Wind turbine yaw control method and system based on lidar wind measurement data

Abstract

The application is suitable for the technical field of wind power generation equipment control, and provides a wind turbine yaw control method and system based on laser radar wind measurement data, which comprises the following steps: constructing a laser radar wind measurement data set and an equipment state data set; preprocessing the laser radar wind measurement data set; constructing a sample set based on the preprocessed wind direction data and wind speed data, and adopting a preset short-step precise prediction model and a long-step trend correction to obtain a future preset-step wind direction prediction data set; calculating a wind alignment deviation based on the wind direction prediction data set and the real-time position of the nacelle; and executing a wind speed grading refined yaw control strategy based on the wind alignment deviation according to the interval in which the current wind speed is located, to generate a yaw control instruction. The application realizes the collaborative optimization of minimizing the wind alignment deviation, reducing equipment loss and improving power generation efficiency in the yaw control of the wind turbine by combining the high precision and comprehensive spatial coverage of the multi-distance laser radar wind measurement data.

Description

Technical Field

[0001] This invention belongs to the field of wind power generation equipment control technology, and particularly relates to a wind turbine yaw control method and system based on lidar wind measurement data. Background Technology

[0002] In the field of wind power generation, yaw control, as one of the core control components of wind turbines, directly determines the wind energy capture efficiency and equipment lifespan. Efficient yaw control ensures that the wind turbine rotor is always precisely aligned with the wind direction, maximizing the utilization of wind energy resources while reducing problems such as uneven equipment load and component wear caused by wind misalignment. This is of great significance to the economic benefits and safe operation of wind farms.

[0003] Existing wind turbine yaw control schemes primarily rely on real-time data collected by wind speed and direction sensors installed at the nacelle. Yaw actions are triggered through simple threshold judgments or traditional control algorithms. However, as wind power technology evolves towards larger capacity, higher altitudes, and more complex wind farm environments, traditional solutions are increasingly revealing significant limitations: Firstly, the nacelle sensors can only collect real-time wind data near the rotor, failing to anticipate distant wind farm changes, resulting in lag in yaw actions and difficulty adapting to rapidly fluctuating wind speed and direction. Secondly, existing solutions suffer from poor data processing adaptability, relying heavily on single models for prediction, leading to error accumulation. Furthermore, the control logic is relatively crude, failing to adequately consider the coordination and adaptation between different wind farm characteristics and equipment operating states. This not only affects wind accuracy but may also increase equipment wear and shorten component lifespan due to frequent or improper yaw actions.

[0004] The development of lidar wind measurement technology has provided new possibilities for solving the above problems. It can collect wind speed and direction data at different distances in front of the wind turbine with high precision over long distances, providing real-time and forward-looking data support for yaw control.

[0005] However, the application of lidar wind measurement data in yaw control still faces bottlenecks, such as data redundancy and complex processing of multi-distance and multi-dimensional wind measurement data. Existing control systems cannot fully leverage their spatial coverage and high-precision advantages. Therefore, there is an urgent need for a simple, efficient, and versatile yaw control method to achieve effective utilization of lidar wind measurement data and a comprehensive improvement in yaw control performance. Summary of the Invention

[0006] The purpose of this invention is to provide a wind turbine yaw control method and system based on lidar wind measurement data, in order to solve the above-mentioned technical problems.

[0007] This invention is implemented as follows: a wind turbine yaw control method based on lidar wind measurement data, comprising the following steps:

[0008] Dataset construction: Extract data in time sequence according to the preset sampling period to construct lidar wind measurement dataset and equipment status dataset;

[0009] Data preprocessing: The lidar wind measurement dataset is preprocessed to obtain preprocessed wind direction and wind speed data;

[0010] Wind direction prediction: A sample set is constructed based on preprocessed wind direction and wind speed data, and a preset short step size accurate prediction model and a long step size trend correction are used to obtain a wind direction prediction dataset for the future preset step size.

[0011] Wind deviation calculation: Based on the wind direction prediction dataset and the real-time position angle of the cabin, the wind deviation is calculated;

[0012] Yaw control: Based on the current wind speed range and the wind deviation, execute a wind speed-level refined yaw control strategy and generate yaw control commands.

[0013] Furthermore, the lidar wind measurement dataset is defined as follows:

[0014] ;

[0015] in, ;T s The sampling period; for Real-time wind speed at a corresponding distance L, collected by the lidar at any given time; for The corresponding distance collected by the constant-time lidar Real-time wind direction at the location; N is the total number of sampling points; i is the index of the sampling point; k is the number of lidar detection distances; For different detection distances;

[0016] The device status dataset is defined as follows:

[0017] ;

[0018] in, This refers to the real-time position angle of the wind turbine nacelle. This refers to the cumulative operating time of the yaw motor.

[0019] Furthermore, the data preprocessing steps include:

[0020] The data from multiple measuring points within the plane at each detection distance L are weighted and integrated to obtain the integrated wind direction and wind speed data for each distance plane.

[0021] The wind direction data is divided into intervals based on the wind speed at each distance, and the integrated wind direction data is corrected based on the statistical characteristics of wind direction at multiple distances at the same time to obtain the corrected wind direction data.

