a spatial voxel. The module assigns each spatial voxel an independent 3D spatial index. And the corresponding geometric center coordinates.

[0110] S203, the interferometric prediction module will use the real wind field data matrix Discrete detection points are mapped to corresponding spatial voxels, and data aggregation operations are performed. The interferometric prediction module retrieves the three-dimensional coordinates of each detection point, assigns them to their corresponding spatial voxels, and counts those falling into the same spatial voxel. The number of effective detection points within the voxel. When the number of effective detection points is greater than zero, the interferometric prediction module calculates the arithmetic mean of the absolute wind speed values ​​of all detection points within the voxel, generating the voxel's value at the current calculation time. Characteristic wind speed Simultaneously, the interferometric prediction module calculates the arithmetic mean of the true wind direction deviation angles of all detection points within the voxel, generating the voxel's position at the current calculation time. Characteristic wind direction deviation .

[0111] S204, the interferometric prediction module performs empty voxel data interpolation and filling operations. When the number of valid detection points within a spatial voxel is zero, the interferometric prediction module determines that the voxel is an empty voxel. The interferometric prediction module sets a spatial interpolation search radius, extracts voxels adjacent to the empty voxel within the spatial interpolation search radius that contain valid data, and calculates the characteristic wind speed and characteristic wind direction deviations of the empty voxel using a spatial inverse distance weighted interpolation algorithm. If no voxel containing valid data is found within the spatial interpolation search radius, the interferometric prediction module extracts the three-dimensional spatial index of that voxel from the previous calculation time. The corresponding historical characteristic wind speed and historical characteristic wind direction deviations are used as the filling data for the empty voxel at the current calculation time. For the application of the spatial inverse distance weighted interpolation algorithm, those skilled in the art can implement it using a general three-dimensional mesh interpolation model. Its spatial data interpolation principle is a well-known technology in this field and will not be elaborated here.

[0112] S205, the interferometry prediction module will calculate the current time. As timestamps, data structures are bound to all spatial voxels, generating a spatiotemporal voxel set marked with spatial coordinates and timestamps. The interferometric prediction module continuously performs partitioning and aggregation operations within a control cycle that progresses over time, storing the continuously generated spatiotemporal voxel sets in a memory queue to construct a three-dimensional wind field evolution matrix in the spatiotemporal dimension. The interferometric prediction module outputs the three-dimensional wind field evolution matrix, providing a structured flow field data source for subsequent time-transfer calculations.

[0113] The specific steps for calculating the transit time under the Taylor frozen turbulence assumption in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0114] S206, the interferometric prediction module in the main control computing unit extracts the characteristic wind speed, characteristic wind direction deviation, and spatial coordinates of each spatial voxel in the three-dimensional wind field evolution matrix. Wind turbine wake control needs to be based on wind field evolution prediction. The interferometric prediction module introduces the Taylor frozen turbulence assumption, assuming that the spatial morphology of the aerodynamic structure within the flow field remains unchanged as it propagates towards the rotor plane, and that it convects downstream at a local characteristic wind speed. The interferometric prediction module obtains the relative geometric offset parameters between the forward-facing lidar installation position and the center of the rotor plane. Based on the geometric center coordinates of the spatial voxels, the interferometric prediction module performs coordinate translation compensation using these relative geometric offset parameters to calculate the absolute longitudinal physical distance from each spatial voxel to the wind turbine rotor plane.

[0115] S207, the interferometric prediction module calculates the longitudinal wind speed component of each spatial voxel in the direction normal to the wind turbine plane. The interferometric prediction module calculates the longitudinal wind speed component perpendicular to the wind turbine plane by multiplying the characteristic wind speed by the cosine of the characteristic wind direction deviation angle. To avoid anomalies caused by zero or negative denominators in subsequent time division operations, the interferometric prediction module sets a minimum effective longitudinal wind speed threshold. This minimum effective longitudinal wind speed threshold is set based on the cut-in wind speed parameters of the wind turbine generator. When the calculated longitudinal wind speed component is less than this threshold, the interferometric prediction module updates the longitudinal wind speed component value of the spatial voxel to this minimum effective longitudinal wind speed threshold.

[0116] S208, the interferometric prediction module performs the transit time calculation for the spatial voxels. The interferometric prediction module divides the absolute longitudinal physical distance from each spatial voxel to the wind turbine plane by the corresponding longitudinal wind speed component to calculate the transit time of the wind field characteristics contained in that spatial voxel to the wind turbine plane.

[0117] S209, the interferometric prediction module calculates the predicted arrival time of the wind field characteristics for each spatial voxel. The module adds the timestamp of the current calculation time to the calculated transit time to generate the predicted arrival time for that spatial voxel. The module then uses the transit time and predicted arrival time as attribute labels for the time dimension and binds them to the data structure of the corresponding spatial voxel. The interferometric prediction module completes the transit time mapping for spatial voxels and outputs a leading wind field sequence with predicted arrival time markers, providing a time alignment reference for subsequent calculations of asymmetric aerodynamic loads on the wind turbine plane.

[0118] The specific execution steps for determining the wake cooperative control trigger threshold in the wind turbine cluster wake control method provided by this invention are as follows:

[0119] S210, the interferometric prediction module in the main control computing unit extracts the predicted equivalent wind speed sequence and the predicted wind direction sequence based on the leading wind field sequence with predicted arrival time markers. The interferometric prediction module performs time-aligned clustering of spatial voxels with the same time label according to the predicted arrival time. The module extracts the characteristic wind speed of each aligned spatial voxel and calculates a predicted equivalent wind speed sequence for the wind turbine's swept surface by performing area-weighted averaging based on the projected area of ​​the spatial voxel on the wind turbine plane. Using the same area-weighted averaging logic, the module calculates and generates a predicted wind direction deviation sequence. The module calculates the environmental turbulence intensity based on the predicted equivalent wind speed sequence. The module calculates the standard deviation of the predicted equivalent wind speed sequence within a set time window and divides this standard deviation by the average value of the predicted equivalent wind speed within that time window to obtain the environmental turbulence intensity. To prevent division by zero anomalies, the module sets the environmental turbulence intensity value to zero when the average value of the predicted equivalent wind speed is less than a preset lower limit.

[0120] S211, the interference prediction module performs the operating wind speed range determination. The wake-coordinated control strategy is applied to the partial load operation condition of the wind turbine generator set when it does not reach its rated output. The interference prediction module extracts the cut-in wind speed parameters and the rated wind speed parameters of the wind turbine generator set, and compares the average value of the predicted equivalent wind speed with the above parameters. To avoid the control state switching caused by fluctuations in the predicted equivalent wind speed near the parameter boundaries, the interference prediction module sets a hysteresis interval at the boundary between the cut-in wind speed and the rated wind speed. For example, the lower limit of the activation wind speed is set to the cut-in wind speed plus a preset tolerance, and the upper limit of the activation wind speed is the rated wind speed minus the preset tolerance. When the average value of the predicted equivalent wind speed remains within the above-mentioned wind speed operating range with hysteresis characteristics for a set period of time, the interference prediction module determines that the wind speed activation condition is met.

[0121] S212, the interferometry prediction module performs wake interference sector determination. The interferometry prediction module reads the spatial coordinates of downstream adjacent wind turbines stored in the wind field topology database, and calculates the relative azimuth angle of the line connecting the two turbines by combining them with the current spatial coordinates of the wind turbine. The interferometry prediction module extracts the current absolute yaw angle of the nacelle, adds the average deviation angle of the predicted wind direction deviation sequence to the absolute yaw angle, and calculates the predicted absolute wind direction angle of the wind field in the absolute coordinate system. The interferometry prediction module estimates the wake spread angle based on the environmental turbulence intensity and the engineering wake model, and widens the wake spread angle to both sides with the relative azimuth angle as the center, generating a wake interference sector. The interferometry prediction module determines whether the predicted absolute wind direction angle is within the wake interference sector. When the predicted absolute wind direction angle is within the wake interference sector, the interferometry prediction module determines that the wind direction interference condition is met. For the estimation of the wake spread angle, those skilled in the art can use a linear wake spread model combined with the thrust coefficient for calculation, which is a well-known technique in the field and will not be elaborated further here.

