Parameter self-adaptive wind turbine wake prediction method and system under complex environment

Abstract

The application discloses a wind turbine wake prediction method and system with parameter self-adaption under complex environment, generates a typical environment scene based on a first data set, optimizes model parameters of a selected wake analysis model based on a second data set under each typical environment scene, obtains an optimal model parameter combination, takes the optimal model parameter combination as an output, takes a first data set corresponding to the typical environment scene as an input, constructs an environment variable-model parameter mapping function, obtains first environment parameters and first unit measured data of a wind turbine of a target offshore wind farm, and performs wake prediction on the wind turbine based on the first environment parameters and the first unit measured data in combination with the environment variable-model parameter mapping function and the wake analysis model, so that adaptive update of key model parameters of the wake analysis model is realized according to actual wind field environment, and thus the wake prediction precision and engineering applicability are improved.

Description

Technical Field

[0001] This invention relates to the field of wake analysis technology, and in particular to a method and system for predicting the wake of wind turbines with adaptive parameters under complex environments. Background Technology

[0002] As the global energy structure accelerates its transition to a low-carbon model, offshore wind power, with its high wind speeds, high utilization hours, and vast development potential, has become a crucial supporting force in building new power systems. With the continuous expansion of offshore wind power installed capacity, the wake effect between wind turbines is becoming increasingly prominent. The wake effect not only significantly reduces the inflow wind speed to downstream turbines but also causes enhanced turbulence and increased turbine loads, resulting in an overall power generation loss for wind farms typically as high as 10%-20%. Against this backdrop, how to quickly and accurately simulate the wake distribution and evolution patterns to provide reliable input for wind farm layout optimization, condition monitoring, and active control has become key to improving overall power generation efficiency and reducing the cost per kilowatt-hour. Therefore, developing a rapid wake calculation method for offshore wind farms that balances high accuracy and high efficiency has become an urgent need to overcome industry bottlenecks and support the operation of intelligent wind farms.

[0003] Analytical wake models, with their advantages of computational efficiency, simple structure, and clear physical meaning of parameters, occupy a core position in wind farm planning, simulation, and operation control. Compared with CFD (Computational Fluid Dynamics) models that require long-term iterative solutions, analytical wake models can quickly estimate wake velocity deficit, wake diffusion, and turbulence changes within seconds, meeting the needs of various scenarios such as wind farm layout optimization, energy assessment, real-time scheduling, and online control. Typical Jensen wake models, Gaussian wake models, and their improved models can describe wake attenuation coefficients, wake expansion trends, and turbulence reconstruction characteristics with a few key parameters, combining computational efficiency and interpretability. In addition, analytical wake models are easy to couple with optimization algorithms, intelligent control strategies, and surrogate models, providing real-time predictions in large-scale wind farm simulation, yaw optimization, wake control, and digital twin systems, making them an important foundational tool for achieving coordinated control of the entire wind farm.

[0004] However, offshore wind farms operate in complex meteorological environments. Their wind profiles, turbulence intensity, stability, and sea surface roughness exhibit strong time-varying characteristics with seasonal and weather events, leading to more complex evolution patterns in single-unit wakes. Existing wake analysis models generally employ fixed empirical parameters, typically derived from onshore wind farm experiments, ideal boundary layers, or steady-state inflow conditions. These models struggle to reflect the impact of seasonal variations, atmospheric stability changes, and extreme wind conditions on offshore wind farms. In this context, fixed parameters not only reduce the model's generalization ability in real marine environments but also significantly amplify wake prediction errors with environmental disturbances. Therefore, the core bottleneck of current wake analysis models lies not in the model structure but in the lack of a parameter calibration mechanism that can dynamically adjust to environmental variables. Thus, it is necessary to propose a wind turbine wake analysis model with adaptive parameters for complex environments to significantly improve engineering applicability and prediction accuracy. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide a method and system for predicting the wake of wind turbines with adaptive parameters under complex environments, which can improve the accuracy of wake prediction and its engineering applicability.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for predicting the wake of wind turbines under complex environments with adaptive parameters, comprising the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

[0007] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A wind turbine wake prediction system with parameter adaptation in complex environments includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

