A wind farm dynamic active power distribution method based on turbulence intensity

Abstract

The application discloses a wind farm dynamic active power distribution method based on turbulence intensity, which calculates the turbulence intensity and establishes the full-field distribution map by collecting the three-dimensional wind speed and wind direction data at the hub height of each unit in real time, and identifies the high / low turbulence area. Combined with the SCADA system data to evaluate the unit fatigue state, the unit with vibration close to the threshold is marked as "high risk". An optimization model with "minimum power tracking error" and "minimum fatigue load increment" as double targets is constructed, and the optimal target power of each unit is solved under the conditions of meeting the total power instruction and single machine constraints. Through priority initial distribution, physical limit correction and smoothing processing, the instruction is issued, and the feedback and emergency rebalancing mechanism is introduced to respond to turbulence mutation in real time. The method realizes the electromechanical coupling collaborative control, ensures the power response accuracy of the power grid, significantly reduces the mechanical load and operation risk of the unit, and improves the economy and safety of the whole life cycle of the wind farm.

Description

Technical Field

[0001] This invention relates to the field of wind farm optimization scheduling, and in particular to a method for dynamic active power allocation in wind farms based on turbulence intensity. Background Technology

[0002] In the actual operation of wind farms, active power allocation strategies are directly related to grid frequency stability and turbine equipment safety. However, traditional allocation methods (such as average allocation and capacity-ratio allocation) mainly focus on balancing electrical performance, often neglecting the complex fluid dynamics environment within the wind farm. This leads to three core contradictions: the contradiction between "equal treatment" and "wind condition differences": traditional strategies usually assume that all turbines are under the same wind conditions, or allocate power based solely on simple wind speed. In reality, due to wake effects and topography, the turbulence intensity varies greatly among turbines at different locations within the wind farm. Turbines in the upstream low-turbulence zone and those in the downstream high-turbulence zone are given the same control objectives. Turbines in the high-turbulence zone already have poor wind energy quality and fluctuating mechanical loads; forcing them to undertake high power generation tasks will exacerbate equipment fatigue damage. The contradiction between "power point tracking" and "equipment lifespan": grid dispatch requires wind farms to respond quickly to power commands (power point tracking), but wind turbines are complex, flexible mechanical structures. Existing methods typically do not consider mechanical fatigue loads as a real-time optimization objective when allocating power. Under conditions of high turbulence intensity, frequent pitch and torque adjustments can lead to significant alternating stresses on the tower, blades, and drive train. Long-term operation can significantly shorten the lifespan of critical components (such as gearboxes and bearings), increasing maintenance costs. There is a contradiction between "static allocation" and the "dynamic environment": wind is constantly changing, and turbulence intensity is highly random and abrupt (e.g., gusts, sudden changes in wind direction). Traditional allocation logic is often static or has a delayed response, lacking emergency avoidance mechanisms for sudden changes in turbulence intensity. When a wind turbine suddenly encounters extreme turbulence (such as wind shear, vortex-induced vibration), if the power command is not adjusted in time, it can easily lead to over-vibration shutdown of the unit, or even structural damage. Summary of the Invention

[0003] A method for dynamic active power allocation in wind farms based on turbulence intensity includes the following steps: S1. Data Acquisition and Turbulence Intensity Calculation: Real-time acquisition of three-dimensional wind speed components and wind direction data at the hub height of each wind turbine is achieved using the lidar of the wind turbine nacelle or the wind measurement tower of the wind farm; turbulence intensity of each turbine is calculated within a 10-minute statistical period, a turbulence intensity distribution map of the entire field is established, and high-turbulence and low-turbulence areas are identified; S2. Unit operating status and fatigue load assessment: Combine SCADA system data to read the current generator speed, pitch angle, and tower vibration acceleration of each unit; assess the current mechanical fatigue state of the unit. When a unit has high turbulence intensity but its vibration value is close to the safety threshold, it is marked as a "high-risk" state. S3. Dynamic Priority Factor Generation: Design a nonlinear mapping function to convert turbulence intensity into power allocation weight coefficients; through a formula, units with lower turbulence intensity are given higher weight coefficients, thereby establishing the logic of priority power generation; S4. Construction of Multi-Objective Optimization Allocation Model: Establish an optimization model with the dual objectives of "minimizing the power tracking error of the entire wind farm" and "minimizing the fatigue load increment of the entire wind farm"; set constraints such as the total output power of the wind farm must be equal to the dispatch command and the power of a single unit cannot exceed the rated power, and use quadratic programming or heuristic algorithms to solve for the optimal target power of each unit; S5. Priority-based initial power allocation: Under the constraints of the optimization model, the power increment is allocated to the units with higher weight coefficient values ​​first; the baseline allocated power for each unit is calculated.

[0004] S6. Power Correction and Limiting: Perform physical feasibility verification on the calculated baseline power allocation; if the baseline power exceeds the rated power, set it to equal the rated power and redistribute the remaining power demand to other units; if the baseline power is lower than the cut-in power, set it to 0 and consider the response rate limits of pitch and torque to smooth the slope of the power command change. S7. Command Issuance and Execution: The final target power command is issued to the main control system of each wind turbine through the wind farm monitoring system or a dedicated wind farm controller; after receiving the command, the turbine tracks the new active power setpoint by adjusting the generator torque and / or pitch angle. S8. Feedback and Rebalancing Mechanism: Repeat the above steps in the next control cycle; when a sudden change in the turbulence intensity of a unit is detected, an emergency redistribution is immediately triggered to rapidly reduce the power setpoint of the unit and prevent equipment damage.