[0022] Based on the correlation between each detection distance and the effect of the wind turbine, differentiated fusion weights are assigned, and multi-distance weighted fusion is performed on the corrected wind direction data and the integrated wind speed data to obtain the fused real-time wind direction and real-time wind speed.

[0023] The rate of change of wind direction is calculated based on the fused real-time wind direction. The size of the sliding window is adaptively adjusted according to the rate of change of wind direction, and the fused real-time wind direction and real-time wind speed are smoothed to obtain preprocessed wind direction and wind speed data.

[0024] Furthermore, the data preprocessing steps specifically include:

[0025] Set the number of measurement points in each distance plane as follows: The coordinates of each measuring point are (x L,j ,y L,j ), j=1,2,...,n; the real-time wind speed and real-time wind direction of the corresponding measuring points are respectively , ;

[0026] Assign differentiated weights to each measuring point ,satisfy The integrated real-time wind speed and direction for a single distance plane are obtained by weighted averaging, as follows:

[0027] ;

[0028] ;

[0029] in, , Distance Real-time wind speed and direction after planar integration;

[0030] After integrating the in-plane data for all distances, categorize by low wind speed segment High wind speed section Divide into intervals, The preset wind speed classification threshold is used, and the baseline data for outlier determination is obtained through the following calculations:

[0031] Calculate the mean of wind direction data at different distances at the same time:

[0032] ;

[0033] in, At the same time The average of wind direction data for each detection range; This refers to the wind direction data after integrating data from various distance planes.

[0034] Calculate the standard deviation of wind direction data at multiple distances at the same time:

[0035] ;

[0036] in, The standard deviation of wind direction data at k detection distances at the same time;

[0037] Based on the mean and standard deviation of the wind direction data, and combined with the segmented characteristics of wind speed, a threshold for differential anomalies is set. The low-wind-speed segment meets the following criteria: High wind speed section meets When this occurs, the wind direction data at that distance is determined to be an outlier. , This represents the outlier threshold coefficient for the corresponding interval; outliers are corrected by interpolating normal data at adjacent distances, as follows:

[0038] ;

[0039] in, This is the corrected wind direction data; , The detection distance between adjacent points of the anomaly distance; , This is normal wind direction data after integrating adjacent distance planes; if there is no abnormal distance... The nearest end or none For the farthest point, the mean of the wind direction data of two adjacent distances on the same side is used for correction;

[0040] The corrected wind direction data and the integrated wind speed data are then subjected to multi-distance weighted fusion, as follows:

[0041] ;

[0042] ;

[0043] in, The real-time wind speed after merging; Differentiated fusion weights for each distance; ; This refers to the wind speed data after integration with the corresponding distance plane; The real-time wind direction after merging; This is the corrected wind direction data for the corresponding distance;

[0044] Based on the fused real-time wind direction, calculate the rate of change of wind direction between adjacent time points:

[0045] ;

[0046] in, for The rate of change of wind direction at any given time; , They are respectively , Real-time wind direction after instantaneous data integration;

[0047] Set wind field fluctuation classification threshold ,when Time to take sliding window ,when Time to take sliding window Then, the merged real-time wind direction and real-time wind speed are smoothed:

[0048] ;

[0049] ;

[0050] in, This refers to the preprocessed wind direction data, i.e., the real-time wind direction after smoothing. To adapt to the sliding window size; for Before the moment Real-time wind direction after fusion of individual sampling points; This refers to the preprocessed wind speed data, i.e., the real-time wind speed after smoothing. for Before the moment Real-time wind direction after fusion of sampling points.

[0051] Furthermore, the wind direction prediction steps specifically include:

[0052] Constructing the sample input set Each sample is ,in The variance of the data fluctuations at each distance. The number of hours at which the sampling time was taken;

[0053] Constructing the sample output set Each sample is Where M is the preset prediction step size, for Always Wind direction forecast at any given time ;

[0054] A lightweight gradient lifter is used as a short-step, accurate prediction model to predict the focusing front. The predicted value of the step, The training is performed with the mean square error between the predicted and the true values ​​as the optimization objective.

[0055] Before output after training of short step size accurate prediction model Step prediction results Combining the wind field trend of the longest-distance wind measurement data, the subsequent... The prediction results from the first step are corrected, and a weighted fusion is performed to obtain the final wind direction prediction value:

[0056] ;

[0057] in, For the first The final wind direction forecast for the step; , To integrate weights, , This is a long-step trend prediction value based on the furthest distance wind measurement data; the final output is an M-step wind direction prediction dataset. .

[0058] Furthermore, the formula for calculating the wind deviation is as follows:

[0059] ;

[0060] in, for Time prediction Constantly deviating from the wind direction; For the first Wind direction forecast, for The real-time position angle of the cabin at any given moment.

