Wake flow optimization method for offshore wind plant

Abstract

The invention provides a wake flow optimization method for an offshore wind plant, and the method comprises the following steps: obtaining an initial wind turbine generator data set; on the basis, a second wind speed multiplier factor, a fourth wind turbine generator data set and turbulence intensity are obtained; acquiring a fifth wind turbine generator data set based on the data; based on the data, constructing a first FLORIS model and calculating a power change rate; based on the data and a first FLORIS model, obtaining a sixth wind turbine generator data set and a predicted total power; and acquiring actual total power based on the data, acquiring a power difference value, and if the actual total power is smaller than or equal to a preset power threshold value, transmitting a plurality of yaw angles of the sixth wind turbine generator data set to the corresponding wind turbine generators so as to realize wake flow optimization of the offshore wind plant. According to the wake flow optimization method for the offshore wind plant, wake flow prediction deviation caused by the wake flow effect of the offshore wind plant is avoided, and the generation power of the wind turbine generator of the offshore wind plant is maximized.

Description

Technical Field

[0001] This invention relates to the field of offshore wind power energy optimization technology, and in particular to a wake optimization method for offshore wind farms. Background Technology

[0002] With the accelerated global energy transition, the installed capacity of offshore wind power continues to grow. However, the wake effect commonly found in offshore wind farms leads to energy losses of 10% to 20%, severely restricting the economic benefits of wind power development. The essence of the wake effect is that upstream wind turbines alter the incoming flow field, causing downstream wind turbines to face the dual challenges of wind speed attenuation and turbulence intensification. Therefore, accurate modeling and optimized control of the wake effect are core technical challenges in the wind power field. In the field of wake modeling, the GB / T18451.2-2021 standard clarifies the methods for wind speed measurement, data screening, and uncertainty assessment in wind turbine power characteristic testing, providing a basic basis for related research. The classic Jensen model describes the wake velocity deficit based on the principle of momentum conservation, but it ignores turbulence anisotropy and lateral wake diffusion, resulting in prediction errors exceeding 15% in complex layout scenarios. The Gaussian model proposed by Bastankhah et al. improves the accuracy of vertical velocity distribution calculation by introducing the influence of yaw angle on wake width. Large eddy simulation (LES) can achieve high-precision flow field simulation, but its classification standards for acoustic anemometers and cup anemometers also provide a basis for sensor selection. Regarding yaw optimization algorithms, game theory-based distributed strategies solve for the optimal yaw angle through local interaction between wind turbines. Some studies attempt to optimize yaw parameters through SCADA data-driven optimization or combine wake models to improve energy efficiency. In the study of turbulence characteristics and wake superposition effects, early research by Frandsen et al. revealed the nonlinear relationship between turbulence intensity and incoming wind speed within the wake. The cumulative-curl model proposed in recent years attempts to improve multi-wake prediction through a nonlinear energy transfer mechanism. Furthermore, when wind turbines exhibit yaw misalignment, their thrust and power characteristics need to be remodeled to meet the accuracy requirements of the wake model.

[0003] Despite current advancements in wake modeling, yaw optimization algorithms, and the study of turbulence characteristics and wake superposition effects, several pressing issues remain in the field of wake modeling. Firstly, the classic Jensen model neglects turbulent anisotropy and lateral wake diffusion, resulting in prediction errors exceeding 15% in complex layout scenarios. While the Gaussian model improves the accuracy of vertical velocity distribution calculations, it suffers from cumulative errors due to the linear superposition assumption when multiple wakes are superimposed. Large eddy simulation (LES) can achieve high-precision flow field simulation, but its computational complexity reaches O(N³), making it difficult to meet real-time optimization requirements. Secondly, in the field of yaw optimization algorithms, game theory-based distributed strategies rely on frequent communication between wind turbines, resulting in convergence times exceeding 10 minutes in scenarios with more than 30 turbines, and insufficient adaptability to power curve differences among heterogeneous turbine models. Traditional optimization models use fixed turbulence intensity parameters, and when atmospheric turbulence intensity fluctuates by more than 20%, the prediction error of wake velocity deficit increases significantly, leading to the failure of the yaw control strategy. Furthermore, there is a lack of a safety backoff mechanism under extreme conditions; when sensors malfunction or models mismatch, it may cause yaw system overload or negative optimization (reduced power generation), and related research has not yet yielded a systematic solution. Moreover, in the field of turbulence characteristics and wake superposition effects research, existing models (such as the cumulative vorticity model) still need improvement in describing nonlinear energy transfer mechanisms, and the prediction accuracy of multiple wake superposition effects is insufficient. When wind turbines exhibit yaw misalignment, their thrust and power characteristics are not accurately modeled, affecting the overall accuracy of the wake model. Summary of the Invention

[0004] The present invention aims to provide a wake optimization method for offshore wind farms to solve the above-mentioned technical problems, avoid wake prediction deviations caused by the wake effect of offshore wind farms, and maximize the power generation of wind turbines in offshore wind farms.

[0005] To address the aforementioned technical problems, this invention provides a wake optimization method for offshore wind farms, comprising the following steps: Obtain the initial wind turbine dataset; Based on the initial wind turbine dataset, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset were obtained. Based on the fourth wind turbine dataset and the turbulence intensity dataset, the fifth wind turbine dataset is obtained; Based on the fifth wind turbine dataset, turbulence intensity dataset, and second wind speed multiplier factor, the first FLORIS model is constructed and the power change rate is calculated. Based on the fifth wind turbine dataset, power change rate and first FLORIS model, obtain the sixth wind turbine dataset and predicted total power; Based on the dataset of the sixth wind turbine, the actual total power was obtained; Based on the predicted total power and the actual total power, the power difference is obtained. If the power difference is less than or equal to a preset power threshold, several yaw angles of the sixth wind turbine dataset are transmitted to the corresponding wind turbine to achieve wake optimization of the offshore wind farm.

[0006] In the above scheme, the fifth wind turbine dataset is obtained by combining the fourth wind turbine dataset and the turbulence intensity dataset. By incorporating relevant information on turbulence characteristics, the dataset is supplemented and optimized, improving its adaptability to the actual operating conditions of the wind farm and laying a data foundation for the subsequent construction of a high-precision first FLORIS model. Next, a first FLORIS model that closely reflects the actual operating conditions is constructed using the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor. Simultaneously, the power change rate is calculated to provide model support and calculation basis for subsequent yaw angle optimization. Then, the yaw angle is optimized and filtered using the fifth wind turbine dataset, the power change rate, and the first FLORIS model to obtain a sixth wind turbine dataset suitable for the current operating conditions. The predicted total power is also calculated, providing a reference standard for verifying the subsequent optimization effect. Finally, by calculating the power difference and transmitting the corresponding yaw angle of the sixth wind turbine dataset to the wind turbine when it is less than or equal to the preset power threshold, negative optimization can be avoided, ensuring the effectiveness of the optimization strategy, avoiding wake prediction deviations caused by the wake effect of offshore wind farms, and maximizing the power generation of wind turbines in offshore wind farms.

[0007] Furthermore, it also includes: If the power difference is greater than the preset power threshold, then several yaw angles of the initial wind turbine dataset will be transmitted to the corresponding wind turbine.

[0008] In the above scheme, if the power difference is greater than the preset power threshold, applying several yaw angles from the sixth wind turbine dataset to the corresponding wind turbine will result in energy loss or equipment damage. Therefore, it is necessary to maintain several yaw angles from the initial wind turbine dataset to the corresponding wind turbine to prevent negative optimization and ensure the economy and reliability of wind turbine operation in offshore wind farms.

