Offshore wind field simulation and wind turbine generator control system and method and wind turbine generator

Abstract

The embodiment of the invention provides an offshore wind field simulation and wind turbine generator control system and method and a wind turbine generator, and the system comprises a data preprocessing module which is used for carrying out format conversion, projection transformation and quality control on multi-source heterogeneous meteorological data; the data fusion module is used for performing intelligent fusion on the preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fusion field data; the mode adjusting module is used for dynamically optimizing a data fusion strategy based on the sea condition change; the resolution conversion module is used for performing temporal-spatial resolution conversion on the three-dimensional fusion field data to generate refined three-dimensional wind field data; and the zone control module is used for performing zone division on the wind field based on the refined three-dimensional wind field data, generating differential yaw control strategies of the corresponding zones according to related data and zone characteristics of each zone, and driving a unit to execute the yaw control strategies. The wind field simulation precision and the adaptability to complex sea conditions are improved, and wind energy capture and equipment loads are balanced through a differential yaw strategy.

Description

Technical Field

[0001] This disclosure relates to the field of offshore wind power technology, and in particular to an offshore wind farm simulation and wind turbine control system, method, and wind turbine. Background Technology

[0002] In the offshore wind power sector, existing wind farm simulation technologies primarily rely on multi-source data fusion and numerical modeling. By processing meteorological data (such as numerical weather prediction, satellite observation, and ground-based detection) using traditional methods, macroscopic wind field analysis can be provided for turbine operation. These technologies achieve data integration and resolution conversion based on fixed patterns, providing basic wind speed and direction distribution characteristics under normal sea conditions to support the formulation of initial turbine control strategies. Regarding wind turbine control, existing systems typically employ sensor-based feedback control logic. By monitoring wind direction and speed in real time to adjust the nacelle orientation, and possessing basic load protection mechanisms, they can achieve stable turbine operation in simple wind field environments.

[0003] However, as the meteorological conditions faced by offshore wind power become increasingly complex, the universality of existing technologies faces challenges. At the wind farm simulation level, traditional data fusion methods have limited ability to handle the spatiotemporal differences of heterogeneous data, are insufficient in capturing dynamic characteristics under complex meteorological scenarios (such as strong convection and sea-land air mass interactions), and conventional resolution conversion techniques are insufficient to meet the needs of refined analysis, leading to deviations between wind farm descriptions and the actual environment. At the turbine control level, traditional strategies mostly employ fixed parameter adjustments, lacking the ability to dynamically identify and differentiate wind farm regional characteristics (such as wind speed gradients and turbulence distribution). Especially in scenarios involving multiple turbines operating collaboratively, the lack of optimization mechanisms for complex interactions such as wake interference makes it difficult to achieve a comprehensive improvement in efficiency and reliability. These technical characteristics reflect the current balance between basic functionality and adaptability to complex scenarios in existing solutions, highlighting the necessity for intelligent upgrades. Summary of the Invention

[0004] This disclosure provides an offshore wind farm simulation and wind turbine control system, method, and wind turbine to address the problems existing in the current technical solutions.

[0005] Based on the above problems, in the first aspect, a control system for offshore wind farm simulation and wind turbine generators is provided, including:

[0006] The data preprocessing module is used for format conversion, projection transformation, and quality control of multi-source heterogeneous meteorological data;

[0007] The data fusion module is used to intelligently fuse preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fused field data.

[0008] The mode adjustment module is used to dynamically optimize the data fusion strategy based on changes in sea state;

[0009] The resolution conversion module is used to perform spatiotemporal resolution conversion on the 3D fused field data to generate refined 3D wind field data.

[0010] The zone control module is used to divide the wind field into zones based on the refined three-dimensional wind field data, generate differentiated yaw control strategies for each zone according to the wind force, wind direction data and regional characteristics, and drive the unit to execute the yaw control strategies.

[0011] In conjunction with the first aspect, in one possible implementation, the partition control module includes:

[0012] The region division unit is used to read the wind speed distribution, turbulence intensity distribution, and wake influence range data from the refined three-dimensional wind field data, and divide the wind field into at least the following control regions based on preset wind speed thresholds, turbulence intensity thresholds, and wake influence distances:

[0013] Low wind speed area, corresponding to wind speeds below the first preset threshold;

[0014] The medium wind speed region corresponds to wind speeds between the first preset threshold and the second preset threshold.

[0015] High turbulence region, corresponding to turbulence intensity higher than the preset turbulence intensity threshold;

[0016] The wake influence area refers to the area within the wake influence distance covered by the wake of the upstream unit.

[0017] In conjunction with the first aspect, in one possible implementation, the partition control module further includes:

[0018] A control strategy generation unit is used to generate a yaw adjustment strategy for each control region by calling a preset yaw control strategy mapping table based on the characteristics of each control region. The yaw control strategy mapping table includes:

[0019] For low wind speed areas, control parameters are generated with a yaw error range of the first range and an adjustment frequency of the first frequency to maximize wind energy capture.

[0020] For medium wind speed areas, control parameters are generated with the yaw error range being the second range and the adjustment frequency being the second frequency, in order to balance energy efficiency and equipment load.

[0021] For the highly turbulent region, control parameters with a yaw error range of the third range and an adjustment frequency of the third frequency are generated to reduce mechanical wear;

[0022] For the wake-affected area, cooperative yaw control parameters based on the yaw angle of the upstream units are generated to reduce wake loss.

[0023] In conjunction with the first aspect, in one possible implementation, the control strategy generation unit is used to correct the yaw control parameters using the following formula when the yaw angle of the wind turbine exceeds the corresponding yaw error range:

[0024]

[0025] Where Δψ represents the yaw angle adjustment vector; A represents the operating condition weight matrix, with the diagonal elements corresponding to the weight coefficients of each control area; α represents the wind speed change rate weight coefficient; β represents the turbulence accumulation effect coefficient. Characterizes the rate of change of wind speed; Offshore wind field simulation and wind turbine control characterize the angle between wind direction and nacelle azimuth; B characterizes the wake transfer matrix, the elements of which are determined by the relative positions of upstream and downstream turbines and the distribution of turbulence intensity; Ψup characterizes the yaw angle matrix of the upstream turbine; t characterizes the current moment, and t′ characterizes the previous moment.

[0026] In conjunction with the first aspect, in one possible implementation, the control strategy generation unit is further configured to enable lidar scanning data in low wind speed areas to obtain the incoming wind direction in advance and compensate for yaw lag; and in the wake influence area, generate a follow yaw command based on the yaw angle of the upstream unit to achieve wind farm-level coordinated yaw.

[0027] In conjunction with the first aspect, in one possible implementation, the data fusion module includes:

[0028] The deep neural network unit receives GFS numerical weather prediction products, GRAPES numerical weather prediction products, satellite first-level data, radar base data, and ground observation data as input. It extracts spatial features of the data through convolutional neural networks, captures time series features of the data through recurrent neural networks, and assigns fusion weights to numerical weather prediction products, satellite data, radar data, and ground observation data through an attention mechanism, and outputs three-dimensional fused field data.