[0008] The beneficial effects of this invention are as follows: A first dataset for classifying typical environmental scenarios and a second dataset for model parameter calibration are obtained from collected multi-source monitoring data. Typical environmental scenarios are generated based on the first dataset. Under each typical environmental scenario, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal model parameter combination. The optimal model parameter combination is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environmental variable-model parameter mapping function. The first environmental parameters and the measured data of the first turbine of the target offshore wind farm are obtained. Based on the first environmental parameters and the measured data of the first turbine, the wake of the wind turbine is predicted using the environmental variable-model parameter mapping function and the wake analysis model to obtain the prediction result. Thus, the environmental variable-model parameter mapping function is constructed using the collected dataset. The environmental variable-model parameter mapping function can output the optimal model parameters according to the input environmental parameters. Wake prediction is performed by combining the environmental variable-model parameter mapping function and the wake analysis model, realizing the adaptive updating of the key model parameters of the wake analysis model with changes in the actual wind farm environment, thereby improving the wake prediction accuracy and engineering applicability. Attached Figure Description

[0009] Figure 1 A flowchart of a wind turbine wake prediction method with parameter adaptation under complex environments, according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a wind turbine wake prediction system with adaptive parameters in a complex environment, according to an embodiment of the present invention. Figure 3 This is a data logic diagram of a wind turbine wake prediction method with parameter adaptation in a complex environment, according to an embodiment of the present invention. Detailed Implementation

[0010] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0011] Before detailing the embodiments of this application, some related concepts will first be explained: Wake analysis model: A major category of wind turbine wake models, it is based on idealized assumptions or fitting experimental data, and uses mathematical expressions to describe the wind speed distribution and attenuation characteristics of the wake region.

[0012] In existing technologies, current wake analysis models generally employ fixed empirical parameters, typically obtained from onshore wind farm experiments, ideal boundary layers, or steady-state inflow conditions. These parameters fail to reflect the impact of environmental factors on offshore wind farms, such as seasonal variations, atmospheric stability changes, and extreme wind conditions. In this context, fixed parameters not only reduce the model's generalization ability in real-world marine environments but also significantly amplify wake prediction errors with environmental disturbances. Therefore, the core bottleneck of current wake analysis models lies not in the model structure, but in the lack of a parameter calibration mechanism that can dynamically adjust to environmental variables.

[0013] To at least solve the above problems, please refer to Figure 1 This invention provides a method for predicting the wake of wind turbines under complex environments with adaptive parameters, including the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

[0014] As can be seen from the above description, the beneficial effects of the present invention are as follows: A first dataset for classifying typical environmental scenarios and a second dataset for model parameter calibration are obtained from the collected multi-source monitoring data. Typical environmental scenarios are generated based on the first dataset. Under each typical environmental scenario, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal model parameter combination. The optimal model parameter combination is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environmental variable-model parameter mapping function. The first environmental parameters and the measured data of the first turbine of the target offshore wind farm are obtained. Based on the first environmental parameters and the measured data of the first turbine, the wake of the wind turbine is predicted using the environmental variable-model parameter mapping function and the wake analysis model to obtain the prediction result. Thus, the environmental variable-model parameter mapping function is constructed using the collected dataset. The environmental variable-model parameter mapping function can output the optimal model parameters according to the input environmental parameters. Wake prediction is performed by combining the environmental variable-model parameter mapping function and the wake analysis model, realizing the adaptive updating of the key model parameters of the wake analysis model with changes in the actual wind farm environment, thereby improving the wake prediction accuracy and engineering applicability.

[0015] Furthermore, the multi-source monitoring data includes a second environmental parameter and second unit measured data; The first dataset for typical environmental scenario segmentation and the second dataset for model parameter calibration are obtained from the multi-source monitoring data, including: The multi-source monitoring data is aligned to obtain aligned multi-source monitoring data; Missing values ​​in the aligned multi-source monitoring data are filled in to obtain filled multi-source monitoring data. The filled multi-source monitoring data is cleaned to obtain cleaned multi-source monitoring data; A first dataset for classifying typical environmental scenarios is generated based on the second environmental parameter in the cleaned multi-source monitoring data. A second dataset for model parameter calibration is generated based on the second environmental parameter in the cleaned multi-source monitoring data and the second unit measured data.

[0016] As described above, the multi-source monitoring data is first aligned, filled, and cleaned, and then the first and second datasets are generated from them. These datasets are used for typical environmental scenario segmentation and model parameter calibration, respectively, which effectively improves data quality and ensures the effectiveness of subsequent scenario segmentation and model parameter calibration.