[0061] Furthermore, the yaw control steps specifically include:

[0062] Set parameters for low and high wind speed ranges: Set a non-action threshold for the low wind speed range. Adjusting the threshold Delay time High wind speed range: set non-action threshold Adjusting the threshold Delay time Simultaneously set the interval between two predictions. M=20 steps, the mean of the prediction deviation from wind. ;

[0063] The yaw start-up decision logic is based on the statistical characteristics of the predicted wind deviation over 20 steps:

[0064] Low wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw;

[0065] High wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw;

[0066] The yaw speed is dynamically set based on the wind speed level and the average deviation from the wind predicted over 20 steps, as shown in the following formula:

[0067] ;

[0068] in, This is the final yaw speed; , Yaw speeds for different deviation levels in low wind speed ranges; , Yaw speeds for different deviation levels in high wind speed ranges;

[0069] Extreme wind condition determination employs multi-distance collaborative logic, when any Wind speed detection at distances of 1 or more Or any step in the 20-step wind speed prediction When this occurs, an emergency yaw is triggered to a safe angle. The formula for calculating the safe angle is as follows:

[0070] ;

[0071] in, For emergency yaw safety angle; This is the first step of wind direction forecast; The safe angle for lateral deflection in the downwind direction;

[0072] Wind speed after smoothing And the duration of stability When the emergency mode is exited, normal control is restored. To restore the wind speed threshold, The threshold for stable duration;

[0073] Yaw Stop Judgment: When yawing in low wind speed conditions, if the average wind deviation over the next 20 steps is predicted... Stop yawing; during high wind speed periods, if the average wind deviation over the next 20 steps is predicted... Stop yawing.

[0074] Another object of the present invention is to provide a wind turbine yaw control system based on lidar wind measurement data, comprising:

[0075] LiDAR is used to collect raw wind measurement data at different distances and measurement points in front of the wind turbine;

[0076] Auxiliary monitoring components are used to collect equipment status data;

[0077] A control computer is used to execute the wind turbine yaw control method based on lidar wind measurement data described above;

[0078] The yaw motor is used to respond to commands from the control computer to drive the nacelle to rotate around the vertical axis to adjust the wind angle, while also providing feedback on the yaw motor's operating status.

[0079] In wind power equipment control technology, in response to the problems of redundant and complex processing of lidar wind measurement data, this invention provides a precise and versatile wind turbine yaw control method by combining the high precision and comprehensive spatial coverage of lidar multi-distance wind measurement data. This method achieves synergistic optimization of minimizing wind deviation, reducing equipment losses, and improving power generation efficiency in wind turbine yaw control. Attached Figure Description

[0080] Figure 1 This is a flowchart illustrating the wind turbine yaw control method based on lidar wind measurement data provided in an embodiment of the present invention.

[0081] Figure 2 A schematic diagram of the structure of a wind turbine yaw control system based on lidar wind measurement data provided in an embodiment of the present invention.

[0082] Figure 3 This is a flowchart illustrating the wind speed classification-based refined yaw control strategy provided in an embodiment of the present invention. Detailed Implementation

[0083] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0084] like Figure 1 As shown, in one embodiment of the present invention, a wind turbine yaw control method based on lidar wind measurement data is provided, comprising the following steps:

[0085] S1. Dataset Construction: Extract data in time sequence according to the preset sampling period to construct the lidar wind measurement dataset and the equipment status dataset;

[0086] Specifically, the lidar wind measurement dataset is defined as follows:

[0087] ;

[0088] in, ;T s The sampling period; for Real-time wind speed (unit: m / s) at a corresponding distance L, collected by the lidar at any time. for The corresponding distance collected by the constant-time lidar Real-time wind direction at the location (unit: °); N is the total number of sampling points; i is the index of the sampling point; k is the number of lidar detection distances; For different detection distances;

[0089] The device status dataset is defined as follows:

[0090] ;

[0091] in, This represents the real-time position angle of the wind turbine nacelle (unit: °). The cumulative operating time of the yaw motor (unit: h).

[0092] S2. Data preprocessing: The lidar wind measurement dataset is preprocessed to obtain preprocessed wind direction and wind speed data.

[0093] Specifically, considering the spatial correlation and high-precision characteristics of lidar wind measurement data, the following core formulas are used to perform preprocessing operations such as data integration, outlier correction, and smoothing, balancing efficiency and accuracy. These include:

[0094] S21. Weighted integration of data from multiple measuring points within the plane at each detection distance L is performed to obtain integrated wind direction and wind speed data for each distance plane.

[0095] Specifically, the lidar can acquire wind speed and direction data from multiple measuring points within a plane at each detection distance L. First, the data within a single distance plane is integrated to obtain representative wind speed and direction at that distance. The number of measuring points within each distance plane is set to... The coordinates of each measuring point are (x L,j ,y L,j ), j=1,2,...,n; the real-time wind speed and real-time wind direction of the corresponding measuring points are respectively , ;

[0096] Considering the positional correlation between different measuring points in the plane and the wind turbine rotor, differentiated weights are assigned to each measuring point. (The measurement points directly in front of the impeller have higher weights, while the measurement points at the edges have lower weights, and this condition is met.) The integrated real-time wind speed and direction of a single-distance plane are obtained by weighted averaging, as shown in formulas (1) and (2) below:

[0097] (1);

[0098] (2);

[0099] in, , Distance Real-time wind speed and direction after planar integration.