[0029] In conjunction with the first aspect, in one possible implementation, the mode adjustment module includes:

[0030] The reinforcement learning unit defines a state space containing sea state type and meteorological element change rate, and an action space containing background field interpolation parameter adjustment and observation data assimilation weight adjustment. The reward function is the matching degree between the fusion result and the measured wind field. Through reinforcement learning training, it generates data fusion strategy adjustment instructions and dynamically adjusts the parameters of the data fusion module.

[0031] In conjunction with the first aspect, in one possible implementation, the resolution conversion module includes:

[0032] Kriging interpolation unit is used to acquire DEM data of offshore wind power area as terrain correction factor. Based on the physical relationship between wind speed and pressure gradient, the correlation constraint of meteorological elements is established. The improved Kriging interpolation algorithm is used to perform spatial interpolation on the three-dimensional fused field data.

[0033] The spatiotemporal joint filtering unit is used to perform Kalman filtering on time-series meteorological data to remove noise. It performs spatiotemporal joint processing on the filtered time-series data and the spatially interpolated three-dimensional fused field data to generate gridded, refined three-dimensional wind field data for offshore wind power with a preset grid resolution.

[0034] Secondly, a method for simulating offshore wind farms and controlling wind turbine generators is provided, including:

[0035] Preprocessing operations for multi-source heterogeneous meteorological data, including format conversion, projection transformation, and quality control;

[0036] Intelligent fusion of preprocessed multi-source heterogeneous meteorological data generates three-dimensional fused field data.

[0037] Dynamically optimize data fusion strategies based on sea state changes;

[0038] Spatiotemporal resolution conversion is performed on the 3D fused field data to generate refined 3D wind field data;

[0039] Based on the refined three-dimensional wind field data, the wind field is divided into regions. Based on the wind force, wind direction data and regional characteristics of each region, a differentiated yaw control strategy is generated for the corresponding region, and the unit is driven to execute the yaw control strategy.

[0040] In conjunction with the second aspect, in one possible implementation, the regional division of the wind field based on the refined three-dimensional wind field data includes:

[0041] Read the wind speed distribution, turbulence intensity distribution, and wake influence range data from the refined three-dimensional wind field data. Based on preset wind speed thresholds, turbulence intensity thresholds, and wake influence distances, divide the wind field into at least the following control regions:

[0042] Low wind speed area, corresponding to wind speeds below the first preset threshold;

[0043] The medium wind speed region corresponds to wind speeds between the first preset threshold and the second preset threshold.

[0044] High turbulence region, corresponding to turbulence intensity higher than the preset turbulence intensity threshold;

[0045] The wake influence area refers to the area within the wake influence distance covered by the wake of the upstream unit.

[0046] In conjunction with the second aspect, in one possible implementation, generating differentiated yaw control strategies for corresponding regions based on wind force, wind direction data, and regional characteristics includes:

[0047] Based on the characteristics of each control area, a preset yaw control strategy mapping table is invoked to generate the corresponding yaw adjustment strategy for that area. The yaw control strategy mapping table includes:

[0048] For low wind speed areas, control parameters are generated with a yaw error range of the first range and an adjustment frequency of the first frequency to maximize wind energy capture.

[0049] For medium wind speed areas, control parameters are generated with the yaw error range being the second range and the adjustment frequency being the second frequency, in order to balance energy efficiency and equipment load.

[0050] For the highly turbulent region, control parameters with a yaw error range of the third range and an adjustment frequency of the third frequency are generated to reduce mechanical wear;

[0051] For the wake-affected area, cooperative yaw control parameters based on the yaw angle of the upstream units are generated to reduce wake loss.

[0052] In conjunction with the second aspect, in one possible implementation, the method further includes: when the yaw angle of the wind turbine exceeds the corresponding yaw error range, correcting the yaw control parameters using the following formula:

[0053]

[0054] Where Δψ represents the yaw angle adjustment vector; A represents the operating condition weight matrix, with the diagonal elements corresponding to the weight coefficients of each control area; α represents the wind speed change rate weight coefficient; β represents the turbulence accumulation effect coefficient. Characterizes the rate of change of wind speed; Offshore wind field simulation and wind turbine control characterize the angle between wind direction and nacelle azimuth; B characterizes the wake transfer matrix, the elements of which are determined by the relative positions of upstream and downstream turbines and the distribution of turbulence intensity; Ψup characterizes the yaw angle matrix of the upstream turbine; t characterizes the current moment, and t′ characterizes the previous moment.

[0055] In conjunction with the second aspect, in one possible implementation, the step of generating differentiated yaw control strategies for corresponding regions based on wind force, wind direction data, and regional characteristics further includes:

[0056] In low-wind-speed areas, lidar scanning data is used to obtain the incoming wind direction in advance and compensate for yaw lag; in wake-affected areas, follow-yaw commands are generated based on the yaw angle of the upstream units to achieve wind farm-level coordinated yaw.

[0057] In conjunction with the second aspect, in one possible implementation, the intelligent fusion of preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fused field data includes:

[0058] It receives GFS numerical weather prediction products, GRAPES numerical weather prediction products, satellite first-level data, radar base data, and ground observation data as input. It extracts spatial features of the data through convolutional neural networks, captures time series features of the data through recurrent neural networks, and assigns fusion weights to numerical weather prediction products, satellite data, radar data, and ground observation data through an attention mechanism, and outputs three-dimensional fused field data.

[0059] In conjunction with the second aspect, in one possible implementation, the data fusion strategy based on dynamic optimization of sea state changes includes:

[0060] Define a state space that includes sea state type and meteorological element change rate, and define an action space that includes background field interpolation parameter adjustment and observation data assimilation weight adjustment. Use the matching degree between the fusion result and the measured wind field as the reward function, and generate data fusion strategy adjustment instructions through reinforcement learning training to dynamically adjust the parameters of the data fusion module.

[0061] In conjunction with the second aspect, in one possible implementation, the spatiotemporal resolution transformation of the three-dimensional fused field data to generate refined three-dimensional wind field data includes:

[0062] DEM data of offshore wind power areas were acquired as topographic correction factors. Based on the physical relationship between wind speed and pressure gradient, correlation constraints of meteorological elements were established. Spatial interpolation of the three-dimensional fused field data was performed by an improved Kriging interpolation algorithm.

[0063] Kalman filtering is applied to time-series meteorological data to remove noise. The filtered time-series data and the spatially interpolated three-dimensional fused field data are then subjected to spatiotemporal joint processing to generate gridded, refined three-dimensional wind field data for offshore wind power with a preset grid resolution.

[0064] Thirdly, a wind turbine generator is provided, including an offshore wind farm simulation and wind turbine generator control system as described in the first aspect, or in any possible embodiment of the first aspect.

[0065] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, performs the steps of a method for simulating an offshore wind farm and controlling a wind turbine as described in the first aspect, or in conjunction with any possible embodiment of the first aspect.