[0005] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S1, the turbulence intensity of each unit within a 10-minute statistical period is calculated, a full-field turbulence intensity distribution map is established, and high-turbulence and low-turbulence regions are identified. The specific steps are as follows: S11. Data Acquisition and Time Window Division: Collect wind speed time series data at each unit from the wind farm SCADA system or the wind measurement equipment (such as nacelle anemometer, lidar) of each unit at a sampling frequency of not less than 1Hz. Divide the continuous wind speed data stream into windows according to a statistical period of 10 minutes, and mark each 10-minute segment with a timestamp as the basic unit for turbulence intensity calculation. S12. Calculation of turbulence intensity for a single unit: Calculate the wind speed time series data for each unit within each 10-minute window, respectively: Average wind speed V: The arithmetic mean of all wind speed samples within this window; Wind speed standard deviation σ: The standard deviation of wind speed samples within this window, reflecting the amplitude of wind speed fluctuations around the average value; The turbulence intensity I of the unit in the 10-minute period is calculated using the standard formula I = σ / V. This calculation is repeated for all units and all time windows to form a time series dataset of turbulence intensity for each unit. S13. Spatial Interpolation and Distribution Mapping of Turbulence Intensity Across the Entire Field: Obtain the precise geographic coordinates (latitude and longitude or planar projection coordinates) of all units in the field. Correlate the turbulence intensity values ​​of each unit during a representative time period (such as a typical 10-minute window or long-term average) calculated in S12 with their coordinates to form spatial scatter data. Import the digital elevation model (DEM) topographic data of the wind farm as the base map. Use spatial interpolation algorithms such as Kriging and inverse distance weighting to calculate the turbulence intensity at any grid point across the entire field based on the discrete turbulence intensity values ​​at each unit location. Generate continuously distributed raster data. Visualize the interpolation results in the form of contour cloud maps or heat maps and overlay them on the topographic map to form a spatial distribution map of turbulence intensity across the entire field. S14. Identification and Threshold Classification of High / Low Turbulence Zones: Setting Classification Standards: Referring to the IEC 61400-1 standard, based on the turbulence intensity at a wind speed of 15 m / s, the site area is divided into Class A (I_ref=0.16, high turbulence), Class B (I_ref=0.14, medium turbulence), and Class C (I_ref=0.12, low turbulence). The threshold can be customized according to the actual turbulence statistical distribution of the site. On the turbulence intensity distribution map of the entire site, the area with the interpolation result higher than the Class A threshold (or the customized high threshold) is designated as the high turbulence zone, and the area with the interpolation result lower than the Class C threshold (or the customized low threshold) is designated as the low turbulence zone, and they are distinguished by different colors or legends on the map; S15. Result Verification and Output: Compare the interpolation results at the location of the wind measurement tower with the measured and calculated turbulence intensity, calculate the error index (such as root mean square error RMSE), verify the accuracy of the spatial interpolation model, and compile and output the results such as the overall turbulence intensity distribution map, the boundary coordinates of high / low turbulence regions, and the turbulence intensity statistics table of each unit as the input basis for the subsequent active power allocation strategy. S16. Dynamic Update and Closed-Loop Feedback: Turbulence intensity is time-varying, so a dynamic update mechanism needs to be established: S11 to S15 are re-executed periodically (e.g., daily, weekly, or according to seasonal changes) to update the full-field turbulence intensity distribution map and the high / low turbulence zone division results with the latest operating data. The updated region division will be fed back to the active power allocation model in a closed loop, so that the allocation strategy can adapt to the changes in turbulence field characteristics under different time periods and different incoming wind directions.

[0006] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S2, data from the SCADA system is used to read the current generator speed, pitch angle, and tower vibration acceleration of each unit; the current mechanical fatigue state of the unit is assessed, as follows; S21. SCADA Data Reading and Synchronization: Read the following key parameters of each unit in real time from the wind farm's SCADA system, with a sampling frequency of no less than 1Hz, and align the timestamps of data from different sources: generator speed (rpm), pitch angle (°), tower front-to-back / lateral vibration acceleration (m / s²), output power (kW, auxiliary parameter), and wind speed (m / s, auxiliary parameter). S22. Data cleaning and operating condition segmentation: Remove sensor dead values ​​and abnormal jumps caused by communication interruptions. Use moving average and low-pass filtering to remove high-frequency noise. Based on generator speed and power threshold, the operating state is segmented into four types of operating condition segments: startup, normal operation, shutdown, and idling. Subsequent fatigue calculations are processed according to the operating conditions. S23. Tower vibration feature extraction: Perform fast Fourier transform on the tower vibration acceleration signal to extract the following features: vibration amplitude corresponding to the first and last modal frequencies of the tower (usually 0.2~0.4Hz), vibration amplitude corresponding to the first and last modal frequencies of the tower, and root mean square value (RMS) of vibration in each frequency band. These amplitudes are directly related to the magnitude of the dynamic bending moment borne by the tower. S24. Component Equivalent Load Estimation: Estimating the load timing of major components based on SCADA indirect parameters: Transmission chain torque: Calculate the electromagnetic torque using the generator speed and power, and then infer the low-speed shaft torque by combining the gearbox ratio; Equivalent bending moment of the tower: The timing of the overturning bending moment at the top of the tower is estimated by combining the thrust coefficient curve corresponding to the pitch angle and the nacelle vibration acceleration. S25 Damage Accumulation: Input the load time sequence from step 4 into the rainflow counting algorithm, count the number of cycles under each stress amplitude, and calculate the cumulative damage amount according to Miner's linear accumulation rule based on the material's SN curve: D = Σ(nᵢ / Nᵢ)+K scf Where nᵢ is the actual number of cycles, Nᵢ is the maximum allowable number of cycles under this stress amplitude, and K scf As a stress concentration correction factor, for geometrically discontinuous parts such as tower welds, flange connection bolt holes, and gear tooth roots, the actual local stress is higher than the nominal stress. Introducing the stress concentration factor to amplify the load time sequence before rainflow counting can more realistically reflect the damage state of key hot spots. S26. Fatigue status assessment and grading: Current cumulative damage value D, the closer D is to 1, the more severe the fatigue; damage rate: the damage increment dD / dt per unit time, a sudden increase in rate indicates deterioration of operating conditions; graded early warning: divided into four levels according to preset thresholds: normal (D<0.6), attention (0.6≤D<0.8), early warning (0.8≤D<0.95), and alarm (D≥0.95); S27. Results Output and Closed-Loop Application: Write the current fatigue status (D value, damage rate, warning level) of each unit into the database and display it visually on the monitoring interface. The evaluation results serve as the key input for the active power allocation strategy: prioritize the allocation of power commands to units with low fatigue damage, delay the damage accumulation of high fatigue units, and achieve collaborative optimization of the overall health status.