[0009] Furthermore, the step of obtaining the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset based on the initial wind turbine dataset includes: Based on the preset wind speed range, preset power threshold and initial wind turbine dataset, obtain the third wind turbine dataset; Based on the preset first segmentation condition and the preset power curve, several power monotonic intervals are obtained; Based on the third wind turbine dataset and several power monotonic intervals, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset are obtained.

[0010] In the above scheme, the initial wind turbine dataset is filtered by pre-set wind speed range and pre-set power threshold to obtain a third wind turbine dataset. Invalid data is filtered out to ensure that the data used in subsequent calculations conforms to the actual operating conditions of the wind turbines, laying a data foundation for accurate parameter solving. Next, the pre-set power curve is divided by a pre-set first segmentation condition to obtain several power monotonic intervals, accurately decomposing the nonlinear characteristics of the power curve and providing a reasonable division basis for subsequent piecewise interpolation calculations. Then, the third wind turbine dataset is matched with several power monotonic intervals to obtain the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset. This provides high-precision input parameters for FLORIS model configuration and yaw optimization algorithm, ensuring the accuracy of wake simulation and optimization control.

[0011] Furthermore, the step of obtaining a third wind turbine dataset based on a preset wind speed range, a preset power threshold, and an initial wind turbine dataset includes: Based on the preset wind speed range and preset power threshold, the initial wind turbine dataset is filtered to obtain the first wind turbine dataset. For any first wind turbine data in the first wind turbine dataset, the following steps are performed: Based on the first wind turbine data, calculate the initial relative wind direction angle corresponding to the first wind turbine data. If there is an initial relative wind direction angle that meets the preset first state or the initial relative wind direction angle exceeds the preset angle range, then the initial relative wind direction angle is repaired to obtain the first relative wind direction angle corresponding to the first wind turbine data. Based on the first wind turbine dataset and the first relative wind direction angles corresponding to several first wind turbine datasets, the second wind turbine dataset is obtained. Data preprocessing is performed on the second wind turbine dataset to obtain the third wind turbine dataset.

[0012] In the above scheme, the initial wind turbine dataset is filtered by setting a preset wind speed range and a preset power threshold to obtain the first wind turbine dataset. This filters out invalid data in the wind turbine dataset that exceeds the preset wind speed range or whose power does not conform to normal operating conditions, ensuring the validity of subsequent calculation data. Next, the initial relative wind direction angle corresponding to the first wind turbine data is calculated, and the initial relative wind direction angle that conforms to the preset first state or exceeds the preset angle range is corrected to obtain the first relative wind direction angle. This prevents outliers in the relative wind direction angle corresponding to the first wind turbine data and ensures that the obtained first relative wind direction angle is accurate and reasonable. Then, the first relative wind direction angles corresponding to several first wind turbine data are matched and integrated with the first wind turbine dataset to form a reliable second wind turbine dataset, providing a basis for subsequent data preprocessing and calculation. Subsequently, the second wind turbine dataset is preprocessed to obtain the third wind turbine dataset. This ensures that the obtained third wind turbine dataset can meet the subsequent calculation requirements for key parameters such as wind speed multiplier factor and turbulence intensity, providing high-precision input data for the FLORIS model dynamic configuration and yaw optimization algorithm.

[0013] Furthermore, the step of obtaining the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset based on the third wind turbine dataset and several power monotonic intervals includes: For any third wind turbine data in the third wind turbine dataset, the following steps are performed: If the power corresponding to the third wind turbine data is within any power monotonic interval, then based on the power monotonic interval and the power corresponding to the third wind turbine data, calculate the estimated wind speed corresponding to the third wind turbine data. Based on the estimated wind speed corresponding to several third wind turbine data and the third wind turbine dataset, calculate the first wind speed multiplier factor; Based on the third wind turbine dataset and the preset power curve, the first wind speed multiplier factor is corrected to obtain the second wind speed multiplier factor. Based on the second wind speed multiplier factor and the third wind turbine dataset, obtain the fourth wind turbine dataset; Based on the dataset of the fourth wind turbine, the corresponding turbulence intensity dataset was obtained.

[0014] In the above scheme, by determining whether the power corresponding to the third wind turbine data falls within any power monotonic interval, and calculating the estimated wind speed corresponding to the third wind turbine data based on the matched power monotonic interval and the corresponding power, the wind speed at different power stages can be accurately mapped. This also aligns with the piecewise characteristics of the power curve, avoiding errors caused by global fitting, and providing basic data for calculating the first wind speed multiplier factor. Next, by correlating the estimated wind speeds corresponding to several third wind turbine data points with the third wind turbine dataset, the first wind speed multiplier factor is obtained, laying the foundation for subsequent secondary corrections. Then, the first wind speed multiplier factor is corrected using the third wind turbine dataset and a preset power curve, making the wind speed multiplier factor more closely match actual operating conditions, resulting in the second wind speed multiplier factor, ensuring the accuracy of the wake model input. Subsequently, the third wind turbine dataset is fused using the second wind speed multiplier factor to form a fourth wind turbine dataset that meets the requirements of subsequent turbulence intensity calculations and dynamic wake model configuration. Finally, the turbulence intensity dataset was calculated using the dataset from the fourth wind turbine, providing key flow field parameters for the dynamic configuration of the FLORIS model and improving the accuracy of wake simulation and yaw optimization.

[0015] Further, obtaining the fifth wind turbine dataset based on the fourth wind turbine dataset and the turbulence intensity dataset includes: For any fourth wind turbine data in the fourth wind turbine dataset, the following steps are performed: obtain the initial wake width corresponding to the fourth wind turbine data; based on the turbulence intensity in the turbulence intensity dataset and the initial wake width corresponding to the fourth wind turbine data, obtain the wake width corresponding to the fourth wind turbine data. Based on the wake width corresponding to several fourth wind turbine data and the fourth wind turbine dataset, the fifth wind turbine dataset is obtained.

[0016] In the above scheme, the wake width corresponding to the fourth wind turbine data is calculated by combining turbulence intensity and initial wake width. This allows the wake width to be dynamically adjusted with turbulence intensity, overcoming the limitation of fixed wake width in traditional models and improving the accuracy of wake modeling for both teams. Next, by fusing the wake widths corresponding to several fourth wind turbine data points with the fourth wind turbine dataset, a fifth wind turbine dataset containing accurate wake width information is formed. This provides reliable flow field characteristic input for subsequent calculations of multiple wake superposition effects and yaw optimization algorithm solutions, ensuring the effectiveness of the yaw optimization strategy.

[0017] Furthermore, the construction of the first FLORIS model and calculation of the power change rate based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor includes: Based on the preset first sampling interval, the dataset of the fifth wind turbine is sampled to obtain several first sampling points; Construct the initial FLORIS model; Based on the turbulence intensity dataset, the second wind speed multiplier factor, and the initial FLORIS model, the first FLORIS model is obtained; The power change rate is calculated based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, the preset second sampling interval, and the preset disturbance step size.