[0017] Furthermore, generating typical environment scenarios based on the first dataset includes: Typical environmental scenarios are generated based on the first dataset using cluster analysis or interval-based time series classification.

[0018] As described above, when generating typical environmental scenarios, cluster analysis or interval-based time series classification can be used to generate them. This can systematically divide the environmental state within the wind farm's operating cycle and obtain several typical environmental scenarios describing different incoming flow properties, thus achieving flexible and reliable division of typical environmental scenarios.

[0019] Furthermore, the second dataset includes wind speed, turbulence intensity, yaw angle, and generator power output; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters, including: In each of the typical environmental scenarios, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to the typical environmental scenario are input into the selected wake analytical model, and the predicted power is output. Construct an error function between the predicted power and the unit's generating power; With the goal of minimizing the error value of the error function, the model parameters of the wake analysis model are optimized using an optimization method to obtain the optimal combination of model parameters corresponding to the typical environmental scenario.

[0020] As described above, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to typical environmental scenarios are input into the selected wake analysis model. The output is the predicted power, and an error function between the predicted power and the generator power is constructed. The optimization objective is to minimize the error value of the error function. The model parameters of the wake analysis model are optimized using optimization methods to obtain the optimal model parameter combination corresponding to typical environmental scenarios. This results in a set of optimal parameter samples covering different environmental scenarios, providing basic data for constructing the mapping relationship between model parameters and wind farm environment.

[0021] Furthermore, based on the first environmental parameters and the measured data of the first turbine unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine unit is predicted, and the prediction results include: The first environmental parameter is input into the environmental variable-model parameter mapping function, and the model parameter is output. Substitute the output model parameters into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the wake analysis model after the parameter update to predict the wake, and the prediction result is output.

[0022] As described above, the first environmental parameter is input into the environmental variable-model parameter mapping function, which outputs model parameters. This mapping function reflects the influence of different incoming flow characteristics on the wake development process, enabling the model parameters to be automatically calculated based on environmental variables in practical applications, rather than taking fixed values. The output model parameters are substituted into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the updated wake analysis model for wake prediction, forming a single-unit wake analysis model that can automatically update key parameters according to incoming flow conditions. This gives the prediction results the dynamic adaptability to consider changes in the actual environment, improving the accuracy of wake prediction.

[0023] Please refer to Figure 2 Another embodiment of the present invention provides a wind turbine wake prediction system with adaptive parameters in complex environments, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

[0024] As can be seen from the above description, the beneficial effects of the present invention are as follows: A first dataset for classifying typical environmental scenarios and a second dataset for model parameter calibration are obtained from the collected multi-source monitoring data. Typical environmental scenarios are generated based on the first dataset. Under each typical environmental scenario, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal model parameter combination. The optimal model parameter combination is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environmental variable-model parameter mapping function. The first environmental parameters and the measured data of the first turbine of the target offshore wind farm are obtained. Based on the first environmental parameters and the measured data of the first turbine, the wake of the wind turbine is predicted using the environmental variable-model parameter mapping function and the wake analysis model to obtain the prediction result. Thus, the environmental variable-model parameter mapping function is constructed using the collected dataset. The environmental variable-model parameter mapping function can output the optimal model parameters according to the input environmental parameters. Wake prediction is performed by combining the environmental variable-model parameter mapping function and the wake analysis model, realizing the adaptive updating of the key model parameters of the wake analysis model with changes in the actual wind farm environment, thereby improving the wake prediction accuracy and engineering applicability.

[0025] Furthermore, the multi-source monitoring data includes a second environmental parameter and second unit measured data; The first dataset for typical environmental scenario segmentation and the second dataset for model parameter calibration are obtained from the multi-source monitoring data, including: The multi-source monitoring data is aligned to obtain aligned multi-source monitoring data; Missing values ​​in the aligned multi-source monitoring data are filled in to obtain filled multi-source monitoring data. The filled multi-source monitoring data is cleaned to obtain cleaned multi-source monitoring data; A first dataset for classifying typical environmental scenarios is generated based on the second environmental parameter in the cleaned multi-source monitoring data. A second dataset for model parameter calibration is generated based on the second environmental parameter in the cleaned multi-source monitoring data and the second unit measured data.