[0066] The beneficial effects of the embodiments disclosed herein include:

[0067] This disclosure provides an offshore wind farm simulation and wind turbine control system, method, and wind turbine, applicable to wind farm analysis and turbine control under complex sea conditions in the offshore wind power field, particularly for refined wind farm analysis and adaptive turbine control under complex sea conditions such as typhoons, alternating sea and land breezes, and high turbulence, supporting multi-unit collaborative operation. The system uses a data preprocessing module to perform format conversion, projection transformation, and quality control on multi-source heterogeneous meteorological data, providing a standardized data source for subsequent processing; a data fusion module utilizes deep neural networks to achieve intelligent fusion of multi-source data, generating three-dimensional fused field data, overcoming the limitations of traditional fixed-weight models; a model adjustment module uses reinforcement learning to perceive sea state changes in real time and dynamically optimize the fusion strategy, improving simulation accuracy under complex scenarios; a resolution conversion module uses improved Kriging interpolation and spatiotemporal joint filtering techniques to generate refined three-dimensional wind farm data specifically for offshore wind power, meeting high-precision analysis requirements; and a zonal control module divides the wind farm into regions based on refined data, calls a yaw control strategy mapping table to generate differentiated parameters, and achieves adaptive control of each region.

[0068] This system, through the coordinated operation of its various modules, forms a complete technical chain from data processing and fusion analysis to control execution, improving the accuracy of wind farm simulation and enhancing its adaptability to complex sea conditions such as typhoons and turbulence. Simultaneously, through differentiated yaw strategies and multi-unit collaborative mechanisms, it effectively balances wind energy capture and equipment load, increasing the annual power generation of wind farms and extending unit lifespan. This provides an intelligent and refined solution for offshore wind power, driving the industry towards higher efficiency and reliability. Attached Figure Description

[0069] Figure 1 A structural diagram of an offshore wind farm simulation and wind turbine control system provided in this disclosure embodiment;

[0070] Figure 2 This is one of the structural diagrams of the partition control module provided in the embodiments of this disclosure;

[0071] Figure 3 This is the second structural diagram of the partition control module provided in the embodiments of this disclosure;

[0072] Figure 4 A structural diagram of the data fusion module provided in this embodiment of the disclosure;

[0073] Figure 5 A structural diagram of the mode adjustment module provided in the embodiments of this disclosure;

[0074] Figure 6 This is a structural diagram of the resolution conversion module provided in an embodiment of the present disclosure;

[0075] Figure 7A flowchart of an offshore wind farm simulation and wind turbine control method provided in this disclosure embodiment. Detailed Implementation

[0076] This disclosure provides an offshore wind farm simulation and wind turbine control system, method, and wind turbine. Preferred embodiments of this disclosure are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit the scope of this disclosure. Furthermore, the embodiments and features described herein can be combined with each other unless otherwise specified.

[0077] This disclosure provides an embodiment of an offshore wind farm simulation and wind turbine control system, such as Figure 1 As shown, it includes:

[0078] Data preprocessing module 101 is used for format conversion, projection transformation and quality control of multi-source heterogeneous meteorological data;

[0079] Data fusion module 102 is used to intelligently fuse preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fused field data;

[0080] The mode adjustment module 103 is used to dynamically optimize the data fusion strategy based on changes in sea state;

[0081] The resolution conversion module 104 is used to perform spatiotemporal resolution conversion on the three-dimensional fused field data to generate refined three-dimensional wind field data.

[0082] The zone control module 105 is used to divide the wind field into regions based on refined three-dimensional wind field data, generate differentiated yaw control strategies for each region according to the wind force, wind direction data and regional characteristics, and drive the unit to execute the yaw control strategies.

[0083] In the offshore wind power sector, existing technologies suffer from several drawbacks. First, the use of a fixed model for fusing multi-source heterogeneous meteorological data leads to insufficient accuracy. Second, the simulation strategy adjustment lags under complex sea conditions. Third, the spatiotemporal resolution is insufficient for refined analysis, and the wind turbine control strategy lacks regional adaptability. These issues restrict wind energy utilization efficiency and equipment reliability. To address these problems, firstly, to resolve the issues of inconsistent multi-source data formats and significant differences in spatiotemporal benchmarks, a data preprocessing module 101 performs format conversion, projection transformation, and quality control on numerical weather prediction, satellite, and radar data to construct a standardized input dataset. Secondly, to address the shortcomings of traditional fixed-weight fusion models in data feature extraction, a deep neural network is introduced into the data fusion module 102. This module uses a convolutional neural network to extract spatial features, a recurrent neural network to capture time-series features, and an attention mechanism to dynamically allocate data weights, achieving intelligent fusion of multi-source data. Finally, to address the issue of the inability to adjust simulation strategies in real time under complex sea conditions, a reinforcement learning framework is embedded in the model adjustment module 103 to adjust sea state type, meteorological element change rate, etc. The state space is defined as the action space, and the background field interpolation parameters and observation data assimilation weights are defined as the action space. The matching degree between the fusion result and the measured wind field is used as the reward function to achieve dynamic optimization of the fusion strategy. Addressing the accuracy bottleneck of traditional interpolation algorithms in spatiotemporal resolution conversion, the Kriging interpolation algorithm is improved in the resolution conversion module 104. Terrain correction factors and physical constraints of meteorological elements are introduced, and spatiotemporal joint filtering technology is combined to achieve the conversion from kilometer-level raw data to hundred-meter-level refined three-dimensional wind field data. Finally, addressing the limitations of the yaw control strategy, control areas are divided based on the refined wind field data. The partition control module 105 calls a preset yaw control strategy mapping table to generate differentiated yaw error ranges, adjustment frequencies, and coordinated yaw commands. This concept, through modular design and technology integration, systematically solves the core contradictions of existing technologies in simulation accuracy, scene adaptability, data resolution, and control strategy flexibility, providing an innovative solution for the efficient and stable operation of offshore wind power.

[0084] In this embodiment of the disclosure, the offshore wind farm simulation and wind turbine control system provided in this embodiment achieves full-process processing from raw data to refined control through multi-module collaboration.

[0085] The data preprocessing module 101 addresses the format differences and spatiotemporal biases of multi-source heterogeneous meteorological data (such as GFS (Global Forecast System), GRAPES (Global / Regional Assimilation and PrEdiction System, China Meteorological Administration's Global / Regional Assimilation and PrEdiction System) numerical forecast products, satellite-level data, radar-based data, etc.). For example, it unifies the latitude and longitude coordinates of different data sources into the same projected coordinate system and uses a quality control mechanism to remove outlier data points, ensuring the standardization and reliability of the input data. For instance, when fusing 10km resolution wind speed data from satellites with minute-level radar scan data, format conversion is needed to eliminate data structure differences, and then statistical methods are used to filter out noise points.

[0086] The data fusion module 102 intelligently fuses the preprocessed multi-source data to generate three-dimensional fused field data. Taking complex sea states as an example, this module can receive large-scale wind field trends from numerical weather prediction and local observation data from satellite radar. Through a deep learning model, it automatically identifies the feature weights of different data sources, such as enhancing the weight of radar echo data in the typhoon eyewall region and suppressing low-confidence numerical prediction error areas, thereby improving the ability of the fused field data to characterize extreme features such as strong winds and turbulence.

[0087] The model adjustment module 103 dynamically optimizes the data fusion strategy based on sea state changes. For example, when a typhoon passes through, the system detects sudden changes in wind speed and increased turbulence intensity, automatically adjusts the parameters of the fusion model, increases the assimilation ratio of real-time observation data, reduces the dependence on the static numerical forecast background field, makes the fusion results closer to the real-time wind field evolution, and improves the simulation response speed in complex scenarios.