[0018] In the above scheme, sampling the dataset of the fifth wind turbine unit by pre-setting a first sampling interval can narrow the calculation range and provide efficient and accurate input samples for the subsequent calculation of the power change rate, thus obtaining several first sampling points. Next, an initial FLORIS model is constructed, establishing the basic framework for wake effect simulation and providing theoretical model support for power calculation. Then, turbulence intensity and the second wind speed multiplier factor are integrated into the initial FLORIS model, enabling the obtained first FLORIS model to accurately reflect the real-time flow field characteristics and the actual operating conditions after wind speed correction, improving the accuracy of wake simulation and power calculation. Finally, through the collaborative calculation of the first FLORIS model, several first sampling points, the dataset of the fifth wind turbine unit, the pre-set second sampling interval, and the pre-set disturbance step size, the power change rate is obtained, providing data support for subsequent yaw angle adjustment and improving the accuracy and efficiency of yaw optimization.

[0019] Furthermore, the calculation of the power change rate based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, a preset second sampling interval, and a preset disturbance step size includes: Based on the first FLORIS model and several first sampling points, the wind turbine power corresponding to several first sampling points is obtained, and the interval where the wind turbine power meets the preset first condition is determined as the first yaw angle feasible region. Based on the preset second sampling interval, the dataset of the fifth wind turbine unit within the first yaw angle feasible region is sampled to obtain several second sampling points; The power change rate is calculated based on the first FLORIS model, the preset perturbation step size, and the second sampling point.

[0020] In the above scheme, the power of wind turbines corresponding to several first sampling points is calculated using the first FLORIS model, and the interval containing the first sampling points that meet the preset first condition is selected as the first yaw angle feasible region, eliminating invalid yaw angle intervals and narrowing the calculation range. Next, the dataset of the fifth wind turbine within the first yaw angle feasible region is sampled using a preset second sampling interval to obtain several second sampling points, providing input samples for accurate calculation of the power change rate. Then, through the collaborative calculation of the first FLORIS model, the preset disturbance step size, and the second sampling points, the power change rate is obtained, ensuring that the optimized yaw angle can guide the offshore wind farm's wind turbines to continuously iterate towards the power maximization goal, thereby optimizing and improving the power generation of the offshore wind farm's wind turbines.

[0021] Furthermore, the step of obtaining the sixth wind turbine dataset and predicted total power based on the fifth wind turbine dataset, power change rate, and the first FLORIS model includes: For any fifth wind turbine data in the fifth wind turbine dataset, the following steps are performed: obtain the first yaw angle corresponding to the fifth wind turbine data; obtain the second yaw angle corresponding to the fifth wind turbine data based on the yaw angle and power change rate corresponding to the fifth wind turbine data; obtain the yaw angle difference based on the first yaw angle and the second yaw angle. Based on the second yaw angle corresponding to the data of the fifth wind turbine and the dataset of the fifth wind turbine, the dataset of the sixth wind turbine is obtained. If the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold, then the predicted total power is obtained based on the first FLORIS model and the sixth wind turbine dataset.

[0022] In the above scheme, the first yaw angle corresponding to the data of the fifth wind turbine is obtained, clarifying the current yaw state of the wind turbine and providing an initial reference for subsequent yaw angle optimization. Next, the yaw angle difference between the first and second yaw angles is calculated to quantify the adjustment range of the yaw angle, facilitating rapid identification of whether the optimized yaw angle has reached a stable state. Then, the second yaw angle corresponding to the data of the fifth wind turbine is fused with the dataset of the fifth wind turbine to obtain the dataset of the sixth wind turbine, providing accurate input for the calculation of the predicted total power. Subsequently, iteration is stopped only when the yaw angle difference is less than a preset yaw angle change threshold and the power change rate is less than a preset power change threshold. At this point, the predicted total power is calculated to obtain the predicted total power of the simulated all wind turbines, avoiding errors in calculating the predicted total power in a non-converged state, ensuring the accuracy and reliability of the predicted total power, and providing a basis for subsequent power verification and triggering of the backoff mechanism.

[0023] Furthermore, it also includes: If the power change rate is greater than or equal to the preset power change threshold and the yaw angle difference is greater than or equal to the preset yaw angle change threshold, then the first FLORIS model is reconstructed based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor, and the power change rate is calculated until the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold.

[0024] In the above scheme, when the power change rate is greater than or equal to the preset power change threshold and the yaw angle difference is greater than or equal to the preset yaw angle change threshold, it indicates that the current yaw angle adjustment has not reached a stable state and there is still room for power adjustment. Therefore, the first FLORIS model is reconstructed based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor, and the power change rate is recalculated to ensure that the goal of maximizing the overall power is always achieved, avoiding getting trapped in a local optimum. Finally, the above iterative process is repeated until the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold, ensuring that the final obtained yaw angle reaches a local optimum or the overall power generation has tended to be maximized, thereby improving the economic efficiency and stability of offshore wind farm wind turbine operation. Attached Figure Description

[0025] Figure 1 A schematic flowchart of a wake optimization method for an offshore wind farm provided in an embodiment of the present invention; Figure 2 A coordinate diagram of wind turbine generators in an offshore wind farm is provided as an embodiment of the present invention; Figure 3 This is a diagram showing the optimization results of the yaw angle of a wind turbine in an offshore wind farm, provided by an embodiment of the present invention. Figure 4 An optimization iterative convergence curve for an offshore wind farm is provided in one embodiment of the present invention; Figure 5 A standard power curve and actual power scatter plot of an offshore wind farm are provided as an embodiment of the present invention; Figure 6 A power boosting box diagram of a single wind turbine in an offshore wind farm is provided as an embodiment of the present invention; Figure 7 A comparison chart of the second wind speed multiplier factor of wind turbine generators in an offshore wind farm, provided in an embodiment of the present invention; Figure 8 A comparison diagram of turbulence intensity of wind turbine generators in an offshore wind farm, provided as an embodiment of the present invention; Figure 9 This is a scatter plot showing the relationship between yaw angle and wind speed in an offshore wind farm, as provided in an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] In this embodiment, yaw optimization is performed by adjusting the yaw angle of the wind turbine to change the wake direction and diffusion range, thereby minimizing the wake loss of the downstream wind turbine and maximizing the power generation of the entire field. (1) The effect of yaw on the wake: When the yaw angle of the wind turbine is At this time, the wake center experiences a lateral shift, and the amount of this shift increases linearly with the downstream distance: Where x represents the downstream distance. Meanwhile, yaw reduces the velocity loss at the wake center and exacerbates its diffusion. For [the following]... The total power generation capacity of a wind farm with typhoon turbines It can be represented as: In the formula For the first The generating capacity of the typhoon turbine generators; Let be the yaw angle of the i-th wind turbine; The effective wind speed of the i-th wind turbine is determined by considering the upstream wake effect. The relationship between the power of a single wind turbine and the effective wind speed is given by the power curve. Sure: Constraints: 1) Yaw angle safety range: According to the design specifications of mainstream wind turbine manufacturers, the yaw angle adjustment range is limited to: 1) To avoid bearing overload and blade fatigue damage. 2) Fixed strategy for faulty wind turbines: For wind turbines that are shut down or have failed, their yaw angle is fixed at 0°.