[0122] S213, the interferometry prediction module performs turbulence intensity threshold determination. When environmental turbulence intensity is high, momentum exchange between the wind turbine wake and the external free flow increases, the wake velocity deficit recovers more quickly, aerodynamic interference to downstream units decreases, and the gain for wake control activation decreases accordingly. The interferometry prediction module reads a preset turbulence intensity suppression threshold and determines whether the calculated environmental turbulence intensity is less than this threshold. When the environmental turbulence intensity is less than this threshold, the interferometry prediction module determines that the turbulence triggering condition is met. This turbulence intensity suppression threshold is calibrated based on historical statistical data from the wind farm's site anemometer towers, with a value range set between 10% and 14%.

[0123] S214, the interferometry prediction module generates a mode trigger command by integrating the above-mentioned judgment conditions. The interferometry prediction module performs a logical AND operation on the wind speed activation condition, wind direction interference condition, and turbulence trigger condition. When all three conditions are met simultaneously, the interferometry prediction module sets the wake coordination control trigger flag to an active state. The interferometry prediction module sends the activated wake coordination control trigger flag to the instruction synthesis module within the main control computing unit to trigger subsequent asymmetric independent pitch control logic. When any judgment condition is not met, the interferometry prediction module keeps the wake coordination control trigger flag in a dormant state.

[0124] If the wake coordination control trigger flag switches from active to dormant, the instruction synthesis module triggers the exit attenuation logic. Within the preset attenuation time window, the additional pitch angle instruction of each blade is linearly reduced to zero at a set rate of change. Then, the wind turbine generator resumes operation according to the conventional maximum power point tracking and unified pitch strategy to avoid mechanical impact on the main shaft and tower caused by asymmetric aerodynamic changes.

[0125] See attached document Figure 3 In the wind turbine cluster wake control method based on radar wind measurement provided by this invention, the specific execution steps of frequency domain conversion of the wind direction deviation sequence are as follows:

[0126] S301, when the wake coordination control trigger flag is active, the frequency decoupling module in the main control computing unit reads the predicted wind direction deviation sequence with predicted arrival time markers output by the interferometric prediction module. The frequency decoupling module sets a sliding data time window and slices the predicted wind direction deviation sequence according to this sliding data time window. The frequency decoupling module sets the length of the data time window to the reciprocal of the preset frequency resolution and sets the overlap rate of adjacent sliding data time windows to 50% to 75% to retain effective information of the time domain sequence boundaries. The frequency decoupling module performs mean removal and linear trend elimination processing on the truncated predicted wind direction deviation sequence, removing the DC component and slowly drifting component in the time domain sequence to generate a zero-mean stationary sequence.

[0127] S302, the frequency decoupling module multiplies the zero-mean stationary sequence by a window function. The frequency decoupling module utilizes the attenuation characteristics of the window function to smooth the data boundaries at both ends of the sequence, thereby suppressing spectral leakage caused by aperiodic truncation in subsequent frequency domain transformation operations. For the specific selection and application of the window function, those skilled in the art can use a Hamming window or a Hanning window; the method of windowing time-domain sequences is well-known in the field and will not be elaborated further here.

[0128] S303, the frequency decoupling module performs a Discrete Fourier Transform (DFT) on the windowed zero-mean stationary sequence. The DFT module converts the time-domain wind direction data to the frequency domain, calculating the complex results of the signal at each discrete frequency point. Based on the real and imaginary parts of the complex results, the DFT module calculates the magnitude and phase angle of each frequency component, generating the wind direction frequency domain amplitude spectrum and wind direction frequency domain phase spectrum.

[0129] S304, the frequency decoupling module performs characteristic frequency extraction in the wind direction frequency domain amplitude spectrum. Due to the mechanical response delay and physical bandwidth limitations of the wind turbine pitch actuator, high-frequency pitch commands cannot be executed. The frequency decoupling module sets an effective low-frequency range based on the physical bandwidth of the pitch system, which is located in the frequency band of 0.01Hz to 0.2Hz. Within this effective low-frequency range, the frequency decoupling module searches for local maxima in the wind direction frequency domain amplitude spectrum and extracts the core frequency component. If no local maxima exist within this effective low-frequency range, the frequency decoupling module extracts the frequency point with the largest absolute amplitude value within this range as the core frequency component to avoid optimization dead zones. The frequency decoupling module simultaneously extracts the amplitude and initial phase information corresponding to this core frequency component in the wind direction frequency domain amplitude and phase spectra. The frequency decoupling module stores the core frequency component, amplitude, and initial phase information in system memory to provide frequency domain excitation reference parameters for subsequent asymmetric pitch control.

[0130] The specific execution steps for separating and reconstructing low-frequency yaw control commands in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0131] S305, the frequency decoupling module in the main control computing unit reads the original, un-mean-removed predicted wind direction deviation sequence. The yaw system of the wind turbine generator has a large mechanical inertia and is mainly used to track low-frequency wind direction changes. The frequency decoupling module sets the low-pass filter cutoff frequency based on the physical response capability of the yaw system. This low-pass filter cutoff frequency is set to 0.005Hz~0.01Hz.

[0132] S306, the frequency decoupling module calls a digital low-pass filter to perform filtering operations on the predicted wind direction deviation sequence, removing high-frequency fluctuation components and separating to obtain the low-frequency wind direction deviation sequence. For the application of the digital low-pass filter, those skilled in the art can use a second-order Butterworth low-pass filter; its digital filtering algorithm is well-known in the field and will not be elaborated upon here.

[0133] S307, the frequency decoupling module extracts the arithmetic mean of the low-frequency wind direction deviation sequence within a set time window and calculates the equivalent yaw error angle. To avoid mechanical fatigue caused by repeated start-stop cycles of the yaw gear, the frequency decoupling module sets a yaw dead zone threshold and a yaw trigger delay parameter. The yaw dead zone threshold is set to 6°–8° according to the operating procedures of the wind turbine generator set, and the yaw trigger delay parameter is set to 30s–60s.

[0134] S308, the frequency decoupling module executes the yaw trigger logic determination. When the absolute value of the equivalent yaw error angle is less than or equal to the yaw dead zone threshold, or greater than the dead zone threshold but the duration does not reach the yaw trigger delay parameter, the frequency decoupling module maintains the current cabin position and does not generate a new yaw action command. When the absolute value of the equivalent yaw error angle is greater than the yaw dead zone threshold, and the duration of this state reaches the yaw trigger delay parameter, the frequency decoupling module determines that the yaw adjustment condition is met.

[0135] S309, when the yaw adjustment condition is met, the frequency decoupling module adds the equivalent yaw error angle to the current absolute yaw angle of the cabin to calculate and obtain the target yaw angle.

[0136] In S310, the frequency decoupling module encapsulates the calculated target yaw angle into a yaw control command. This command is then sent to the yaw actuator of the wind turbine to drive the nacelle to align with the low-frequency trend wind direction. This process decouples the low-frequency wind direction changes in the wind field from the yaw system's response, combining this with the asymmetric pitch system's independent response to high-frequency wind direction fluctuations, thus achieving physical decoupling of the control frequency band.

[0137] In the wind turbine cluster wake control method based on radar wind measurement provided by this invention, the specific execution steps of high-frequency lateral thrust demand aerodynamic mapping are as follows:

[0138] In step S311, the frequency decoupling module in the main control computing unit reads the core frequency components, amplitude, and initial phase information stored in the system memory. The module extracts the spatial coordinates of the downstream adjacent wind turbine and the geometric boundary of the wake interference sector. It determines whether the predicted absolute wind direction angle is located in the left or right half of the wake interference sector and selects the sector boundary closest to the current predicted absolute wind direction angle as the target escape boundary. The module calculates the angular deviation between the predicted absolute wind direction angle and this target escape boundary, adds a preset deflection safety margin angle to this deviation, and generates the target wake deflection angle required to deflect the wake center to the non-interference region. This preset deflection safety margin angle is set to 2°–3° based on the wind direction measurement error range. This process transforms the geometric anti-interference requirements into target parameters for the wind turbine's planar motion.