[0088] The resolution conversion module 104 performs spatiotemporal resolution conversion on the 3D fused field data. In offshore wind power scenarios, this module can combine terrain correction factors (such as sea surface roughness parameters) to interpolate and correct wind speed and direction data at different height levels (such as wind turbine hub height), generating 3D wind field data that meets the requirements of unit load analysis.

[0089] The zone control module 105 divides the wind field into zones based on refined three-dimensional wind field data. According to the wind force, direction, and characteristics of each zone, it calls the strategy mapping table to generate differentiated yaw control strategies. For example, in low-wind-speed zones, the focus is on efficient wind capture, while in high-turbulence zones, the emphasis is on reducing load. The system then drives the turbine units to execute these strategies, achieving refined wind field management and adaptive control of the turbine units.

[0090] In another possible implementation, the data preprocessing module 101 can precisely filter data sources based on geographical coordinate ranges. By pre-setting spatial identifiers for offshore wind power areas, it extracts only multi-source data within that area, such as wind field data from meteorological satellite remote sensing, real-time monitoring data from offshore wind measurement towers, and operational data from the wind turbine SCADA (Supervisory Control and Data Acquisition) system. Irrelevant data from outside the area is filtered out, thereby effectively limiting the amount of data processed. This targeted processing method avoids interference from redundant information, ensuring that the module focuses on key data within the offshore wind power area, improving data processing efficiency and the accuracy of subsequent wind field simulation and control strategy formulation.

[0091] In summary, the system provided by this invention, through the collaborative cooperation of its various modules, not only solves the problems of insufficient accuracy in multi-source data fusion and delayed response in complex scenarios, but also achieves differentiated control based on the characteristics of wind farm zones. Overall, it provides a modular solution for the efficient and reliable operation of offshore wind farms.

[0092] In yet another embodiment provided in this disclosure, such as Figure 2 As shown, the partition control module 105 includes:

[0093] The region division unit 1051 is used to read the wind speed distribution, turbulence intensity distribution, and wake influence range data from the refined three-dimensional wind field data, and divide the wind field into at least the following control regions based on preset wind speed thresholds, turbulence intensity thresholds, and wake influence distances:

[0094] Low wind speed area, corresponding to wind speeds below the first preset threshold;

[0095] The medium wind speed region corresponds to wind speeds between the first preset threshold and the second preset threshold.

[0096] High turbulence region, corresponding to turbulence intensity higher than the preset turbulence intensity threshold;

[0097] The wake influence area refers to the area within the wake influence distance covered by the wake of the upstream unit.

[0098] In this embodiment of the disclosure, the region division unit 1051 of the partition control module 105 realizes refined spatial subdivision of the offshore wind field through multi-dimensional data analysis and dynamic threshold matching mechanism. The unit first reads refined three-dimensional wind field data, extracts key parameters such as wind speed distribution, turbulence intensity distribution and wake influence range, and divides the region based on preset physical thresholds. For example, regarding wind speed, a first preset threshold of 6 m / s and a second preset threshold of 12 m / s can be set to divide the wind field into low-wind-speed areas (<6 m / s, requiring yaw control to maximize alignment with the wind direction to improve wind energy capture efficiency), medium-wind-speed areas (6-12 m / s, requiring a control strategy that balances energy efficiency and unit load), and high-wind-speed areas (≥12 m / s, prioritizing load unloading strategies to protect equipment). In terms of turbulence intensity, a high-turbulence area is defined with a threshold of 15% (because turbulent airflow can easily exacerbate mechanical wear, the yaw error range needs to be widened and the adjustment frequency reduced). The wake-affected area is defined based on the upstream unit's rotor diameter D, with a distance range of 3D-5D behind it as the area affected by wake obstruction (requiring coordinated yaw of multiple units to reduce energy loss).

[0099] In practical applications, the region division unit 1051 can optimize threshold settings by combining machine learning algorithms. For example, by analyzing historical operational data to train models, it can automatically identify the optimal zoning boundaries under different sea conditions. For instance, during typhoons, the system can automatically increase the wind speed threshold based on real-time wind speed change data to adapt to extreme weather. During the summer when sea and land breezes dominate, the threshold parameters are fine-tuned based on the diurnal variation characteristics of turbulence intensity. In addition, the unit can also introduce multi-dimensional parameters such as wind shear and atmospheric stability to achieve more accurate region division. For example, high-wind-speed areas can be further subdivided into strong wind shear sub-regions and uniform strong wind sub-regions, allowing for targeted adjustments to yaw strategies. Through the organic combination of physical rules and intelligent algorithms, this unit transforms the continuous three-dimensional wind field into sub-regions with clear control objectives, providing accurate spatial coordinates and control logic benchmarks for the generation of differentiated yaw strategies. At the same time, through dynamic threshold adjustment and multi-source data fusion, it continuously improves zoning accuracy and strategy adaptability, ensuring deep coordination between the unit's control strategy and the complex and ever-changing wind field environment.

[0100] In yet another embodiment provided in this disclosure, such as Figure 3 As shown, the partition control module 105 also includes:

[0101] The control strategy generation unit 1052 is used to generate a yaw adjustment strategy for each control region by calling a preset yaw control strategy mapping table based on the characteristics of each control region. The yaw control strategy mapping table includes:

[0102] For low wind speed areas, control parameters are generated with a yaw error range of the first range and an adjustment frequency of the first frequency to maximize wind energy capture.

[0103] For medium wind speed areas, control parameters are generated with the yaw error range being the second range and the adjustment frequency being the second frequency, in order to balance energy efficiency and equipment load.

[0104] For the highly turbulent region, control parameters with a yaw error range of the third range and an adjustment frequency of the third frequency are generated to reduce mechanical wear;

[0105] For the wake-affected area, cooperative yaw control parameters based on the yaw angle of the upstream units are generated to reduce wake loss.

[0106] In this embodiment, the control strategy generation unit 1052 of the zone control module 105 generates differentiated yaw control parameters based on the zone division results and the physical characteristics and operational requirements of different wind field areas, by calling a preset strategy library. For example, in low wind speed areas (wind speed < 6 m / s), due to the low wind energy density, the control focus is on improving wind capture efficiency by accurately aligning with the wind direction. The unit generates parameters with a yaw error range of ±5° and an adjustment frequency of 1 time / minute, enabling the nacelle to track the prevailing wind direction in real time. Actual measurement data from a certain sea area shows that this strategy can increase the unit's power generation by about 10% during low wind speed periods. In medium wind speed areas (6-12 m / s), the strategy needs to balance energy capture and equipment wear. Therefore, parameters with an error range of ±8° and an adjustment frequency of 1 time / 2 minutes are generated. While ensuring wind capture efficiency, this reduces mechanical wear caused by frequent yaw system actions. In typical sea and land wind scenarios, this can reduce unit load fluctuations by about 15%.

[0107] In highly turbulent regions with turbulence intensity >15%, airflow disturbances can exacerbate fatigue in the blades and transmission system. The unit utilizes parameters with an error range of ±10° and an adjustment frequency of once every 5 minutes. By relaxing alignment accuracy and extending adjustment intervals, the alternating load caused by turbulence is reduced. Testing has verified that this strategy can reduce the peak gearbox load by 15%-20%. In the wake influence area (3D-5D distance behind the upstream unit), the unit calculates coordinated control parameters based on the real-time yaw angle of the upstream unit. For example, when the upstream unit yaws 10° to avoid strong winds, the downstream unit automatically adjusts by 5° to reduce wake obstruction.