[0026] As described above, the multi-source monitoring data is first aligned, filled, and cleaned, and then the first and second datasets are generated from them. These datasets are used for typical environmental scenario segmentation and model parameter calibration, respectively, which effectively improves data quality and ensures the effectiveness of subsequent scenario segmentation and model parameter calibration.

[0027] Furthermore, generating typical environment scenarios based on the first dataset includes: Typical environmental scenarios are generated based on the first dataset using cluster analysis or interval-based time series classification.

[0028] As described above, when generating typical environmental scenarios, cluster analysis or interval-based time series classification can be used to generate them. This can systematically divide the environmental state within the wind farm's operating cycle and obtain several typical environmental scenarios describing different incoming flow properties, thus achieving flexible and reliable division of typical environmental scenarios.

[0029] Furthermore, the second dataset includes wind speed, turbulence intensity, yaw angle, and generator power output; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters, including: In each of the typical environmental scenarios, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to the typical environmental scenario are input into the selected wake analytical model, and the predicted power is output. Construct an error function between the predicted power and the unit's generating power; With the goal of minimizing the error value of the error function, the model parameters of the wake analysis model are optimized using an optimization method to obtain the optimal combination of model parameters corresponding to the typical environmental scenario.

[0030] As described above, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to typical environmental scenarios are input into the selected wake analysis model. The output is the predicted power, and an error function between the predicted power and the generator power is constructed. The optimization objective is to minimize the error value of the error function. The model parameters of the wake analysis model are optimized using optimization methods to obtain the optimal model parameter combination corresponding to typical environmental scenarios. This results in a set of optimal parameter samples covering different environmental scenarios, providing basic data for constructing the mapping relationship between model parameters and wind farm environment.

[0031] Furthermore, based on the first environmental parameters and the measured data of the first turbine unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine unit is predicted, and the prediction results include: The first environmental parameter is input into the environmental variable-model parameter mapping function, and the model parameter is output. Substitute the output model parameters into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the wake analysis model after the parameter update to predict the wake, and the prediction result is output.

[0032] As described above, the first environmental parameter is input into the environmental variable-model parameter mapping function, which outputs model parameters. This mapping function reflects the influence of different incoming flow characteristics on the wake development process, enabling the model parameters to be automatically calculated based on environmental variables in practical applications, rather than taking fixed values. The output model parameters are substituted into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the updated wake analysis model for wake prediction, forming a single-unit wake analysis model that can automatically update key parameters according to incoming flow conditions. This gives the prediction results the dynamic adaptability to consider changes in the actual environment, improving the accuracy of wake prediction.

[0033] The wind turbine wake prediction method and system with adaptive parameters under complex environments described above are applicable to wind turbine wake prediction scenarios. The following detailed implementation methods illustrate these methods: Please refer to Figure 1 One embodiment of the present invention is as follows: A method for predicting the wake of wind turbines under complex environments with adaptive parameters, comprising the following steps: S1. Collect multi-source monitoring data of the target offshore wind farm.

[0034] Among them, such as Figure 3 As shown, the multi-source monitoring data includes a second environmental parameter and second unit measured data. The second environmental parameter includes wind speed, wind direction, turbulence intensity, and wind shear index provided by a wind tower or lidar, as well as air temperature, humidity, air pressure, and sea surface temperature provided by a meteorological and sea state monitoring system, which are used to estimate atmospheric stability. The second unit measured data includes hub height wind speed, wind direction, power generation, yaw angle, and unit operating status recorded by the wind turbine data acquisition and monitoring control (SCADA) system.

[0035] The time resolution of the multi-source monitoring data is typically 1-10 minutes, and it may come from different sampling systems, so a unified time alignment process is required.

[0036] S2. Obtain a first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration from the multi-source monitoring data, such as... Figure 3 As shown, specifically including S21-S25: S21. Align the multi-source monitoring data to obtain aligned multi-source monitoring data.

[0037] Specifically, the multi-source monitoring data is aligned according to a uniform time step to obtain aligned multi-source monitoring data.

[0038] S22. Fill in the missing values ​​in the aligned multi-source monitoring data to obtain the filled multi-source monitoring data.

[0039] Specifically, time series interpolation methods are used to fill in missing values ​​in the aligned multi-source monitoring data that have time jumps or measurement gaps, to obtain filled multi-source monitoring data.