[0007] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S3, a nonlinear mapping function is designed to convert turbulence intensity into power allocation weighting coefficients. The specific steps are as follows: S31. Standardization of turbulence intensity: The specific turbulence intensity values ​​of each unit calculated in S1 are normalized. Specifically, the average turbulence level of all operating units in the field is calculated as a baseline, and then the turbulence value of a single unit is divided by this baseline to obtain a relative ratio. S32. Introduce a penalty mechanism for fatigue state: The health status of the unit assessed in step two is converted into a braking signal for weight calculation. An attenuation coefficient is designed. When the vibration value of the unit is within the safe range, the coefficient is close to one and has almost no impact on the weight. When the vibration acceleration is detected to exceed the preset warning threshold, the coefficient will decay non-linearly and rapidly approach zero. This means that when calculating the final weight, the score of the unit will be forced to be significantly reduced. S33. Generate final priority weights: Combine the above two factors and use the reciprocal of the smoothed turbulence ratio as the base weight: the smaller the turbulence, the larger the reciprocal, and the higher the base score. Multiply by the fatigue penalty coefficient, and then normalize the calculated result across the entire field to ensure that the sum of the weights of all units equals 1. The final output weight factor represents the priority capability of each unit in the next stage of power allocation.

[0008] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S4, an optimization model is established with the dual objectives of minimizing the full-field power tracking error and minimizing the full-field fatigue load increment. The specific steps are as follows: S41. Determine the variables to be solved: The variables to be solved are: how much electricity each unit should increase or decrease in power generation during the next control cycle. This is an allocation problem of adjustment amount, not an allocation of absolute value. S42. Constructing a dual-core objective logic: The model simultaneously considers two seemingly contradictory metrics: The primary objective is to ensure accuracy, meaning that the total power output of all units after adjustments must be exactly equal to the command value issued by the power grid dispatch center, and the difference between the actual total calculated by the model and the command value must be minimized. The second objective is to reduce losses and avoid subjecting high-turbulence, high-vibration units to large power fluctuations. The objective function will penalize the power adjustments assigned to high-risk units, which are required to perform drastic pitch and torque adjustments, resulting in very poor performance and high penalties. Conversely, assigning power fluctuations to stable units will result in very low penalties. S43 sets a hard boundary condition: the sum of the total output must meet the target; this is a mandatory condition. Output range constraints: The power of each unit cannot be lower than the minimum grid-connected power (if it is too low, the unit will be shut down directly), nor can it be higher than the maximum power that the unit can actually capture under the current turbulent wind conditions (this upper limit will be dynamically adjusted when the turbulence is strong). Adjustment rate constraint: Considering the response inertia of the pitch mechanism and generator, the power change per second cannot exceed the specified upper limit of the grade rate to prevent mechanical shock; S44. Solution and Engineering Processing: The objective function includes an absolute value penalty for the power variation range, transforming it into a standard solvable form. Under the premise of satisfying the equality constraints, the solver will automatically find the optimal path that minimizes the total adjustment range of each unit and avoids high-risk units. The final output is the most reasonable power adjustment command for each unit at the current moment.

[0009] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S6, the physical feasibility of the calculated baseline power allocation is verified, as follows; The theoretically allocated baseline power is sequentially verified by multiple constraints, including the upper limit of rated capacity, the lower limit of minimum operating power, the pitch / torque ramp rate limit, the turbulence intensity derating boundary, and the vibration protection threshold. The intersection of all limiting conditions is taken as the final feasible target power command. When a power gap occurs due to limiting, it is re-allocated among the remaining healthy units according to priority weight. At the same time, the cutoff amount and the reason for limiting are fed back to the next control cycle as optimization prior information.

[0010] The beneficial effects of this invention are as follows: By prioritizing power increments to turbines with low turbulence intensity and high wind energy quality, these turbines in a "favorable position" can operate in a more efficient range, reducing energy loss due to wake interference and thus improving the overall wind energy utilization rate. The constructed multi-objective optimization model aims to minimize the overall power tracking error, ensuring that the wind farm can respond accurately and quickly to grid dispatch commands, avoiding the overshoot or undershoot phenomena that may occur in traditional methods, and improving the grid-friendliness of the wind farm as a controllable power source. This method introduces "turbulence intensity" as the core weight for power allocation. By reducing the output of turbines in the high turbulence zone, the mechanical stress and vibration levels of these turbines under severe wind conditions are actively reduced, significantly mitigating fatigue accumulation in mechanical components (such as towers, blades, and drive trains). Combining SCADA vibration data with turbulence intensity calculations, "high-risk" turbines are identified in step S2. This dual criterion ensures that the control strategy is based not only on environmental wind conditions but also on the actual health status of the equipment, preventing the unit from shutting down or being damaged due to forced power generation when it is close to the safety threshold. Attached Figure Description

