Wind measurement lidar wake tracking method based on dynamic modal decomposition

Abstract

The present application relates to the field of radar technology, and more particularly to a wind measurement laser radar wake tracking method based on dynamic modal decomposition. First, a global unified time reference is set, and the space-time matrix of the wind field viewing wind speed, the actual cabin yaw angle and the real-time current of the yaw motor are synchronously collected. Then, the wind speed matrix is subjected to dynamic modal decomposition, the fluid modal is extracted, and the wake dynamic evolution matrix is generated. At the same time, the remaining available yaw rate margin and real-time fatigue torque are calculated based on the yaw angle and the current to evaluate the yaw execution capability. Further, the kinematic tracking conflict residual and the dynamic load risk index are calculated, and when the tracking demand exceeds the execution bandwidth, the adaptive execution modal order reduction operation is performed to remove high-frequency data to generate a smooth target yaw angle command. Finally, the underlying limiting logic is triggered in combination with the load risk index to close-loop drive the cabin yaw. The wake tracking accuracy and equipment operation safety are effectively balanced.

Description

Technical Field

[0001] This invention relates to the field of radar technology, and in particular to a wake tracking method for wind-measuring lidar based on dynamic mode decomposition. Background Technology

[0002] With the continuous development of wind power generation technology, wake tracking technology for horizontal axis wind turbines is crucial for improving the overall power generation of wind farms. Currently, the industry mainly relies on forward-looking wind measurement lidar to obtain the wind field status and adjust the nacelle orientation accordingly to align with the wind direction. However, existing wake tracking methods typically generate yaw commands directly based on raw wind measurement data, failing to fully integrate with the real-time operating status of the yaw actuator for dynamic adaptation. This unidirectional control logic has significant drawbacks: on the one hand, when wind conditions change drastically, the rate of change of the generated commands often exceeds the actual response capability of the yaw mechanism, leading to tracking lag, increased deviation, and consequently reduced wind energy capture efficiency; on the other hand, blindly pursuing high-frequency command tracking will cause the yaw equipment to operate at full load or even overload for extended periods, exacerbating mechanical wear and fatigue losses in the yaw drive chain and severely shortening the equipment's service life. Therefore, a wake tracking solution that can balance tracking performance and equipment execution capability is urgently needed to achieve efficient and safe operation. Summary of the Invention

[0003] The main objective of this invention is to provide a wake tracking method for wind-measuring lidar based on dynamic mode decomposition. By establishing information interaction between the sensing end and the execution end, it overcomes the limitations of traditional one-way command issuance. By using dynamic mode decomposition to decompose the complex wake evolution into independent oscillation modes, and combining the real-time rate margin and load risk of the yaw system for adaptive order reduction and amplitude limiting adjustment, it not only effectively avoids tracking deviation and mechanical shock caused by the command change rate exceeding the equipment response capability, thus improving wind energy capture efficiency, but also actively reduces drive output under high load conditions, significantly reducing fatigue loss of the yaw drive chain. Therefore, without increasing hardware costs, it achieves a balance between tracking smoothness and equipment lifespan.

[0004] The technical solution of the present invention is as follows: Firstly, a wake tracking method for wind-measuring lidar based on dynamic mode decomposition is proposed, which includes the following steps: Collect the actual wind field state, nacelle spatial position and yaw motor electromagnetic state, and obtain the line-of-sight wind speed spatiotemporal matrix, actual nacelle yaw angle and yaw motor real-time current; Based on the initially set modal truncation order, dynamic modal decomposition calculation is performed on the line-of-sight wind speed spatiotemporal matrix. The fluid modes with the highest energy proportion of the aforementioned modal truncation order are selected to generate a wake dynamic evolution matrix containing angular frequency and target angular amplitude. The actual yaw rate is obtained by time derivative of the actual cabin yaw angle. The remaining available yaw rate margin is obtained by subtracting the current actual yaw rate from the rated maximum yaw rate. The real-time fatigue torque is obtained by multiplying the real-time current of the yaw motor by the motor torque constant, and the yaw execution capability state point is generated. Extract the sum of angular velocity requirements of all retained modes in the wake dynamic evolution matrix, subtract the remaining available yaw rate margin to calculate the kinematic tracking conflict residual, and divide the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index, and output the bandwidth and load conflict state vector. When the kinematic tracking conflict residual is greater than zero, a modal order reduction operation is performed to successively reduce the modal truncation order and remove the highest angular frequency data until the recalculated kinematic tracking conflict residual is less than or equal to zero. Based on all fluid modes retained after order reduction, a time-domain reconstruction is performed to generate a smooth target yaw angle command. The difference between the smoothed target yaw angle command and the actual cabin yaw angle is input into the proportional-integral controller to output the yaw motor drive signal. When the dynamic load risk index is greater than 0.8, the underlying limiting logic is triggered to reduce the output amplitude of the drive signal, thereby driving the physical cabin attitude to change.

[0005] A further improvement of this invention is that the steps of collecting the actual wind field state, the nacelle spatial position, and the yaw motor electromagnetic state include: setting a globally unified time reference t to keep the sampling period consistent with the control period; the lidar optical receiver scanning and analyzing the actual wind field state according to the time reference t to obtain wind speed data at different spatial coordinate points; the yaw gear absolute encoder sampling the nacelle spatial position at the time reference t to obtain the real-time orientation data of the nacelle; the Hall sensor inside the yaw driver sampling the current of the yaw motor electromagnetic state according to the time reference t to obtain the real-time current data of the motor windings; and outputting the line-of-sight wind speed spatiotemporal matrix, the actual nacelle yaw angle, and the real-time current of the yaw motor, wherein the spatial resolution of the line-of-sight wind speed spatiotemporal matrix is ​​set to 2 meters, and the detection range covers -200 meters to 200 meters along the x-axis and 0 meters to 400 meters along the y-axis.