[0108] The control strategy generation unit 1052 supports a dynamic optimization mechanism. By analyzing the correlation between energy efficiency and load in historical operating data, it automatically corrects parameter thresholds using algorithms such as linear regression. For example, based on long-term data, a wind farm can optimize the adjustment frequency in high-turbulence areas from once every 5 minutes to once every 4 minutes, improving wind capture efficiency during certain periods while controlling the load. Simultaneously, the unit can interface with feedforward sensors such as lidar. When the rate of change in incoming wind direction exceeds 5° / min, it adjusts control parameters in advance. For instance, in low-wind-speed areas, the adjustment frequency can be temporarily increased to once every 30 seconds to compensate for mechanical response lag and further enhance control real-time performance. By integrating regional physical characteristics, operational objectives, and parameter configurations, this unit forms a set of strategies covering different scenarios. Through data-driven optimization mechanisms and feedforward sensing capabilities, it continuously improves the adaptability of control strategies to complex wind farm environments, providing refined strategy support for the efficient and reliable operation of offshore wind turbines under multiple operating conditions.

[0109] In another embodiment provided in this disclosure, the control strategy generation unit 1052 is used to correct the yaw control parameters using the following formula when the yaw angle of the wind turbine exceeds the corresponding yaw error range:

[0110]

[0111] Where Δψ represents the yaw angle adjustment vector; A represents the operating condition weight matrix, with the diagonal elements corresponding to the weight coefficients of each control area; α represents the wind speed change rate weight coefficient; β represents the turbulence accumulation effect coefficient. Characterizes the rate of change of wind speed; Offshore wind field simulation and wind turbine control characterize the angle between wind direction and nacelle azimuth; B characterizes the wake transfer matrix, the elements of which are determined by the relative positions of upstream and downstream turbines and the distribution of turbulence intensity; Ψup characterizes the yaw angle matrix of the upstream turbine; t characterizes the current moment, and t′ characterizes the previous moment.

[0112] In this embodiment, the control strategy generation unit 1052 intelligently corrects the yaw control parameters using a formula when the yaw angle of the wind turbine exceeds the corresponding error range. In the formula, Δψ serves as the yaw angle adjustment vector, integrating the combined influence of multiple physical quantities. The diagonal elements of the operating condition weight matrix A correspond to the weight coefficients of each control region. For example, in low-wind-speed regions, since wind energy capture is the priority target, its corresponding weight coefficient increases, making the adjustment more focused on the characteristics of that region. If a surge in turbulence is detected in real time, the weight of low-wind-speed regions in A can be dynamically reduced to avoid excessive adjustment caused by turbulence, which could exacerbate mechanical wear. The wind speed change rate weight coefficient α and... The relationship with the wind speed change rate is significant. If the wind speed changes rapidly (e.g., in a sudden gust), α amplifies the impact of the wind speed change rate on the adjustment amount, prompting the unit to respond promptly to wind direction changes. Combined with a short-term wind speed prediction model, Δψ can be corrected in advance, rather than relying solely on current and previous data, enhancing the forward-looking nature of the control. The angle between the wind direction and the nacelle azimuth (θ-φ) is calculated using a cosine function to ensure the rationality of the adjustment direction. For example, when the angle is large, the adjustment amount will be increased to quickly align with the wind direction.

[0113] The turbulence cumulative effect coefficient β, combined with the integral term, reflects the long-term impact of turbulence on the unit. For example, in high-turbulence regions with longer operating times, the integral term increases, and the adjustment tends to reduce mechanical wear. The elements of the wake transfer matrix B are determined by the relative positions of the upstream and downstream units and the turbulence intensity distribution. The yaw angle matrix Ψup of the upstream unit transmits influence through B. For instance, the downstream unit adjusts collaboratively based on the upstream yaw angle, calculated using the B matrix, to reduce wake obstruction losses. In practical applications, real-time scanning of turbulence distribution using lidar and online updating of the B matrix parameters can improve the adjustment accuracy in the wake region. This formula, through multi-parameter coupling, accurately adapts to complex wind field conditions. For example, in low wind speeds with drastic wind speed changes, it quickly adjusts the yaw angle to improve wind capture efficiency. In the wake region, collaborative control is achieved based on the yaw of the upstream unit, reducing energy loss, ensuring the effective achievement of control objectives in each region, and improving the overall energy efficiency and operational reliability of the unit.

[0114] In another embodiment provided in this disclosure, the control strategy generation unit 1052 is also used to enable lidar scanning data in low wind speed areas to obtain the incoming wind direction in advance and compensate for yaw lag; in the wake influence area, it generates a follow yaw command based on the yaw angle of the upstream unit to achieve wind farm-level coordinated yaw.

[0115] In this embodiment, the control strategy generation unit 1052 employs differentiated control logic tailored to the characteristics of different wind farm areas to improve the operating efficiency of wind turbine units. In low-wind-speed areas, where wind energy resources are relatively scarce, timely and accurate wind energy capture becomes crucial. This unit utilizes lidar scanning data to obtain incoming wind direction information in advance, effectively compensating for the mechanical response lag of the yaw system. It can also deeply integrate lidar data with short-term wind speed prediction models, not only compensating for lag but also predicting wind direction changes and planning yaw paths in advance. For example, if the model predicts continuous changes in future wind direction, the unit adjusts the yaw rate in advance to avoid mechanical losses caused by frequent start-stop operations. For instance, in a low-wind-speed period at a certain offshore wind farm, lidar detects wind direction changes in advance, allowing the unit to adjust the yaw angle accordingly, which can improve power generation performance under low-wind-speed conditions. In the wake-affected area, to reduce energy loss caused by the wake effect, the unit generates a follow-yaw command based on the yaw angle of the upstream turbine, achieving wind farm-level coordinated control. Turbulence intensity is introduced as a dynamic adjustment factor; when turbulence is strong, the follow-yaw angle is reduced to minimize instability impact, while when turbulence is weak, the angle is increased to fully utilize the adjustment effect. For example, if the upstream turbine yaws by 15° to optimize the wind capture direction, the downstream turbine adjusts accordingly after receiving the command, reducing wake obstruction. Actual measurements show that power generation in this area can increase by approximately 12%, effectively improving the overall energy efficiency of the wind farm. By integrating extended implementation methods into the strategy, the control strategy generation unit 1052 can more accurately adapt to complex wind farm environments, improve the overall energy efficiency and operational stability of the wind farm, ensure the effective achievement of control objectives in each area, enhance the system's adaptability to changing operating conditions, and promote the upgrade of wind turbine control from local optimization to global coordinated optimization.

[0116] In yet another embodiment provided in this disclosure, such as Figure 4 As shown, the data fusion module 102 includes:

[0117] The deep neural network unit 1021 is used to receive GFS numerical weather prediction products, GRAPES numerical weather prediction products, satellite first-level data, radar base data and ground observation data as input. It extracts spatial features of the data through convolutional neural networks, captures time series features of the data through recurrent neural networks, and assigns fusion weights to numerical weather prediction products, satellite data, radar data and ground observation data through attention mechanisms, and outputs three-dimensional fused field data.