[0011] Figure 1 A flowchart of a dynamic active power allocation method for wind farms based on turbulence intensity. Detailed Implementation A method for dynamic active power allocation in wind farms based on turbulence intensity, such as... Figure 1 The process, as shown, includes the following steps: S1. Data Acquisition and Turbulence Intensity Calculation: Real-time acquisition of three-dimensional wind speed components and wind direction data at the hub height of each wind turbine is achieved using the lidar of the wind turbine nacelle or the wind measurement tower of the wind farm; turbulence intensity of each turbine is calculated within a 10-minute statistical period, a turbulence intensity distribution map of the entire field is established, and high-turbulence and low-turbulence areas are identified; S2. Unit operating status and fatigue load assessment: Combine SCADA system data to read the current generator speed, pitch angle, and tower vibration acceleration of each unit; assess the current mechanical fatigue state of the unit. If the turbulence intensity of a unit is high, but its vibration value is close to the safety threshold, it is marked as a "high-risk" state. S3. Dynamic Priority Factor Generation: Design a nonlinear mapping function to convert turbulence intensity into power allocation weight coefficients; through a formula, units with lower turbulence intensity are given higher weight coefficients, thereby establishing the logic of priority power generation; S4. Construction of Multi-Objective Optimization Allocation Model: Establish an optimization model with the dual objectives of "minimizing the power tracking error of the entire wind farm" and "minimizing the fatigue load increment of the entire wind farm"; set constraints such as the total output power of the wind farm must be equal to the dispatch command and the power of a single unit cannot exceed the rated power, and use quadratic programming or heuristic algorithms to solve for the optimal target power of each unit; S5. Priority-based initial power allocation: Under the constraints of the optimization model, the power increment is allocated to the units with higher weight coefficient values ​​first; the baseline allocated power for each unit is calculated.

[0012] S6. Power Correction and Limiting: Perform physical feasibility verification on the calculated baseline power allocation; if the baseline power exceeds the rated power, set it to equal the rated power and redistribute the remaining power demand to other units; if the baseline power is lower than the cut-in power, set it to 0 and consider the response rate limits of pitch and torque to smooth the slope of the power command change. S7. Command Issuance and Execution: The final target power command is issued to the main control system of each wind turbine through the wind farm monitoring system or a dedicated wind farm controller; after receiving the command, the turbine tracks the new active power setpoint by adjusting the generator torque and / or pitch angle. S8. Feedback and Rebalancing Mechanism: Repeat the above steps in the next control cycle; when a sudden change in the turbulence intensity of a unit is detected, an emergency redistribution is immediately triggered to rapidly reduce the power setpoint of the unit and prevent equipment damage.

[0013] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S1, the turbulence intensity of each unit within a 10-minute statistical period is calculated, a full-field turbulence intensity distribution map is established, and high-turbulence and low-turbulence regions are identified. The specific steps are as follows: S11. Data Acquisition and Time Window Division: Collect wind speed time series data at each unit from the wind farm SCADA system or the wind measurement equipment (such as nacelle anemometer, lidar) of each unit at a sampling frequency of not less than 1Hz. Divide the continuous wind speed data stream into windows according to a statistical period of 10 minutes, and mark each 10-minute segment with a timestamp as the basic unit for turbulence intensity calculation. S12. Calculation of turbulence intensity for a single unit: Calculate the wind speed time series data for each unit within each 10-minute window, respectively: Average wind speed V: The arithmetic mean of all wind speed samples within this window; Wind speed standard deviation σ: The standard deviation of wind speed samples within this window, reflecting the amplitude of wind speed fluctuations around the average value; The turbulence intensity I of the unit in the 10-minute period is calculated using the standard formula I = σ / V. This calculation is repeated for all units and all time windows to form a time series dataset of turbulence intensity for each unit. S13. Spatial Interpolation and Distribution Mapping of Turbulence Intensity Across the Entire Field: Obtain the precise geographic coordinates (latitude and longitude or planar projection coordinates) of all units in the field. Correlate the turbulence intensity values ​​of each unit during a representative time period (such as a typical 10-minute window or long-term average) calculated in S12 with their coordinates to form spatial scatter data. Import the digital elevation model (DEM) topographic data of the wind farm as the base map. Use spatial interpolation algorithms such as Kriging and inverse distance weighting to calculate the turbulence intensity at any grid point across the entire field based on the discrete turbulence intensity values ​​at each unit location. Generate continuously distributed raster data. Visualize the interpolation results in the form of contour cloud maps or heat maps and overlay them on the topographic map to form a spatial distribution map of turbulence intensity across the entire field. S14. Identification and Threshold Classification of High / Low Turbulence Zones: Setting Classification Standards: Referring to the IEC 61400-1 standard, based on the turbulence intensity at a wind speed of 15 m / s, the site area is divided into Class A (I_ref=0.16, high turbulence), Class B (I_ref=0.14, medium turbulence), and Class C (I_ref=0.12, low turbulence). The threshold can be customized according to the actual turbulence statistical distribution of the site. On the turbulence intensity distribution map of the entire site, the area with the interpolation result higher than the Class A threshold (or the customized high threshold) is designated as the high turbulence zone, and the area with the interpolation result lower than the Class C threshold (or the customized low threshold) is designated as the low turbulence zone, and they are distinguished by different colors or legends on the map; S15. Result Verification and Output: Compare the interpolation results at the location of the wind measurement tower with the measured and calculated turbulence intensity, calculate the error index (such as root mean square error RMSE), verify the accuracy of the spatial interpolation model, and compile and output the results such as the overall turbulence intensity distribution map, the boundary coordinates of high / low turbulence regions, and the turbulence intensity statistics table of each unit as the input basis for the subsequent active power allocation strategy. S16. Dynamic Update and Closed-Loop Feedback: Turbulence intensity is time-varying, so a dynamic update mechanism needs to be established: S11 to S15 are re-executed periodically (e.g., daily, weekly, or according to seasonal changes) to update the full-field turbulence intensity distribution map and the high / low turbulence zone division results with the latest operating data. The updated region division will be fed back to the active power allocation model in a closed loop, so that the allocation strategy can adapt to the changes in turbulence field characteristics under different time periods and different incoming wind directions.