[0139] S312, the frequency decoupling module performs analytical calculations of the lateral thrust coefficient. To achieve wake deflection, asymmetric pitch control periodically adjusts the local pitch angle of each blade, altering the aerodynamic lift and drag distribution in the rotor plane, synthesizing a net lateral force in the rotor plane, thereby forcing a shift in the overall trajectory of the downstream wake. Based on the average value of the predicted equivalent wind speed, the frequency decoupling module obtains the axial thrust coefficient under the current operating condition by querying the preset aerodynamic parameter table of the wind turbine. The frequency decoupling module multiplies the wake sensitivity coefficient, the target wake deflection angle, and the axial thrust coefficient to calculate the lateral thrust coefficient required by the rotor plane. This wake sensitivity coefficient is set to a range of 0.5–1.5 based on aerodynamic theory. The query of the aerodynamic parameter table can be implemented using a one-dimensional linear interpolation algorithm, and its table lookup operation is a well-known technique in the field, which will not be elaborated here.

[0140] S313, the frequency decoupling module calculates the absolute lateral thrust amplitude. The module sequentially multiplies the lateral thrust coefficient, air density, pi, the square of the rotor radius, and the square of the predicted equivalent wind speed average of the rotor swept surface, then divides the product by two to calculate the required high-frequency lateral thrust amplitude for the rotor. Air density is measured by environmental sensors located outside the nacelle. To prevent lateral thrust from causing fatigue overload on the rotor shaft, the frequency decoupling module sets a safe lateral thrust limit. This limit is calculated based on the maximum allowable bending moment of the wind turbine's main shaft. When the calculated high-frequency lateral thrust amplitude exceeds this safe limit, the module sets the high-frequency lateral thrust amplitude to the safe limit.

[0141] S314, the frequency decoupling module performs the reconstruction of the high-frequency lateral thrust time-domain signal. The frequency decoupling module uses the high-frequency lateral thrust amplitude after safety limiting as the amplitude parameter, the core frequency component as the frequency parameter, and combines it with the initial phase information to reconstruct and generate the time-varying target lateral thrust time-domain command. The calculation formula for the target lateral thrust time-domain command is:

[0142] ;

[0143] In the formula, Indicates the calculation time The target lateral thrust time-domain command; This indicates the amplitude of the high-frequency lateral thrust after safety limiting processing; This represents the core frequency component extracted during frequency domain transformation; This represents the time variable, i.e., the current calculation time. This indicates the initial phase information corresponding to the core frequency component. The target lateral thrust time-domain command maps high-frequency wind field fluctuation parameters to a dynamic aerodynamic load requirement benchmark for the wind turbine plane, providing a tracking target for physical quantity alignment for downstream independent pitch actuators.

[0144] The specific execution steps for constructing the three-dimensional safety load envelope boundary in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0145] S401, the load constraint module in the main control computing unit reads the real-time predicted equivalent wind speed, rotor speed, and reference pitch angle of the wind turbine generator set, and simultaneously acquires the flapping moment data and yaw moment data of each blade collected in real time by the blade root load sensor. Under unified pitch control, affected by atmospheric wind shear and tower shadow effect, the rotor rotation plane is subjected to uneven background airflow, generating background load. The load constraint module calculates the reference axial thrust currently borne by the wind turbine generator set based on aerodynamic theory. At the same time, the load constraint module uses Coleman coordinate transformation (dq transformation) to project the flapping moment data and yaw moment data of each blade rotation coordinate system to a fixed nacelle coordinate system. After data fusion and low-pass filtering, the module calculates and obtains the pitch reference moment and yaw reference moment currently actually borne by the wind turbine generator set. For the theoretical calculation of the reference axial thrust and the spatial coordinate transformation of the blade root moment, those skilled in the art can use the blade element momentum theory model combined with the multi-blade coordinate system transformation matrix to achieve this. The calculation method is a well-known technology in the field and will not be described in detail here.

[0146] S402, the load constraint module reads the mechanical limit parameters of the wind turbine generator's main shaft. These mechanical limit parameters include the main shaft's ultimate axial thrust and ultimate overturning moment. The load constraint module multiplies the ultimate axial thrust and ultimate overturning moment by a preset safety derating factor to calculate the axial thrust protection threshold and total bending moment protection threshold, respectively. This preset safety derating factor is dynamically adjusted based on the environmental turbulence intensity of the wind turbine generator's operating conditions, with a value range of 0.8 to 0.95. The greater the environmental turbulence intensity, the smaller the safety derating factor, in order to reserve a mechanical safety margin.

[0147] S403, the load constraint module performs dynamic load margin calculation. The load constraint module subtracts the reference axial thrust from the axial thrust protection threshold to calculate the dynamic axial thrust margin. Asymmetric independent pitch changes the aerodynamic distribution of each sector on the rotor swept surface, generating additional pitch and yaw moments at the rotor center. The load constraint module adds the square of the pitch reference moment and the square of the yaw reference moment and then takes the square root to calculate the current composite reference moment. The load constraint module subtracts this composite reference moment from the total moment protection threshold to obtain the total moment dynamic margin. To ensure that the composite moment on the main bearing does not exceed the limit under the superposition of additional moments caused by independent pitch, the load constraint module divides the total moment dynamic margin by the square root of two as the single-axis limit, and calculates the pitch and yaw moment dynamic margins respectively. To avoid algorithmic dead zones, when any dynamic margin value calculated by the load constraint module is less than zero, the load constraint module forcibly sets that dynamic margin to zero to prevent the continued superposition of asymmetric load commands under gust conditions.

[0148] S404, the load constraint module constructs a three-dimensional safe load envelope boundary. Using axial thrust, pitch moment, and yaw moment as orthogonal coordinate axes in three-dimensional space, the load constraint module uses the dynamic margins of axial thrust, pitch moment, and yaw moment as the maximum envelope amplitudes of the corresponding coordinate axes to construct a three-dimensional safe load envelope space. The load constraint module stores the boundary coordinate system parameters of this three-dimensional safe load envelope space in system memory for boundary limiting constraints on the synthetic aerodynamic forces generated by subsequent independent pitch control commands, preventing mechanical structural fatigue damage caused by wake deflection control in wind turbine generators.

[0149] The specific execution steps for predicting aeroelastic additional loads in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0150] In S405, the load constraint module in the main control computing unit reads the target lateral thrust time-domain command and simultaneously extracts the current predicted equivalent wind speed, rotor speed, and reference pitch angle of the wind turbine generator. When performing asymmetric pitch control to generate lateral thrust, the periodic changes in the local pitch angle of each blade not only alter the lateral aerodynamic forces in the rotor plane but also cause changes in the airflow angle of attack of each blade section, resulting in unbalanced changes in the axial aerodynamic thrust of the blades in different orientations. Combined with the flexible deformation effect of the blades, a corresponding aerodynamic load is formed at the center of the main shaft, resulting in cross-coupling of aerodynamic loads on each coordinate axis. The load constraint module predicts the three-dimensional additional load induced at the center of the rotor by the target lateral thrust time-domain command based on the current operating parameters.

[0151] S406, the load constraint module calls the offline aeroelastic sensitivity matrix table pre-stored in the system memory. This offline aeroelastic sensitivity matrix table is generated by applying a step asymmetric pitch command to the wind turbine generator model in multibody dynamics simulation software and recording the load response data at each steady-state operating point. The load constraint module uses the predicted equivalent wind speed, rotor speed, and reference pitch angle as indexes to query the offline aeroelastic sensitivity matrix table through multidimensional linear interpolation to obtain the thrust cross-sensitivity coefficient, pitch moment cross-sensitivity coefficient, and yaw moment cross-sensitivity coefficient under the current operating condition. For the application of the multidimensional linear interpolation algorithm, those skilled in the art can use the table lookup calculation method to implement it. The interpolation operation is a well-known technology in the field and will not be described in detail here.

[0152] S407, the load constraint module multiplies the target lateral thrust time domain command with the thrust cross sensitivity coefficient, pitch moment cross sensitivity coefficient and yaw moment cross sensitivity coefficient respectively, and calculates the axial additional thrust, additional pitch moment and additional yaw moment at the current calculation time.