[0006] A further improvement of this invention is that the step of performing dynamic mode decomposition calculation on the line-of-sight wind speed spatiotemporal matrix based on an initially set modal truncation order includes: receiving the line-of-sight wind speed spatiotemporal matrix, wherein the initially set modal truncation order is denoted as k; the sensing end controller constructs a snapshot matrix from the spatial wind speed data in the line-of-sight wind speed spatiotemporal matrix at different times, and obtains the oscillation frequency and corresponding amplitude of all fluid modes by solving the matrix through singular value decomposition and linear mapping; combining the modal truncation order k, selecting the top k fluid modes with the highest energy proportion and retaining their corresponding oscillation parameters; and outputting the wake dynamic evolution matrix, wherein the wake dynamic evolution matrix includes the angular frequency and the target angular amplitude of the i-th order fluid mode.

[0007] A further improvement of the present invention is that the step of generating the yaw execution capability state point includes: receiving the actual cabin yaw angle and the real-time current of the yaw motor, and obtaining the preset rated maximum yaw rate and the motor torque constant; the execution controller performs first-order time derivative of the actual cabin yaw angle and takes the absolute value to obtain the current actual yaw rate; subtracting the current actual yaw rate from the rated maximum yaw rate to calculate the remaining available yaw rate margin; simultaneously, multiplying the real-time current of the yaw motor by the motor torque constant to calculate the real-time fatigue torque; and outputting the yaw execution capability state point, which is composed of the current actual yaw rate and the real-time fatigue torque.

[0008] A further improvement of the present invention is that the step of outputting the bandwidth and load conflict state vector includes: receiving the wake dynamic evolution matrix and the yaw execution capability state point, and obtaining the preset ultimate fatigue torque; the sensing end controller extracts the product of the target angular amplitude and the angular frequency of all retained modes in the wake dynamic evolution matrix, takes the absolute value and sums them to obtain the total angular velocity requirement; subtracts the remaining available yaw rate margin from the total angular velocity requirement to calculate the kinematic tracking conflict residual; simultaneously, divides the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index; and outputs the bandwidth and load conflict state vector composed of the kinematic tracking conflict residual and the dynamic load risk index.

[0009] A further improvement of this invention is that the step of generating a smooth target yaw angle command includes: receiving the bandwidth and load conflict state vector, the wake dynamic evolution matrix, and the modal truncation order k; determining whether the kinematic tracking conflict residual is greater than zero; if it is greater than zero, performing the modal order reduction operation, successively reducing the value of the modal truncation order k by 1 each time, and simultaneously removing the row of data with the highest angular frequency from the wake dynamic evolution matrix until the recalculated kinematic tracking conflict residual is less than or equal to zero, thus obtaining the updated modal truncation order; if the kinematic tracking conflict residual is less than or equal to zero in the initial state, keeping the original modal truncation order unchanged; based on all fluid modes retained after order reduction, performing time-domain reconstruction according to the angular frequency of each mode and the target angle amplitude, retaining the phase information of each mode during the reconstruction process; and outputting the smooth target yaw angle command and the updated modal truncation order.

[0010] A further improvement of the present invention is that the step of driving the physical cabin attitude change includes: receiving the smoothed target yaw angle command, the actual cabin yaw angle, and the dynamic load risk index; the execution controller subtracts the smoothed target yaw angle command from the actual cabin yaw angle to obtain the yaw angle error, inputs the yaw angle error into the proportional-integral controller for adjustment, and outputs the drive signal of the yaw motor; continuously monitors the dynamic load risk index, and if the value is greater than 0.8, triggers the underlying limiting logic to proportionally reduce the output amplitude of the drive signal and reduce the output torque of the motor; drives the physical cabin attitude change so that the actual cabin yaw angle gradually approaches the smoothed target yaw angle command, thus completing the control.

[0011] Secondly, a computer-readable storage medium is proposed, on which a computer program is stored. When the computer program is executed by a processor, it implements the above-mentioned wind-measuring lidar wake tracking method based on dynamic mode decomposition.

[0012] Thirdly, an electronic device is proposed, including a memory for storing instructions and a processor for executing the instructions, causing the device to perform the above-described wind-measuring lidar wake tracking method based on dynamic mode decomposition.

[0013] The technical effects of this invention are as follows: A wake tracking method for wind-measuring lidar based on dynamic mode decomposition was constructed. By establishing information interaction between the sensing end and the execution end, the limitations of traditional one-way command issuance are overcome. Dynamic mode decomposition is used to decompose the complex wake evolution into independent oscillation modes. Adaptive order reduction and amplitude limiting adjustment are performed by combining the real-time rate margin and load risk of the yaw system. This not only effectively avoids tracking deviation and mechanical shock caused by the command change rate exceeding the equipment response capability and improves wind energy capture efficiency, but also actively reduces drive output under high load conditions, significantly reducing fatigue loss of the yaw drive chain. Thus, a balance between tracking smoothness and equipment lifespan is achieved without increasing hardware costs. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the wake tracking method of wind-measuring lidar based on dynamic mode decomposition according to Embodiment 1 of the present invention. Detailed Implementation

[0015] Example 1: This example is applied to the wake tracking control scenario of a forward-looking lidar for a horizontal axis wind turbine. The entire processing flow is completed collaboratively by a lidar sensing unit deployed on the top of the nacelle, a nacelle attitude acquisition unit, a yaw motor status acquisition unit, a sensing-end controller, and an execution-end controller. Specifically, as follows... Figure 1 As shown, the wake tracking method for wind-measuring lidar based on dynamic mode decomposition proposed in this embodiment includes the following specific steps: The process involves collecting data on the actual wind field flow state, nacelle spatial position, and yaw motor electromagnetic state to obtain the line-of-sight wind speed spatiotemporal matrix, actual nacelle yaw angle, and real-time yaw motor current. The steps include: setting a globally unified time reference t to ensure the sampling period matches the control period; a lidar optical receiver scanning and analyzing the actual wind field flow state according to the time reference t to obtain wind speed data at different spatial coordinate points; a yaw gear absolute encoder sampling the nacelle spatial position angle according to the time reference t to obtain real-time nacelle orientation data; and a Hall sensor inside the yaw driver sampling the yaw motor electromagnetic state current according to the time reference t to obtain real-time current data of the motor windings. The output includes the line-of-sight wind speed spatiotemporal matrix, the actual nacelle yaw angle, and the real-time yaw motor current. The spatial resolution of the line-of-sight wind speed spatiotemporal matrix is ​​set to 2 meters, with a detection range covering -200 meters to 200 meters along the x-axis and 0 meters to 400 meters along the y-axis.