[0014] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S2, data from the SCADA system is used to read the current generator speed, pitch angle, and tower vibration acceleration of each unit; the current mechanical fatigue state of the unit is assessed, as follows; S21. SCADA Data Reading and Synchronization: Read the following key parameters of each unit in real time from the wind farm's SCADA system, with a sampling frequency of no less than 1Hz, and align the timestamps of data from different sources: generator speed (rpm), pitch angle (°), tower front-to-back / lateral vibration acceleration (m / s²), output power (kW, auxiliary parameter), and wind speed (m / s, auxiliary parameter). S22. Data cleaning and operating condition segmentation: Remove sensor dead values ​​and abnormal jumps caused by communication interruptions. Use moving average and low-pass filtering to remove high-frequency noise. Based on generator speed and power threshold, the operating state is segmented into four types of operating condition segments: startup, normal operation, shutdown, and idling. Subsequent fatigue calculations are processed according to the operating conditions. S23. Tower vibration feature extraction: Perform fast Fourier transform on the tower vibration acceleration signal to extract the following features: vibration amplitude corresponding to the first and last modal frequencies of the tower (usually 0.2~0.4Hz), vibration amplitude corresponding to the first and last modal frequencies of the tower, and root mean square value (RMS) of vibration in each frequency band. These amplitudes are directly related to the magnitude of the dynamic bending moment borne by the tower. S24. Component Equivalent Load Estimation: Estimating the load timing of major components based on SCADA indirect parameters: Transmission chain torque: Calculate the electromagnetic torque using the generator speed and power, and then infer the low-speed shaft torque by combining the gearbox ratio; Equivalent bending moment of the tower: The timing of the overturning bending moment at the top of the tower is estimated by combining the thrust coefficient curve corresponding to the pitch angle and the nacelle vibration acceleration. S25 Damage Accumulation: Input the load time sequence from step 4 into the rainflow counting algorithm, count the number of cycles under each stress amplitude, and calculate the cumulative damage amount according to Miner's linear accumulation rule based on the material's SN curve: D = Σ(nᵢ / Nᵢ)+K scf Where nᵢ is the actual number of cycles, Nᵢ is the maximum allowable number of cycles under this stress amplitude, and K scf As a stress concentration correction factor, for geometrically discontinuous parts such as tower welds, flange connection bolt holes, and gear tooth roots, the actual local stress is higher than the nominal stress. Introducing the stress concentration factor to amplify the load time sequence before rainflow counting can more realistically reflect the damage state of key hot spots. S26. Fatigue status assessment and grading: Current cumulative damage value D, the closer D is to 1, the more severe the fatigue; damage rate: the damage increment dD / dt per unit time, a sudden increase in rate indicates deterioration of operating conditions; graded early warning: divided into four levels according to preset thresholds: normal (D<0.6), attention (0.6≤D<0.8), early warning (0.8≤D<0.95), and alarm (D≥0.95); S27. Results Output and Closed-Loop Application: Write the current fatigue status (D value, damage rate, warning level) of each unit into the database and display it visually on the monitoring interface. The evaluation results serve as the key input for the active power allocation strategy: prioritize the allocation of power commands to units with low fatigue damage, delay the damage accumulation of high fatigue units, and achieve collaborative optimization of the overall health status.

[0015] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S3, a nonlinear mapping function is designed to convert turbulence intensity into power allocation weighting coefficients. The specific steps are as follows: S31. Standardization of turbulence intensity: The specific turbulence intensity values ​​of each unit calculated in S1 are normalized. Specifically, the average turbulence level of all operating units in the field is calculated as a baseline, and then the turbulence value of a single unit is divided by this baseline to obtain a relative ratio. S32. Introduce a penalty mechanism for fatigue state: The health status of the unit assessed in step two is converted into a braking signal for weight calculation. An attenuation coefficient is designed. When the vibration value of the unit is within the safe range, the coefficient is close to one and has almost no impact on the weight. When the vibration acceleration is detected to exceed the preset warning threshold, the coefficient will decay non-linearly and rapidly approach zero. This means that when calculating the final weight, the score of the unit will be forced to be significantly reduced. S33. Generate final priority weights: Combine the above two factors and use the reciprocal of the smoothed turbulence ratio as the base weight: the smaller the turbulence, the larger the reciprocal, and the higher the base score. Multiply by the fatigue penalty coefficient, and then normalize the calculated result across the entire field to ensure that the sum of the weights of all units equals one. The final output weight factor represents the priority capability of each unit in the next stage of power allocation.