[0153] In step S408, the load constraint module invokes a digital low-pass filter to filter the aforementioned axial additional thrust, additional pitch moment, and additional yaw moment, eliminating high-frequency numerical noise introduced by model linearization calculations. The cutoff frequency of this digital low-pass filter is set to 1.5Hz–2.5Hz based on the mechanical resonance frequency of the wind turbine's main shaft. To prevent the filtering process from filtering out transient overload peak values, which could lead to subsequent protection logic failure, the load constraint module synchronously extracts the absolute value envelope of the load in the time domain before filtering as an extreme value compensation component. This extreme value compensation component is then fused with the filtered data to generate a three-dimensional additional load prediction sequence. This sequence is provided to the subsequent load limiting stage to assess whether the total load after superimposing asymmetric pitch commands exceeds the safety boundary.

[0154] The specific execution steps of the lateral thrust saturation cutoff calculation in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0155] S409, the load constraint module in the main control computing unit extracts the three-dimensional additional load prediction sequence at the current calculation time. This sequence includes axial additional thrust, additional pitch moment, and additional yaw moment. The load constraint module performs absolute value calculations on the above three additional load parameters to obtain the predicted load increment amplitude borne by the wind turbine generator at the current calculation time.

[0156] In step S410, the load constraint module calls the boundary coordinate system parameters of the three-dimensional safe load envelope space stored in the system memory. These parameters include the dynamic margins of axial thrust, pitch moment, and yaw moment. The three-dimensional additional loads caused by asymmetric pitch are coupled along the spatial axes, and exceeding the limits in any dimension will lead to mechanical fatigue. Therefore, the minimum safe proportion among the multiple dimensions is extracted as a global constraint. To prevent the algorithm dead zone of division by zero overflow in subsequent division operations, the load constraint module adds a preset anti-zero offset constant to each of the three predicted load increment amplitudes (absolute values) obtained in step S409. This anti-zero offset constant is set to 0.001. The load constraint module divides the dynamic margin of axial thrust by the amplitude of the additional axial thrust after adding the zero offset constant to obtain the thrust margin ratio; it divides the dynamic margin of pitch moment by the amplitude of the additional pitch moment after adding the zero offset constant to obtain the pitch margin ratio; and it divides the dynamic margin of yaw moment by the amplitude of the additional yaw moment after adding the zero offset constant to obtain the yaw margin ratio. The load constraint module then compares the thrust margin ratio, pitch margin ratio, yaw margin ratio, and the constant 1.0, extracting the minimum value as the global cutoff coefficient. This comparison process, by comparing multiple safety constraints in parallel, selects the physical dimension with the smallest margin as the global limiting benchmark, ensuring the completeness of the cutoff action.

[0157] In step S411, the load constraint module multiplies the original target lateral thrust time-domain command with the global truncation coefficient to calculate the safe lateral thrust command. This safe lateral thrust command is an executable control command that has undergone multi-dimensional structural safety boundary limiting processing. Through this multiplication operation, the load constraint module proportionally scales the amplitude of the asymmetric pitch action, outputting the lateral aerodynamic force under load constraints while ensuring that no mechanical structural nodes of the wind turbine generator experience overload.

[0158] S412, the load constraint module executes the dead-zone detection logic for the saturation state. When the wind turbine is in a highly turbulent wind condition, the background load increases, causing a decrease in various dynamic margins. If the global cutoff coefficient calculated by the load constraint module is less than the saturation exit threshold, and the duration of this state reaches the saturation detection time, the load constraint module determines that the lateral thrust request enters the deep saturation zone. The saturation exit threshold is set to 0.1–0.2, and the saturation detection time is set to 3–5 seconds.

[0159] S413: When it is determined that the lateral thrust request has entered the deep saturation region, the load constraint module triggers the protection logic. The load constraint module forcibly sets the current safe lateral thrust command to zero and sends the deep saturation flag to the command synthesis module in the main control computing unit to suspend the reconstruction process of the lateral thrust time domain command, maintain the basic state operation of the unit, and avoid the control system frequently switching in and out at the saturation boundary, which would cause mechanical wear of the pitch actuator.

[0160] See attached document Figure 4 The specific execution steps for solving the local additional pitch angle command for each blade in the wind turbine group wake control method based on radar wind measurement provided by this invention are as follows:

[0161] In S501, the instruction synthesis module in the main control computing unit extracts the safe lateral thrust command and reads the real-time rotor azimuth angle signal measured by the encoder inside the nacelle. Since the wind turbine generator set is equipped with three evenly distributed blades, and the mechanical installation phase difference between adjacent blades is 120°, the instruction synthesis module calculates the absolute azimuth angle of each blade at the current calculation moment based on this mechanical installation phase difference. The calculation of the absolute azimuth angle of each blade can be achieved by an angle offset accumulation method, which is a well-known technique in the field and will not be elaborated upon here.

[0162] S502, the command synthesis module determines the thrust pitch angle sensitivity. Using the predicted equivalent wind speed and reference pitch angle as indexes, the module queries the aerodynamic sensitivity table in system memory via two-dimensional linear interpolation to obtain the lateral thrust pitch angle sensitivity coefficient under the current operating condition. This coefficient characterizes the lateral thrust increment caused by a unit local pitch angle change within the rotor plane. To avoid division-by-zero overflow caused by sensitivity approaching zero, the command synthesis module sets a minimum sensitivity lower limit threshold. When the absolute value of the lateral thrust pitch angle sensitivity coefficient obtained from the table is less than this lower limit threshold, the module clamps it to this threshold. This minimum sensitivity lower limit threshold is set to 0.1 kN / deg. The module divides the safe lateral thrust command by the clamped lateral thrust pitch angle sensitivity coefficient to calculate the total independent pitch amplitude at the current calculation moment. Through this division operation, the module transforms the mechanical requirement reference into a pitch angle control reference.

[0163] S503, the command synthesis module performs spatial azimuth projection allocation. Asymmetric independent pitch control changes the aerodynamic angle of attack of the blades as they rotate to the target spatial azimuth, synthesizing the target lateral force at the rotor center. Due to the spatial periodicity of the rotor's swept surface aerodynamic characteristics, the command synthesis module calculates the cosine value by adding the absolute azimuth angle of each blade to the phase compensation angle, and then multiplies this cosine value by the total independent pitch amplitude to obtain the local additional pitch angle command for each blade. This projection process causes the blades to generate the maximum pitch angle deviation when sweeping through the horizontal azimuth, thereby synthesizing the target lateral aerodynamic force. Because there is a time delay between the aerodynamic response mechanism of the wind turbine generator and the pitch actuator, the command synthesis module introduces a phase compensation angle to perform phase lead correction on the allocation weights, compensating for the lag in the pitch angle response of the blades when sweeping through the target azimuth. The phase compensation angle is calculated based on the product of the first-order inertial time constant of the pitch system and the current rotor speed, and the product result is converted into a radian or angular value of the corresponding dimension to ensure the consistency of physical dimensions when performing spatial orientation projection allocation.

[0164] S504, the command synthesis module performs action limiting and command output. The module limits the amplitude and rate of change of the local additional pitch angle command for each blade. To prevent bearing wear caused by pitch control, the amplitude limiting threshold is set to -3deg to 3deg, and the rate of change limiting threshold is set to 2deg / s to 4deg / s. The module adds the limited local additional pitch angle command to the current reference pitch angle to generate the absolute pitch command for each blade. To prevent the superimposed command from exceeding the physical stroke of the mechanical actuator, the module uses preset mechanical physical limit parameters to perform a secondary saturation truncation on the absolute pitch command. The absolute pitch command, after the above superposition and limiting processing, is sent to each independent pitch driver of the wind turbine to drive each blade to perform asymmetric pitch control, thereby achieving lateral deflection of the wake.

[0165] The specific execution steps for setting the wake excitation frequency based on the Strouhal number in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0166] In S505, the command synthesis module in the main control computing unit extracts the predicted equivalent wind speed and the wind turbine rotor diameter parameters. During the wake field evolution, periodic aerodynamic disturbances are applied at the rotor to excite the hydrodynamic instability characteristics of the wake shear layer, accelerating momentum exchange between the external free flow and the wake region, and promoting wake recovery. The command synthesis module solves for the excitation frequency based on this aerodynamic principle.