[0118] In this embodiment, the data fusion module 102 achieves efficient fusion of multi-source heterogeneous data with the help of the deep neural network unit 1021. Its inputs cover diverse data sources such as GFS, GRAPES numerical weather prediction products, satellite-level data, radar-based data, and ground observation data. The deep neural network unit 1021 employs a hierarchical processing architecture: the Convolutional Neural Network (CNN) scans through multiple layers of convolutional kernels to extract spatial features such as wind speed gradients in satellite cloud images and spatial distribution of precipitation particles in radar echoes. For example, it can identify the spatial morphology of strong wind areas of typhoon eyewalls from satellite infrared cloud images. Furthermore, it can be combined with multi-scale convolutional kernel technology to more comprehensively capture spatial features at different scales and improve the perception of complex meteorological spatial distributions, such as being more accurate in identifying large-scale frontal cloud systems and small-scale convective cloud clusters. The Recurrent Neural Network (RNN) performs time-series modeling on the hourly wind speed sequences of numerical weather predictions and the minute-level wind direction data from ground observation stations to capture time-dimensional features such as diurnal variations of sea and land winds and cumulative growth of typhoon wind speeds. In addition, gating mechanisms such as LSTM (Long Short-Term Memory) or GRU (Gated Recurrent Unit) can be introduced during implementation to effectively solve the gradient vanishing or exploding problem of traditional RNNs, and enhance the ability to model long-term time series data. For example, it can more accurately capture trends when processing long-term monsoon variation data.

[0119] The attention mechanism, as a core component, can dynamically calibrate the fusion weights of different data sources. In winter cold wave scenarios, radar-based data captures strong wind shear at the leading edge of cold fronts more accurately, and the attention mechanism automatically increases its weight to over 60%, while suppressing the weight of error regions caused by insufficient model resolution in numerical forecasts. In summer stable sea states, the high precision advantage of ground observation data gives it a weight of up to 40%, ensuring the spatiotemporal consistency of the fused data. Furthermore, in implementation, the attention mechanism can not only be based on real-time data characteristics but also dynamically adjust the weight allocation strategy based on historical fusion accuracy and application scenario requirements. For example, during periods of stable meteorological conditions, the weight of numerical forecast products can be increased based on historical data statistical patterns to enhance the stability of the fusion results. When processing long-term time-series data, the Transformer architecture can be introduced to replace the traditional RNN, utilizing its self-attention mechanism to capture dependencies across multiple time steps in parallel. For instance, when analyzing multi-day forecast data of typhoon tracks, the Transformer can more efficiently extract long-term evolution features from the time series, improving the fusion model's ability to represent complex meteorological processes. Ultimately, the deep neural network unit 1021 outputs fused field data containing three-dimensional space (x,y,z) and time dimensions through a composite mechanism of "spatial feature extraction - temporal dynamic modeling - intelligent weight allocation". This provides a highly reliable input basis for subsequent mode adjustment and resolution conversion, significantly improving the spatiotemporal accuracy and dynamic response capability of offshore wind field simulation.

[0120] In yet another embodiment provided in this disclosure, such as Figure 5 As shown, the mode adjustment module 103 includes:

[0121] The reinforcement learning unit 1031 is used to define a state space including sea state type and meteorological element change rate, and an action space including background field interpolation parameter adjustment and observation data assimilation weight adjustment. The matching degree between the fusion result and the measured wind field is used as the reward function. Through reinforcement learning training, data fusion strategy adjustment instructions are generated to dynamically adjust the parameters of the data fusion module 102.

[0122] In this embodiment of the disclosure, the mode adjustment module 103 constructs a dynamic data fusion strategy optimization mechanism through the reinforcement learning unit 1031. This unit first defines a state space containing multi-dimensional features such as sea state type (e.g., calm sea state, typhoon, sea breeze), meteorological element change rate (e.g., wind speed change rate > 5 m / s / h, wind direction shear > 10° / km), turbulence intensity (e.g., > 15%), and atmospheric stability, as well as an action space containing background field interpolation parameter adjustment (e.g., LAPS model horizontal diffusion coefficient), observation data assimilation weight adjustment (e.g., radar data assimilation ratio), and differential parameter optimization for different height layers (e.g., wind turbine hub height 100-120 meters). The root mean square error (RMSE) between the fusion result and the buoy measured wind field is used as the reward function. For example, in a scenario where a typhoon is passing through and accompanied by high turbulence (turbulence intensity 20%), the state space simultaneously detects the "typhoon" sea state, wind speed change rate of 8 m / s / h, and turbulence intensity exceeding the standard. The reinforcement learning unit 1031 increases the radar data assimilation weight from 30% to 50% through the action space, and at the same time increases the weight of the lidar data to 25% separately for the wheel hub height layer, so that the RMSE of the fusion result of this height layer decreases from 2.8 m / s to 1.6 m / s, triggering positive feedback of the reward function.

[0123] Reinforcement learning algorithms can be implemented as deep reinforcement learning algorithms such as DQN (Deep Q-Network) and PPO (Proximal Policy Optimization). These algorithms fit high-dimensional state-action value functions through neural networks. For example, in complex scenarios involving typhoon path changes, sea-land breeze circulation, and complex terrain, deep reinforcement learning models can converge to the optimal adjustment strategy within 100 iterations. This mechanism, which incorporates turbulence intensity and height layer characteristics into the state-action system and optimizes strategy search through advanced algorithms, enables the system to accurately identify the adjustment needs of fusion parameters in extreme weather or multi-level complex flow field environments. For instance, in scenarios dominated by summer sea-land breezes and with low-level jet streams, it automatically adjusts the numerical forecast interpolation parameters near the boundary layer height, improving the accuracy of wind field simulation in that region. This upgrades the data fusion strategy from static experience-driven to dynamic intelligent-driven, significantly enhancing the adaptability of the fused field data to spatiotemporally heterogeneous sea conditions and laying a reliable foundation for subsequent refined wind field modeling and component-based zonal control.

[0124] In yet another embodiment provided in this disclosure, such as Figure 6 As shown, the resolution conversion module 104 includes:

[0125] Kriging interpolation unit 1041 is used to acquire DEM data of offshore wind power areas as terrain correction factors. Based on the physical relationship between wind speed and pressure gradient, meteorological element correlation constraints are established, and spatial interpolation of three-dimensional fused field data is performed through an improved Kriging interpolation algorithm.

[0126] The spatiotemporal joint filtering unit 1042 is used to perform Kalman filtering on time series meteorological data to remove noise, and to perform spatiotemporal joint processing on the filtered time series data and the spatially interpolated three-dimensional fused field data to generate gridded and refined three-dimensional wind field data for offshore wind power with a preset grid resolution.