[0016] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S4, an optimization model is established with the dual objectives of minimizing the full-field power tracking error and minimizing the full-field fatigue load increment. The specific steps are as follows: S41. Determine the variables to be solved: The variables to be solved are: how much electricity each unit should increase or decrease in power generation during the next control cycle. This is an allocation problem of adjustment amount, not an allocation of absolute value. S42. Constructing a dual-core objective logic: The model simultaneously considers two seemingly contradictory metrics: The primary objective is to ensure accuracy, meaning that the total power output of all units after adjustments must be exactly equal to the command value issued by the power grid dispatch center, and the difference between the actual total calculated by the model and the command value must be minimized. The second objective is to reduce losses and avoid subjecting high-turbulence, high-vibration units to large power fluctuations. The objective function will penalize the power adjustments assigned to high-risk units, which are required to perform drastic pitch and torque adjustments, resulting in very poor performance and high penalties. Conversely, assigning power fluctuations to stable units will result in very low penalties. S43 sets a hard boundary condition: the sum of the total output must meet the target; this is a mandatory condition. Output range constraints: The power of each unit cannot be lower than the minimum grid-connected power (if it is too low, the unit will be shut down directly), nor can it be higher than the maximum power that the unit can actually capture under the current turbulent wind conditions (this upper limit will be dynamically adjusted when the turbulence is strong). Adjustment rate constraint: Considering the response inertia of the pitch mechanism and generator, the power change per second cannot exceed the specified upper limit of the grade rate to prevent mechanical shock; S44. Solution and Engineering Processing: The objective function includes an absolute value penalty for the power variation range, transforming it into a standard solvable form. Under the premise of satisfying the equality constraints, the solver will automatically find the optimal path that minimizes the total adjustment range of each unit and avoids high-risk units. The final output is the most reasonable power adjustment command for each unit at the current moment.

[0017] Furthermore, a method for dynamic active power allocation in wind farms based on turbulence intensity is proposed. In step S6, the physical feasibility of the calculated baseline power allocation is verified, as follows; The theoretically allocated baseline power is sequentially verified by multiple constraints, including the upper limit of rated capacity, the lower limit of minimum operating power, the pitch / torque ramp rate limit, the turbulence intensity derating boundary, and the vibration protection threshold. The intersection of all limiting conditions is taken as the final feasible target power command. When a power gap occurs due to limiting, it is re-allocated among the remaining healthy units according to priority weight. At the same time, the cutoff amount and the reason for limiting are fed back to the next control cycle as optimization prior information.

[0018] Example 2 This embodiment takes an offshore wind farm as an example. The wind farm has an installed capacity of 300MW and consists of 50 permanent magnet direct-drive wind turbine units, each with a capacity of 6MW, arranged in a regular array. The wind farm is equipped with an energy management platform that communicates with the main control system of each unit and the integrated automation system of the substation via a fiber optic ring network, with a control cycle set to 1 minute.

[0019] At a certain moment, the power grid dispatch center issued an active power command of 210MW to the wind farm. The current actual output of the entire farm is 200MW, requiring an additional 10MW to meet the dispatch requirements. The wind speed distribution within the wind farm is uneven; the wind speed of the front-row turbines is approximately 10m / s, while the wind speed of the rear-row turbines, affected by wake turbulence, is approximately 7-8m / s, and there are significant turbulent fluctuations in some areas.

[0020] Specific implementation process S1. Data Acquisition and Turbulence Intensity Calculation: The energy management platform uses lidar on the top of each unit's nacelle to collect three-dimensional wind speed components at hub height at a sampling frequency of 4Hz. Using the current moment as the endpoint, it extracts wind speed time-series data from the past 10 minutes, calculates the average wind speed and standard deviation for each unit, and then calculates the turbulence intensity value for each unit using the formula "turbulence intensity equals wind speed standard deviation divided by average wind speed." After summarizing the calculation results, a spatial distribution map of the overall turbulence intensity was generated using the inverse distance weighted interpolation method. Analysis showed that the area containing approximately eight units in the middle of the third row, due to the superposition effect of the wake from the preceding row, had a turbulence intensity exceeding 0.16 and was designated as a high-turbulence zone; while the turbulence intensity of the units on the windward side and edge of the first row was below 0.10 and was designated as a low-turbulence zone. S2. Unit Operating Status and Fatigue Load Assessment: The energy management platform synchronously reads real-time operating data of each unit from the SCADA system. Comparison revealed: The effective value of the front and rear vibration acceleration of tower WT-15 (located in the high turbulence zone) has reached 0.85 m / s², which is close to the warning threshold of 0.95 m / s² for this model. The vibration value of WT-08 (located in the low turbulence region) is 0.32 m / s², indicating good condition. According to preset rules, the system marked WT-15 as a high-risk state and the other units as normal states; S3. Dynamic Priority Factor Generation: The turbulence intensity value of each unit is input into a nonlinear mapping function. This function is designed such that the smaller the turbulence intensity of the unit, the larger the output base weight. On this basis, a fatigue penalty coefficient is introduced. For WT-15, which is marked as high-risk, its base weight is forcibly multiplied by a decay coefficient of 0.2, putting it at a disadvantage in power allocation. After calculation and full-field normalization, the priority factor of the low-turbulence zone and normal vibration unit (such as WT-08) was about 0.035, while the priority factor of the high-risk unit WT-15 was only 0.006. S4. Construction and solution of multi-objective optimization allocation model: When constructing the optimization model, the first objective is to minimize the deviation between the total output of the entire field and the dispatch command of 210MW, and the second objective is to minimize the total cost of the power adjustment of each unit and its turbulence intensity weighted together. The constraints include: the total output of the entire site is equal to 210MW, the output of a single unit does not exceed 6MW, and the ramp rate limit is 0.6MW per minute for power variation. The model uses a quadratic programming solver to complete the calculation within 50 milliseconds and outputs the initial target power of each unit. S5. Priority-based initial power allocation: During the solution process, since the optimization model incorporates the priority factor into the weight matrix of the objective function, the system automatically allocates the required 10MW power increment to units with higher priority factors. Units such as WT-08 in the low-turbulence zone are allocated power increments ranging from 0.3 to 0.5MW, while the power command of the high-risk unit WT-15 is maintained at its original level and does not undertake any incremental tasks. S6: Power Correction and Limiting: Physical feasibility verification is performed on each unit based on the initial allocation results. The base power calculation result for WT-22 is 6.15MW, which exceeds the rated value. The system truncates it to 6MW, and the excess 0.15MW gap is redistributed to WT-08. The base power of unit WT-37 is 0.8MW, which is lower than its minimum stable operating power of 1.2MW. The system corrects its command to 0, and the unit is taken out of operation. Its original output is taken over by a nearby healthy unit. The slope of the power change of all units is checked. Among them, the command change of WT-08 exceeds the upper limit of the ramp rate. The system performs smoothing and limiting processing on its command. S7. Command Issuance and Execution: The final power command after verification and correction is issued to the main control system of each unit through the fiber optic ring network. After receiving the power increase command, the main control system of units such as WT-08 in the low turbulence region increases the output by adjusting the generator torque; WT-37 executes the normal shutdown procedure after receiving the zero power command; WT-15 maintains the current power unchanged. Within 10 seconds of the command being issued, the total power output of the field steadily increased from 200MW to 210MW, with the tracking error controlled within 1%. S8. Feedback and Rebalancing Mechanism: Upon entering the next 1-minute control cycle, the system repeats steps S1 to S7. During this cycle, the wind measurement system detects that the turbulence intensity at location WT-07 suddenly increases from 0.09 to 0.18 within 30 seconds, triggering the abrupt change judgment condition; The system immediately initiated an emergency redistribution procedure: the priority factor of the unit was forcibly lowered, and after the optimization model was re-solved, the power command of WT-07 was rapidly reduced by 1.2MW. The power reduction was shared by the surrounding low-turbulence units. This effectively avoided the overload impact that WT-07 might be caused by sudden changes in turbulence. The entire process was completed smoothly within two control cycles.