[0167] S506, the command synthesis module determines the Strouhal number under the current operating condition. The Strouhal number characterizes the dimensionless relationship between eddy shedding frequency and flow velocity and characteristic length in an unsteady flow field. Based on the environmental turbulence intensity provided by the feedforward wind radar, the command synthesis module obtains the corresponding optimal Strouhal number by querying the frequency mapping table in the system memory through one-dimensional linear interpolation. In low turbulence intensity environments, wake dissipation is slow, so a high dimensionless frequency is configured to excite large-scale eddies; in high turbulence intensity environments, environmental mixing is strong, so a low Strouhal number is configured to reduce the actuation loss of the pitch mechanism. The optimal Strouhal number is set to a range of 0.25–0.35. The construction of the mapping table between turbulence intensity and Strouhal number can be achieved by those skilled in the art using offline wind tunnel experiments or computational fluid dynamics simulation identification methods. This mapping identification is a well-known technique in the field and will not be elaborated upon here.

[0168] S507, the command synthesis module uses the optimal Strouhal number, predicted equivalent wind speed, and rotor diameter to calculate the initial wake excitation frequency. The module multiplies the optimal Strouhal number by the predicted equivalent wind speed to obtain a reference value for aerodynamic disturbance velocity; then, it divides this reference value by the rotor diameter of the wind turbine to obtain the current initial wake excitation frequency. Since the rotor diameter is a non-zero constant, this division operation is numerically stable and does not have a division-to-zero overflow dead zone. Through this calculation, the command synthesis module transforms dimensionless hydrodynamic empirical parameters into physical frequencies under the current wind speed conditions.

[0169] S508, the command synthesis module performs resonance avoidance and frequency band limiting processing on the initial wake excitation frequency. Since the wind turbine tower and blades have mechanical resonant frequencies, if the aerodynamic excitation frequency coincides with these resonant frequencies, the resulting periodic aerodynamic forces can cause structural resonance fatigue. The command synthesis module calls the system's preset first-order natural frequency forbidden zone parameters for the tower and the out-of-plane natural frequency forbidden zone parameters for the blades to determine whether the initial wake excitation frequency falls within these frequency forbidden zones. If it does, the command synthesis module clamps the frequency to the nearest edge value of the forbidden zone. Simultaneously, limited by the response performance of the pitch drive motor, the command synthesis module extracts the upper limit frequency of the physical response bandwidth of the pitch system. If the frequency after resonance avoidance processing exceeds this upper limit frequency, the command synthesis module truncates it to this upper limit frequency, generating a safe wake excitation frequency. This wake excitation frequency is directly output to the synthesis stage of the local additional pitch angle command, used to modulate the dynamic fluctuation period of the additional pitch angle of each blade, superimposing this excitation frequency during spatial azimuth projection allocation, thereby inducing wake vortex shedding.

[0170] The specific execution steps of frequency domain interferometry and adaptive frequency shift synthesis in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0171] S509, the instruction synthesis module in the main control computing unit reads the real-time rotor speed of the wind turbine generator. During rotation, the wind turbine is affected by wind shear and tower shadow effects, generating periodic aerodynamic loads. The instruction synthesis module converts the real-time rotor speed into rotational frequency and calculates its first and third harmonic values ​​to obtain the background load frequency distribution under the current operating conditions. The conversion from speed to harmonics can be calculated using physical conversion formulas by those skilled in the art; such calculations are well-known in the field and will not be elaborated upon here.

[0172] S510, the command synthesis module performs frequency domain interferometry comparison. If the wake excitation frequency is close to the background load frequency, they will produce a beat frequency effect in the time domain, causing low-frequency oscillations and superposition amplification of the aerodynamic load amplitude, leading to mechanical fatigue. The command synthesis module uses the calculated first and third harmonics as centers, expanding outwards to both sides by a preset interference bandwidth ratio to construct the first and third harmonic interference exclusion zones. This interference bandwidth ratio is set to 10%–15%. The command synthesis module compares the wake excitation frequency generated in the previous step with the aforementioned interference exclusion zones to determine whether the wake excitation frequency falls within the interference exclusion zone.

[0173] S511: When the wake excitation frequency does not fall within any interference forbidden zone, the instruction synthesis module determines that there is no interference in the frequency domain and directly outputs the wake excitation frequency as a candidate frequency. When the wake excitation frequency falls within the first harmonic interference forbidden zone or the third harmonic interference forbidden zone, the instruction synthesis module determines that there is a risk of interference in the frequency domain and triggers the adaptive frequency shift logic.

[0174] In S512, during the adaptive frequency shift process, the command synthesis module calculates the absolute distance between the wake excitation frequency and the upper and lower boundaries of the interference exclusion zone. To reduce the deviation from the wake excitation effect, the command synthesis module selects the boundary value with the closer absolute distance as the frequency shift target. If the absolute distance between the wake excitation frequency and the upper and lower boundaries is equal, the command synthesis module selects the lower frequency boundary value as the frequency shift target to reduce the operating frequency of the subsequent pitch actuator. To avoid structural risks introduced by the frequency shift operation, the command synthesis module calls the aforementioned mechanical resonant frequency exclusion zone parameters for secondary verification. If the frequency shift target does not fall into the mechanical resonant frequency exclusion zone, the command synthesis module confirms it as a candidate frequency; if the frequency shift target falls into the mechanical resonant frequency exclusion zone, the command synthesis module selects the other boundary value of the interference exclusion zone for secondary verification. If the other boundary value also falls into the mechanical resonant frequency exclusion zone, it indicates that the interference exclusion zone and the mechanical resonant exclusion zone overlap under the current operating condition. The command synthesis module triggers the backoff protection mechanism, forcibly setting the excitation amplitude of the dynamic wake disturbance to zero, suspending the dynamic wake disturbance in the current calculation cycle, and ensuring the structural safety of the unit. Through this multi-level avoidance logic, the instruction synthesis module pushes the frequency away from the beat frequency region.

[0175] In step S513, the instruction synthesis module performs time-domain smoothing on the generated candidate frequencies. Changes in wind speed or fluctuations in engine speed can cause the interference exclusion zone to drift, leading to a step abrupt change in the candidate frequencies at the boundaries. The instruction synthesis module uses discrete first-order low-pass filtering logic to process the candidate frequencies. Specifically, it weights and sums the candidate frequencies of the current calculation cycle with the final wake excitation frequency of the previous calculation cycle using preset filtering coefficients to obtain the final wake excitation frequency of the current calculation cycle. The cutoff frequency corresponding to this first-order low-pass filtering logic is set to 0.1Hz to 0.5Hz. This filtering process avoids the electrical signal impact on the pitch actuator caused by frequency abrupt changes, ensuring the continuity of hardware operation. Subsequently, the instruction synthesis module generates a high-frequency micro-motion disturbance wave based on the final wake excitation frequency and assigns a preset excitation disturbance amplitude. The command synthesis module directly superimposes the high-frequency micro-motion disturbance wave onto the local additional pitch angle command of each blade obtained in step S503 in the form of algebraic addition, so that while the blade is performing spatial azimuth projection allocation, it will generate additional high-frequency aerodynamic micro-motion that excites wake vortex shedding.

[0176] The specific execution steps for monitoring the cumulative micro-motion stroke status of pitch control in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0177] S601, the wear suppression module in the main control computing unit reads the real-time pitch angle feedback signal of each blade measured by the pitch encoder inside the wind turbine hub. Under asymmetric independent pitch control, the blades need to perform periodic micro-motion movements. High-frequency, small-amplitude reciprocating motion can damage the lubricating film inside the bearing, accelerating raceway fretting wear. The wear suppression module extracts this real-time pitch angle feedback signal to quantitatively assess the mechanical load state of the pitch system. For noise reduction preprocessing of the pitch encoder signal, those skilled in the art can use a first-order low-pass filter method, which is a well-known technique in the field and will not be elaborated here.

[0178] S602, the wear suppression module calculates the sliding cumulative pitch stroke of each blade. The wear suppression module calculates the difference between the real-time pitch angle feedback signal at the current sampling time and the previous sampling time, and takes the absolute value of this difference to obtain the single-step pitch increment. Subsequently, the wear suppression module discretely accumulates the single-step pitch increments at each sampling time within a set sliding time window to obtain the sliding cumulative pitch stroke. The length of this sliding time window is set according to the thermal balance constant of the pitch system and the grease renewal cycle, and ranges from 60s to 300s.