[0016] In this embodiment, during data acquisition and synchronization processing, three types of data are simultaneously acquired: wind field flow field state, nacelle spatial position, and yaw motor electromagnetic state. After timestamp alignment, standardized time-series data is output, providing an input basis for subsequent wake evolution analysis and yaw control. This process receives a synchronization trigger signal from a global time reference, driving the lidar optical receiver, yaw gear absolute encoder, and yaw driver internal Hall sensor to synchronously complete sampling. All sampled data are bound to the same timestamp and transmitted to the corresponding controller. The lidar uses the coherent Doppler detection principle, emitting narrow-linewidth laser pulses and receiving large-linewidth laser pulses. The backscattered echoes of aerosol particles in the air are analyzed, and the Doppler frequency shift of the echo signals is used to calculate the line-of-sight wind speed. A planar position indication scanning mode is employed, completing one fan-shaped area scan within each sampling period, covering a preset detection range. The lidar optical receiver scans and analyzes the forward region along the wind turbine's axial direction to acquire line-of-sight wind speed data at different spatial coordinate points. The wind speed data from all spatial points are arranged according to their spatial coordinates to form a single-moment wind speed distribution vector. Wind speed distribution vectors from multiple consecutive moments together form the line-of-sight wind speed spatiotemporal matrix. The spatial resolution of the line-of-sight wind speed spatiotemporal matrix is ​​set to 2 meters, and the detection range is along the nacelle axis. The axial direction covers -200 meters to 200 meters, along the horizontal direction perpendicular to the nacelle axis. The axis covers a range from 0 meters to 400 meters, of which The positive axis points towards the incoming flow area in front of the wind turbine. At each sampling moment, the lidar completes a full-area scan and outputs the spatial wind speed distribution for that moment. This distribution is stored in the form of a two-dimensional array, with rows and columns corresponding to... shaft and The spatial coordinates of the axis are represented by array elements, with each element representing the line-of-sight wind speed in meters per second. The yaw gear absolute encoder is a multi-turn photoelectric encoder with a preferred resolution of 0.01 degrees per pulse. It can continuously record the absolute yaw angle of the nacelle throughout the entire lifespan of the unit without requiring a power-on homing operation. It is installed on the yaw gear disk and operates based on a time reference. The real-time orientation of the nacelle is sampled to output the actual nacelle yaw angle in degrees. This angle is based on true north, with clockwise rotation as the positive direction. The value range covers 0 to 360 degrees. The Hall sensor inside the yaw drive is a linear Hall element with a sampling accuracy preferably of 0.01 amperes. The sampling frequency is consistent with the global sampling period, which can reflect the current change of the motor windings in real time. It is connected in series in the three-phase winding circuit of the yaw motor and is based on a time reference. The real-time current of the motor windings is sampled, and the real-time current of the yaw motor is output in amperes. During the sampling process, the effective values ​​of the three-phase currents are converted to obtain the single-value motor operating current for subsequent calculations. After the acquisition of the three types of data is completed, the sensing controller and the execution controller receive the corresponding data content respectively. The line-of-sight wind speed spatiotemporal matrix is ​​transmitted to the sensing controller, and the actual nacelle yaw angle and the real-time current of the yaw motor are synchronously transmitted to the execution controller. All data are based on a time reference. The data are cached sequentially in the local storage unit, with a preferred cache depth of 100 sampling periods to meet the data length requirements of subsequent time-series analysis. In a further implementation, outlier removal processing can be added during the data acquisition stage. The amplitude of the line-of-sight wind speed data obtained from a single sampling is verified. If the wind speed value of a certain spatial point exceeds the preset reasonable wind speed range, the interpolation result of the adjacent spatial point is used for replacement. The reasonable wind speed range can be set to 0 m / s to 40 m / s to avoid abnormal data caused by fog and rain affecting the accuracy of subsequent analysis.

[0017] Based on the initially set modal truncation order, dynamic modal decomposition calculation is performed on the line-of-sight wind speed spatiotemporal matrix. The fluid modes with the highest energy proportions according to the aforementioned modal truncation order are selected to generate a wake dynamic evolution matrix containing angular frequencies and target angular amplitudes. The step of performing dynamic modal decomposition calculation on the line-of-sight wind speed spatiotemporal matrix based on the initially set modal truncation order includes: receiving the line-of-sight wind speed spatiotemporal matrix, wherein the initially set modal truncation order is denoted as k; the sensing end controller constructs a snapshot matrix from the spatial wind speed data in the line-of-sight wind speed spatiotemporal matrix at different times, and obtains the oscillation frequencies and corresponding amplitudes of all fluid modes through singular value decomposition and linear mapping; combined with the modal truncation order k, the top k fluid modes with the highest energy proportions are selected, and their corresponding oscillation parameters are retained; the wake dynamic evolution matrix is ​​output, wherein the wake dynamic evolution matrix contains the angular frequency and the target angular amplitude of the i-th order fluid mode.