[0127] In this embodiment, the resolution conversion module 104 achieves an upgrade from coarse-resolution fused data to refined three-dimensional wind field data through the collaborative operation of the Kriging interpolation unit 1041 and the spatiotemporal joint filtering unit 1042. The Kriging interpolation unit 1041 first acquires DEM (Digital Elevation Model) data (such as a 50-meter resolution DEM) of the offshore wind power area as a terrain correction factor. For example, in sea areas with islands or abrupt changes in seabed topography, the DEM data can quantify the blocking or accelerating effect of topographic undulations on wind speed. Combined with the physical relationship between wind speed and pressure gradient (such as the geostrophic wind equation), the correlation constraints of meteorological elements are established to ensure that the interpolation process conforms to the principles of fluid dynamics. The improved Kriging interpolation algorithm combines traditional geostatistical methods with physical constraints. For example, when calculating the semivariogram, a pressure gradient term is introduced as a covariate, so that the interpolation result not only reflects the spatial distribution of the data, but also follows the physical law that "wind speed is proportional to pressure gradient". In addition, during implementation, machine learning models (such as random forests) can be introduced to optimize the interpolation parameters, and the optimal semivariogram model under different sea conditions can be automatically identified by training historical data.

[0128] The spatiotemporal joint filtering unit 1042 performs Kalman filtering on time-series meteorological data (such as wind speed sequences scanned by radar every minute). Through iterative updates of the state equation and observation equation, it effectively filters out sensor noise and outliers. For example, it reduces the root mean square error of the lidar measured wind speed sequence from 0.8 m / s to 0.3 m / s. Adaptive Kalman filtering can also be used to dynamically adjust the process noise covariance matrix according to sea state complexity. For instance, it increases the noise weight during typhoon passage to adapt to drastic data fluctuations and improves filtering robustness in extreme scenarios. The filtered time-series data and the spatially interpolated 3D fused field data are matched with timestamps and spatial grids using a spatiotemporal cross-validation algorithm (such as dynamic time warping technology) to ensure consistency of wind speed and direction data at the same spatiotemporal point. Ultimately, the module generates detailed 3D wind field data at the 100-meter level (e.g., 100m×100m×10m grid), which can accurately characterize key parameters such as wind shear and turbulence intensity at the wind turbine hub height (e.g., 120 meters). This provides a high-precision data foundation for unit load analysis and control strategy formulation. Compared with traditional interpolation methods, it achieves significant improvements in both spatiotemporal resolution and physical consistency. In particular, after integrating machine learning optimization and adaptive filtering mechanisms, the system's ability to characterize complex sea conditions and extreme weather is further enhanced.

[0129] Based on the same disclosed concept, this disclosure also provides an offshore wind farm simulation and wind turbine control method, a wind turbine, and a storage medium. Since the principles by which these methods, wind turbines, and storage media solve the problems are similar to those of the aforementioned offshore wind farm simulation and wind turbine control system, the implementation of these methods, wind turbines, and storage media can refer to the implementation of the aforementioned system, and the repeated parts will not be described again.

[0130] With the above Figure 1 Corresponding to the offshore wind farm simulation and wind turbine control system shown, this disclosure also provides an offshore wind farm simulation and wind turbine control method, such as... Figure 7 As shown, it includes:

[0131] S701. Preprocessing operations for format conversion, projection transformation and quality control of multi-source heterogeneous meteorological data;

[0132] S702. Intelligent fusion of preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fused field data;

[0133] S703, Dynamically Optimize Data Fusion Strategy Based on Sea State Changes;

[0134] S704. Perform spatiotemporal resolution conversion on the three-dimensional fused field data to generate refined three-dimensional wind field data;

[0135] S705 divides the wind field into regions based on refined three-dimensional wind field data, generates differentiated yaw control strategies for each region based on wind force, wind direction data and regional characteristics, and drives the unit to execute the yaw control strategies.

[0136] In another embodiment provided in this disclosure, the wind field is divided into regions based on refined three-dimensional wind field data, including:

[0137] Read the wind speed distribution, turbulence intensity distribution, and wake influence range data from the refined 3D wind field data. Based on preset wind speed thresholds, turbulence intensity thresholds, and wake influence distances, divide the wind field into at least the following control regions:

[0138] Low wind speed area, corresponding to wind speeds below the first preset threshold;

[0139] The medium wind speed region corresponds to wind speeds between the first preset threshold and the second preset threshold.

[0140] High turbulence region, corresponding to turbulence intensity higher than the preset turbulence intensity threshold;

[0141] The wake influence area refers to the area within the wake influence distance covered by the wake of the upstream unit.

[0142] In another embodiment provided in this disclosure, a differentiated yaw control strategy for each region is generated based on wind force and direction data and regional characteristics, including:

[0143] Based on the characteristics of each control area, a preset yaw control strategy mapping table is invoked to generate the corresponding yaw adjustment strategy for that area. The yaw control strategy mapping table includes:

[0144] For low wind speed areas, control parameters are generated with a yaw error range of the first range and an adjustment frequency of the first frequency to maximize wind energy capture.

[0145] For medium wind speed areas, control parameters are generated with the yaw error range being the second range and the adjustment frequency being the second frequency, in order to balance energy efficiency and equipment load.

[0146] For the highly turbulent region, control parameters with a yaw error range of the third range and an adjustment frequency of the third frequency are generated to reduce mechanical wear;

[0147] For the wake-affected area, cooperative yaw control parameters based on the yaw angle of the upstream units are generated to reduce wake loss.

[0148] In another embodiment provided in this disclosure, the method further includes: when the yaw angle of the wind turbine exceeds the corresponding yaw error range, correcting the yaw control parameters using the following formula:

[0149]

[0150] Where Δψ represents the yaw angle adjustment vector; A represents the operating condition weight matrix, with the diagonal elements corresponding to the weight coefficients of each control area; α represents the wind speed change rate weight coefficient; β represents the turbulence accumulation effect coefficient. Characterizes the rate of change of wind speed; Offshore wind field simulation and wind turbine control characterize the angle between wind direction and nacelle azimuth; B represents the wake transfer matrix, whose elements are determined by the relative positions of upstream and downstream turbines and the distribution of turbulence intensity; Ψ up The matrix represents the yaw angle of the upstream unit; t represents the current time, and t′ represents the previous time.

[0151] In another embodiment provided in this disclosure, generating a differentiated yaw control strategy for a corresponding region based on wind force and direction data and regional characteristics of each region further includes:

[0152] In low-wind-speed areas, lidar scanning data is used to obtain the incoming wind direction in advance and compensate for yaw lag; in wake-affected areas, follow-yaw commands are generated based on the yaw angle of the upstream units to achieve wind farm-level coordinated yaw.

[0153] In another embodiment provided in this disclosure, preprocessed multi-source heterogeneous meteorological data are intelligently fused to generate three-dimensional fused field data, including:

[0154] It receives GFS numerical weather prediction products, GRAPES numerical weather prediction products, satellite first-level data, radar base data, and ground observation data as input. It extracts spatial features of the data through convolutional neural networks, captures time series features of the data through recurrent neural networks, and assigns fusion weights to numerical weather prediction products, satellite data, radar data, and ground observation data through an attention mechanism, and outputs three-dimensional fused field data.

[0155] In another embodiment provided in this disclosure, a data fusion strategy is dynamically optimized based on sea state changes, including:

[0156] Define a state space that includes sea state type and meteorological element change rate, and define an action space that includes background field interpolation parameter adjustment and observation data assimilation weight adjustment. Use the matching degree between the fusion result and the measured wind field as the reward function, and generate data fusion strategy adjustment instructions through reinforcement learning training to dynamically adjust the parameters of data fusion module 102.