[0179] S603, the wear suppression module determines the pitch micro-motion duty cycle. Since the pitch activity frequency varies under different operating conditions, simple cumulative stroke is insufficient to reflect the relative load level of the pitch hardware. The wear suppression module extracts the rated maximum pitch rate of the pitch actuator motor, multiplies this rated maximum pitch rate by the length of the sliding time window, and obtains the theoretical maximum pitch stroke within that time window. The wear suppression module divides the calculated sliding cumulative pitch stroke by this theoretical maximum pitch stroke to obtain the pitch micro-motion duty cycle for each blade. Since both the time length and the rated maximum pitch rate are non-zero constants, this division operation possesses numerical stability. This operation transforms the angular physical quantity into a dimensionless load index, used to characterize the current load level of the pitch system.

[0180] S604, the wear suppression module outputs status monitoring commands based on the pitch micro-motion duty cycle. The wear suppression module calls the preset pitch health threshold within the system and compares the pitch micro-motion duty cycle of each blade with this threshold. To avoid high-frequency switching of the status flag bit caused by duty cycle fluctuations at threshold critical points, the wear suppression module uses hysteresis comparison logic for judgment. Specifically, the preset pitch health threshold includes a trigger upper threshold and a recovery lower threshold. The trigger upper threshold is set to 0.25, and the recovery lower threshold is set to 0.15. When the pitch micro-motion duty cycle of any blade is greater than or equal to the trigger upper threshold, the wear suppression module determines that the pitch system is in a high-load wear state, outputs a fatigue degradation flag bit, and maintains this flag bit state; until the pitch micro-motion duty cycle of all blades falls back to below the recovery lower threshold, the wear suppression module determines that the pitch system has returned to the normal load range and outputs a normal operation flag bit. The generated status flags will be stored in the system memory and sent as a prerequisite to the subsequent grease redistribution monitoring logic to proactively improve the operating conditions of the pitch bearing and ensure the service life of the actuator.

[0181] In the wind turbine cluster wake control method based on radar wind measurement provided by this invention, the specific execution steps of the grease redistribution sweep pulse superposition trigger are as follows:

[0182] S605, the wear suppression module in the main control calculation unit reads the fatigue degradation flag output from the aforementioned steps. In continuous, small-amplitude independent pitch control, the grease in the contact area between the pitch bearing rollers and raceways is squeezed to both sides of the motion trajectory, resulting in a lean lubrication state at the contact surface and causing mechanical wear. Applying a large pitch angle can encourage the grease on both sides to re-cover the contact area. The wear suppression module records the cumulative activation time of the fatigue degradation flag. To avoid interruption of timing due to brief resets of the flag caused by operating condition fluctuations, the wear suppression module calculates the effective activation time using an integral accumulation method. When the cumulative activation time reaches a preset sweep trigger time threshold, the wear suppression module triggers the grease redistribution logic. This sweep trigger time threshold is set to 3600s–7200s.

[0183] S606, the wear suppression module generates a sweep pulse command. The wear suppression module sets the target amplitude and execution period of the sweep pulse. To avoid aerodynamic imbalance loads on the wind turbine rotor caused by large independent pitch movements, the wear suppression module configures the sweep pulse command to synchronize with a uniform pitch offset, meaning the command acts in the same direction and with equal amplitude on all three blades. The sweep pulse command adopts a half-sine wave shape, with a target amplitude set to 3-5 degrees and an execution period set to 10-20 seconds. For the discretization generation of the sweep pulse waveform, those skilled in the art can use basic signal generator algorithms; such discretization is a well-known technique in the field and will not be elaborated upon here.

[0184] S607, the wear suppression module performs boundary limiting and rate verification on the generated sweep pulse command. The wear suppression module extracts the unified pitch reference angle under the current operating conditions of the wind turbine generator. The wear suppression module calls the preset blade mechanical limit angle range and calculates the stroke margin for both positive and negative superimposed pulses. The wear suppression module selects the direction with a stroke margin greater than the target amplitude as the actual direction of action of the sweep pulse. If both bidirectional stroke margins are less than the target amplitude, the wear suppression module selects the side with the larger margin and truncates the target amplitude to the maximum allowable remaining stroke on that side. Simultaneously, the wear suppression module calculates the maximum command change rate of the sweep pulse command within the execution cycle. If this maximum command change rate exceeds the upper limit of the operating speed of the pitch actuator motor, the wear suppression module reduces the command change rate by extending the execution cycle. Through bidirectional selection and verification logic, the system's shutdown protection mechanism is avoided from being triggered by large-angle sweep actions, eliminating command execution dead zones caused by operating condition limitations.

[0185] S608, the wear suppression module superimposes the verified sweep pulse command onto the original independent pitch command of each blade and executes it. Through the large-angle motion triggered by this sweep pulse command, the contact area between the pitch bearing rollers and raceways can complete the grease redistribution process. The wear suppression module maintains this command output and enters the real-time execution status monitoring stage.

[0186] The specific execution steps for closed-loop state reset and normal command recovery in the wind turbine cluster wake control method based on radar wind measurement provided by this invention are as follows:

[0187] In S609, the wear suppression module in the main control computing unit monitors the execution progress of the sweep pulse command in real time. When the sweep pulse command's execution time reaches the set execution cycle, and the actual pitch angle of each blade falls back to the tolerance range of the unified pitch reference angle, the wear suppression module determines that the grease redistribution process has ended and generates a wake recovery coordinated control trigger flag. This tolerance range is set to 0.1deg to 0.3deg. To prevent program deadlock caused by the actual pitch angle failing to meet the tolerance condition for an extended period due to operating condition disturbances, the wear suppression module simultaneously introduces a timeout exit mechanism. If the sweep pulse command's execution time reaches 1.5 to 2 times the set execution cycle, the wear suppression module directly determines that the grease redistribution process has ended and generates a wake recovery coordinated control trigger flag.

[0188] S610, the controller in the main control computing unit executes state reset logic. During grease redistribution sweep, the independent pitch command output of wake control is suspended. If closed-loop control is directly restored, the algorithm's internal buffer and historical state values ​​will cause abrupt output changes due to accumulated deviations, triggering aerodynamic load shocks. After receiving the wake coordination control restoration trigger flag, the controller clears the relevant internal filtered historical values ​​or state accumulation terms to zero and synchronously overwrites the previous cycle's output historical value with the current unified pitch reference angle. This reset operation is used to suppress controller integral saturation and reduce command abrupt changes.

[0189] In S611, the instruction synthesis module in the main control computing unit employs a time-weighted function to ensure smooth transition of instructions. Since the pitch actuator is a large-inertia mechanical system, direct switching would cause mechanical oscillations. This time-weighted method allows independent pitch command components to be smoothly integrated. Within a preset recovery time window, the instruction synthesis module multiplies the local additional pitch angle command by a transition weight that increases over time, and then superimposes it onto the unified pitch reference angle. The formula for this smooth transition calculation is as follows:

[0190] ;

[0191] In the formula, Indicates the recovery phase The final pitch control command for the support blades; Number the blades, with values ​​of 1, 2, or 3; This indicates the unified pitch reference angle of the wind turbine generator at the current sampling time; This indicates the number of samples generated by the wake control algorithm at the current sampling time. Additional pitch angle command for the support blades; This indicates the transition weight.

[0192] S612, the instruction synthesis module calculates the aforementioned transition weights. The instruction synthesis module employs linear ramp logic, incrementing the transition weight from 0 to 1 within the recovery time window. Specifically, the calculation process is as follows: the instruction synthesis module calculates the time difference between the current calculation time and the start time of receiving the recovery wake co-control trigger flag, and divides this time difference by the preset recovery time window length to obtain the transition weight for the current calculation cycle. The recovery time window length is set to a range of 5s to 10s. To avoid weight out-of-bounds dead zones caused by program execution cycles, when the calculated time difference is greater than or equal to the recovery time window length, the instruction synthesis module uses amplitude limiting logic to adjust the transition weight. Fixed at 1. This linear climb process helps reduce system overshoot when independent pitch commands are engaged.