[0018] In this embodiment, during the construction of the wake dynamic evolution matrix, the sensing-end controller performs dynamic mode decomposition calculations on the line-of-sight wind speed spatiotemporal matrix based on an initially set modal truncation order. This extracts the dominant fluid oscillation modes and generates the wake dynamic evolution matrix to quantify the wake's motion characteristics. This process uses the continuously sampled line-of-sight wind speed spatiotemporal matrix as input, and through three steps—time-series snapshot construction, linear mapping solution, and mode selection—obtains a set of parameters characterizing the wake oscillation properties. The output wake dynamic evolution matrix can be directly used for subsequent yaw rate calculations. The initially set modal truncation order is denoted as... The initial preferred value range is 8 to 15, which can be configured according to the wind field turbulence intensity and wake complexity. In this embodiment, the initial... The value is set to 10. Because the original line-of-sight wind speed spatiotemporal matrix has a high spatial dimension, directly constructing a snapshot matrix would lead to excessive computation. Therefore, before constructing the snapshot vector, the spatial wind speed distribution can be uniformly downsampled to reduce the spatial resolution from 2 meters to 5 meters. The downsampling method uses the neighborhood averaging method, which reduces the data dimensionality while preserving the spatial distribution characteristics of the flow field. The number of spatial points after downsampling can be controlled to the thousands, making the dimension of the snapshot matrix compatible with the computing power of conventional industrial controllers, ensuring that the decomposition calculation can be completed within a single sampling period. The sensing controller first expands the spatial wind speed data in the line-of-sight wind speed spatiotemporal matrix at different times into a one-dimensional vector. Each time point corresponds to a one-dimensional snapshot vector, and the dimension of the vector is equal to the total number of spatial points in the downsampled line-of-sight wind speed spatiotemporal matrix. The snapshot vectors at time step i are arranged column-wise to construct a snapshot matrix, denoted as the snapshot vector at time step i. The snapshot vector at each time point is ,in The value range is 1 to The snapshot matrix can be expressed in the following form: ;in For the number of snapshots, The preferred value range is 50 to 200, in this embodiment A value of 100 corresponds to 10 seconds of continuous sampling data. The snapshot matrix is ​​divided into the first... Matrix composed of columns After Matrix composed of columns The two satisfy a linear mapping relationship, that is and Linear operators can be used to connect them. To implement the mapping, in order to solve for this linear operator, we first perform... Perform singular value decomposition, the decomposition form is as follows: ;in Let be a unitary matrix composed of left singular vectors. It is a diagonal matrix composed of singular values. Let be a unitary matrix composed of right singular vectors, with superscript... The conjugate transpose operation, based on the singular value decomposition result, can be used to transform linear operators. Projecting onto a lower-dimensional subspace yields a reduced-order linear operator. The calculation method is as follows: For the reduced-order linear operator Solve for the eigenvalues ​​and eigenvectors, where the eigenvalues ​​are... Each eigenvalue is denoted as The corresponding eigenvectors are denoted as The angular frequencies of the corresponding fluid modes can be calculated based on the eigenvalues. The calculation method is to take the natural logarithm of the eigenvalues, then take the imaginary part, and finally divide by the sampling period. ,Right now: ;in The unit is radians per second, representing the oscillation rate of the corresponding fluid mode. The target angular amplitude corresponding to each fluid mode is denoted as . Its value is determined by the modulus and singular values ​​of the corresponding eigenvector, and is expressed in degrees. It characterizes the peak amplitude of the wake deflection in this mode. After solving all fluid modes, it is combined with the modal cutoff order. Sort by energy percentage from highest to lowest for each mode, and select the mode with the highest energy percentage. Each fluid mode is selected, retaining its corresponding angular frequency and target angular amplitude parameters. The energy percentage is calculated as the ratio of the square of the singular value corresponding to a single mode to the sum of the squares of all singular values. A higher energy percentage indicates a greater contribution of that mode to the flow field evolution. Ultimately, the retained... The parameters of each fluid mode are arranged in rows to generate a wake dynamic evolution matrix. Each row of the matrix corresponds to a first-order fluid mode, and each column corresponds to the angular frequency and the target angular amplitude, respectively. The wake dynamic evolution matrix is ​​output to the subsequent calculation unit of the sensing end controller to evaluate the angular velocity requirements for wake tracking. The design idea of ​​dynamic mode decomposition in this step is that the spatiotemporal evolution of complex wake fields can be approximated by the superposition of a few dominant oscillation modes. By extracting low-dimensional modes, the amount of computation can be greatly reduced while retaining the main motion characteristics of the wake, which meets the computing power requirements of real-time control of wind turbine units. At the same time, the modal expression form facilitates the subsequent quantitative evaluation of the angular velocity requirements for wake tracking, providing a data basis for the introduction of execution capability constraints.

[0019] The process involves: deriving the actual cabin yaw angle over time to obtain the current actual yaw rate; subtracting the current actual yaw rate from the rated maximum yaw rate to obtain the remaining usable yaw rate margin; and multiplying the real-time current of the yaw motor by the motor torque constant to obtain the real-time fatigue torque, thus generating a yaw execution capability state point. The steps for generating the yaw execution capability state point include: receiving the actual cabin yaw angle and the real-time current of the yaw motor, and obtaining the preset rated maximum yaw rate and the motor torque constant; the execution controller derives the actual cabin yaw angle over time and takes the absolute value to obtain the current actual yaw rate; subtracting the current actual yaw rate from the rated maximum yaw rate to calculate the remaining usable yaw rate margin; simultaneously, multiplying the real-time current of the yaw motor by the motor torque constant to calculate the real-time fatigue torque; and outputting the yaw execution capability state point, which is composed of the current actual yaw rate and the real-time fatigue torque.