[0028] This embodiment provides a wake optimization method for offshore wind farms. For detailed steps, please refer to [link to relevant documentation]. Figure 1 ,include: Step S1: Obtain the initial wind turbine dataset; Step S2: Based on the initial wind turbine dataset, obtain the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset; Step S3: Based on the fourth wind turbine dataset and the turbulence intensity dataset, obtain the fifth wind turbine dataset; Step S4: Based on the fifth wind turbine dataset, turbulence intensity dataset, and second wind speed multiplier factor, construct the first FLORIS model and calculate the power change rate; Step S5: Based on the fifth wind turbine dataset, power change rate, and the first FLORIS model, obtain the sixth wind turbine dataset and predicted total power; Step S6: Obtain the actual total power based on the sixth wind turbine dataset; Step S7: Based on the predicted total power and the actual total power, obtain the power difference. If the power difference is less than or equal to the preset power threshold, then transmit several yaw angles of the sixth wind turbine dataset to the corresponding wind turbine to achieve wake optimization of the offshore wind farm.

[0029] In this embodiment, key parameters such as wind speed, wind direction, and power of each wind turbine in the offshore wind farm are acquired minute-by-minute using Supervisory Control and Data Acquisition (SCADA) software. Real-time SCADA data is accessed at a 1-minute frequency to provide real-time operating condition information and obtain an initial wind turbine dataset. Simultaneously, power curves, thrust coefficients, and specific coordinates of different wind turbine models within the offshore wind farm are loaded via YAML configuration files. A fifth wind turbine dataset is obtained by combining the fourth wind turbine dataset and the turbulence intensity dataset. By incorporating relevant information on turbulence characteristics, the dataset is supplemented and optimized, improving its adaptability to the actual operating conditions of the wind farm and laying the data foundation for the subsequent construction of a high-precision first FLORIS model. Next, a first FLORIS model closely reflecting the actual operating conditions is constructed using the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor. The power change rate is also calculated to provide model support and calculation basis for subsequent yaw angle optimization adjustments. Then, by optimizing and filtering the yaw angle using the fifth wind turbine dataset, power change rate, and the first FLORIS model, a sixth wind turbine dataset suitable for the current operating conditions was obtained, and the predicted total power was calculated. This provides a reference standard for verifying the subsequent optimization effects. Finally, the power difference is calculated. : , Indicates the actual total power. Indicates the rated total power, 0.01 That is, the preset power threshold (0.01 in this embodiment). Based on historical SCADA data analysis, it was determined that when the power drop exceeds 1%, the probability of wake model mismatch or sensor malfunction exceeds 95%. At this point, the preset power threshold set has a high confidence level for fault identification; in this embodiment, 0.01 is set... This represents the upper limit of natural fluctuations in typical wind farm operation. Values ​​below this threshold may be considered normal fluctuations, while values ​​above it indicate a systematic deviation requiring immediate intervention. In this embodiment, 0.01 is set as the upper limit. This conforms to the safety margin requirements for wind farm control in IEC 61400-25-2, ensuring that backoff actions are triggered only when there is a clear negative optimization, thus avoiding frequent malfunctions. In this embodiment... The system dynamically updates with the operating cycle to ensure the backoff strategy adapts to changes in the flow field. When the yaw angle of the sixth wind turbine dataset is less than or equal to the preset power threshold, it transmits the corresponding yaw angle to the wind turbine. This avoids negative optimization, ensures the effectiveness of the optimization strategy, and prevents wake prediction errors caused by the wake effect of offshore wind farms, thereby maximizing the power generation of wind turbines in offshore wind farms. Finally, based on the operating characteristics of the wind turbines, three wind speed ranges are defined: low wind speed range (u < 3 m / s): the wind turbines are below the start-up threshold, yaw adjustment has no energy gain, and the fixed yaw angle is 0° to reduce mechanical losses; medium wind speed range ( ): In the maximum power point tracking range, the wake effect is significant. Full-range yaw optimization is activated to increase power generation by turning the wake; high wind speed area ( The wind turbine has entered constant power control mode, maintaining the current yaw angle to avoid bearing overload. In this embodiment, the operating range is divided by wind speed threshold. Only the wind turbine data of the sixth wind turbine unit that belongs to the medium wind speed range is transmitted to the corresponding wind turbine unit for yaw angle optimization. This avoids invalid yaw at low wind speeds and equipment risks at high wind speeds, thereby achieving wake optimization of offshore wind farms.

[0030] Furthermore, it also includes: If the power difference is greater than the preset power threshold, then several yaw angles of the initial wind turbine dataset will be transmitted to the corresponding wind turbine.

[0031] In this embodiment, when the power difference is greater than the preset power threshold, if several yaw angles from the sixth wind turbine dataset are applied to the corresponding wind turbine, it will result in energy loss or equipment damage. Therefore, a safety backoff is triggered to maintain several yaw angles from the initial wind turbine dataset transmitted to the corresponding wind turbine, preventing negative optimization and ensuring the economy and reliability of wind turbine operation in offshore wind farms.

[0032] Furthermore, the step of obtaining the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset based on the initial wind turbine dataset includes: Based on the preset wind speed range, preset power threshold and initial wind turbine dataset, obtain the third wind turbine dataset; Based on the preset first segmentation condition and the preset power curve, several power monotonic intervals are obtained; Based on the third wind turbine dataset and several power monotonic intervals, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset are obtained.

[0033] In this embodiment, the initial wind turbine dataset is filtered using a preset wind speed range and a preset power threshold to obtain a third wind turbine dataset. Invalid data is filtered out to ensure that the data used in subsequent calculations conforms to the actual operating conditions of the wind turbines, laying a data foundation for accurate parameter solving. Next, a first segmentation condition is preset (using the monotonically decreasing power point in the preset standard power curve as the segmentation point). , The preset power curve is divided into several monotonic power intervals (total number of segments), and the nonlinear characteristics of the power curve are accurately decomposed to provide a reasonable basis for subsequent piecewise interpolation calculations. Then, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset are obtained by matching the third wind turbine dataset with several monotonic power intervals. This provides high-precision input parameters for the FLORIS model configuration and yaw optimization algorithm, ensuring the accuracy of wake simulation and optimization control. The FLORIS (Farm Layout Optimization and Rotor Interaction Simulator) model is the mainstream engineering model for wind farm wake effect simulation. Its core is based on the law of conservation of momentum and the Gaussian turbulence assumption. By coupling the aerodynamic characteristics of wind turbines with the wake propagation law, it realizes dynamic flow field modeling.

[0034] Furthermore, the step of obtaining a third wind turbine dataset based on a preset wind speed range, a preset power threshold, and an initial wind turbine dataset includes: Based on the preset wind speed range and preset power threshold, the initial wind turbine dataset is filtered to obtain the first wind turbine dataset. For any first wind turbine data in the first wind turbine dataset, the following steps are performed: Based on the first wind turbine data, calculate the initial relative wind direction angle corresponding to the first wind turbine data. If there is an initial relative wind direction angle that meets the preset first state or the initial relative wind direction angle exceeds the preset angle range, then the initial relative wind direction angle is repaired to obtain the first relative wind direction angle corresponding to the first wind turbine data. Based on the first wind turbine dataset and the first relative wind direction angles corresponding to several first wind turbine datasets, the second wind turbine dataset is obtained. Data preprocessing is performed on the second wind turbine dataset to obtain the third wind turbine dataset.