[0100] S22. Divide the wind direction data into intervals based on the wind speed at each distance, and correct the integrated wind direction data based on the statistical characteristics of wind direction at multiple distances at the same time (such as mean and standard deviation) to obtain the corrected wind direction data.

[0101] Specifically, after integrating the in-plane data for all distances, it is categorized by low wind speed segment. High wind speed section Divide into intervals, The preset wind speed classification threshold is used, and the baseline data for outlier determination is obtained through the following calculations:

[0102] Calculate the mean of wind direction data at different distances at the same time:

[0103] (3);

[0104] in, At the same time The average of wind direction data for each detection range; The wind direction data is integrated from various distance planes and used to characterize the overall wind direction trend of the wind field at the current moment;

[0105] The standard deviation of wind direction data at multiple distances at the same time is calculated as shown in formula (4):

[0106] (4);

[0107] in, The standard deviation is the wind direction data of k detection distances at the same time. The larger the standard deviation, the more violent the wind direction fluctuation in the current wind field, and the more stringent the outlier judgment needs to be.

[0108] Based on the mean and standard deviation of the wind direction data, and combined with the segmented characteristics of wind speed, a threshold for differential anomalies is set. The low-wind-speed segment meets the following criteria: High wind speed section meets When this occurs, the wind direction data at that distance is determined to be an outlier. , This is the abnormal threshold coefficient for the corresponding interval; abnormal values ​​are corrected by interpolating normal data at adjacent distances, as shown in the following formula (5):

[0109] (5);

[0110] in, This is the corrected wind direction data; , The detection distance between adjacent points of the anomaly distance; , This is normal wind direction data after integrating adjacent distance planes; if there is no abnormal distance... The nearest end or none For the farthest point, the average of the wind direction data of two adjacent distances on the same side is used for correction to ensure that the corrected data matches the actual trend of the wind field.

[0111] S23. Based on the correlation between each detection distance and the effect of the wind turbine, assign differentiated fusion weights, and perform multi-distance weighted fusion on the corrected wind direction data and the integrated wind speed data to obtain the fused real-time wind direction and real-time wind speed.

[0112] Specifically, the logic for setting the differentiated fusion weights is as follows: the closer the wind measurement data is to the wind turbine, the more directly it represents the actual wind state of the impeller, so it is given a higher weight. The data from a distance is mainly used to perceive the wind field trend, so the weight is relatively low. The final weight is obtained by matching the actual characteristics of the specific wind field and must meet the constraint that the sum of the weights of each distance is 1. The formulas for multi-distance weighted fusion of the corrected wind direction data and the integrated wind speed data are shown in (6)-(7):

[0113] (6);

[0114] (7);

[0115] in, The real-time wind speed after merging; Differentiated fusion weights for each distance; ; This refers to the wind speed data after integration with the corresponding distance plane; The real-time wind direction after merging; This is the corrected wind direction data for the corresponding distance.

[0116] S24. Calculate the wind direction change rate based on the fused real-time wind direction, adaptively adjust the sliding window size according to the wind direction change rate, and smooth the fused real-time wind direction and real-time wind speed to obtain preprocessed wind direction and wind speed data.

[0117] Specifically, based on the fused real-time wind direction, the rate of change of wind direction between adjacent moments is calculated using formula (8) to determine the characteristics of wind field fluctuations:

[0118] (8);

[0119] in, for The rate of change of wind direction at any given time; , They are respectively , Real-time wind direction after instantaneous data integration;

[0120] In addition, set a threshold for classifying wind field fluctuations. ,when (Rapid fluctuation wind) Take sliding window ,when (Stable / moderately fluctuating wind) use a sliding window Then, the real-time wind direction and real-time wind speed after fusion are smoothed using formulas (9)-(10):

[0121] (9);

[0122] (10);

[0123] in, This refers to the preprocessed wind direction data, i.e., the real-time wind direction after smoothing. To adapt to the sliding window size; for Before the moment Real-time wind direction after fusion of individual sampling points; This refers to the preprocessed wind speed data, i.e., the real-time wind speed after smoothing. for Before the moment Real-time wind direction after fusion of sampling points.

[0124] S3. Wind Direction Forecasting: Based on preprocessed wind direction and speed data, a sample set is constructed. A pre-defined short-step accurate prediction model and a long-step trend correction model are used to obtain a wind direction forecast dataset for the future preset step size. Specifically, this includes:

[0125] S31. Constructing the sample input set Each sample is ,in The variance of the data fluctuations at each distance (characterizing the stability of the wind field) is given. The number of hours at which the sampling time was taken;

[0126] Constructing the sample output set Each sample is Where M is the preset prediction step size, and in this embodiment of the invention, M=20. for Always Wind direction forecast at any given time .

[0127] S32. A lightweight gradient booster (LGBM) is used as a short-step, accurate prediction model to predict the focusing front. The predicted value of the step, The training is performed with the mean squared error (MSE) between the predicted and true values ​​as the optimization objective, and the objective function is shown in Equation (11):

[0128] (11);

[0129] in, The objective function (mean squared error) of the LGBM model; For short-step accurate prediction models, the first The predicted wind direction for the step; For the first The actual wind direction value of the step.