[0020] In this embodiment, during the yaw capability status assessment, the actuator controller calculates the kinematic margin and dynamic load level of the current yaw system based on the actual cabin yaw angle and the real-time current of the yaw motor, generating a yaw capability status point to provide constraints for subsequent conflict judgment. This process simultaneously receives cabin attitude data and motor current data, and obtains two core state parameters through rate conversion and torque conversion, respectively, comprehensively reflecting the remaining capability and load level of the yaw system. The actuator controller first obtains two preset parameters: the rated maximum yaw rate and the motor torque constant, where the rated maximum yaw rate is denoted as... The unit is degrees per second, and its value is determined by the rated speed of the yaw motor and the transmission ratio of the yaw gear. The preferred value range is 0.5 degrees per second to 1.0 degrees per second. In this embodiment... The value is taken as 0.8 degrees per second, and the motor torque constant is denoted as... The unit is Newton-meters per ampere (Nm / Ampere), and its value is determined by the winding parameters of the yaw motor and the performance of the permanent magnet. The preferred value range is 1.2 Nm / Ampere to 2.5 Nm / Ampere. In this embodiment... The value is set to 1.8 Nm per ampere. Subsequently, the actuator controller performs a first-order time derivative calculation on the actual cabin yaw angle. The derivative process is implemented using a backward difference method, that is, by dividing the difference between the yaw angle at the current time and the previous time by the sampling period. To obtain the current yaw rate, take the absolute value of the current rate to get the current actual yaw rate, denoted as . The unit is degrees per second, using the rated maximum yaw rate. Subtract the current actual yaw rate The remaining available yaw rate margin is calculated and denoted as . The unit is degrees per second. Its value characterizes the additional yaw rate capability that the yaw system can provide under the current motion state. A larger margin value indicates a more sufficient rate margin for the yaw system to track wake oscillations. Simultaneously, the actuator controller records the real-time current of the yaw motor as... Multiply by the motor torque constant The real-time fatigue torque is calculated and denoted as . The unit is Newton-meter (Nm), and its value represents the current output torque of the yaw motor. It can reflect the load level borne by the yaw gear and transmission mechanism. The higher the torque value, the faster the fatigue damage of the transmission mechanism accumulates. The current actual yaw rate and the real-time fatigue torque are combined to generate the yaw execution capability state point. This state point is stored in the execution end controller in the form of a two-dimensional array and is synchronously transmitted to the sensing end controller for subsequent conflict state calculation. In a further implementation, the real-time fatigue torque can be filtered by a moving average. The filter window length is preferably 5 to 20 sampling periods to smooth the noise fluctuations caused by current sampling, improve the stability of load evaluation, and avoid frequent jumps in the load index that cause the limiting logic to repeatedly switch in and out.

[0021] The process involves extracting the sum of angular velocity requirements for all retained modes in the wake dynamic evolution matrix, subtracting the remaining available yaw rate margin to calculate the kinematic tracking conflict residual, and dividing the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index. The output bandwidth and load conflict state vector is then generated. The steps include: receiving the wake dynamic evolution matrix and the yaw execution capability state point, and obtaining the preset ultimate fatigue torque; the sensing controller extracting the product of the target angular amplitude and the angular frequency of all retained modes in the wake dynamic evolution matrix, taking the absolute value, and summing the products to obtain the sum of angular velocity requirements; subtracting the remaining available yaw rate margin from the sum of angular velocity requirements to calculate the kinematic tracking conflict residual; simultaneously, dividing the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index; and outputting the bandwidth and load conflict state vector composed of the kinematic tracking conflict residual and the dynamic load risk index.

[0022] In this embodiment, during the conflict state calculation and modal order reduction adjustment process, the tracking conflict degree at the kinematic level and the load risk level at the dynamic level are calculated by combining the wake dynamic evolution matrix and the yaw execution capability state point. This generates bandwidth and load conflict state vectors, and modal order reduction is performed when kinematic conflict exists. Finally, a smooth target yaw angle command is reconstructed. This process first completes the quantitative calculation of two conflict indicators, then adjusts the number of modes retained through closed-loop order reduction iterations to match the wake tracking requirements with the yaw system's execution capability. Finally, time-domain signal reconstruction is performed based on the adapted modes. The sensing-end controller first obtains the preset ultimate fatigue torque parameter, denoted as... The unit is Newton-meters (Nm), and its value is determined by the fatigue life design threshold of the yaw transmission mechanism. A preferred value is 1.2 to 1.5 times the rated torque of the yaw motor. In this embodiment... The value is set to 240 Nm. Subsequently, the sensing controller extracts the target angular amplitude and angular frequency of all retained modes from the wake dynamic evolution matrix. First, the modal angular frequency is... Convert to degrees per second. The transformation relationship is as follows: The instantaneous maximum yaw rate amplitude generated during a single-mode oscillation is the product of the target angular amplitude and the converted angular frequency. This product, after radian-degree conversion, is standardized to degrees per second. The absolute values ​​of the single-mode maximum angular rate amplitudes corresponding to all retained modes are summed to obtain the total angular rate requirement, denoted as . The unit is degrees per second. The formula for calculating the total angular velocity requirement is as follows: In the formula For the first Peak wake deflection angle of a first-order fluid mode, in degrees. The first unit after conversion to degrees per second The first modal angular frequency, and the product of the two terms, correspond to the instantaneous maximum yaw rate achievable by a single-mode oscillation. After unit conversion, the overall dimension is unified to degrees per second. Given the current modal cutoff order, the total angular velocity requirement represents the total yaw angular velocity amplitude required by the yaw system to track all retained wake oscillation modes. This value integrates the oscillation velocities and deflection amplitudes of all retained modes, and can intuitively reflect the bandwidth requirements of wake tracking on the yaw system. The total angular velocity requirement... Subtract the remaining available yaw rate margin The kinematic tracking conflict residuals are calculated and denoted as... The unit is degrees per second. When this residual is greater than zero, it indicates that the tracking angular velocity required by the currently retained wake mode exceeds the remaining execution capacity of the yaw system, resulting in a kinematic tracking conflict. Forcibly tracking high-frequency modes will lead to yaw motor saturation and increased tracking error, which will negatively impact wake tracking performance. The larger the value of the kinematic tracking conflict residual, the greater the gap between the current wake tracking requirement and the yaw execution capacity, and the more high-frequency modes need to be eliminated. At the same time, the real-time fatigue torque will be... Divide by the ultimate fatigue torque The dynamic load risk index is calculated and denoted as . The index is a dimensionless value, ranging from 0 to 1. The higher the value, the closer the load on the yaw transmission mechanism is to the design limit, and the higher the risk of fatigue damage. The kinematic tracking conflict residual and the dynamic load risk index are combined to output the bandwidth and load conflict state vector. This vector is synchronously transmitted to the modal reduction unit and the subsequent control limiting unit.

