Adaptive Yaw Optimization Control Method for Wind Turbines Driven by Multi-Source Meteorological Data
By combining multi-beam lidar and numerical weather prediction with deep learning flow field reconstruction technology, a multi-objective optimization model was constructed, enabling adaptive yaw control of wind turbines under complex wind conditions. This solved the problems of insufficient perception accuracy and adaptive decision-making ability in existing technologies, and improved the operating efficiency and reliability of wind turbines.
Patent Information
- Application Number
- CN202611086456.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing wind turbine yaw control technology has significant limitations in terms of sensing accuracy, adaptive decision-making ability, and multi-source information fusion, making it difficult to meet the requirements of high efficiency, high reliability, and low operation and maintenance costs. In particular, it is difficult to achieve accurate wind alignment and stable operation under complex wind conditions.
A multi-beam lidar system is used to acquire real-time three-dimensional wind field data. Combined with macro-meteorological information from a numerical weather prediction system, the flow field is reconstructed using a deep learning network. A multi-objective optimization function is constructed and an improved particle swarm optimization algorithm is used for optimization. Adaptive yaw control is achieved by combining the fuzzy proportional-integral-derivative control algorithm.
It improves the coverage and prediction accuracy of wind farm perception, enhances the adaptability of yaw decision-making, reduces fatigue wear of yaw actuators, extends the service life of the unit, and improves the reliability and overall benefits of wind turbine units in complex operating scenarios.
Smart Images

Figure CN123109943A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wind power generation technology, and in particular to an adaptive yaw optimization control method and device for wind turbines driven by multi-source meteorological data. Background Technology
[0002] With the transformation of the global energy structure, wind power, as a clean and renewable energy source, has occupied an increasingly important strategic position in modern power systems. As the core equipment for wind energy capture and conversion, the operating efficiency and safety of wind turbines directly affect the economic benefits of wind farms and the stable operation of the power grid. In complex and ever-changing natural environments, wind turbines require sophisticated control systems to cope with diverse weather conditions in order to maximize wind energy utilization while effectively reducing fatigue loads on the mechanical structure and ensuring reliable operation throughout the unit's entire life cycle.
[0003] The yaw control system is a crucial component in wind turbines' ability to regulate wind capture. Its core function is to drive the yaw motor to rotate the nacelle, ensuring the rotor's swept surface remains perpendicular to the incoming flow direction. This maximizes wind energy capture and reduces asymmetric aerodynamic loads caused by yaw errors. However, existing yaw control technologies have revealed several shortcomings in engineering practice. Specifically, information sensing methods are limited, relying excessively on single-point wind speed and direction data obtained from anemometers on the nacelle. This method is significantly affected by turbulence interference caused by blade rotation, and measurement errors and time lags are difficult to effectively avoid. Furthermore, control strategies often trigger start-stop actions based on fixed yaw error thresholds, lacking dynamic adaptive capabilities to rapid wind direction changes or flow field evolution under complex terrain conditions. This results in frequent actions and wind deviations, exacerbating fatigue damage to the drivetrain and increasing energy consumption in the yaw drive mechanism. In addition, the multi-source heterogeneous sensing data fusion and processing mechanism is not yet sound, making it difficult to effectively integrate multi-dimensional information such as lidar wind measurement, numerical weather forecasting, and collaborative observation of surrounding units. This results in insufficient accuracy in predicting the evolution trend of incoming flow, making it difficult to support forward-looking yaw decisions.
[0004] It is evident that existing yaw control methods for wind turbines have significant limitations in terms of sensing accuracy, adaptive decision-making capabilities, and the fusion and utilization of multi-source information, making it difficult to meet the comprehensive requirements of modern large-scale wind turbines for high efficiency, high reliability, and low operation and maintenance costs. Therefore, there is an urgent need for a yaw control method that can deeply integrate multi-source meteorological data and possess adaptive optimization capabilities to overcome the bottlenecks of traditional technologies and achieve accurate wind alignment and stable operation of wind turbines under complex wind conditions. Summary of the Invention
[0005] The main objective of this invention is to provide an adaptive yaw optimization control method for wind turbines driven by multi-source meteorological data.
[0006] Another objective of this invention is to propose an adaptive yaw optimization control device for wind turbines driven by multi-source meteorological data.
[0007] The third objective of this invention is to provide an electronic device.
[0008] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.
[0009] To achieve the above objectives, a first aspect of the present invention proposes a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines, comprising:
[0010] The system uses a multi-beam lidar system to acquire real-time three-dimensional wind field data within a predetermined distance range, and collects physical parameters reflecting the operating status of the wind turbine through a data acquisition and monitoring control system. At the same time, it synchronously accesses macro-meteorological forecast messages issued by an external numerical weather prediction system. The real-time three-dimensional wind field data and the physical parameters are preprocessed to remove abnormal pulse points and perform multi-dimensional linear completion. At the same time, meteorological data with different spatial resolutions are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine. A deep learning network is used to extract spatiotemporal features from the preprocessed real-time three-dimensional wind field data and physical parameters. The continuity constraints and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane. Based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator. An improved particle swarm optimization algorithm is used to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions. Based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor, and the changes in yaw error are monitored in real time. The control parameters are dynamically adjusted through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
[0011] Optionally, a multi-beam lidar system is used to acquire real-time three-dimensional wind field data within a predetermined distance range, and a data acquisition and monitoring control system is used to collect physical parameters reflecting the operating status of the wind turbine. Simultaneously, macro-meteorological forecasts from an external numerical weather prediction system are accessed, including: The real-time three-dimensional wind field data is acquired using a multi-beam lidar system. The multi-beam lidar system adopts a multi-beam scanning mode and sets up multiple detection surfaces at different predetermined distances in front of the wind turbine. Each detection surface is distributed with a preset number of sampling points. The wind speed component of each sampling point in the laser line of sight is calculated by measuring the Doppler frequency shift of the laser echo at each sampling point, and the wind speed component is converted into three-dimensional vector wind field data based on geometric relationships. The data acquisition and monitoring control system extracts physical parameters that reflect the operating status of the wind turbine, including generator speed, blade pitch angle, nacelle position angle, transmission chain vibration frequency, and current output power. The system synchronously accesses macro-meteorological forecast messages published by external numerical weather prediction systems. These messages include data on horizontal wind speed, wind direction, atmospheric pressure, temperature, and humidity for a predetermined time period.