[0130] S33, Short-Step Accurate Prediction Model Output After Training Step prediction results Combining the wind field trend of the longest-distance wind measurement data, the subsequent... The prediction results of step (M1+M2=20) are corrected, and the final wind direction prediction value is obtained by weighted fusion using formula (12):

[0131] (12);

[0132] in, For the first The final wind direction forecast for the step; , To integrate weights, The error is dynamically allocated based on the real-time prediction error. This is a long-step trend forecast based on the longest-distance wind measurement data. Specifically, it can be predicted using the current time and a short-step accurate forecasting model. Based on the wind direction data, trend change characteristics (wind direction change rate, moving average, consistency of wind direction change direction, etc.) are constructed. Then, a linear regression trend prediction model is used to fit the farthest distance wind direction sequence to predict the future. The step length trend prediction value; the final output is a 20-step wind direction prediction dataset. .

[0133] S4. Calculation of Wind Deviation: Based on the wind direction prediction dataset and the real-time position angle of the cabin, the wind deviation is calculated; specifically, the formula for calculating the wind deviation is:

[0134] (13);

[0135] in, for Time prediction Wind deviation at any time (unit: °); For the first Wind direction forecast, for The real-time position angle of the cabin at any given moment is taken as the minimum value to ensure that the deviation calculation conforms to the periodic characteristics of the wind direction of 0° and 360°.

[0136] S5. Yaw control: Based on the current wind speed range and the wind deviation, execute a wind speed graded refined yaw control strategy and generate yaw control commands.

[0137] In practical applications, the wind speed classification-based refined yaw control strategy is based on wind speed classification and wind deviation, and achieves differentiated yaw control through formulaic logic, while retaining emergency protection for extreme wind conditions. Specifically, it includes:

[0138] S51, according to low wind speed section ( ) and high wind speed section ( Setting parameters: Set the non-action threshold for low wind speed range. Adjusting the threshold Delay time High wind speed range: set non-action threshold Adjusting the threshold Delay time Simultaneously set the interval between two predictions. M=20 steps, the mean of the prediction deviation from wind. ;

[0139] The yaw start-up decision logic is based on the statistical characteristics of the predicted wind deviation over 20 steps:

[0140] Low wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw;

[0141] High wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw.

[0142] S52. Yaw speed is dynamically set based on wind speed level and the average deviation from wind prediction over 20 steps, using the following formula:

[0143] (14);

[0144] in, This is the final yaw speed; , Yaw speeds for different deviation levels in low wind speed ranges; , Yaw speeds for different deviation levels in high wind speed ranges.

[0145] S53. Extreme wind condition determination adopts multi-distance collaborative logic; when any... Wind speed detection at distances of 1 or more Or any step in the 20-step wind speed prediction When this occurs, an emergency yaw is triggered to a safe angle. The formula for calculating the safe angle is as follows:

[0146] (15);

[0147] in, For emergency yaw safety angle; This is the first step of wind direction forecast; The safe angle for lateral deflection in the downwind direction;

[0148] Wind speed after smoothing And the duration of stability When the emergency mode is exited, normal control is restored. To restore the wind speed threshold, This is the threshold for the stable duration.

[0149] S54. Yaw Stop Judgment: When yawing in low wind speed conditions, if the average wind deviation over the next 20 steps is predicted... Stop yawing; during high wind speed periods, if the average wind deviation over the next 20 steps is predicted... If the above conditions are not met, stop yawing; if not, continue yawing and repeat the judgment process of steps S4-S5.

[0150] like Figure 2 As shown, in another embodiment of the present invention, a wind turbine yaw control system based on lidar wind measurement data is also provided. This system serves as the hardware support for the above-described method, forming a functional closed loop of data acquisition, computational decision-making, action execution, and data interaction. Specifically, it includes:

[0151] LiDAR (core data acquisition component) is used to collect raw wind measurement data from different distances and measuring points in front of the wind turbine in a long-distance, multi-dimensional manner.

[0152] The auxiliary monitoring component works in conjunction with the core data acquisition component to collect equipment status data, including nacelle position sensors (collecting real-time nacelle position angles), timers (recording motor running time), etc., providing supplementary equipment status data for the control computer's logical operations and improving control accuracy;

[0153] The control computer (core computing and decision-making component) is used to execute the wind turbine yaw control method based on lidar wind measurement data. Specifically, it is responsible for receiving lidar wind measurement data and equipment status data, performing preprocessing operations such as data integration, anomaly correction, and smoothing, running the wind direction prediction model and hierarchical control logic, and finally generating control commands such as yaw start, stop, and speed adjustment.

[0154] The yaw motor (the core motion execution component) is used to respond to the instructions of the control computer, drive the nacelle to rotate around the vertical axis to adjust the wind angle, and at the same time provide feedback on the yaw motor's operating status (cumulative running time, real-time speed) to ensure precise and stable yaw actions.