[0193] S613, when the transition weight reaches 1, the smooth transition phase ends. The instruction synthesis module switches the system state to the normal asymmetric independent pitch control mode. At this time, the wear suppression module clears the cumulative pitch stroke value recorded in the system memory to zero, resets the fatigue degradation flag to the normal operation flag, and synchronously clears the aforementioned cumulative activation timer. Subsequently, the wear suppression module restarts reading the real-time pitch angle feedback signal and enters the next round of acquisition and discrete accumulation calculation of the cumulative pitch stroke. The closed-loop system completes the state reset and enters the normal wake control and mechanical condition monitoring cycle.

[0194] Specific application examples:

[0195] Assume the current location is a wind farm, and the upstream wind turbine (WT1) is a 5MW unit with a rotor diameter of... The wheel hub height is 90m; directly behind downstream. There is another unit of the same model (WT2) at that location;

[0196] Current free-flow wind speed The rotor speed of the wind turbine generator is 12 rpm, and the ambient turbulence intensity is 12%.

[0197] The cabin attitude calculation and natural frequency extraction: The cabin IMU detects a continuous lateral acceleration signal; the coordinate compensation module sets a sliding time window. The transient lateral translational displacement of the cabin is calculated by integration. Yaw rotation deviation angle ;

[0198] Simultaneously, a Hanning window is applied to the lateral acceleration sequence and an FFT is performed to optimize within the prior interval of 0.1Hz to 0.5Hz, thereby extracting the current first-order natural frequency of the tower. ;

[0199] Forward radar detection space is divided into grids with a resolution of .

[0200] Based on the Taylor frozen turbulence assumption, the wind direction deviation voxel 200m in front of the radar, which is transmitted at a speed of 10m / s, is predicted to take 20s to reach the wind turbine plane.

[0201] The interferometry prediction module calculated that the wake center trajectory, which includes the 4° wind direction deviation, will cover 65% of the downstream WT2 swept area (exceeding the set 50% interferometry threshold), and the turbulence intensity is 12% < 14%, thus activating the wake cooperative control flag.

[0202] Perform FFT on the predicted wind direction deviation sequence;

[0203] Low-frequency separation: Slow wind drift below 0.01Hz (average 2°) is handled by the yaw servo system for slow wind counter-steering;

[0204] High-frequency separation: Extracting the core wind field fluctuation frequency To make the wake deflect from WT2, the target wake deflection angle needs to be calculated to be 6°.

[0205] The command synthesis module determines the target lateral force based on the required deflection angle. Based on well-known aerodynamic conversion logic (i.e., combining lateral thrust coefficient, air density, rotor sweep area, and predicted equivalent wind speed for conventional physical conversion), the system calculates that the current required lateral aerodynamic force is approximately 40.7 kN. Therefore, the original target lateral thrust time-domain command amplitude is directly set. ;

[0206] The load constraint module reads the current spindle mechanical limits: the ultimate overturning moment threshold is 3000kNm; the background synthetic reference moment under the current wind speed is 1500kNm, and the dynamic margin is 1500kNm;

[0207] Aeroelastic predictions show that applying a 40.7 kN lateral thrust will induce an additional pitching moment of approximately 800 kNm.

[0208] The calculated margin ratio is 1500 / 800 = 1.875 > 1, indicating that this thrust command is safe. The global cutoff factor is set to 1.0, and no weakening command is needed.

[0209] According to the table, the thrust pitch angle sensitivity is 20 kN / deg.

[0210] The total independent pitch amplitude is 40.7 / 20≈2.0°;

[0211] Wake excitation frequency calculation: Optimal Strouhal number obtained from a table. ;

[0212] Initial wake excitation frequency ;

[0213] Interferometric comparison: background load The tower frequency is 0.25Hz; 0.023Hz is far from the interference forbidden zone, so no frequency shift is required;

[0214] The single-blade local additional pitch angle command is ultimately expressed as: +Minor disturbances;

[0215] After 100 seconds of controlled operation (sliding time window), due to the high-frequency 2.0° reciprocating fine adjustment, the cumulative pitch stroke of a single blade reaches 280°;

[0216] The motor has a rated maximum pitch rate of 8° / s and a theoretical maximum stroke of 800°.

[0217] Calculate the pitch micro-motion duty cycle: 280 / 800 = 0.35;

[0218] If the value exceeds the trigger upper limit threshold (0.25), it is determined that the system has entered a high-load wear state. After the accumulated time reaches the set value (e.g., 3600s), the IPC command is paused.

[0219] Superimposed synchronous sweep pulse: A half-sine wave with an amplitude of 4° and a period of 15s is superimposed on a uniform reference angle to promote the redistribution of bearing grease;

[0220] After execution, the internal state is reset, and the weights are adjusted using an 8-second recovery time window. The independent pitch command is linearly increased from 0 to 1, and then smoothly recovered to 2.0°.

[0221] Experimental data and logic verification explanation:

[0222] See attached document Figure 5This step visually demonstrates the physical logic of the mid-frequency decoupling in step S30. The waveform data in Figure (a) shows that the system successfully decouples and separates the original wind direction deviation signal containing high-frequency fluctuations: slow, large-scale wind direction drift is extracted and handled by the yaw system, while rapid wind direction fluctuations are separated and converted into control inputs for the asymmetric pitch system. Combined with the accurately extracted core frequency in Figure (b), this invention successfully avoids the yaw mechanism from ineffectively following high-frequency gusts, providing a precise high-frequency aerodynamic reference for the dynamic lateral deflection of the wake.

[0223] See attached document Figure 6 This step corresponds to the saturation truncation calculation logic in step S40. The time-domain load curve in the figure clearly shows that under complex wind conditions, the unconstrained predicted additional load will exceed the horizontally set dynamic load safety boundary. The protection mechanism of this invention can identify these dangerous commands in real time and perform saturation truncation calculations on them using a global truncation coefficient. As shown by the solid line in the figure, the load commands exceeding the boundary are successfully flattened, and the final executed load is forcibly and precisely limited within the safety boundary, effectively preventing transient mechanical overload damage to the unit.

[0224] See attached document Figure 7 This section perfectly demonstrates the logic of monitoring the cumulative micro-motion state of pitch control and redistributing grease in step S60. Experimental recording curves show that during the conventional wake control phase, a single blade needs to continuously perform high-frequency, small-amplitude pitch movements (the dense micro-motion waveforms on both sides of the figure). When the algorithm detects that the micro-motion load reaches the target, it actively triggers a large-angle synchronous sweep command (the large-amplitude peak in the 80s-100s range of the figure). This intuitive large-angle pulse action successfully breaks the local micro-motion dead zone of the bearing raceway, prompting grease to re-cover the contact surface. Subsequently, the system smoothly returns to normal wake fine-tuning control, thus demonstrating from the execution action that this solution has the ability to actively avoid bearing dry friction.

[0225] See attached document Figure 8 This section provides a final summary of the overall system's effectiveness. The comparison of the data distribution in the bar chart clearly demonstrates the significant success of this invention:

[0226] In terms of power generation efficiency, the black bar of the present invention is close to the 5% mark, approximately 4.8%, and its gain is more than twice that of the traditional yaw control group.

[0227] Regarding mechanical losses, traditional yaw control methods, while achieving a slight increase in power generation, result in a significant rise in the corresponding gray column, approaching the 20% mark, leading to a surge in main shaft fatigue load of approximately 18.5%, at a high cost. In contrast, the solution of this invention, while significantly outperforming the power generation, only slightly exceeds the 5% mark in its gray column, successfully limiting the increase in main shaft fatigue load to an extremely low level of approximately 6.2%. This directly demonstrates that the solution of this invention achieves an excellent optimal solution between the two mutually constraining engineering objectives of improving the overall wind farm power generation efficiency and ensuring the safe operation of individual units.

Claims

1. A method for wind turbine cluster wake control based on radar wind measurements, characterized in that, Includes the following steps: Acquire real wind field data collected by radar and dynamically compensated, predict the interference state between the wake trajectory and the swept area of ​​the downstream unit, and generate wake cooperative control trigger flag. Based on the wake coordination control trigger flag, high-frequency wind direction change data is separated from the real wind field data, and the high-frequency wind direction change data is converted into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary. Based on the aforementioned safe lateral thrust command, a local additional pitch angle command is calculated, and combined with a safe wake excitation frequency that avoids the inherent frequency of the unit structure, a comprehensive pitch command is synthesized. During the execution of the integrated pitch control command, the cumulative micro-motion stroke integral value of the pitch control action is monitored. If the cumulative micro-motion stroke integral value is greater than the preset wear threshold, a synchronous sweep command is superimposed on the unified pitch reference angle of the unit's current operating condition to perform grease redistribution. After the execution is completed, the issuance of the integrated pitch control command is resumed.