[0012] Optionally, the real-time three-dimensional wind field data and the physical parameters are preprocessed to remove abnormal pulse points and perform multi-dimensional linear completion. Simultaneously, meteorological data with different spatial resolutions are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine, including: Abnormal pulse points in the real-time three-dimensional wind field data and physical parameters are identified using statistical criteria based on the standard deviation multiple. When the deviation of a data point from the moving average exceeds the standard deviation multiple of the preset multiple, it is judged as abnormal and removed. Multidimensional linear completion is performed on the missing data generated after removal. The real-time three-dimensional wind field data, the physical parameters, and the meteorological data in the macro-meteorological forecast report are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine, thus completing the spatial reference alignment of data with different spatial resolutions.
[0013] Optionally, a deep learning network is used to extract spatiotemporal features from the preprocessed real-time 3D wind field data and physical parameters. The continuity and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane, including: The long short-term memory model in deep learning network is used to extract spatiotemporal features from preprocessed real-time three-dimensional wind field data and physical parameters. The long short-term memory model includes an input layer, multiple hidden layers and an output layer. Each hidden layer contains a preset number of neurons and uses a preset activation function. The continuity equation and momentum equation in the fluid dynamics equations are embedded as physical constraints of the deep learning network in the spatiotemporal feature extraction process. The real-time three-dimensional wind field data is transformed into a three-dimensional vector matrix that conforms to physical laws, and the equivalent wind speed distribution and inflow angle field covering the wind turbine swept surface grid are reconstructed. When reconstructing the inflow angle field, the wind shear index is obtained in real time according to the wind speed fitting formula in the height direction, and the wind speed at different heights is corrected vertically based on the wind shear pointer. At the same time, the compensation coefficient of the tower shielding effect is obtained through the preset correction table, and the grid wind speed near the tower area is nonlinearly corrected.
[0014] Optionally, based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that balances maximizing capture power and minimizing fatigue load on the yaw actuator. An improved particle swarm optimization algorithm is then used to perform iterative optimization within a constraint space defined by yaw speed limits and maximum yaw angle deviation, searching for the optimal yaw target angle under the current wind conditions, including: Based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator. The multi-objective optimization function includes a power gain term calculated based on the wind energy utilization coefficient curve and a load loss term evaluated based on the equivalent fatigue load model. The power gain term and the load loss term are weighted and combined into a single optimization objective through weighting coefficients. An improved particle swarm optimization algorithm is used to perform optimization iterations within a constraint space consisting of yaw speed limits and maximum yaw angle deviations. The improved particle swarm optimization algorithm adopts an adaptive weight adjustment strategy, in which the inertial weight decreases linearly from a preset starting value to a preset ending value during the optimization process. A maximum number of iterations and a preset optimization accuracy are set. In each iteration, the position and velocity of the particles are updated according to the evaluation results of the multi-objective optimization function until the convergence condition is met, and the optimal yaw target angle under the current wind conditions is searched. During the optimization and iteration process, the yaw rate of the wind turbine is limited to a preset angular velocity threshold, and the yaw acceleration is limited to a preset angular acceleration threshold. The minimum trigger threshold for a single yaw action is set to a preset angle value, and the maximum yaw error is limited to a preset error range.
[0015] Optionally, based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor, and changes in yaw error are monitored in real time. The control parameters are dynamically adjusted using a fuzzy proportional-integral-derivative control algorithm through a closed-loop feedback mechanism, including: Based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor. The yaw inverter uses vector control technology to adjust the output of the yaw motor. The yaw error is monitored in real time, and a fuzzy proportional-integral-derivative control algorithm is adopted. Based on the real-time yaw error and its rate of change, the proportional coefficient, integral coefficient and derivative coefficient are adjusted online through a preset number of control rules. The closed-loop feedback is double-verified using data from the nacelle position encoder and the anemometer. When the deviation between the angle fed back by the nacelle position encoder and the command angle exceeds the preset deviation threshold, the compensation calibration program is automatically started to achieve adaptive yaw control of the wind turbine.
[0016] To achieve the above objectives, a second aspect of the present invention provides a multi-source meteorological data-driven adaptive yaw optimization control device for wind turbines, comprising: The data acquisition module is used to acquire real-time three-dimensional wind field data within a predetermined distance range using a multi-beam lidar system, and to acquire physical parameters reflecting the operating status of the wind turbine through the data acquisition and monitoring control system, while simultaneously accessing macro-meteorological forecast messages issued by an external numerical weather prediction system. The data preprocessing module is used to preprocess the real-time three-dimensional wind field data and the physical parameters, remove abnormal pulse points and perform multi-dimensional linear completion, and at the same time map meteorological data with different spatial resolutions to a three-dimensional Cartesian coordinate system centered on the wind turbine. The wind field reconstruction module is used to extract spatiotemporal features from preprocessed real-time three-dimensional wind field data and physical parameters using a deep learning network. It also embeds the continuity constraints and momentum constraints of the fluid dynamics equations as physical constraint terms into the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane. The optimization decision module is used to construct a multi-objective optimization function that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator based on the reconstructed equivalent wind speed distribution and inflow angle field. An improved particle swarm optimization algorithm is used to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions. The execution control module is used to generate control commands based on the optimal yaw target angle, drive the yaw inverter to adjust the speed and torque of the yaw motor, monitor the changes in yaw error in real time, and dynamically adjust the control parameters through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
[0017] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.