[0155] Example 1: This example provides an application instance of the above-described wind turbine yaw control method. It is only an illustrative example and is not limited thereto. Specifically, it includes the following steps:

[0156] Step 1: Set the sampling period Continuous collection of wind farm operation data for 24 hours, total number of sampling points .

[0157] LiDAR wind measurement dataset ,in , Record the real-time wind speed at each distance (unit: m / s). Record the real-time wind direction at each distance (unit: °);

[0158] Device Status Core Dataset ,in Real-time position angle (unit: °) is obtained through cabin position sensors. The cumulative running time of the yaw motor is recorded using a timer (unit: h).

[0159] Step 2: Data Preprocessing

[0160] Step 2.1: Number of measurement points in each distance plane The weights of each measuring point in the plane are calibrated based on their positional correlation with the impeller as follows: weight of the measuring point directly in front of the impeller. The weights of the four measuring points directly in front are respectively: The weights of the four edge measurement points are as follows: (satisfy Wind speed classification thresholds Low wind speed section anomaly threshold coefficient Anomaly threshold coefficient in high wind speed section Wind field fluctuation classification threshold Adaptive sliding window parameters (Rapidly fluctuating wind) (Stable / moderately fluctuating wind), the fusion weights for each distance are calibrated according to the wind field characteristics. , , , Predict the total number of steps Short step size prediction steps Long step length predicts the number of steps. .

[0161] Step 2.2: Each distance There are 9 measuring points in the plane. According to the set measuring point weights, the integrated wind speed at each distance from the plane is calculated by formulas (1) and (2). ,wind direction .

[0162] Then, outlier correction was performed on the multi-distance integrated data: for each sampling time... Calculate the mean of the wind direction data at the four distances according to formula (3). Calculate the standard deviation of wind direction according to formula (4). Low wind speed range ( When the wind direction at a certain distance meets the condition: When this occurs, the outlier is corrected using adjacent distance wind direction data according to formula (5); high wind speed section ( When the wind direction at a certain distance meets the condition: When the time comes, outlier correction is completed according to formula (5).

[0163] Step 2.3: Perform weighted fusion of the corrected wind speed and wind direction data according to formulas (6) and (7) to obtain... and ; Calculate the rate of change of wind direction according to formula (8) Determine the characteristics of wind field fluctuations: when At that time, take Data smoothing is performed according to formulas (9) and (10); when At that time, take Smoothing is performed according to formulas (9) and (10) to obtain the result. and .

[0164] Step 2.4: Sample Input Set ,in , The number of hours for the sampling time (0~23); Sample output set ,in .

[0165] Step 3: Training and prediction of the multi-step wind direction prediction model:

[0166] Step 3.1: Divide the sample set into a training set and a test set in a 7:3 ratio. The training set is used to optimize the LGBM model parameters, with the mean squared error (MSE) shown in formula (11) as the objective function, focusing on the first 8 steps ( Accurate prediction; outputs the prediction results of the first 8 steps after training. .

[0167] Step 3.2: Based on the current moment and the wind direction data predicted in the previous 8 steps in Step 3.1, construct trend change characteristics (wind direction change rate, moving average, consistency of wind direction change direction, etc.), and use a linear regression trend prediction model to fit the farthest distance wind direction sequence to predict the wind direction change trend in the next 12 steps; superimpose the wind direction change output of the trend model onto the current wind direction benchmark value to generate the long-step trend prediction value corresponding to each step. The weighted fusion is performed according to formula (12), where the fusion weight is... Dynamically allocated by real-time prediction error (in this embodiment) The final output is a 20-step wind direction prediction dataset. .

[0168] Step 4: Calculate the wind deviation:

[0169] Calculate the wind deviation for each prediction step using formula (13). Considering the periodic characteristics of wind direction, take Ensure the deviation is within the range of 0°-180°, and calculate the mean of the wind deviation from the 20-step prediction. .

[0170] Step 5: As Figure 3 As shown, refined yaw control is executed:

[0171] Step 5.1: Setting control parameters:

[0172] Low wind speed section ( ): Inaction threshold Adjusting the threshold Delay time Yaw speed (The deviation is between 10° and 18°) (deviation > 18°);

[0173] High wind speed section ( ): Inaction threshold Adjusting the threshold Delay time Yaw speed (The deviation is between 8° and 15°) (deviation > 15°);

[0174] Extreme wind parameters: maximum wind speed Restore wind speed threshold Stable duration threshold Downwind side deflection safety angle ;

[0175] Other parameters: Interval between two predictions .

[0176] Step 5.2: Yaw Initiation Detection:

[0177] Low wind speed range: If Do not act; if Yaw starts after a 150-second delay, at a speed of 0.3° / s; if Yaw starts after a 150-second delay, at a speed of 0.5° / s;

[0178] High wind speed section: If Do not act; if Yaw starts after a 10-second delay, at a speed of 0.4° / s; if Yaw starts after a 10-second delay, at a speed of 0.6° / s.