[0035] In this embodiment, the initial wind turbine dataset is filtered by setting a preset wind speed range (preset wind speed range is [0,25] m / s) and a preset power threshold (invalid points with power less than or equal to NAN). This filters out invalid points in the initial wind turbine dataset where the measured power is NAN, or invalid points where the measured wind speed is NAN, or anomalies where the initial wind turbine data is concentrated in the standard power curve where the power is less than or equal to 0, or anomalies where the initial wind turbine data is concentrated in the standard power curve where the wind speed is less than or equal to 0. This process obtains the first wind turbine dataset, effectively filtering out invalid data in the wind turbine dataset that exceeds the preset wind speed range and whose power does not conform to normal operating conditions, ensuring the validity of subsequent calculation data. Next, the initial relative wind direction angle corresponding to the first wind turbine data is calculated: using geographical north as the reference direction, the initial relative wind direction angle of the i-th first wind turbine data is calculated. , The northerly wind direction for the i-th data point of the first wind turbine (obtained from the data set of the first wind turbine). The actual wind direction angle for the i-th first wind turbine data (obtained from the first wind turbine dataset). Periodic ambiguity of the angle is eliminated through 360° modulo operation to ensure the initial relative wind direction angle is within the range [0°, 360°]. Subsequently, initial relative wind direction angles that meet the preset first state or exceed the preset angle range are repaired to obtain the first relative wind direction angle: The preset first state represents the case where the initial relative wind direction angle is missing, and the preset angle range is [0°, 360°]. A bidirectional filling strategy is used to repair the initial relative wind direction angle of the i-th first wind turbine data that meets the preset first state: For the initial relative wind direction angle of the i-th first wind turbine data that exceeds the preset angle range, it is first marked as an outlier NaN, and then repaired by linear interpolation: Here, j and k are the indices of the latest valid data points before and after the outliers. Through repair, outliers in the relative wind direction angles corresponding to the first wind turbine data are prevented, ensuring that the obtained first relative wind direction angles are accurate and reasonable. Furthermore, the first relative wind direction angles corresponding to several processed first wind turbine data sets are complete, meeting the requirements of real-time optimization. Then, the first relative wind direction angles corresponding to several first wind turbine data sets are matched and integrated with the first wind turbine dataset to form a reliable second wind turbine dataset, providing a basis for subsequent data preprocessing and calculation. Subsequently, the second wind turbine dataset is preprocessed to obtain a third wind turbine dataset, ensuring that the obtained third wind turbine dataset can meet the subsequent calculation requirements for key parameters such as wind speed multiplier factor and turbulence intensity, providing high-precision input data for the FLORIS model dynamic configuration and yaw optimization algorithm.

[0036] Furthermore, the step of obtaining the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset based on the third wind turbine dataset and several power monotonic intervals includes: For any third wind turbine data in the third wind turbine dataset, the following steps are performed: If the power corresponding to the third wind turbine data is within any power monotonic interval, then based on the power monotonic interval and the power corresponding to the third wind turbine data, calculate the estimated wind speed corresponding to the third wind turbine data. Based on the estimated wind speed corresponding to several third wind turbine data and the third wind turbine dataset, calculate the first wind speed multiplier factor; Based on the third wind turbine dataset and the preset power curve, the first wind speed multiplier factor is corrected to obtain the second wind speed multiplier factor. Based on the second wind speed multiplier factor and the third wind turbine dataset, obtain the fourth wind turbine dataset; Based on the dataset of the fourth wind turbine, the corresponding turbulence intensity dataset was obtained.

[0037] In this embodiment, the estimated wind speed corresponding to the third wind turbine data is calculated based on whether the power corresponding to the third wind turbine data falls within any power monotonic interval, and the matched power monotonic interval and the corresponding power: for each monotonic segment ,like (In the formula, This refers to the v-th data point of the third wind turbine in the third wind turbine dataset. This represents the k-th power data point in the standard power curve. (For the (k+1)th power data point in the standard power curve), then linear interpolation is used to calculate the estimated wind speed corresponding to the data of the third wind turbine. : In the formula, This represents the k-th wind speed data point in the standard power curve. This represents the (k+1)th wind speed data point in the standard power curve, enabling accurate mapping of wind speed across different power stages. It also aligns with the piecewise characteristics of the power curve, avoiding errors from global fitting and providing fundamental data for calculating the first wind speed multiplier factor. Next, the first wind speed multiplier factor is obtained by correlating the estimated wind speeds corresponding to several third wind turbine data points with the third wind turbine dataset. : In the formula, This represents the v-th data point of the third wind turbine in the third wind turbine dataset, laying the foundation for subsequent secondary corrections. After obtaining the first wind speed multiplier factor, The uncertainty is limited to 0.9~1.1 and meets the management principles for measurement uncertainty in GB / T 18451.2-2021 Wind Turbine Generator Power Characteristic Test: The uncertainty of wind speed measuring equipment is usually controlled within... Within. Therefore, Limited to Within this range, corrections are used to control the magnitude of the correction to avoid spurious errors introduced by sensor malfunctions or overcorrection, especially in the range above the rated wind speed (where power output tends to saturate), and to avoid non-physical corrections to invalid data segments. Secondly, in actual wind farm operation, the wind speed-power relationship tends to saturate above the rated wind speed, and overcorrection (such as...) is detrimental. This could lead to spurious optimizations in subsequent models within invalid data segments (such as the constant power region), or even cause control oscillations. Furthermore, the defined range set in this embodiment, verified by extensive experimental data, effectively balances correction sensitivity and system stability. Finally, analysis of historical SCADA data shows that over 95% of the wind speed multiplier factor falls within the [0.95, 1.05] range, while this embodiment extends it to [0.9, 1.1], providing a buffer for extreme conditions (such as sudden turbulence changes and blade contamination) and preventing model mismatch due to individual anomalies. Although the initial correction of the first wind speed multiplier factor addresses the static deviation of the power curve, real-time disturbances (such as sudden atmospheric turbulence changes, sensor noise, and short-term climate change) still exist in the operation of offshore wind farms, leading to dynamic deviations between the simulated and actual power in subsequent models. Moreover, traditional wake models rely on fixed parameters (such as turbulence intensity and thrust coefficient) and cannot respond to these dynamic changes in real time. Without secondary correction, the accumulated error will increase over time, causing the yaw control strategy to fail (e.g., misjudging the wake direction, underestimating speed deficit), and even leading to negative optimization (reduced power generation). Therefore, to address dynamic disturbances such as sudden changes in atmospheric turbulence, this embodiment designs an adaptive adjustment strategy based on power deviation. The first wind speed multiplier factor is corrected using the third wind turbine dataset and a preset power curve, making the wind speed multiplier factor more closely match actual operating conditions, thus obtaining the second wind speed multiplier factor. In the formula, This represents the first wind speed multiplier factor for the v-th wind turbine. This represents the difference between the actual power of the v-th wind turbine and its corresponding simulated power on the standard power curve. This indicates that the step size is adjusted to 0.01 to ensure the accuracy of the wake model input. Subsequently, the third wind turbine dataset is fused using the second wind speed multiplier factor to form the fourth wind turbine dataset, which meets the requirements for subsequent turbulence intensity calculations and dynamic wake model configuration. The corrected fourth wind turbine data is as follows: In the formula, This represents the corrected data for any fourth wind turbine unit. This represents the data for the third wind turbine corresponding to the data for the fourth wind turbine. An initial correction addresses static modeling biases, while a second correction addresses dynamic operational biases, forming a progressive correction system of calibration followed by fine-tuning. This brings the difference between simulated and actual wake power to within 1%, providing high-precision input for yaw optimization. The two corrections address static inconsistencies in model parameters and dynamic uncertainties in the operating environment, respectively, resolving the insufficient model accuracy in traditional wake optimization methods. The initial correction uses piecewise interpolation to achieve accurate mapping of physical laws, while the second correction uses closed-loop feedback to achieve dynamic adaptation to real-time operating conditions. Together, they construct a complete data-driven model correction feedback optimization chain, providing both theoretical and engineering guarantees for the efficient and reliable operation of wind farms. Finally, turbulence intensity is calculated using the fourth wind turbine dataset: turbulence intensity is dynamically calculated based on the sliding window technique, i.e., the ratio of the standard deviation to the average wind speed corresponding to the fourth wind turbine data within 10 minutes, reflecting the wind speed fluctuation characteristics. , , In the formula, Indicates turbulence intensity. This represents the standard deviation of the wind speed in the data set of the fourth wind turbine unit. This represents the average wind speed of the fourth wind turbine unit. This represents the wind speed corresponding to the data from the fourth wind turbine, where n represents the number of data points for the fourth wind turbine (i.e., the length of the sliding window). Finally, according to the IEC 61400-1 standard, the turbulence intensity is... The range is limited to [0.01, 0.25] to avoid extreme values ​​interfering with subsequent models. This provides key flow field parameters for the dynamic configuration of the FLORIS model, improving the accuracy of wake simulation and yaw optimization.