[0018] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, for implementing a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines as described in the first aspect embodiment.
[0019] To achieve the above objectives, the fourth aspect of this application proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines as described in the first aspect embodiment.
[0020] The embodiments of the present invention have the following beneficial effects: (1) This invention integrates the three-dimensional wind measurement data of lidar, the real-time operating parameters of the unit and the macro-meteorological information of numerical weather forecast. Through the deep learning flow field reconstruction technology under spatiotemporal alignment and physical constraints, it greatly expands the coverage of wind field perception, effectively improves the prediction accuracy of the incoming wind field, and provides data support for yaw decision that is more in line with the actual flow field characteristics.
[0021] (2) The present invention constructs a multi-objective optimization model that takes into account both maximizing power capture and minimizing fatigue load. By improving the particle swarm algorithm, it dynamically seeks optimization in the yaw operation constraint space. Combined with fuzzy PID adaptive control and dual-verification closed-loop feedback mechanism, the yaw control parameters can be dynamically adjusted according to real-time wind conditions, which effectively improves the adaptability of yaw response under different operating scenarios.
[0022] (3) In the optimization process, the present invention clearly sets boundary constraints for yaw angular velocity and angular acceleration, and embeds a yaw action anti-shaking trigger mechanism in the execution logic. While improving the energy capture level of the wind turbine, it can effectively reduce the asymmetric load and fatigue loss of the yaw actuator, reduce the frequency of operation and maintenance of the yaw system, and extend the overall service life of the unit.
[0023] (4) This invention realizes the localized deployment of data processing and algorithm reasoning through a multi-core heterogeneous edge computing gateway, and combines a hardware neural network acceleration unit to realize low-latency processing of flow field reconstruction and optimization calculation. At the same time, it can access the medium and long-term wind trend data of numerical weather forecast, realize the coordination of short-term wind response and long-term operation scheduling, and further improve the overall reliability and comprehensive benefits of wind turbine in complex operation scenarios. Attached Figure Description
[0024] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines provided in an embodiment of the present invention; Figure 2 This is a diagram of the overall hardware architecture for adaptive yaw control of a multi-source weather-driven wind turbine provided in an embodiment of the present invention. Figure 3A schematic diagram illustrating the principle framework of physical constraint deep learning for 3D flow field reconstruction provided in this embodiment of the invention; Figure 4 A flowchart of the multi-objective optimization adaptive yaw closed-loop control method provided in this embodiment of the invention; Figure 5 This is a structural diagram of a wind turbine adaptive yaw optimization control device driven by multi-source meteorological data, provided in an embodiment of the present invention. Detailed Implementation
[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0027] The following description, with reference to the accompanying drawings, describes a multi-source meteorological data-driven adaptive yaw optimization control method and apparatus for wind turbines according to embodiments of the present invention.
[0028] Example 1 This invention provides a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines. Figure 1 This is a flowchart illustrating a multi-source meteorological data-driven adaptive yaw optimization control method for wind turbines, as provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S1: Use a multi-beam lidar system to acquire real-time three-dimensional wind field data within a predetermined distance range, and collect physical parameters reflecting the operating status of the wind turbine through a data acquisition and monitoring control system, while simultaneously accessing macro-meteorological forecast messages issued by an external numerical weather prediction system.
[0029] This step is the fundamental step for data acquisition and input in the adaptive yaw control method of this application, and corresponds to, as follows: Figure 2 The core process of inputting multi-source heterogeneous data from three data sources—a multi-beam lidar wind measurement device, a main control system database, and a numerical weather prediction system—to the edge computing gateway, while simultaneously matching... Figure 3 The front-end "multi-source heterogeneous data input" unit provides comprehensive and reliable raw data support for subsequent data preprocessing, flow field reconstruction, yaw optimization, and closed-loop control. For example... Figure 2As shown, the complete hardware system built in this application includes seven core modules: a multi-beam lidar wind measurement device, a numerical weather prediction system, a main control system database, an edge computing gateway, a wind turbine main control system, a yaw drive inverter group, and a yaw actuator motor. These modules cooperate to form a complete data acquisition, processing control, and signal feedback transmission link. In this embodiment, step S1 mainly completes the synchronous acquisition and real-time transmission of three types of front-end core data. The overall signal transmission logic is clear and hierarchical: the raw three-dimensional wind field data collected by the multi-beam lidar wind measurement device is unidirectionally and in real-time input to the edge computing gateway; the macro-meteorological forecast messages output by the numerical weather prediction system are stably connected to the edge computing gateway; and the historical operating data and real-time status parameters of the unit stored in the main control system database are also synchronously transmitted to the edge computing gateway. The three types of data sources maintain a parallel synchronous acquisition mode and are uniformly merged into the edge computing gateway for caching and storage, preparing for subsequent multi-source data fusion processing.
[0030] In terms of hardware selection and communication interaction, this application embodiment has made targeted optimizations to ensure high accuracy and real-time performance of data acquisition. This solution uses a 1550nm wavelength fiber laser as the core light source for the multi-beam lidar wind measurement device. This light source has strong coherence and excellent resistance to environmental interference, making it suitable for the complex outdoor operating environment of wind farms. The device is equipped with a high-precision MEMS scanning mirror, enabling stable synchronous scanning of four beams. The laser pulse repetition frequency is fixed at 10kHz, effectively covering a large airspace 50 to 500 meters in front of the wind turbine. It can divide the airspace in front of the turbine into multiple virtual detection sections, capturing inflow wind field information from all directions. The radar equipment establishes a communication link with an edge computing gateway via Gigabit Ethernet combined with industrial Ethernet protocol, enabling high-speed, lossless transmission of the original Doppler shift detection signal, avoiding data transmission delay and packet loss issues.