[0040] S23. Perform data cleaning on the filled multi-source monitoring data to obtain cleaned multi-source monitoring data.

[0041] Specifically, data from the supplemented multi-source monitoring data that are abnormal due to sensor malfunctions or measurements that deviate significantly from the normal range are removed to obtain cleaned multi-source monitoring data.

[0042] S24. Generate a first dataset for typical environmental scenario segmentation based on the second environmental parameter in the cleaned multi-source monitoring data, such as... Figure 3 As shown.

[0043] The first dataset includes wind speed, wind direction, turbulence intensity, temperature, humidity, wind shear index, atmospheric stability, wind direction statistical characteristics, wind speed distribution pattern, and weather process parameters.

[0044] S25. Generate a second dataset for model parameter calibration based on the second environmental parameter in the cleaned multi-source monitoring data and the second unit measured data, such as... Figure 3 As shown.

[0045] The second dataset includes wind speed, turbulence intensity, yaw angle, and generator power output. In an optional implementation, the second dataset may further include wind direction and wind shear index. That is, the second dataset is composed of data extracted from the second environmental parameters and the measured data of the second generator unit.

[0046] S3. Generate typical environment scenarios based on the first dataset.

[0047] Specifically, typical environmental scenarios are generated based on the first dataset using clustering analysis methods (such as K-means, hierarchical clustering) or interval-based time series classification methods.

[0048] In the process of generating typical environmental scenarios using cluster analysis, the first dataset is standardized to ensure that data of different dimensions can serve as clustering features in the same space. After generating typical environmental scenarios, each set of scenario samples is statistically analyzed to extract indicators such as typical wind speed profiles, wind direction shift characteristics, average turbulence levels, wind shear intensity, and stability types. In this way, long-term wind farm operation data can be divided into several typical environmental scenarios with representative environmental characteristics.

[0049] In one alternative implementation, it further includes: For each typical environmental scenario, representative operating condition information is constructed, including the frequency of scenario occurrence, typical inflow wind speed range, and power statistics of wind turbine units, to provide the environmental basis and data input required for subsequent parameter calibration.

[0050] S4. In each of the typical environmental scenarios, the model parameters of the selected wake flow analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters, specifically including S41-S43: S41. In each of the typical environmental scenarios, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to the typical environmental scenario are input into the selected wake analysis model, and the predicted power is output.

[0051] S42. Construct an error function between the predicted power and the generating power of the unit.

[0052] S43. With minimizing the error value of the error function as the optimization objective, the model parameters of the wake analysis model are optimized using an optimization method to obtain the optimal combination of model parameters corresponding to the typical environmental scenario.

[0053] The optimization methods include gradient descent, genetic algorithms, particle swarm optimization, and other optimization methods suitable for continuous parameters. The model parameters are key parameters controlling wake diffusion and velocity decay within the model; different wake analysis models have different model parameters.

[0054] Taking the Gaussian wake model (EPFL model) as an example: The basic formulas and parameters of the Gaussian wake model are as follows: ; ; In the formula, Indicates a speed loss. Indicates the incoming wind speed. This indicates the wind speed at a point downstream. Indicates the thrust coefficient. This represents the standard deviation (wake width) of the Gaussian distribution in the transverse direction. Indicates the impeller diameter, Indicates lateral distance (vertical wind direction). Indicates the downstream distance (from the unit's axis along the wind direction). This represents the standard deviation of the Gaussian distribution in the transverse direction (wake width). The wake width increases linearly with downstream distance, and is generally written as: ; In the formula, Indicates the wake diffusion coefficient. Indicates the starting position. Indicates the initial wake width; ; ; In the formula, , , , All of these are model parameters. Indicates turbulence intensity; The Gaussian model for wake deflection under yaw conditions is as follows: ; ; ; In the formula, The effective thrust coefficient indicates a reduction. This indicates the centerline of the wake that causes the deflection. The yaw angle represents the wind turbine's rotation angle, used to describe the angle of deflection of the turbine's rotor axis relative to the direction of the incoming wind. , , All of these are model parameters. This represents the basic wake deflection calculated from the model's built-in physical relationships. It does not contain any empirical correction terms and is used to characterize the theoretical deflection behavior of the wake centerline under yaw.