[0038] Further, obtaining the fifth wind turbine dataset based on the fourth wind turbine dataset and the turbulence intensity dataset includes: For any fourth wind turbine data in the fourth wind turbine dataset, the following steps are performed: obtain the initial wake width corresponding to the fourth wind turbine data; based on the turbulence intensity in the turbulence intensity dataset and the initial wake width corresponding to the fourth wind turbine data, obtain the wake width corresponding to the fourth wind turbine data. Based on the wake width corresponding to several fourth wind turbine data and the fourth wind turbine dataset, the fifth wind turbine dataset is obtained.

[0039] In this embodiment, the wake width varies linearly with downstream distance. Therefore, the wake width corresponding to the fourth wind turbine data is calculated by combining the turbulence intensity and the initial wake width. In the formula: The initial wake width corresponding to the data of the fourth wind turbine (and the rotor diameter) Related, ), Wake width, The turbulence diffusion coefficient is derived from turbulence intensity and is jointly determined by atmospheric stability and the yaw state of the wind turbine. It allows the wake width to be dynamically adjusted with turbulence intensity, overcoming the limitation of a fixed wake width in traditional models and improving the accuracy of wake modeling for both wind turbines. Next, by fusing the wake widths corresponding to several fourth wind turbine data points with the fourth wind turbine dataset, a fifth wind turbine dataset containing accurate wake width information is formed: (The last sentence appears to be incomplete and possibly refers to a different topic.) Horizontal wind speed component Follows the Gaussian curve: In the formula: For incoming air velocity; This is the speed loss factor, compared to the preset thrust factor of the wind turbine. and downstream distance Related; The wake width describes the diffusion characteristics of the lateral velocity distribution. The obtained dataset from the fifth wind turbine provides reliable flow field characteristics input for subsequent calculations of multiple wake superposition effects and yaw optimization algorithms, ensuring the effectiveness of the yaw optimization strategy. When the wakes of multiple wind turbines overlap, the total velocity deficit is calculated using the principle of linear superposition: In the formula: U represents the number of upstream wind turbines affecting the target wind turbine. It is the speed loss generated by the r-th wind turbine at that wind turbine.

[0040] Furthermore, the construction of the first FLORIS model and calculation of the power change rate based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor includes: Based on the preset first sampling interval, the dataset of the fifth wind turbine is sampled to obtain several first sampling points; Construct the initial FLORIS model; Based on the turbulence intensity dataset, the second wind speed multiplier factor, and the initial FLORIS model, the first FLORIS model is obtained; The power change rate is calculated based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, the preset second sampling interval, and the preset disturbance step size.

[0041] In this embodiment, the yaw optimization objective function and constraints are as follows: In the formula, Let n be the yaw angle of the nth wind turbine. This is the state coefficient of the wind turbine (0 for a wind turbine that is shut down, and 1 for a wind turbine that is operating normally). To account for the wake effect on wind turbine power, the goal is to maximize the total active power while considering the shutdown of wind turbines and yaw angle constraints. A two-stage sequence optimization algorithm is then used to calculate the power change rate. First, the dataset of the fifth wind turbine is sampled within a preset first sampling interval (the feasible region of yaw angle). )Sampling 5 points evenly, i.e. This process narrows the computational scope, providing efficient and accurate input samples for subsequent power change rate calculations, and yields several first sampling points. Next, an initial FLORIS model is constructed, establishing the basic framework for wake effect simulation and providing theoretical model support for power calculation. Then, turbulence intensity and the second wind speed multiplier factor are integrated into the initial FLORIS model, enabling the obtained first FLORIS model to accurately reflect real-time flow field characteristics and actual operating conditions after wind speed correction, improving the accuracy of wake simulation and power calculation. Finally, through the collaborative calculation of the first FLORIS model, several first sampling points, the fifth wind turbine dataset, a preset second sampling interval, and a preset disturbance step size, the power change rate is obtained, providing data support for subsequent yaw angle adjustments and improving the accuracy and efficiency of yaw optimization.

[0042] Furthermore, the calculation of the power change rate based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, a preset second sampling interval, and a preset disturbance step size includes: Based on the first FLORIS model and several first sampling points, the wind turbine power corresponding to several first sampling points is obtained, and the interval where the wind turbine power meets the preset first condition is determined as the first yaw angle feasible region. Based on the preset second sampling interval, the dataset of the fifth wind turbine unit within the first yaw angle feasible region is sampled to obtain several second sampling points; The power change rate is calculated based on the first FLORIS model, the preset perturbation step size, and the second sampling point.

[0043] In this embodiment, the power of wind turbines corresponding to several first sampling points is calculated using the first FLORIS model, and the interval containing the first sampling points that meet the preset first condition is selected as the first yaw angle feasible region. The first FLORIS model is used to quickly evaluate the total power of each sampling point combination, locate the 5° interval where the optimal solution is located, eliminate invalid yaw angle intervals, narrow the calculation range, and ensure that the optimal solution interval is covered with the minimum number of samples. Next, the fifth wind turbine dataset within the first yaw angle feasible region is sampled using a preset second sampling interval to obtain several second sampling points. The optimal point obtained by coarse search is then used to further refine the calculation. Nearby, that is Four points are sampled within the preset second sampling interval to further refine the yaw angle, providing input samples for accurate calculation of the power change rate. This effectively handles non-convex optimization problems caused by multiple wake superpositions and improves convergence accuracy. Then, through the collaborative calculation of the first FLORIS model, the preset perturbation step size, and the second sampling points, the power change rate is obtained: at the current yaw angle... At this point, calculate the partial derivative of power with respect to the yaw angle (i.e., the rate of change of power): ;in The power value is obtained through simulation using the first FLORIS model, with a preset perturbation step size. Then, the yaw angle is updated using the gradient ascent method. ;in The learning rate is used to control the update step size. The power change rate here is the real-time power change rate caused by the yaw angle change, not the rate of change in the standard power curve, to ensure the optimization process matches the current dynamic response of the flow field. Through the above steps, this embodiment can reduce computational complexity from... Down to It deeply integrates two-level optimization with task-level (Python multi-process) and data-level (OpenMP multi-thread) parallel computing, reducing the time for a single optimization from 42.83 minutes serially to 0.83 minutes in a 36-unit scenario (a speedup of 51.4 times), and ensures that the optimized yaw angle can guide the offshore wind farm turbines to continuously iterate towards the goal of maximizing power, thereby optimizing and improving the power generation of offshore wind farm turbines.