[0031] Meanwhile, the edge computing gateway, as the core computing power carrier of this system, undertakes the entire process of data processing and algorithm calculation. In this embodiment, the gateway adopts a multi-core heterogeneous processor with a main frequency of 1.8GHz, paired with 16GB DDR4 high-speed memory, providing sufficient hardware computing power to meet the real-time computing needs of large-scale wind farm data. The gateway also features a dedicated hardware neural network acceleration unit, which can adapt to subsequent complex deep learning inference tasks for flow field reconstruction, significantly improving algorithm efficiency. In terms of communication, the gateway establishes a stable interactive channel with the wind turbine main control system via a dual-redundant CAN bus, ensuring reliable bidirectional transmission of control commands and status data. Simultaneously, it connects to the wind farm's backbone fiber optic network through a secure isolation gateway, stably and continuously acquiring meteorological data output by the numerical weather prediction system while ensuring network security isolation.
[0032] Regarding the three-dimensional wind field data acquisition logic of the multi-beam lidar system, this application embodiment adopts a full-coverage multi-beam scanning detection mode. Specifically, the system sequentially deploys multiple independent wind field detection surfaces at different predetermined distances in front of the wind turbine, and a preset number of wind speed sampling points are evenly distributed within each detection surface to ensure no blind spots in the detection airspace and uniform distribution of sampling data. During normal operation, the system continuously measures the Doppler frequency shift value generated by the laser echo signal at each sampling point in real time. Based on the Doppler wind measurement principle, the one-dimensional wind speed component along the laser line of sight propagation direction at the corresponding sampling point is first calculated using the frequency shift data. Then, combined with the spatial geometric conversion relationship of laser emission and reflection, the one-dimensional wind speed component is vector decomposed and synthesized, ultimately converting it into complete three-dimensional vector wind field data, accurately covering the real wind field information of multiple layers of space in front of the turbine.
[0033] Regarding the collection of physical parameters of wind turbine operation, this application relies on the turbine's supporting data acquisition and monitoring control system to achieve automated real-time data collection and aggregation. The system can accurately extract five core types of turbine operation physical parameters: real-time generator speed, uniform average blade pitch angle of the three blades, current absolute position angle of the nacelle, real-time vibration frequency of the drive train, and current real-time output power of the turbine. These parameters comprehensively reflect the real-time operating conditions, transmission status, and power generation performance of the wind turbine. All collected turbine operation physical parameters are stored in real time. Figure 2 The main control system database shows the archived data, and the edge computing gateway can retrieve historical and real-time data from the database in real time according to the algorithm's operation requirements, so as to realize the fusion analysis of multi-time series data.
[0034] For the access and processing of external macro-meteorological data, the edge computing gateway can continuously receive standardized forecast messages in GRIB2 format from external numerical weather prediction systems around the clock. These messages contain refined meteorological evolution data for a predetermined time period within the wind farm area, corely covering five key meteorological indicators: horizontal average wind speed at 100 meters altitude, wind direction gradient, regional atmospheric pressure, air temperature, and air humidity. Compared to the close-range real-time micro-wind field data collected by lidar, the large-scale macro-meteorological data provided by numerical weather prediction systems can reflect the overall trend of regional wind field changes. It can serve as background field data for distant wind fields, effectively compensating for the limitations of close-range detection data. This provides important supplementary constraints for subsequent flow field trend prediction and wind field reconstruction optimization, improving the environmental adaptability of the overall control algorithm.
[0035] Step S2 involves preprocessing the real-time three-dimensional wind field data and the physical parameters, removing abnormal pulse points and performing multi-dimensional linear completion, and simultaneously mapping meteorological data of different spatial resolutions to a three-dimensional Cartesian coordinate system centered on the wind turbine.
[0036] This step corresponds to Figure 3 The data preprocessing and spatiotemporal alignment module in the core principle framework of the flow field reconstruction shown is a key intermediate link connecting the raw data acquisition and wind field reconstruction calculations. All processing calculations are deployed in... Figure 2 This process is completed within the edge computing gateway shown, leveraging the gateway's high-speed memory and parallel computing capabilities to achieve standardized calibration of massive amounts of multi-source data. The lidar 3D wind field data, unit operating physical parameters, and macro-meteorological forecast data acquired through S1 suffer from issues such as inconsistent sampling frequencies, inconsistent spatial references, environmental noise interference, and missing local data. These issues prevent direct input into deep learning models for computation. Therefore, this step is necessary to perform anomaly cleaning, data completion, time synchronization, and spatial registration to achieve standardization and homogenization of all data sources.
[0037] Regarding abnormal data cleaning and missing data completion, this application employs a statistical discrimination criterion based on standard deviation multiples to identify abnormal pulse points in the original data, adapting to the noise suppression requirements under complex wind farm conditions. The system calculates the moving average of each data sequence in real time, comparing the deviation of a single frame of data with the moving average point by point. When the deviation of a data point exceeds a preset standard deviation multiple, it can be determined that the data is abnormal pulse data caused by bird obstruction, rain and snow interference, or instantaneous equipment vibration, and the abnormal point is then removed. For the local data gaps caused by abnormal removal, this application uses a multidimensional linear completion algorithm, combining the numerical change patterns of adjacent effective sampling points in the spatiotemporal dimension for fitting and completion, ensuring the continuity of temporal data and the integrity of spatial data, and avoiding data loss that leads to distortion in subsequent flow field reconstruction.