[0157] In another embodiment provided in this disclosure, spatiotemporal resolution conversion is performed on the three-dimensional fused field data to generate refined three-dimensional wind field data, including:

[0158] DEM data of offshore wind power areas were acquired as topographic correction factors. Based on the physical relationship between wind speed and pressure gradient, correlation constraints of meteorological elements were established. Spatial interpolation of the three-dimensional fused field data was performed by an improved Kriging interpolation algorithm.

[0159] Kalman filtering is applied to time-series meteorological data to remove noise. The filtered time-series data and the spatially interpolated three-dimensional fused field data are then subjected to spatiotemporal joint processing to generate gridded, refined three-dimensional wind field data for offshore wind power with a preset grid resolution.

[0160] This disclosure provides a wind turbine generator set, and any embodiment of this disclosure provides an offshore wind farm simulation and a wind turbine generator set control system.

[0161] This disclosure provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program performs the steps of a method for simulating an offshore wind farm and controlling a wind turbine generator, as provided in any embodiment of this disclosure. The storage medium can be volatile or non-volatile computer-readable storage.

[0162] Through the above description of the embodiments, those skilled in the art can clearly understand that the embodiments of this disclosure can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this disclosure.

[0163] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes in the drawings are not necessarily essential for implementing this disclosure.

[0164] Those skilled in the art will understand that the modules in the system of the embodiments can be distributed in the system of the embodiments as described in the embodiments, or they can be located in one or more systems different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0165] The sequence numbers of the embodiments disclosed above are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0166] Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A control system for offshore wind farm simulation and wind turbine generator, characterized in that, include: The data preprocessing module is used for format conversion, projection transformation, and quality control of multi-source heterogeneous meteorological data; The data fusion module is used to intelligently fuse preprocessed multi-source heterogeneous meteorological data to generate three-dimensional fused field data. The mode adjustment module is used to dynamically optimize the data fusion strategy based on changes in sea state; The resolution conversion module is used to perform spatiotemporal resolution conversion on the 3D fused field data to generate refined 3D wind field data. The zone control module is used to divide the wind field into zones based on the refined three-dimensional wind field data, generate differentiated yaw control strategies for each zone according to the wind force, wind direction data and regional characteristics, and drive the unit to execute the yaw control strategies.

2. The system as described in claim 1, characterized in that, The partition control module includes: The region division unit is used to read the wind speed distribution, turbulence intensity distribution, and wake influence range data from the refined three-dimensional wind field data, and divide the wind field into at least the following control regions based on preset wind speed thresholds, turbulence intensity thresholds, and wake influence distances: Low wind speed area, corresponding to wind speeds below the first preset threshold; The medium wind speed region corresponds to wind speeds between the first preset threshold and the second preset threshold. High turbulence region, corresponding to turbulence intensity higher than the preset turbulence intensity threshold; The wake influence area refers to the area within the wake influence distance covered by the wake of the upstream unit.

3. The system as described in claim 1, characterized in that, The partition control module also includes: A control strategy generation unit is used to generate a yaw adjustment strategy for each control region by calling a preset yaw control strategy mapping table based on the characteristics of each control region. The yaw control strategy mapping table includes: For low wind speed areas, control parameters are generated with a yaw error range of the first range and an adjustment frequency of the first frequency to maximize wind energy capture. For medium wind speed areas, control parameters are generated with the yaw error range being the second range and the adjustment frequency being the second frequency, in order to balance energy efficiency and equipment load. For the highly turbulent region, control parameters with a yaw error range of the third range and an adjustment frequency of the third frequency are generated to reduce mechanical wear; For the wake-affected area, cooperative yaw control parameters based on the yaw angle of the upstream units are generated to reduce wake loss.

4. The system as described in claim 3, characterized in that, The control strategy generation unit is used to correct the yaw control parameters using the following formula when the yaw angle of the wind turbine exceeds the corresponding yaw error range: Where Δψ represents the yaw angle adjustment vector; A represents the operating condition weight matrix, with the diagonal elements corresponding to the weight coefficients of each control area; α represents the wind speed change rate weight coefficient; β represents the turbulence accumulation effect coefficient. Characterizes the rate of change of wind speed; Offshore wind field simulation and wind turbine control characterize the angle between wind direction and nacelle azimuth; B characterizes the wake transfer matrix, the elements of which are determined by the relative positions of upstream and downstream turbines and the distribution of turbulence intensity; Ψup characterizes the yaw angle matrix of the upstream turbine; t characterizes the current moment, and t′ characterizes the previous moment.

5. The system as described in claim 3, characterized in that, The control strategy generation unit is also used to enable lidar scanning data in low wind speed areas to obtain the incoming wind direction in advance and compensate for yaw lag; in the wake influence area, it generates follow yaw commands based on the yaw angle of the upstream unit to achieve wind farm-level coordinated yaw.

6. The system as described in claim 1, characterized in that, The data fusion module includes: The deep neural network unit receives GFS numerical weather prediction products, GRAPES numerical weather prediction products, satellite first-level data, radar base data, and ground observation data as input. It extracts spatial features of the data through convolutional neural networks, captures time series features of the data through recurrent neural networks, and assigns fusion weights to numerical weather prediction products, satellite data, radar data, and ground observation data through an attention mechanism, and outputs three-dimensional fused field data.

7. The system as described in claim 1, characterized in that, The mode adjustment module includes: The reinforcement learning unit defines a state space containing sea state type and meteorological element change rate, and an action space containing background field interpolation parameter adjustment and observation data assimilation weight adjustment. The reward function is the matching degree between the fusion result and the measured wind field. Through reinforcement learning training, it generates data fusion strategy adjustment instructions and dynamically adjusts the parameters of the data fusion module.

8. The system as described in claim 1, characterized in that, The resolution conversion module includes: Kriging interpolation unit is used to acquire DEM data of offshore wind power area as terrain correction factor. Based on the physical relationship between wind speed and pressure gradient, the correlation constraint of meteorological elements is established. The improved Kriging interpolation algorithm is used to perform spatial interpolation on the three-dimensional fused field data. The spatiotemporal joint filtering unit is used to perform Kalman filtering on time-series meteorological data to remove noise. It performs spatiotemporal joint processing on the filtered time-series data and the spatially interpolated three-dimensional fused field data to generate gridded, refined three-dimensional wind field data for offshore wind power with a preset grid resolution.

9. A method for simulating offshore wind farms and controlling wind turbine generators, characterized in that, include: Preprocessing operations for multi-source heterogeneous meteorological data, including format conversion, projection transformation, and quality control; Intelligent fusion of preprocessed multi-source heterogeneous meteorological data generates three-dimensional fused field data. Dynamically optimize data fusion strategies based on sea state changes; Spatiotemporal resolution conversion is performed on the 3D fused field data to generate refined 3D wind field data; Based on the refined three-dimensional wind field data, the wind field is divided into regions. Based on the wind force, wind direction data and regional characteristics of each region, a differentiated yaw control strategy is generated for the corresponding region, and the unit is driven to execute the yaw control strategy.

10. A wind turbine generator set, characterized in that, Includes the offshore wind farm simulation and wind turbine control system as described in any one of claims 1-7.

Images