[0021] Results: Compared with the traditional method of allocating capacity proportionally, the method described in this embodiment achieves the following results: The power fluctuation amplitude of high-risk units was reduced by about 45%, effectively delaying the accumulation of fatigue damage to key components; The tracking accuracy of the dispatch instructions was not affected by the overall power output, and remained above 98%. During the turbulent sudden change event, the system responded promptly, and no unit shutdown event triggered by excessive vibration occurred.

Claims

1. A method for dynamic active power allocation in wind farms based on turbulence intensity, characterized in that, Includes the following steps: S1. Data Acquisition and Turbulence Intensity Calculation: Real-time acquisition of three-dimensional wind speed components and wind direction data at the hub height of each wind turbine is achieved using the lidar of the wind turbine nacelle and the wind measurement tower of the wind farm; the turbulence intensity of each turbine is calculated within a 10-minute statistical period, a full-field turbulence intensity distribution map is established, and high-turbulence and low-turbulence areas are identified; S2. Unit operating status and fatigue load assessment: Combine SCADA system data to read the current generator speed, pitch angle and tower vibration acceleration of each unit; assess the current mechanical fatigue state of the unit. When a unit has high turbulence intensity but its vibration value is close to the safety threshold, it is marked as a high-risk state. S3. Dynamic Priority Factor Generation: Design a nonlinear mapping function to convert turbulence intensity into power allocation weighting coefficients; The formula assigns a higher weighting coefficient to units with lower turbulence intensity, thus establishing the logic of priority power generation. S4. Construction of Multi-Objective Optimization Allocation Model: Establish an optimization model with the dual objectives of minimizing the power tracking error of the entire wind farm and minimizing the fatigue load increment of the entire wind farm; set constraints such as the total output power of the wind farm must be equal to the dispatch command and the power of a single unit cannot exceed the rated power, and use quadratic programming and heuristic algorithms to solve for the optimal target power of each unit; S5. Priority-based initial power allocation: Under the constraints of the optimization model, power increments are preferentially allocated to units with higher weight coefficient values; Calculate the baseline power allocation for each unit. S6. Power Correction and Limiting: Perform physical feasibility verification on the calculated baseline power allocation; if the baseline power exceeds the rated power, set it to equal the rated power and redistribute the remaining power demand to other units; if the baseline power is lower than the cut-in power, set it to 0 and consider the response rate limits of pitch and torque to smooth the slope of the power command change. S7. Command Issuance and Execution: The final target power command is issued to the main control system of each wind turbine through the wind farm monitoring system and dedicated wind farm controller; after receiving the command, the turbine tracks the new active power setpoint by adjusting the generator torque and / or pitch angle. S8. Feedback and Rebalancing Mechanism: Repeat the above steps in the next control cycle; when a sudden change in the turbulence intensity of a unit is detected, an emergency redistribution is immediately triggered to rapidly reduce the power setpoint of the unit and prevent equipment damage.