[0038] Regarding multi-source data time alignment, this application employs the IEEE 1588 Precision Time Protocol to unify the clocks of all hardware devices within the system, synchronizing the clocks of the lidar, main control system database, edge computing gateway, and numerical weather prediction system to microsecond-level precision, thus completely eliminating timing misalignment issues caused by sampling clock deviations between different devices. Simultaneously, considering the inherent physical delay in the propagation of the remotely detected wind field from the lidar detection surface to the wind turbine plane, this application introduces a dedicated delay compensation calculation formula to achieve dynamic timing calibration. The specific calculation formula is as follows:
[0039] In the formula, This represents the wind field propagation delay time, in seconds. This represents the straight-line distance between the lidar detection surface and the wind turbine plane, in meters. This represents the horizontal component of the inflow wind speed, measured in meters per second. The system dynamically calculates the delay time based on real-time wind speed detection, shifting the remote detection data backward by the corresponding delay time on the time axis. This achieves precise timing matching between long-distance wind field information and the current operating status of the wind turbine, eliminating control deviations caused by leading data.
[0040] At the level of multi-source data spatial registration, due to significant differences in spatial resolution and spatial reference between lidar detection data, turbine status data, and numerical weather prediction data, this application achieves unified mapping of the entire data domain through a three-dimensional kriging interpolation algorithm. Specifically, all wind field data and meteorological data are uniformly converted to a three-dimensional Cartesian coordinate system with the center of the wind turbine hub as the origin, and a uniform grid resolution of 2m×2m is set. Discrete detection point data are interpolated into regular grid data covering the entire area in front of the wind turbine, completely solving the problem of inconsistent spatial scales among different data sources and laying a spatial data foundation for subsequent full-coverage wind field reconstruction.
[0041] Step S3: A deep learning network is used to extract spatiotemporal features from the preprocessed real-time three-dimensional wind field data and physical parameters. The continuity constraints and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane.
[0042] This step is the core innovative computational component of the method in this application, and it fully corresponds to... Figure 3 The overall computational flow of the core principle framework for flow field reconstruction shown is based on... Figure 2 The hardware neural network acceleration unit on the edge computing gateway enables high-speed parallel computing, and the time for a single complete wind field reconstruction can be controlled within 50 milliseconds, meeting the response speed requirements of real-time yaw control of wind turbines. This step follows the standardized and spatiotemporally aligned multi-source data completed by S2 preprocessing, mines the spatiotemporal characteristics of the wind field through a physically constrained deep learning model, and corrects the data-driven results by combining real physical laws of fluid mechanics, ultimately outputting high-precision, full-coverage wind turbine swept surface flow field information.
[0043] At the spatiotemporal feature extraction level, this embodiment selects a Long Short-Term Memory (LSTM) model as the core infrastructure of the deep learning network, adapting to the characteristics of wind fields with large temporal fluctuations and strong spatiotemporal correlations. The LTM model consists of an input layer, multiple hidden layers, and an output layer. Each hidden layer contains a predetermined number of neurons, coupled with a predetermined activation function to complete nonlinear feature mapping, effectively fitting the dynamic changes of complex wind fields. Relying on the unique forget gate, input gate, and output gate structure of the LTM network, the model can autonomously select, retain, and forget temporal data features, effectively mining the short-term turbulent fluctuation characteristics and long-term wind speed change trends of three-dimensional wind field data, and accurately extracting the spatiotemporal correlation information of wind fields under multi-source data coupling.
[0044] To avoid distorted results that violate aerodynamic laws in purely data-driven models, this application incorporates the core fluid dynamics equations of continuity and momentum as dedicated physical constraints, fully embedding them into the loss function calculation process of the Long Short-Term Memory (LSTM) network. Through hard constraints based on physical rules, the network output is forcibly corrected, ensuring that the wind field data obtained from model inference strictly satisfies the laws of fluid mass and momentum conservation. This transforms the preprocessed discrete real-time three-dimensional wind field data into a continuous three-dimensional vector matrix that conforms to real physical laws, completely covering the entire swept surface grid area of the wind turbine rotor.
[0045] Building upon this foundation, this application introduces a dual environmental compensation and correction mechanism for the inflow angle field reconstruction results, further enhancing the accuracy of flow field reconstruction under complex operating conditions. The first mechanism is wind shear index compensation. The system calculates the wind shear index of the corresponding height layer in real time based on the wind speed fitting formula in the vertical direction of the unit. Using this wind shear index, the system performs vertical gradient correction on the wind speed at grid points at different heights, restoring the true wind shear characteristics of the atmospheric boundary layer. The second mechanism is tower shielding effect compensation. The system retrieves a pre-stored tower shielding correction coefficient lookup table and matches the corresponding nonlinear compensation coefficient based on the relative spatial position of the grid points and the tower. This corrects the airflow speed and inflow angle in the tower-shielded area, eliminating flow field distortion errors caused by tower bypass and airflow obstruction. After these dual corrections, the system ultimately outputs accurate equivalent wind speed distribution and inflow angle field in the rotor sweep plane, providing reliable flow field input conditions for subsequent yaw optimization decisions.
[0046] Step S4: Based on the reconstructed equivalent wind speed distribution and inflow angle field, construct a multi-objective optimization function that balances maximizing capture power and minimizing fatigue load on the yaw actuator. Use an improved particle swarm optimization algorithm to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions.
[0047] This step corresponds to Figure 4 The flow field analysis, optimization modeling, and algorithm optimization stages in this application utilize high-precision reconstructed flow field data output from S3. Multi-objective optimization modeling and constraint optimization calculations are performed within the edge computing gateway, ultimately outputting the optimal yaw target angle adapted to the current real-time wind conditions, achieving a bidirectional balance between power generation efficiency and equipment lifespan. Traditional yaw control often focuses solely on wind energy capture efficiency, easily leading to frequent yaw mechanism movements and severe fatigue load accumulation. This application constructs a dual-objective balance optimization function to achieve a dynamic balance between power generation revenue and equipment losses.