[0044] Furthermore, the step of obtaining the sixth wind turbine dataset and predicted total power based on the fifth wind turbine dataset, power change rate, and the first FLORIS model includes: For any fifth wind turbine data in the fifth wind turbine dataset, the following steps are performed: obtain the first yaw angle corresponding to the fifth wind turbine data; obtain the second yaw angle corresponding to the fifth wind turbine data based on the yaw angle and power change rate corresponding to the fifth wind turbine data; obtain the yaw angle difference based on the first yaw angle and the second yaw angle. Based on the second yaw angle corresponding to the data of the fifth wind turbine and the dataset of the fifth wind turbine, the dataset of the sixth wind turbine is obtained. If the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold, then the predicted total power is obtained based on the first FLORIS model and the sixth wind turbine dataset.

[0045] In this embodiment, the first yaw angle corresponding to the data of the fifth wind turbine is obtained, clarifying the current yaw state of the wind turbine and providing an initial reference for subsequent optimization and adjustment of the yaw angle. Next, the yaw angle difference between the first and second yaw angles is calculated to quantify the adjustment range of the yaw angle, facilitating rapid identification of whether the optimized yaw angle has reached a stable state. Then, the second yaw angle corresponding to the data of the fifth wind turbine is fused with the dataset of the fifth wind turbine to obtain the dataset of the sixth wind turbine, providing accurate input for the calculation of predicted total power. Subsequently, it is ensured that the yaw angle difference is less than a preset yaw angle change threshold and the power change rate is less than a preset power change threshold (the preset yaw angle change threshold is set to...). (The preset power change threshold is 0.01%). At this point, the iteration stops, and the predicted total power is calculated to obtain the predicted total power of the simulated wind turbine. This avoids errors caused by calculating the predicted total power in a non-converged state, ensuring the accuracy and reliability of the predicted total power, and providing a basis for judgment for subsequent power verification and triggering of the backoff mechanism.

[0046] Furthermore, it also includes: If the power change rate is greater than or equal to the preset power change threshold and the yaw angle difference is greater than or equal to the preset yaw angle change threshold, then the first FLORIS model is reconstructed based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor, and the power change rate is calculated until the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold.

[0047] In this embodiment, when the power change rate is greater than or equal to a preset power change threshold and the yaw angle difference is greater than or equal to a preset yaw angle change threshold, it indicates that the current yaw angle adjustment has not reached a stable state and there is still room for power adjustment. Therefore, the first FLORIS model is reconstructed based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor, and the power change rate is recalculated to ensure that the goal of maximizing the overall power is always achieved, avoiding getting trapped in a local optimum. Finally, the above iterative process is repeated until the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold, ensuring that the final obtained yaw angle reaches a local optimum or the overall power generation has tended to be maximized, thereby improving the economic efficiency and stability of the offshore wind farm's wind turbine operation.

[0048] This embodiment employs an acceleration strategy combining task-level and data-level parallelism: task-level parallelism (Python multiprocessing) can allocate evaluation tasks with different yaw angle combinations to independent processes, utilizing multi-core CPUs for parallel computation, with a working process count of... Under these circumstances, ideal acceleration ratio Data-level parallelism (OpenMP multithreading): The wake superposition calculation is cyclically parallelized within the FLORIS model kernel. Based on Gustafson's law, when the serial portion accounts for a default 5%, the number of OpenMP threads is [missing information]. Under these circumstances, acceleration ratio upper limit .

[0049] To demonstrate the effectiveness of this embodiment, a Chinese offshore wind farm is used as the research object. This wind farm is equipped with 36 5.5MW wind turbines, with a total installed capacity of 198MW, and includes a 220kV offshore substation and an onshore control center. The wind turbine coordinates are as follows: Figure 2 As shown. By Figure 2 It can be seen that the wind turbine coordinates are distributed in a rectangular array (x∈[-4000,10000]m, y∈[0,8000]m). Based on actual measurements using SCADA data at 10-minute intervals, the total power output increased from 95.4MW to 101.0MW at a typical moment, representing a 5.8% increase in power generation, thus verifying the engineering effectiveness of the optimization strategy. Please refer to [link / reference]. Figure 3 The upstream main turbines (1-10) have an average yaw angle of 12.7°, which reduces wind speed attenuation by 0.8 m / s in 30% of the downstream area through wake deflection. The edge turbines (29-36) have a fixed yaw angle of -15° (compliant with the IEC61400-1 safety threshold) due to the superposition of multiple wakes. After optimization, the standard deviation of turbulence intensity across the entire field is reduced by 18.5% compared to the traditional strategy, demonstrating the ability of yaw control to regulate wake diffusion. Please refer to [link / reference]. Figure 4 The two-stage sequential optimization algorithm converged within 30 iterations, reducing the power difference from the initial 18MW to 250W, better than the 0.1% convergence threshold. The first 15 coarse searches (5° yaw intervals) were used to quickly narrow down the solution space, while the latter 15 fine searches (1° intervals) were used to handle nonlinear wake interactions. Combined with hybrid parallel computing (OpenMP + Python multi-process), a single optimization took 0.83 minutes, meeting IEC real-time control requirements. Please see [link / reference]. Figure 5 The measured power of this model showed a goodness of fit R=0.91 (>0.9) to the standard curve, meeting the uncertainty requirements of GB / T18451.2-2021 (error ≤1%). Power increased linearly in the low wind speed range (<11m / s), and the measured power fluctuation in the constant power range (≥11m / s) was ≤300kW, verifying the adaptability of the second-order closed-loop correction to sudden turbulence changes. Power curve inversion reduced the wake model input error to 0.7%. Please refer to [link / reference]. Figure 6 Before optimization, the average power output per unit was 2650.47 kW, which increased to 2805.19 kW after optimization, a 5.8% increase. The interquartile range (IQR) decreased from 720 kW to 580 kW, demonstrating the adaptability of the wake optimization strategy in this embodiment to wind turbines. Please refer to... Figure 7 The wind speed multiplier factor ranges from 0.947 to 1.066, which conforms to the range of 0.9 to 1.1 specified in GB / T18451.2-2021. Upstream wind turbines (Nos. 1-5) (such as wind turbine No. 1 and) The measured wind speed was 6.6% higher than the model prediction, and the downstream wind turbines (Nos. 30-36) were also affected. (such as wind turbine No. 36) This indicates that the wake attenuation is effectively compensated. Compared to traditional fixed... This correction system effectively reduces wake velocity prediction errors. Please see [link / reference]. Figure 8 The turbulence intensity (TI) values ​​range from 0.06 to 0.225, conforming to the IEC 61400-1 standard (0.01 to 0.25). The upstream inflow region (sections 1-5) has a TI of 0.06 to 0.11, while the downstream wake region (sections 20-36) has a TI of 0.15 to 0.225, consistent with the theory of nonlinear turbulence relationships. The combination of a 10-minute sliding window and a 1-minute real-time update ensures that the dynamic response delay of turbulence intensity is less than 2 minutes, reducing the error caused by multiple wake superposition by 42%. Please refer to [link / reference]. Figure 9 When the wind speed is 7-9 m / s, the yaw angle is concentrated (0-15°), reflecting the effectiveness and stability of yaw control in this wind speed range. At low wind speeds (5-6 m / s) and high wind speeds (close to 10 m / s), the yaw angle dispersion is slightly higher, reflecting the strategy's adaptive adjustment to different wind speed conditions. This directly verifies the optimization effect of yaw control in the medium-speed range (7-9 m / s), providing empirical support for operating condition matching in offshore wind turbine wake control and power stability improvement.