[0179] Step 5.3: Emergency Response to Extreme Wind Conditions

[0180] When the wind speed detected at any two or more distances is ≥25m / s or any step of the 20-step predicted wind speed is ≥25m / s, the emergency yaw safety angle is calculated according to formula (15). The emergency yaw is triggered to that angle; when the wind speed is ≤18m / s after smoothing and remains stable for more than 60s, the emergency mode is exited and normal control is restored.

[0181] Step 5.4: Yaw Stop Determination

[0182] When yawing in low wind speed conditions, if the predicted average wind deviation over the next 20 steps is less than 10°, stop yawing.

[0183] If the average wind deviation is predicted to be less than 8° when yawing in high wind speeds, stop yawing.

[0184] If the above conditions are not met, continue to yaw and repeat the judgment process of steps 4-5.

[0185] Furthermore, single-condition or short-term operating data is insufficient to fully verify the comprehensive advantages of the yaw control method based on multi-distance wind measurement data from lidar in this invention. Therefore, a comparative experiment was conducted using 100,000 sets of complete historical operating data from a wind farm (including multiple scenarios such as complex and extreme wind conditions). The results are shown in Table 1. As can be seen from Table 1, compared with the traditional yaw control method (without lidar wind measurement data), the method provided by this embodiment of the invention reduces the cumulative wind deviation by 39.28%, the cumulative operating time of the yaw motor by 67.42%, and the misjudgment rate of extreme wind conditions by 85.71%. This fully demonstrates the significant effect of this method in improving wind accuracy, reducing equipment wear and tear, and ensuring operational safety.

[0186] Table 1

[0187] Comparison indicators Method of the present invention Traditional yaw control methods Optimization range Cumulative sum of wind deviations (°) 682431.5 1156789.2 39.28% Cumulative running time of the yaw motor (h) 486.2 1492.5 67.42% Extreme wind condition misjudgment rate (%) 0.3 2.1 85.71%

[0188] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0189] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0190] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A wind turbine yaw control method based on lidar wind measurement data, characterized in that, Includes the following steps: Dataset construction: Extract data in time sequence according to the preset sampling period to construct lidar wind measurement dataset and equipment status dataset; Data preprocessing: The lidar wind measurement dataset is preprocessed to obtain preprocessed wind direction and wind speed data; Wind direction prediction: A sample set is constructed based on preprocessed wind direction and wind speed data, and a preset short step size accurate prediction model and a long step size trend correction are used to obtain a wind direction prediction dataset for the future preset step size. Wind deviation calculation: Based on the wind direction prediction dataset and the real-time position angle of the cabin, the wind deviation is calculated; Yaw control: Based on the current wind speed range and the wind deviation, execute a wind speed graded fine yaw control strategy and generate yaw control commands. The lidar wind measurement dataset is defined as follows: ； in, ;T s The sampling period; for Real-time wind speed at a corresponding distance L, collected by the lidar at any given time; for The corresponding distance collected by the constant-time lidar Real-time wind direction at the location; N is the total number of sampling points; i is the index of the sampling point; k is the number of lidar detection distances; For different detection distances; The device status dataset is defined as follows: ； in, This refers to the real-time position angle of the wind turbine nacelle. The cumulative operating time of the yaw motor; The data preprocessing steps include: The data from multiple measuring points within the plane at each detection distance L are weighted and integrated to obtain the integrated wind direction and wind speed data for each distance plane. The wind direction data is divided into intervals based on the wind speed at each distance, and the integrated wind direction data is corrected based on the statistical characteristics of wind direction at multiple distances at the same time to obtain the corrected wind direction data. Based on the correlation between each detection distance and the effect of the wind turbine, differentiated fusion weights are assigned, and multi-distance weighted fusion is performed on the corrected wind direction data and the integrated wind speed data to obtain the fused real-time wind direction and real-time wind speed. The rate of change of wind direction is calculated based on the fused real-time wind direction. The size of the sliding window is adaptively adjusted according to the rate of change of wind direction. The fused real-time wind direction and real-time wind speed are then smoothed to obtain pre-processed wind direction and wind speed data. The data preprocessing steps specifically include: Set the number of measurement points in each distance plane as follows: The coordinates of each measuring point are (x L,j ,y L,j ), j=1,2,...,n; the real-time wind speed and real-time wind direction of the corresponding measuring points are respectively , ; Assign differentiated weights to each measuring point ,satisfy The integrated real-time wind speed and direction for a single distance plane are obtained by weighted averaging, as follows: ； ； in, , Distance Real-time wind speed and direction after planar integration; After integrating the in-plane data for all distances, categorize by low wind speed segment High wind speed section Divide into intervals, The preset wind speed classification threshold is used, and the baseline data for outlier determination is obtained through the following calculations: Calculate the mean of wind direction data at different distances at the same time: ； in, At the same time The average of wind direction data for each detection range; This refers to the wind direction data after integrating data from various distance planes. Calculate the standard deviation of wind direction data at multiple distances at the same time: ； in, The standard deviation of wind direction data at k detection distances at the same time; Based on the mean and standard deviation of the wind direction data, and combined with the segmented characteristics of wind speed, a threshold for differential anomalies is set. The low-wind-speed segment meets the following criteria: High wind speed section meets When this occurs, the wind direction data at that distance is determined to be an outlier. , This represents the outlier threshold coefficient for the corresponding interval; outliers are corrected by interpolating normal data at adjacent distances, as follows: ； in, This is the corrected wind direction data; , The detection distance between adjacent points of the anomaly distance; , This is normal wind direction data after integrating adjacent distance planes; if there is no abnormal distance... The nearest end or none For the farthest point, the mean of the wind direction data of two adjacent distances on the same side is used for correction; The corrected wind direction data and the integrated wind speed data are then subjected to multi-distance weighted fusion, as follows: ； ； in, The real-time wind speed after merging; Differentiated fusion weights for each distance; ; This refers to the wind speed data after integration with the corresponding distance plane; The real-time wind direction after merging; This is the corrected wind direction data for the corresponding distance; Based on the fused real-time wind direction, calculate the rate of change of wind direction between adjacent time points: ； in, for The rate of change of wind direction at any given time; , They are respectively , Real-time wind direction after instantaneous data integration; Set wind field fluctuation classification threshold ,when Time to take sliding window ,when Time to take sliding window Then, the merged real-time wind direction and real-time wind speed are smoothed: ； ； in, This refers to the preprocessed wind direction data, i.e., the real-time wind direction after smoothing. To adapt to the sliding window size; for Before the moment Real-time wind direction after fusion of individual sampling points; This refers to the preprocessed wind speed data, i.e., the real-time wind speed after smoothing. for Before the moment Real-time wind speed obtained by merging the data from each sampling point.