[0048] The complete expression of the multi-objective optimization function constructed in this application is as follows:
[0049] In the formula, To comprehensively optimize the objective function evaluation value, the smaller the value, the better the overall control performance of the current yaw angle scheme; This is the power gain weighting coefficient, used to adjust the weight ratio of the power generation efficiency target in the optimization process; This is the power gain term, calculated in real time based on the wind turbine's inherent wind energy utilization coefficient curve. It directly reflects the wind energy capture capability under the current yaw state, and the higher the value, the higher the power generation efficiency. This is the load loss weighting coefficient, used to adjust the weighting ratio of the fatigue loss target of the yaw mechanism; As a load loss term, the mechanical fatigue loss caused by yaw motion is quantitatively evaluated using an equivalent fatigue load model; For real-time operation of the unit's blade tip speed ratio; To ensure real-time uniformity of the blade pitch angle; This refers to the real-time yaw error angle of the cabin. The gyroscopic torque caused by yaw error; This refers to the aerodynamic torque generated by the asymmetric airflow of the wind turbine. The two weighting coefficients can be flexibly adjusted according to the wind farm's operation and maintenance strategy to adapt to different operational needs, such as prioritizing power generation or extending service life.
[0050] To improve optimization accuracy and convergence stability, this application employs an improved particle swarm optimization algorithm to perform iterative optimization within the constrained space. In this embodiment, the particle population size is set to 50, with each particle corresponding to a set of independent yaw angle candidate solutions, ensuring the completeness of the optimization space traversal. The algorithm introduces an adaptive inertia weight adjustment strategy. During the optimization iteration process, the inertia weight decreases linearly from a preset starting value to a preset ending value. In the early stages of iteration, a larger inertia weight is used to expand the global search range and avoid getting trapped in local optima; in the later stages of iteration, a smaller inertia weight is used to improve the local fine-grained search capability and accelerate the convergence speed of the optimal solution. Simultaneously, the algorithm pre-configures a maximum number of iterations and a preset optimization accuracy threshold. In each iteration, the position and velocity of all particles are updated based on the evaluation value of the multi-objective optimization function until the convergence accuracy or maximum iteration condition is met, at which point the computation terminates, locking in the optimal yaw angle solution.
[0051] To ensure the safety and smoothness of yaw actions and avoid damage to equipment due to operation exceeding thresholds, this application applies hard boundary constraints throughout the iterative optimization process. Specific constraints include: limiting the yaw angular velocity of the wind turbine to a preset threshold of no more than 0.5° / s, and limiting the yaw angular acceleration to a preset threshold of no more than 0.2° / s²; simultaneously setting a minimum trigger angle threshold for a single yaw action to prevent invalid yaw caused by minute angular deviations; and limiting the maximum permissible yaw error range of the system to control the steady-state yaw deviation within a reasonable range, thereby minimizing the frequency of yaw mechanism start-up and shutdown and load impact while ensuring power generation accuracy.
[0052] Step S5: Based on the optimal yaw target angle, a control command is generated to drive the yaw inverter to adjust the speed and torque of the yaw motor, and the change of yaw error is monitored in real time. The control parameters are dynamically adjusted through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
[0053] This step is the final execution and closed-loop feedback stage of the entire control method, corresponding to... Figure 2 The hardware architecture's control drive and signal feedback links are fully covered. Figure 4 The illustrated adaptive control, mechanism execution, state feedback, and closed-loop iteration process, through a combination of hardware execution and software control, achieves adaptive and precise yaw control of the wind turbine under all operating conditions, forming a complete closed-loop system of perception, calculation, control, and feedback. The optimal yaw target angle obtained through S4 optimization is sent to the wind turbine's main control system via a dual-redundant CAN bus through an edge computing gateway. The main control system, equipped with a real-time operating system PLC controller, completes instruction parsing and logic scheduling.
[0054] The main control system generates precise drive commands based on the optimal yaw target angle and sends the commands to the yaw drive frequency converter group. This application is configured with 4 or 6 AC asynchronous yaw actuator motors and corresponding frequency converters. The frequency converters adopt vector control technology, which can accurately adjust the real-time output speed and torque of the yaw motor. Combined with pulse width modulation technology, it realizes smooth start, uniform speed steering and smooth braking in the yaw process, completely avoiding the speed fluctuation and torque impact problems of traditional control methods, and reducing mechanical vibration and fatigue wear.
[0055] To adapt to the dynamic control requirements under complex wind conditions, this application employs a fuzzy proportional-integral-derivative (FID) control algorithm to achieve adaptive adjustment of control parameters. The system collects two core feedback quantities in real time: the cabin yaw error value and the error change rate. Based on preset fuzzy control rules, it dynamically adjusts the proportional, integral, and derivative coefficients online. When the actual cabin angle deviates significantly from the target angle, the algorithm automatically increases the proportional coefficient, rapidly improving the system response speed and shortening the adjustment time. When the cabin position approaches the optimal target angle, the derivative coefficient is increased, effectively suppressing angle overshoot and oscillation. The integral coefficient continuously eliminates static deviations, ensuring steady-state control accuracy and achieving full-range dynamic adaptive control.
[0056] This application establishes a dual-verification closed-loop feedback mechanism to ensure the accuracy and reliability of yaw control. The system synchronously collects real-time data from the nacelle position encoder and the anemometer on the top of the nacelle for bidirectional verification. When the deviation between the actual nacelle angle reported by the nacelle position encoder and the angle commanded by the main control exceeds a preset deviation threshold, the system automatically initiates a built-in compensation calibration program to correct the drive command in real time and eliminate static control errors. Simultaneously, the photoelectric encoder integrated at the tail of the yaw motor transmits motor angle data back in real time. If the system detects a static error that continuously exceeds 0.2 degrees, it will automatically trigger compensation logic to further improve control accuracy.
[0057] To address the frequent start-stop issues of the yaw mechanism caused by short-term turbulence and minor wind fluctuations, and to reduce equipment fatigue, this application incorporates software-based anti-shake control logic. The system only initiates the yaw action when the difference between the optimal yaw target angle and the current actual angle of the nacelle remains stable for more than 3 seconds, and the angle deviation exceeds a preset trigger threshold of 3 degrees. This effectively filters out invalid yaw commands caused by transient wind disturbances, significantly improving the mechanism's operational stability and service life.