[0023] When the kinematic tracking conflict residual is greater than zero, a modal order reduction operation is performed to successively reduce the modal truncation order and remove the highest angular frequency data until the recalculated kinematic tracking conflict residual is less than or equal to zero. Based on all fluid modes retained after order reduction, a time-domain reconstruction is performed to generate a smooth target yaw angle command. The step of generating the smooth target yaw angle command includes: receiving the bandwidth and load conflict state vector, the wake dynamic evolution matrix, and the modal truncation order k; determining whether the kinematic tracking conflict residual is greater than zero; if it is greater than zero, then performing the modal order reduction operation to successively reduce the highest angular frequency data. The modal truncation order k is decreased by 1 each time, and the row of data with the highest angular frequency is removed from the wake dynamic evolution matrix until the recalculated kinematic tracking conflict residual is less than or equal to zero, thus obtaining the updated modal truncation order. If the kinematic tracking conflict residual is less than or equal to zero in the initial state, the original modal truncation order remains unchanged. Based on all fluid modes retained after order reduction, time-domain reconstruction is performed according to the angular frequency of each mode and the target angular amplitude, while retaining the phase information of each mode during the reconstruction process. The smoothed target yaw angle command and the updated modal truncation order are output.

[0024] In this embodiment, after receiving the bandwidth and load conflict state vector, the sensing controller first determines whether the kinematic tracking conflict residual is greater than zero. If the kinematic tracking conflict residual is greater than zero, it performs a mode reduction operation, successively reducing the mode truncation order. The value, in each iteration will be The value is decreased by 1, and the row of data with the highest angular frequency is removed from the wake dynamic evolution matrix, i.e., the fastest oscillating first-order fluid mode is removed. After completing one order reduction, the kinematic tracking conflict residual is recalculated based on the remaining fluid modes according to the aforementioned calculation method of the total angular velocity requirement. It is then checked again whether the residual is greater than zero. If the recalculated residual is still greater than zero, the order reduction operation continues until the recalculated kinematic tracking conflict residual is less than or equal to zero. At this point, the order reduction stops, and the updated modal truncation order and the corresponding set of retained fluid modes are obtained. By successively reducing the order, the maximum number of modes retained that satisfy the execution capability constraint can be found, thus... To maximize the wake tracking effect within the constraints and avoid excessive reduction in order at once that would lead to a loss of tracking accuracy, if the kinematic tracking conflict residual is less than or equal to zero in the initial state, it means that the wake tracking requirement under the current modal truncation order does not exceed the execution capability of the yaw system. The original modal truncation order is kept unchanged. The design idea of ​​this reduction logic is that the high-frequency wake mode has a fast oscillation speed but a low energy proportion and contributes little to the overall wake motion. Prioritizing the elimination of high-frequency modes can make the tracking requirement adapt to the bandwidth limit of the yaw system while preserving the wake tracking effect as much as possible, avoiding the actuator from entering a saturation state, and also reducing the mechanical load impact caused by high-frequency tracking. After adapting and adjusting the modal order, time-domain reconstruction is performed based on all fluid modes retained after order reduction to generate a smooth target yaw angle command. During the time-domain reconstruction, the phase information of each mode is retained. The phase information comes from the real part of the eigenvalue and the projection result of the initial snapshot during dynamic mode decomposition. The time-domain contribution of each mode is the target angle amplitude multiplied by the cosine function of the corresponding angular frequency and the phase. The time-domain contributions of all retained modes are superimposed to obtain a continuous target yaw angle time-series signal. The calculation formula for time-domain reconstruction is as follows: ;in The updated modal truncation order. For the first The initial phase of the first fluid mode, in radians. for The target yaw angle command at any given time, in degrees, is reconstructed in this way. It retains the dominant oscillation characteristics of the wake while eliminating high-frequency oscillation components that exceed the execution capability. The signal is smooth and continuous overall, without any abrupt changes, and can adapt to the response bandwidth of the yaw system. The generated smooth target yaw angle command and the updated modal truncation order are output to the execution controller. The updated modal truncation order can be used as the initial truncation order reference for the next control cycle, realizing adaptive iterative adjustment of the truncation order. In another optional implementation, the step size of the modal order reduction can be set to a value greater than 1, such as a step size of 2, to accelerate the convergence speed of the order reduction iteration, which is suitable for wind field scenarios with sudden changes in turbulence intensity. At the same time, a lower limit value of the modal truncation order can be set, preferably 2 to 3, to avoid excessive order reduction that would cause a significant decrease in wake tracking performance and ensure basic wake following capability.

[0025] The difference between the smoothed target yaw angle command and the actual cabin yaw angle is input into the proportional-integral controller to output the yaw motor drive signal. When the dynamic load risk index is greater than 0.8, the underlying limiting logic is triggered to reduce the output amplitude of the drive signal, thereby driving a change in the attitude of the physical cabin. The steps of driving a change in the attitude of the physical cabin include: receiving the smoothed target yaw angle command, the actual cabin yaw angle, and the dynamic load risk index; the execution controller calculates the yaw angle error by subtracting the smoothed target yaw angle command from the actual cabin yaw angle, inputs the yaw angle error into the proportional-integral controller for adjustment, and outputs the drive signal of the yaw motor; continuously monitoring the dynamic load risk index, if the value is greater than 0.8, triggering the underlying limiting logic to proportionally reduce the output amplitude of the drive signal and reduce the output torque of the motor; driving a change in the attitude of the physical cabin, so that the actual cabin yaw angle gradually approaches the smoothed target yaw angle command, thus completing the control.