2. The wind turbine yaw control method based on lidar wind measurement data according to claim 1, characterized in that, The steps for wind direction prediction specifically include: Constructing the sample input set Each sample is ,in The variance of the data fluctuations at each distance. The number of hours at which the sampling time was taken; Constructing the sample output set Each sample is Where M is the preset prediction step size, for Always The wind direction forecast at any given time. ; A lightweight gradient lifter is used as a short-step, accurate prediction model to predict the pre-focusing phase. The predicted value of the step, The training is performed with the mean square error between the predicted and the true values ​​as the optimization objective. Before output after training of short step size accurate prediction model Step prediction results Combining the wind field trend of the longest-distance wind measurement data, the subsequent... The prediction results from the first step are corrected, and a weighted fusion is performed to obtain the final wind direction prediction value: ； in, For the first The final wind direction forecast for the step; , To integrate weights, , This is a long-step trend prediction value based on the furthest distance wind measurement data; the final output is an M-step wind direction prediction dataset. .

3. The wind turbine yaw control method based on lidar wind measurement data according to claim 2, characterized in that, The formula for calculating the wind deviation is: ； in, for Time prediction Constantly deviating from the wind direction; For the first Wind direction forecast, for The real-time position angle of the cabin at any given moment.

4. The wind turbine yaw control method based on lidar wind measurement data according to claim 3, characterized in that, The yaw control steps specifically include: Set parameters for low and high wind speed ranges: Set a non-action threshold for the low wind speed range. Adjusting the threshold Delay time High wind speed range: set non-action threshold Adjusting the threshold Delay time Simultaneously set the interval between two predictions. M=20 steps, the mean of the prediction deviation from wind. ; The yaw start-up decision logic is based on the statistical characteristics of the predicted wind deviation over 20 steps: Low wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw; High wind speed range: the average value of the wind deviation predicted over 20 steps. Do not act at times; when When, delay Initiate yaw; when When, delay Initiate yaw; The yaw speed is dynamically set based on the wind speed level and the average deviation from the wind predicted over 20 steps, as shown in the following formula: ； in, This is the final yaw speed; , Yaw speeds for different deviation levels in low wind speed ranges; , Yaw speeds for different deviation levels in high wind speed ranges; Extreme wind condition determination employs multi-distance collaborative logic, when any Wind speed detection at distances of 1 or more Or any step in the 20-step wind speed prediction When this occurs, an emergency yaw is triggered to a safe angle. The formula for calculating the safe angle is as follows: ； in, For emergency yaw safety angle; This is the first step of wind direction forecast; The safe angle for lateral deflection in the downwind direction; Wind speed after smoothing And the duration of stability When the emergency mode is exited, normal control is restored. To restore the wind speed threshold, The threshold for stable duration; Yaw Stop Judgment: When yawing in low wind speed conditions, if the average wind deviation over the next 20 steps is predicted... Stop yawing; during high wind speeds, if the average wind deviation over the next 20 steps is predicted... Stop yawing.

5. A wind turbine yaw control system based on lidar wind measurement data, characterized in that, include: LiDAR is used to collect raw wind measurement data at different distances and measurement points in front of the wind turbine; Auxiliary monitoring components are used to collect equipment status data; A control computer is used to execute the wind turbine yaw control method based on lidar wind measurement data as described in any one of claims 1-4; The yaw motor is used to respond to commands from the control computer to drive the nacelle to rotate around the vertical axis to adjust the wind angle, while also providing feedback on the yaw motor's operating status.

Images