[0055] It can be seen that when the selected wake analysis model is the Gaussian wake model, the model parameters to be optimized are: , , , , , , .

[0056] S5. Using the optimal model parameter combination as the output and the first dataset corresponding to the typical environment scenario as the input, construct an environment variable-model parameter mapping function.

[0057] Specifically, the optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input. An environmental variable-model parameter mapping function is constructed using a multivariate regression model, a nonlinear fitting method, a kernel function method, or a function approximation method based on machine learning.

[0058] In the process of constructing the function, an appropriate function structure is selected based on the complexity of parameter changes, the optimal combination of model parameters is used for training, and methods such as cross-validation and error analysis are used to determine its suitability. The hyperparameters of the model are also adjusted as needed.

[0059] S6. Obtain the first environmental parameters and the measured data of the first unit of the wind turbine of the target offshore wind farm.

[0060] The first environmental parameter, including air temperature, humidity, and wind shear index, is sourced from a wind measurement tower, lidar, and sea state monitoring system. The first unit's measured data is sourced from the SCADA system, including hub-height wind speed, wind direction, power generation, yaw angle, and unit operating status. Both the first environmental parameter and the first unit's measured data are collected during operation, while the second environmental parameter and the second unit's measured data are historical data. In other words, they share the same data source but differ in timeliness. Figure 3 As shown.

[0061] S7. Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result, specifically including S71-S73: S71. Input the first environmental parameter into the environmental variable-model parameter mapping function and output the model parameter.

[0062] S72. Substitute the output model parameters into the wake analysis model to obtain the updated wake analysis model.

[0063] S73. Input the first environmental parameters and the measured data of the first unit into the wake analysis model after parameter update to perform wake prediction, and output the prediction result.

[0064] The prediction results include the velocity deficit field caused by the wake and the corresponding deflection position.

[0065] In summary, the wind turbine wake prediction method for complex environments described above in this invention obtains a first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration from collected multi-source monitoring data. Typical environmental scenarios are generated based on the first dataset. Under each typical environmental scenario, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal model parameter combination. The optimal model parameter combination is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environmental variable-model parameter mapping function. The first environmental parameters and measured data of the first turbine of the target offshore wind farm are obtained. Based on the first environmental parameters and the measured data of the first turbine, combined with the environmental variable-model parameter mapping function and the wake analysis model, wake prediction of the wind turbine is performed to obtain the prediction result. This method constructs an environmental variable-model parameter mapping function using the collected dataset. The environmental variable-model parameter mapping function can output optimal model parameters based on the input environmental parameters. Combining this function with the wake analysis model for wake prediction enables adaptive updates of key model parameters as the actual wind farm environment changes, thus improving wake prediction accuracy and engineering applicability. Furthermore, by inputting wind speed, turbulence intensity, and yaw angle corresponding to typical environmental scenarios into the selected wake analysis model, the predicted power is output. An error function between the predicted power and the generator's output power is constructed. Minimizing the error value of this error function is the optimization objective. Optimization methods are used to optimize the model parameters of the wake analysis model, obtaining the optimal model parameter combination corresponding to typical environmental scenarios. This yields a set of optimal parameter samples covering different environmental scenarios, providing fundamental data for constructing the mapping relationship between model parameters and the wind farm environment.

[0066] According to another aspect of the invention, Figure 2 This is a schematic diagram illustrating a wind turbine wake prediction system with adaptive parameters in a complex environment according to an embodiment of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the wind turbine wake prediction method with adaptive parameters in a complex environment as described above.

[0067] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting the wake of a wind turbine under complex environments with adaptive parameters, characterized in that, Including the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

2. The wind turbine wake prediction method with parameter adaptation under complex environments according to claim 1, characterized in that, The multi-source monitoring data includes a second environmental parameter and measured data from the second unit. The first dataset for typical environmental scenario segmentation and the second dataset for model parameter calibration are obtained from the multi-source monitoring data, including: The multi-source monitoring data is aligned to obtain aligned multi-source monitoring data; Missing values ​​in the aligned multi-source monitoring data are filled in to obtain filled multi-source monitoring data. The filled multi-source monitoring data is cleaned to obtain cleaned multi-source monitoring data; A first dataset for classifying typical environmental scenarios is generated based on the second environmental parameter in the cleaned multi-source monitoring data. A second dataset for model parameter calibration is generated based on the second environmental parameter in the cleaned multi-source monitoring data and the second unit measured data.