[0058] At the closed-loop iteration level, all operational status data during yaw execution, including the nacelle's real-time position, motor operating torque, unit output power, and drivetrain vibration, are transmitted back to the wind turbine's main control system in real time and simultaneously pushed to the edge computing gateway. This serves as the input data source for the next round of S1 multi-source data acquisition, enabling continuous iterative optimization of wind farm perception, flow field reconstruction, yaw optimization, control execution, and status feedback. Simultaneously, all operational data, control parameters, and flow field reconstruction results are stored in the main control system's database in real time, providing data support for subsequent algorithm iteration optimization, unit status monitoring, and operation and maintenance analysis. The entire system can adaptively adapt to complex and ever-changing wind farm operating conditions over the long term.
[0059] Example 2 This invention provides a multi-source meteorological data-driven adaptive yaw optimization control device for wind turbines. Figure 5 This is a schematic diagram of a multi-source meteorological data-driven adaptive yaw optimization control device for wind turbines, provided in an embodiment of the present invention. Figure 5 As shown, the system includes: The data acquisition module 100 is used to acquire real-time three-dimensional wind field data within a predetermined distance range using a multi-beam lidar system, and to acquire physical parameters reflecting the operating status of the wind turbine through the data acquisition and monitoring control system, while simultaneously accessing macro-meteorological forecast messages issued by an external numerical weather prediction system. The data preprocessing module 200 is used to preprocess the real-time three-dimensional wind field data and the physical parameters, remove abnormal pulse points and perform multi-dimensional linear completion, and at the same time map meteorological data with different spatial resolutions to a three-dimensional Cartesian coordinate system centered on the wind turbine. The wind field reconstruction module 300 is used to extract spatiotemporal features from preprocessed real-time three-dimensional wind field data and physical parameters using a deep learning network. The continuity constraints and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane. The optimization decision module 400 is used to construct a multi-objective optimization function that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator based on the reconstructed equivalent wind speed distribution and inflow angle field. An improved particle swarm optimization algorithm is used to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions. The execution control module 500 is used to generate control commands based on the optimal yaw target angle, drive the yaw frequency converter to adjust the speed and torque of the yaw motor, monitor the changes in yaw error in real time, and dynamically adjust the control parameters through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
[0060] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.
[0061] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.
[0062] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.
[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0064] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0065] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
Claims
1. A multi-source meteorological data-driven adaptive yaw optimization control method for wind turbine generators, characterized in that, include: The system uses a multi-beam lidar system to acquire real-time three-dimensional wind field data within a predetermined distance range, and collects physical parameters reflecting the operating status of the wind turbine through a data acquisition and monitoring control system. At the same time, it synchronously accesses macro-meteorological forecast messages issued by an external numerical weather prediction system. The real-time three-dimensional wind field data and the physical parameters are preprocessed to remove abnormal pulse points and perform multi-dimensional linear completion. At the same time, meteorological data with different spatial resolutions are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine. A deep learning network is used to extract spatiotemporal features from the preprocessed real-time three-dimensional wind field data and physical parameters. The continuity constraints and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane. Based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator. An improved particle swarm optimization algorithm is used to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions. Based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor, and the changes in yaw error are monitored in real time. The control parameters are dynamically adjusted through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
2. The method according to claim 1, characterized in that, A multi-beam lidar system is used to acquire real-time three-dimensional wind field data within a predetermined distance range. A data acquisition and monitoring control system collects physical parameters reflecting the operating status of the wind turbine generators. Simultaneously, macro-meteorological forecasts from an external numerical weather prediction system are accessed, including: The real-time three-dimensional wind field data is acquired using a multi-beam lidar system. The multi-beam lidar system adopts a multi-beam scanning mode and sets up multiple detection surfaces at different predetermined distances in front of the wind turbine. Each detection surface is distributed with a preset number of sampling points. The wind speed component of each sampling point in the laser line of sight is calculated by measuring the Doppler frequency shift of the laser echo at each sampling point, and the wind speed component is converted into three-dimensional vector wind field data based on geometric relationships. The data acquisition and monitoring control system extracts physical parameters that reflect the operating status of the wind turbine, including generator speed, blade pitch angle, nacelle position angle, transmission chain vibration frequency, and current output power. The system synchronously accesses macro-meteorological forecast messages published by external numerical weather prediction systems. These messages include data on horizontal wind speed, wind direction, atmospheric pressure, temperature, and humidity within a predetermined time period.
3. The method according to claim 2, characterized in that, The real-time three-dimensional wind field data and the physical parameters are preprocessed to remove abnormal pulse points and perform multi-dimensional linear completion. Simultaneously, meteorological data with different spatial resolutions are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine, including: An abnormal pulse point in the real-time three-dimensional wind field data and physical parameters is identified by a statistical criterion based on the standard deviation multiple. When the deviation of a data point from the moving average exceeds the standard deviation multiple of the preset multiple, it is judged as abnormal and removed. The missing data generated after removal is filled by multidimensional linear completion. The real-time three-dimensional wind field data, the physical parameters, and the meteorological data in the macro-meteorological forecast report are uniformly mapped to a three-dimensional Cartesian coordinate system centered on the wind turbine, thus completing the spatial reference alignment of data with different spatial resolutions.