[0050] This embodiment also compares the performance of serial computing strategies, task-level parallelism only, data-level parallelism only, and the hybrid parallelism strategy adopted in this embodiment. Please refer to Table 1 to verify the effectiveness of the hybrid parallel architecture of this embodiment. The synergistic effect of task-level parallelism and data-level parallelism adopted in this embodiment achieves a speedup of up to 51.4 times, far exceeding the single parallelism strategy (6.6 times for task-level only, 6.2 times for data-level only), breaking through the bottleneck of real-time optimization computing for large-scale wind farms.

[0051] Table 1 Comparison of Parallel Acceleration Effects The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A wake optimization method for offshore wind farms, characterized in that, Includes the following steps: Obtain the initial wind turbine dataset; Based on the initial wind turbine dataset, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset were obtained. Based on the fourth wind turbine dataset and the turbulence intensity dataset, the fifth wind turbine dataset is obtained; Based on the fifth wind turbine dataset, turbulence intensity dataset, and second wind speed multiplier factor, the first FLORIS model is constructed and the power change rate is calculated. Based on the fifth wind turbine dataset, power change rate and first FLORIS model, obtain the sixth wind turbine dataset and predicted total power; Based on the dataset of the sixth wind turbine, the actual total power was obtained; Based on the predicted total power and the actual total power, the power difference is obtained. If the power difference is less than or equal to a preset power threshold, several yaw angles of the sixth wind turbine dataset are transmitted to the corresponding wind turbine to achieve wake optimization of the offshore wind farm.

2. The wake optimization method for offshore wind farms according to claim 1, characterized in that, Also includes: If the power difference is greater than the preset power threshold, then several yaw angles of the initial wind turbine dataset will be transmitted to the corresponding wind turbine.

3. The wake optimization method for offshore wind farms according to claim 1, characterized in that, The process of obtaining the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset based on the initial wind turbine dataset includes: Based on the preset wind speed range, preset power threshold and initial wind turbine dataset, obtain the third wind turbine dataset; Based on the preset first segmentation condition and the preset power curve, several power monotonic intervals are obtained; Based on the third wind turbine dataset and several power monotonic intervals, the second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset are obtained.

4. The wake optimization method for offshore wind farms according to claim 3, characterized in that, The process of obtaining the third wind turbine dataset based on a preset wind speed range, a preset power threshold, and an initial wind turbine dataset includes: Based on the preset wind speed range and preset power threshold, the initial wind turbine dataset is filtered to obtain the first wind turbine dataset. For any first wind turbine data in the first wind turbine dataset, the following steps are performed: Based on the first wind turbine data, calculate the initial relative wind direction angle corresponding to the first wind turbine data. If there is an initial relative wind direction angle that meets the preset first state or the initial relative wind direction angle exceeds the preset angle range, then the initial relative wind direction angle is repaired to obtain the first relative wind direction angle corresponding to the first wind turbine data. Based on the first wind turbine dataset and the first relative wind direction angles corresponding to several first wind turbine datasets, the second wind turbine dataset is obtained. Data preprocessing is performed on the second wind turbine dataset to obtain the third wind turbine dataset.

5. The wake optimization method for offshore wind farms according to claim 3, characterized in that, The second wind speed multiplier factor, the fourth wind turbine dataset, and the turbulence intensity dataset are obtained based on the third wind turbine dataset and several power monotonic intervals; including: For any third wind turbine data in the third wind turbine dataset, the following steps are performed: If the power corresponding to the third wind turbine data is within any power monotonic interval, then based on the power monotonic interval and the power corresponding to the third wind turbine data, calculate the estimated wind speed corresponding to the third wind turbine data. Based on the estimated wind speed corresponding to several third wind turbine data and the third wind turbine dataset, calculate the first wind speed multiplier factor; Based on the third wind turbine dataset and the preset power curve, the first wind speed multiplier factor is corrected to obtain the second wind speed multiplier factor. Based on the second wind speed multiplier factor and the third wind turbine dataset, obtain the fourth wind turbine dataset; Based on the dataset of the fourth wind turbine, the corresponding turbulence intensity dataset was obtained.

6. The wake optimization method for offshore wind farms according to claim 1, characterized in that, The process of obtaining the fifth wind turbine dataset based on the fourth wind turbine dataset and the turbulence intensity dataset includes: For any fourth wind turbine data in the fourth wind turbine dataset, the following steps are performed: obtain the initial wake width corresponding to the fourth wind turbine data; based on the turbulence intensity in the turbulence intensity dataset and the initial wake width corresponding to the fourth wind turbine data, obtain the wake width corresponding to the fourth wind turbine data. Based on the wake width corresponding to several fourth wind turbine data and the fourth wind turbine dataset, the fifth wind turbine dataset is obtained.

7. The wake optimization method for offshore wind farms according to claim 1, characterized in that, The first FLORIS model is constructed and the power change rate is calculated based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor; including: Based on the preset first sampling interval, the dataset of the fifth wind turbine is sampled to obtain several first sampling points; Construct the initial FLORIS model; Based on the turbulence intensity dataset, the second wind speed multiplier factor, and the initial FLORIS model, the first FLORIS model is obtained; The power change rate is calculated based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, the preset second sampling interval, and the preset disturbance step size.

8. The wake optimization method for an offshore wind farm according to claim 7, characterized in that, The calculation of the power change rate is based on the first FLORIS model, several first sampling points, the fifth wind turbine dataset, a preset second sampling interval, and a preset disturbance step size; including: Based on the first FLORIS model and several first sampling points, the wind turbine power corresponding to several first sampling points is obtained, and the interval where the wind turbine power meets the preset first condition is determined as the first yaw angle feasible region. Based on the preset second sampling interval, the dataset of the fifth wind turbine unit within the first yaw angle feasible region is sampled to obtain several second sampling points; The power change rate is calculated based on the first FLORIS model, the preset perturbation step size, and the second sampling point.

9. The wake optimization method for an offshore wind farm according to claim 1, characterized in that, The process of obtaining the sixth wind turbine dataset and predicted total power based on the fifth wind turbine dataset, power change rate, and the first FLORIS model includes: For any fifth wind turbine data in the fifth wind turbine dataset, the following steps are performed: obtain the first yaw angle corresponding to the fifth wind turbine data; obtain the second yaw angle corresponding to the fifth wind turbine data based on the yaw angle and power change rate corresponding to the fifth wind turbine data; obtain the yaw angle difference based on the first yaw angle and the second yaw angle. Based on the second yaw angle corresponding to the data of the fifth wind turbine and the dataset of the fifth wind turbine, the dataset of the sixth wind turbine is obtained. If the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold, then the predicted total power is obtained based on the first FLORIS model and the sixth wind turbine dataset.

10. The wake optimization method for an offshore wind farm according to claim 9, characterized in that, Also includes: If the power change rate is greater than or equal to the preset power change threshold and the yaw angle difference is greater than or equal to the preset yaw angle change threshold, then the first FLORIS model is reconstructed based on the fifth wind turbine dataset, the turbulence intensity dataset, and the second wind speed multiplier factor, and the power change rate is calculated until the power change rate is less than the preset power change threshold or any yaw angle difference is less than the preset yaw angle change threshold.

Images