[0026] In this embodiment, during the yaw control output and load limiting protection process, the actuator controller generates a yaw motor drive signal based on the target yaw angle command and the actual cabin yaw angle. This signal, combined with the dynamic load risk index, triggers the underlying limiting logic, ultimately driving a change in cabin attitude and completing the closed-loop control for wake tracing. This process uses the yaw angle error as input, generates the motor drive command through closed-loop adjustment, and simultaneously introduces a load protection mechanism to prevent overload of the transmission mechanism. This ensures the safe operation of the yaw system while tracking the wake tracing. The actuator controller first subtracts the smoothed target yaw angle command from the current actual cabin yaw angle to obtain the yaw angle error, denoted as yaw angle . The unit is degrees. The yaw angle error is input into the proportional-integral controller for adjustment. The proportional coefficient of the proportional-integral controller is denoted as... The unit is amperes per second per degree, and the integral coefficient is denoted as . The units are amperes per square second per degree. Both parameters can be determined through model identification and debugging of the wind turbine yaw system. The preferred range for the proportional coefficient is 0.2 to 1.0, and the preferred range for the integral coefficient is 0.05 to 0.3. The output of the proportional-integral controller is the drive signal for the yaw motor. This drive signal is transmitted to the yaw driver in the form of a current command to control the output torque and direction of the yaw motor. During the control process, the actuator controller continuously monitors the value of the dynamic load risk index. If the value is greater than 0.8, the underlying limiting logic is triggered, proportionally reducing the output amplitude of the drive signal. The threshold of the dynamic load risk index is set to 0.8 to consider the fatigue damage accumulation characteristics of the yaw transmission mechanism. When the load exceeds 80% of the limit value, the growth rate of fatigue damage will significantly accelerate. Therefore, early intervention in limiting protection can effectively extend the service life of the mechanism. The limiting ratio is related to the value of the dynamic load risk index; the higher the risk index, the larger the limiting ratio. The maximum output amplitude of the drive signal is limited to a value inversely proportional to the risk index. Specifically, when... When the value is greater than 0.8 and less than or equal to 1, the limiting factor can be expressed as: The final output value of the drive signal is the original output of the proportional-integral controller multiplied by the limiting coefficient, thereby reducing the motor output torque and alleviating the load pressure on the yaw transmission mechanism. When the dynamic load risk index falls to 0.8 or below, the limiting logic automatically exits, and the drive signal returns to normal proportional-integral regulation output. The execution priority of the underlying limiting logic is higher than that of the proportional-integral controller output, meaning that the limiting logic will post-process the output result of the proportional-integral controller. No matter how large the controller output command is, the signal ultimately output to the driver will not exceed the limiting value. The limiting process adopts a smooth transition method. When the risk index crosses the 0.8 threshold, the limiting coefficient gradually changes within 5 sampling periods to avoid sudden changes in the drive signal when the limiting logic enters and exits, reducing the impact on the nacelle attitude. After receiving the drive signal, the device controls the yaw motor to output the corresponding torque, which drives the entire nacelle to rotate through the yaw gear disk, so that the actual nacelle yaw angle gradually approaches the smooth target yaw angle command, thereby realizing the real-time tracking of the nacelle orientation to the wake oscillation. When the difference between the actual nacelle yaw angle and the target yaw angle command is less than the preset steady-state error threshold, the system enters the steady-state adjustment stage. The preferred value range of the steady-state error threshold is 0.1 degrees to 0.5 degrees. In this embodiment, the value is 0.2 degrees. In a further embodiment, the proportional-integral controller can be replaced with a proportional-integral-derivative controller. The introduction of the derivative element improves the dynamic response speed of the system, which is suitable for mountain wind field scenarios where the wake changes rapidly. At the same time, a rate limiting element can be added to the controller output to further constrain the rate of change of yaw rate and reduce mechanical shock.

[0027] Example 2: This example provides an electronic device, including a processor and a memory, wherein the memory stores a computer program that can be called by the processor; the processor executes the above-described wind-measuring lidar wake tracking method based on dynamic mode decomposition by calling the computer program stored in the memory.

[0028] The electronic device can vary considerably depending on its configuration or performance. It may include one or more Central Processing Units (CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the wind-measuring lidar wake tracking method based on dynamic mode decomposition provided in the above-described embodiment. The electronic device may also include other components for implementing its functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Details will not be elaborated upon in this embodiment.

[0029] Those skilled in the art will recognize that this invention can be implemented as a system, method, or computer program product. Therefore, this invention can be implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "circuit," "module," or "system." Furthermore, in some embodiments, this invention can also be implemented as a computer program product contained in one or more computer-readable media, which includes computer-readable program code.

[0030] Any combination of one or more computer-readable media may be used. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0031] This invention is described with reference to flowchart illustrations and block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and block diagrams, as well as combinations of blocks in the flowchart illustrations and block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0032] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and boxes Figure 1 The steps of the function specified in one or more boxes.

[0033] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A wake tracking method for wind-measuring lidar based on dynamic mode decomposition, characterized in that: The specific steps include the following: Collect the actual wind field state, nacelle spatial position and yaw motor electromagnetic state, and obtain the line-of-sight wind speed spatiotemporal matrix, actual nacelle yaw angle and yaw motor real-time current; Based on the initially set modal cutoff order, dynamic modal decomposition calculation is performed on the line-of-sight wind speed spatiotemporal matrix. The fluid modes with the highest energy proportion of the aforementioned modal cutoff order are selected to generate a wake dynamic evolution matrix containing angular frequency and target angular amplitude. The actual yaw rate is obtained by time derivative of the actual cabin yaw angle. The remaining available yaw rate margin is obtained by subtracting the current actual yaw rate from the rated maximum yaw rate. The real-time fatigue torque is obtained by multiplying the real-time current of the yaw motor by the motor torque constant, and the yaw execution capability state point is generated. Extract the sum of angular velocity requirements of all retained modes in the wake dynamic evolution matrix, subtract the remaining available yaw rate margin to calculate the kinematic tracking conflict residual, and divide the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index, and output the bandwidth and load conflict state vector. When the kinematic tracking conflict residual is greater than zero, a modal order reduction operation is performed to successively reduce the modal truncation order and remove the highest angular frequency data until the recalculated kinematic tracking conflict residual is less than or equal to zero. Based on all fluid modes retained after order reduction, a time-domain reconstruction is performed to generate a smooth target yaw angle command. The difference between the smoothed target yaw angle command and the actual cabin yaw angle is input into the proportional-integral controller to output the yaw motor drive signal. When the dynamic load risk index is greater than 0.8, the underlying limiting logic is triggered to reduce the output amplitude of the drive signal, thereby driving the physical cabin attitude to change.

2. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 1, characterized in that: The steps for collecting the actual wind field state, nacelle spatial position, and yaw motor electromagnetic state include: setting a global unified time reference t to keep the sampling period consistent with the control period; the lidar optical receiver scanning and analyzing the actual wind field state according to the time reference t to obtain wind speed data at different spatial coordinate points; the yaw gear absolute encoder sampling the nacelle spatial position at the time reference t to obtain the real-time orientation data of the nacelle; the Hall sensor inside the yaw driver sampling the current of the yaw motor electromagnetic state according to the time reference t to obtain the real-time current data of the motor windings; and outputting the line-of-sight wind speed spatiotemporal matrix, the actual nacelle yaw angle, and the real-time current of the yaw motor, wherein the spatial resolution of the line-of-sight wind speed spatiotemporal matrix is ​​set to 2 meters, and the detection range covers -200 meters to 200 meters along the x-axis and 0 meters to 400 meters along the y-axis.

3. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 2, characterized in that: The step of performing dynamic mode decomposition calculation on the line-of-sight wind speed spatiotemporal matrix based on the initially set modal truncation order includes: receiving the line-of-sight wind speed spatiotemporal matrix, wherein the initially set modal truncation order is denoted as k; the sensing end controller constructs a snapshot matrix from the spatial wind speed data in the line-of-sight wind speed spatiotemporal matrix at different times, and obtains the oscillation frequency and corresponding amplitude of all fluid modes through singular value decomposition and linear mapping; combining the modal truncation order k, selecting the top k fluid modes with the highest energy proportion and retaining their corresponding oscillation parameters; and outputting the wake dynamic evolution matrix, wherein the wake dynamic evolution matrix includes the angular frequency and the target angular amplitude of the i-th order fluid mode.

4. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 3, characterized in that: The steps for generating the yaw execution capability state point include: receiving the actual cabin yaw angle and the real-time current of the yaw motor, and obtaining the preset rated maximum yaw rate and the motor torque constant; the actuator controller performs first-order time derivative of the actual cabin yaw angle and takes the absolute value to obtain the current actual yaw rate; subtracting the current actual yaw rate from the rated maximum yaw rate to calculate the remaining available yaw rate margin; simultaneously, multiplying the real-time current of the yaw motor by the motor torque constant to calculate the real-time fatigue torque; and outputting the yaw execution capability state point, which is composed of the current actual yaw rate and the real-time fatigue torque.

5. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 4, characterized in that: The steps for outputting the bandwidth and load conflict state vector include: receiving the wake dynamic evolution matrix and the yaw execution capability state point, and obtaining the preset ultimate fatigue torque; the sensing end controller extracts the product of the target angular amplitude and the angular frequency of all retained modes in the wake dynamic evolution matrix, takes the absolute value, and sums them to obtain the total angular velocity requirement; subtracts the remaining available yaw rate margin from the total angular velocity requirement to calculate the kinematic tracking conflict residual; simultaneously, divides the real-time fatigue torque by the ultimate fatigue torque to calculate the dynamic load risk index; and outputs the bandwidth and load conflict state vector composed of the kinematic tracking conflict residual and the dynamic load risk index.

6. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 5, characterized in that: The steps for generating a smooth target yaw angle command include: receiving the bandwidth and load conflict state vector, the wake dynamic evolution matrix, and the modal truncation order k; determining whether the kinematic tracking conflict residual is greater than zero; if it is greater than zero, performing the modal order reduction operation, successively reducing the value of the modal truncation order k by 1 each time, and simultaneously removing the row of data with the highest angular frequency from the wake dynamic evolution matrix until the recalculated kinematic tracking conflict residual is less than or equal to zero, thus obtaining the updated modal truncation order; if the kinematic tracking conflict residual is less than or equal to zero in the initial state, keeping the original modal truncation order unchanged; based on all fluid modes retained after order reduction, performing time-domain reconstruction according to the angular frequency of each mode and the target angle amplitude, retaining the phase information of each mode during the reconstruction process; and outputting the smooth target yaw angle command and the updated modal truncation order.

7. The wake tracking method for wind-measuring lidar based on dynamic mode decomposition according to claim 6, characterized in that: The steps for changing the attitude of the physical cabin include: receiving the smoothed target yaw angle command, the actual cabin yaw angle, and the dynamic load risk index; the actuator controller subtracts the smoothed target yaw angle command from the actual cabin yaw angle to obtain the yaw angle error, inputs the yaw angle error into the proportional-integral controller for adjustment, and outputs the drive signal of the yaw motor; continuously monitoring the dynamic load risk index, if the value is greater than 0.8, triggering the underlying limiting logic, proportionally reducing the output amplitude of the drive signal, and reducing the output torque of the motor; driving the attitude of the physical cabin to change, so that the actual cabin yaw angle gradually approaches the smoothed target yaw angle command, thus completing the control.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the wind lidar wake tracking method based on dynamic mode decomposition as described in any one of claims 1-7.

9. An electronic device, characterized in that, Includes a memory for storing instructions; and a processor for executing the instructions, causing the device to perform the wake tracking method for wind-measuring lidar based on dynamic mode decomposition as described in any one of claims 1 to 7.

Images