4. The method according to claim 3, characterized in that, A deep learning network is used to extract spatiotemporal features from preprocessed real-time 3D wind field data and physical parameters. The continuity and momentum constraints of the fluid dynamics equations are embedded as physical constraint terms in the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane, including: The long short-term memory model in deep learning network is used to extract spatiotemporal features from preprocessed real-time three-dimensional wind field data and physical parameters. The long short-term memory model includes an input layer, multiple hidden layers and an output layer. Each hidden layer contains a preset number of neurons and uses a preset activation function. The continuity equation and momentum equation in the fluid dynamics equations are embedded as physical constraints of the deep learning network in the spatiotemporal feature extraction process. The real-time three-dimensional wind field data is transformed into a three-dimensional vector matrix that conforms to physical laws, and the equivalent wind speed distribution and inflow angle field covering the wind turbine swept surface grid are reconstructed. When reconstructing the inflow angle field, the wind shear index is obtained in real time according to the wind speed fitting formula in the height direction, and the wind speed at different heights is corrected vertically based on the wind shear pointer. At the same time, the compensation coefficient of the tower shielding effect is obtained through the preset correction table, and the grid wind speed near the tower area is nonlinearly corrected.
5. The method according to claim 4, characterized in that, Based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that balances maximizing capture power and minimizing fatigue load on the yaw actuator. An improved particle swarm optimization algorithm is then used to perform iterative optimization within a constraint space defined by yaw speed limits and maximum yaw angle deviation, searching for the optimal yaw target angle under the current wind conditions, including: Based on the reconstructed equivalent wind speed distribution and inflow angle field, a multi-objective optimization function is constructed that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator. The multi-objective optimization function includes a power gain term calculated based on the wind energy utilization coefficient curve and a load loss term evaluated based on the equivalent fatigue load model. The power gain term and the load loss term are weighted and combined into a single optimization objective through weighting coefficients. An improved particle swarm optimization algorithm is used to perform optimization iterations within a constraint space consisting of yaw speed limits and maximum yaw angle deviations. The improved particle swarm optimization algorithm adopts an adaptive weight adjustment strategy, in which the inertial weight decreases linearly from a preset starting value to a preset ending value during the optimization process. A maximum number of iterations and a preset optimization accuracy are set. In each iteration, the position and velocity of the particles are updated according to the evaluation results of the multi-objective optimization function until the convergence condition is met, and the optimal yaw target angle under the current wind conditions is searched. During the optimization and iteration process, the yaw rate of the wind turbine is limited to a preset angular velocity threshold, and the yaw acceleration is limited to a preset angular acceleration threshold. The minimum trigger threshold for a single yaw action is set to a preset angle value, and the maximum yaw error is limited to a preset error range.
6. The method according to claim 5, characterized in that, Based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor, and the changes in yaw error are monitored in real time. The control parameters are dynamically adjusted through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm, including: Based on the optimal yaw target angle, control commands are generated to drive the yaw inverter to adjust the speed and torque of the yaw motor. The yaw inverter uses vector control technology to adjust the output of the yaw motor. The yaw error is monitored in real time, and a fuzzy proportional-integral-derivative control algorithm is adopted. Based on the real-time yaw error and its rate of change, the proportional coefficient, integral coefficient and derivative coefficient are adjusted online through a preset number of control rules. The closed-loop feedback is double-verified using data from the nacelle position encoder and the anemometer. When the deviation between the angle fed back by the nacelle position encoder and the command angle exceeds the preset deviation threshold, the compensation calibration program is automatically started to achieve adaptive yaw control of the wind turbine.
7. A multi-source meteorological data-driven adaptive yaw optimization control device for wind turbine generators, characterized in that, include: The data acquisition module is used to acquire real-time three-dimensional wind field data within a predetermined distance range using a multi-beam lidar system, and to acquire physical parameters reflecting the operating status of the wind turbine through the data acquisition and monitoring control system, while simultaneously accessing macro-meteorological forecast messages issued by an external numerical weather prediction system. The data preprocessing module is used to preprocess the real-time three-dimensional wind field data and the physical parameters, remove abnormal pulse points and perform multi-dimensional linear completion, and at the same time map meteorological data with different spatial resolutions to a three-dimensional Cartesian coordinate system centered on the wind turbine. The wind field reconstruction module is used to extract spatiotemporal features from preprocessed real-time three-dimensional wind field data and physical parameters using a deep learning network. It also embeds the continuity constraints and momentum constraints of the fluid dynamics equations as physical constraint terms into the feature extraction process to reconstruct the equivalent wind speed distribution and inflow angle field of the wind turbine sweep plane. The optimization decision module is used to construct a multi-objective optimization function that takes into account both maximizing capture power and minimizing fatigue load on the yaw actuator based on the reconstructed equivalent wind speed distribution and inflow angle field. An improved particle swarm optimization algorithm is used to perform optimization iteration within the constraint space consisting of yaw speed limit and maximum yaw angle deviation to search for the optimal yaw target angle under the current wind conditions. The execution control module is used to generate control commands based on the optimal yaw target angle, drive the yaw inverter to adjust the speed and torque of the yaw motor, monitor the changes in yaw error in real time, and dynamically adjust the control parameters through a closed-loop feedback mechanism using a fuzzy proportional-integral-derivative control algorithm to achieve adaptive yaw control of the wind turbine.
8. The apparatus according to claim 7, characterized in that, The data acquisition module is also used for: The real-time three-dimensional wind field data is acquired using a multi-beam lidar system. The multi-beam lidar system adopts a multi-beam scanning mode and sets up multiple detection surfaces at different predetermined distances in front of the wind turbine. Each detection surface is distributed with a preset number of sampling points. The wind speed component of each sampling point in the laser line of sight is calculated by measuring the Doppler frequency shift of the laser echo at each sampling point, and the wind speed component is converted into three-dimensional vector wind field data based on geometric relationships. The data acquisition and monitoring control system extracts physical parameters that reflect the operating status of the wind turbine, including generator speed, blade pitch angle, nacelle position angle, transmission chain vibration frequency, and current output power. The system synchronously accesses macro-meteorological forecast messages published by external numerical weather prediction systems. These messages include data on horizontal wind speed, wind direction, atmospheric pressure, temperature, and humidity within a predetermined time period.
9. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-6.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.