2. The radar wind measurement based wind turbine cluster wake control method of claim 1, wherein, The radar collects and acquires dynamically compensated real wind field data, including: Perform time definite integral on the acceleration and angular velocity data output by the cabin inertial measurement unit to obtain transient translational displacement and transient rotational deviation angle; Using the transient translational displacement and transient rotational deviation angle, an instantaneous homogeneous coordinate transformation matrix is ​​constructed to remap the original spatial wind field point cloud data matrix collected by the forward lidar to the reference static coordinate system, thereby generating the real wind field data. The discrete Fourier transform is performed synchronously on the acceleration data to locate the point of maximum vibration amplitude in the effective frequency band sequence preserved by bandpass truncation, and the corresponding natural frequency of the unit structure is extracted.

3. The wind turbine cluster wake control method based on radar wind measurement according to claim 2, characterized in that, Predict the interference state between the wake trajectory and the swept area of ​​the downstream unit and generate wake cooperative control trigger flags, including: Based on the real wind field data, three-dimensional wind field spatiotemporal voxels are divided and constructed. The transmission time is calculated by dividing the absolute longitudinal physical distance from the spatial voxel to the wind turbine plane by the corresponding longitudinal wind speed component. An area-weighted average is applied to the time-aligned spatial voxel feature wind speeds to generate a predicted equivalent wind speed sequence, and the environmental turbulence intensity is calculated and extracted based on the predicted equivalent wind speed sequence. When the predicted equivalent wind speed sequence meets the wind speed operating range and the environmental turbulence intensity is lower than the turbulence intensity suppression threshold, and the predicted absolute wind direction corner falls within the wake interference sector estimated based on the environmental turbulence intensity, the wind direction interference condition is determined to be met and the wake cooperative control trigger flag is generated in an active state.

4. The wind turbine cluster wake control method based on radar wind measurement according to claim 3, characterized in that, Before converting the high-frequency wind direction change data into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary, the process also includes: The high-frequency wind direction change data corresponding to the wake coordination control trigger flag is converted into a frequency domain to generate a wind direction frequency domain amplitude spectrum. The core frequency component with the largest amplitude and the initial phase information of the core frequency component are extracted in the effective low-frequency range of the wind direction frequency domain amplitude spectrum. Calculate the angular deviation between the current predicted absolute wind direction angle and the target escape boundary, and generate the target wake deflection angle by combining it with the preset deflection safety margin angle; The lateral thrust coefficient is obtained by multiplying the wake sensitivity coefficient, the target wake deflection angle, and the axial thrust coefficient, and the high-frequency lateral thrust amplitude is calculated based on the lateral thrust coefficient. Using the limited high-frequency lateral thrust amplitude as the amplitude parameter, and combining it with the core frequency component and the initial phase information, a target lateral thrust time-domain command is reconstructed for aerodynamic mapping.

5. The wind turbine cluster wake control method based on radar wind measurement according to claim 4, characterized in that, Converting the high-frequency wind direction change data into a safe lateral thrust command constrained by the three-dimensional safe load envelope boundary includes: Based on the current operating parameters of the wind turbine generator set, calculate the dynamic margin of axial thrust, dynamic margin of pitch moment and dynamic margin of yaw moment, and construct the orthogonal three-dimensional safety load envelope boundary. Query the offline aeroelastic sensitivity matrix table, and combine it with the target lateral thrust time domain command to predict and generate the axial additional thrust, additional pitch moment amplitude and additional yaw moment at the current moment; Divide the dynamic margins of each dimension by the corresponding additional amplitude after adding the zero offset constant to obtain the thrust margin ratio, pitch margin ratio, and yaw margin ratio. Extract the minimum value among the thrust margin ratio, pitch margin ratio, and yaw margin ratio as the global truncation coefficient and multiply it with the target lateral thrust time domain command to generate the safe lateral thrust command.

6. The wind turbine cluster wake control method based on radar wind measurement according to claim 5, characterized in that, Calculating the local additional pitch angle command based on the aforementioned safe lateral thrust command includes: Divide the safe lateral thrust command by the lateral thrust pitch angle sensitivity coefficient after the lower limit clamping process to calculate the total independent pitch amplitude. The cosine value is obtained by adding the absolute azimuth angle of each blade to the phase compensation angle calculated using the first-order inertial time constant, and then using the cosine value to modulate the total independent pitch amplitude, and executing the spatial azimuth projection allocation to generate the local additional pitch angle command for each blade.

7. The wind turbine cluster wake control method based on radar wind measurement according to claim 6, characterized in that, By combining a safe wake excitation frequency that avoids the inherent frequency of the generator structure, a comprehensive pitch command is synthesized, including: The predicted equivalent wind speed and environmental turbulence intensity under the current operating conditions of the unit are obtained. The initial wake excitation frequency is calculated by combining the optimal Strouhal number matched according to the environmental turbulence intensity with the predicted equivalent wind speed and the rotor diameter. When the initial wake excitation frequency falls into the interference forbidden zone constructed by multiplying the rotor speed frequency, an adaptive frequency shift is performed towards the low-frequency side or the boundary value of the side with the closer absolute distance to obtain the candidate frequency; After verifying that the candidate frequency does not fall into the mechanical resonant frequency forbidden zone corresponding to the inherent frequency of the unit structure, the candidate frequency is subjected to time-domain smoothing processing with first-order low-pass filtering to confirm it as the safe wake excitation frequency. The safe wake excitation frequency is used to generate a high-frequency micro-motion disturbance wave, which is then superimposed onto the local additional pitch angle command in an algebraic addition form to output the comprehensive pitch command.

8. The wind turbine cluster wake control method based on radar wind measurement according to claim 1, characterized in that, The cumulative micro-stroke integral value for monitoring pitch control includes: Within a set sliding time window, the single-step pitch increment of the real-time pitch angle feedback signal is extracted and discretely accumulated to obtain the sliding cumulative pitch stroke. Divide the cumulative sliding pitch stroke by the sliding time window and combine it with the rated maximum pitch rate to calculate the theoretical maximum pitch stroke, and then calculate the pitch micro-motion duty cycle. When the pitch micro-motion duty cycle of any blade reaches the trigger upper limit threshold, a fatigue degradation flag is sent to the command synthesis stage. When the cumulative activation time of the flag reaches the sweep trigger time threshold, it is determined that the cumulative micro-motion stroke integral value is greater than the preset wear threshold, thereby suspending the integrated pitch command.

9. The wind turbine cluster wake control method based on radar wind measurement according to claim 8, characterized in that, Synchronous sweep commands are superimposed on the unified pitch reference angle under the current operating conditions of the unit to perform grease redistribution, including: Extract the unified pitch reference angle of the unit under the current operating conditions, and calculate the positive and negative stroke margins respectively in combination with the preset blade mechanical limit angle range; Select an action direction in which the stroke margin is greater than the target amplitude, and construct a sweep pulse command in half-sine wave shape based on the action direction; After performing the maximum command change rate check to ensure that the rate limit is not exceeded, the sweep pulse command is applied to the three blades in the same direction and with the same amplitude as the synchronous sweep command, forcing the internal grease to re-cover the contact area.

10. The wind turbine cluster wake control method based on radar wind measurement according to claim 9, characterized in that, After execution is complete, resume the issuance of the integrated pitch command, including: When the execution time of the synchronous sweep command reaches the execution cycle and the actual pitch angle of each blade falls back to the tolerance range, a recovery wake cooperative control trigger flag is generated. Based on the recovery wake coordination control trigger flag, the historical state accumulation item is cleared to zero, and the linear climb logic is started within the preset recovery time window to generate a transition weight that increases over time. The local additional pitch angle command is multiplied by the transition weight and then superimposed on the unified pitch reference angle to perform a smooth cut-in. After the transition weight climbs to the peak value, the cumulative micro-motion stroke integral value is cleared and the full-function closed loop is restored.