3. The wind turbine wake prediction method with parameter adaptation under complex environments according to claim 1, characterized in that, Typical environment scenarios generated based on the first dataset include: Typical environmental scenarios are generated based on the first dataset using cluster analysis or interval-based time series classification.

4. The wind turbine wake prediction method with parameter adaptation under complex environments according to claim 1, characterized in that, The second dataset includes wind speed, turbulence intensity, yaw angle, and generator power output; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters, including: In each of the typical environmental scenarios, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to the typical environmental scenario are input into the selected wake analytical model, and the predicted power is output. Construct an error function between the predicted power and the unit's generating power; With the goal of minimizing the error value of the error function, the model parameters of the wake analysis model are optimized using an optimization method to obtain the optimal combination of model parameters corresponding to the typical environmental scenario.

5. The wind turbine wake prediction method with parameter adaptation under complex environments according to claim 1, characterized in that, Based on the first environmental parameters and the measured data of the first turbine unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine unit is predicted, and the prediction results include: The first environmental parameter is input into the environmental variable-model parameter mapping function, and the model parameter is output. Substitute the output model parameters into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the wake analysis model after the parameter update to predict the wake, and the prediction result is output.

6. A wind turbine wake prediction system with parameter adaptation under complex environments, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Collect multi-source monitoring data of the target offshore wind farm; A first dataset for typical environmental scenario segmentation and a second dataset for model parameter calibration are obtained from the multi-source monitoring data. Generate typical environmental scenarios based on the first dataset; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters; The optimal combination of model parameters is used as the output, and the first dataset corresponding to the typical environmental scenario is used as the input to construct an environment variable-model parameter mapping function. Obtain the first environmental parameters and the measured data of the first turbine of the target offshore wind farm; Based on the first environmental parameters and the measured data of the first unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine is predicted to obtain the prediction result.

7. The wind turbine wake prediction system with parameter adaptation under complex environments according to claim 6, characterized in that, The multi-source monitoring data includes a second environmental parameter and measured data from the second unit. The first dataset for typical environmental scenario segmentation and the second dataset for model parameter calibration are obtained from the multi-source monitoring data, including: The multi-source monitoring data is aligned to obtain aligned multi-source monitoring data; Missing values ​​in the aligned multi-source monitoring data are filled in to obtain filled multi-source monitoring data. The filled multi-source monitoring data is cleaned to obtain cleaned multi-source monitoring data; A first dataset for classifying typical environmental scenarios is generated based on the second environmental parameter in the cleaned multi-source monitoring data. A second dataset for model parameter calibration is generated based on the second environmental parameter in the cleaned multi-source monitoring data and the second unit measured data.

8. The wind turbine wake prediction system with parameter adaptation under complex environments according to claim 6, characterized in that, Typical environment scenarios generated based on the first dataset include: Typical environmental scenarios are generated based on the first dataset using cluster analysis or interval-based time series classification.

9. The wind turbine wake prediction system with parameter adaptation under complex environments according to claim 6, characterized in that, The second dataset includes wind speed, turbulence intensity, yaw angle, and generator power output; In each of the typical environmental scenarios, the model parameters of the selected wake analysis model are optimized based on the second dataset to obtain the optimal combination of model parameters, including: In each of the typical environmental scenarios, the wind speed, turbulence intensity, and yaw angle from the second dataset corresponding to the typical environmental scenario are input into the selected wake analytical model, and the predicted power is output. Construct an error function between the predicted power and the unit's generating power; With the goal of minimizing the error value of the error function, the model parameters of the wake analysis model are optimized using an optimization method to obtain the optimal combination of model parameters corresponding to the typical environmental scenario.

10. The wind turbine wake prediction system with parameter adaptation under complex environments according to claim 6, characterized in that, Based on the first environmental parameters and the measured data of the first turbine unit, combined with the environmental variable-model parameter mapping function and the wake analysis model, the wake of the wind turbine unit is predicted, and the prediction results include: The first environmental parameter is input into the environmental variable-model parameter mapping function, and the model parameter is output. Substitute the output model parameters into the wake analysis model to obtain the updated wake analysis model. The first environmental parameter and the measured data of the first unit are input into the wake analysis model after the parameter update to predict the wake, and the prediction result is output.

Images