2. The method for dynamic active power allocation in a wind farm based on turbulence intensity as described in claim 1, characterized in that, In step S1, the turbulence intensity of each unit within a 10-minute statistical period is calculated, a full-field turbulence intensity distribution map is established, and high-turbulence and low-turbulence regions are identified. The specific steps are as follows: S11. Data Acquisition and Time Window Division: Collect wind speed time series data at each unit from the wind farm SCADA system or the wind measurement equipment of each unit at a sampling frequency of not less than 1Hz. Divide the continuous wind speed data stream into windows according to a statistical period of 10 minutes, and mark each 10-minute segment with a timestamp as the basic unit for turbulence intensity calculation. S12. Calculation of turbulence intensity for a single unit: Calculate the wind speed time series data for each unit within each 10-minute window, respectively: Average wind speed V: The arithmetic mean of all wind speed samples within this window; Wind speed standard deviation σ: The standard deviation of wind speed samples within this window, reflecting the amplitude of wind speed fluctuations around the average value; The turbulence intensity I of the unit in the 10-minute period is calculated using the standard formula I = σ / V. This calculation is repeated for all units and all time windows to form a time series dataset of turbulence intensity for each unit. S13. Spatial Interpolation and Distribution Mapping of Turbulence Intensity Across the Entire Field: Obtain the precise geographic coordinates (latitude and longitude) of all units in the field. Correlate the turbulence intensity values ​​of each unit calculated in S12 for a representative period (long-term average) with their coordinates to form spatial scatter data. Import the topographic data of the wind farm's digital elevation model as the base map. Using the Kriging spatial interpolation algorithm, based on the discrete turbulence intensity values ​​at each unit location, calculate the turbulence intensity at any grid point across the entire field, generating continuously distributed raster data. Visualize the interpolation results in the form of contour cloud maps and overlay them onto the topographic map to form a spatial distribution map of turbulence intensity across the entire field. S14. Identification and Threshold Classification of High / Low Turbulence Zones: Setting Classification Standards: Referring to the IEC 61400-1 standard, the site area is divided into Class A (I_ref=0.16, high turbulence), Class B (I_ref=0.14, medium turbulence), and Class C (I_ref=0.12, low turbulence) based on the actual turbulence statistical distribution of the site. On the overall turbulence intensity distribution map, areas with interpolation results higher than the Class A threshold are designated as high turbulence zones, and areas with interpolation results lower than the Class C threshold are designated as low turbulence zones. These are distinguished by different colors and legends on the map. S15. Result Verification and Output: Compare the interpolation results at the location of the wind measurement tower with the measured and calculated turbulence intensity, calculate the error index: root mean square error RMSE, verify the accuracy of the spatial interpolation model, and compile and output the results of the overall turbulence intensity distribution map, the boundary coordinates of high / low turbulence regions, and the turbulence intensity statistics table of each unit as the input basis for the subsequent active power allocation strategy. S16. Dynamic Update and Closed-Loop Feedback: Turbulence intensity is time-varying, so a dynamic update mechanism needs to be established: Periodically re-execute S11 to S15, update the full-field turbulence intensity distribution map and the division results of high / low turbulence zones with the latest operating data, and feed the updated region division into closed loops to the active power allocation model, so that the allocation strategy can adapt to the changes in turbulence field characteristics under different time periods and different incoming wind directions.

3. The method for dynamic active power allocation in a wind farm based on turbulence intensity as described in claim 1, characterized in that, In step S2, data from the SCADA system is used to read the current generator speed, pitch angle, and tower vibration acceleration of each unit; the current mechanical fatigue state of the unit is assessed, as follows; S21. SCADA Data Reading and Synchronization: Read the following parameters of each unit in real time from the wind farm's SCADA system, with a sampling frequency of not less than 1Hz, and align the timestamps of data from different sources: generator speed, pitch angle, tower front-to-back / left-to-right vibration acceleration, output power, and wind speed; S22. Data cleaning and operating condition segmentation: Remove sensor dead values ​​and abnormal jumps caused by communication interruptions. Use moving average and low-pass filtering to remove high-frequency noise. Based on generator speed and power threshold, the operating state is segmented into four types of operating condition segments: startup, normal operation, shutdown, and idling. Subsequent fatigue calculations are processed according to the operating conditions. S23. Tower vibration feature extraction: Perform fast Fourier transform on the tower vibration acceleration signal to extract the following features: vibration amplitude corresponding to the first and last first-order modal frequencies of the tower, vibration amplitude corresponding to the first and last first-order modal frequencies of the tower, and root mean square value of vibration in each frequency band. S24. Component Equivalent Load Estimation: Estimating the load timing of major components based on SCADA indirect parameters: Transmission chain torque: Calculate the electromagnetic torque using the generator speed and power, and then infer the low-speed shaft torque by combining the gearbox ratio; Equivalent bending moment of the tower: The timing of the overturning bending moment at the top of the tower is estimated by combining the thrust coefficient curve corresponding to the pitch angle and the nacelle vibration acceleration. S25 Damage Accumulation: Input the load time sequence of S24 into the rainflow counting algorithm, count the number of cycles under each stress amplitude, and combine it with the material's SN curve to calculate the cumulative damage according to Miner's linear accumulation rule: D = Σ(nᵢ / Nᵢ) + K scf Where nᵢ is the actual number of cycles, Nᵢ is the maximum allowable number of cycles under this stress amplitude, and K scf This is the stress concentration correction factor; S26. Fatigue status assessment and grading: Current cumulative damage value D, the closer D is to 1, the more severe the fatigue; damage rate: the damage increment dD / dt per unit time, a sudden increase in rate indicates deterioration of operating conditions; graded early warning: divided into four levels according to preset thresholds: normal D<0.6, attention 0.6≤D<0.8, early warning 0.8≤D<0.95, and alarm D≥0.95; S27. Results Output and Closed-Loop Application: The current fatigue status of each unit, including D-value, damage rate, and warning level, is written into the database and displayed visually on the monitoring interface. The evaluation results serve as a key input for the active power allocation strategy: power commands are preferentially allocated to units with low fatigue damage to delay the accumulation of damage in units with high fatigue, thereby achieving collaborative optimization of the overall health status.

4. The method for dynamic active power allocation in a wind farm based on turbulence intensity as described in claim 1, characterized in that, In step S3, a nonlinear mapping function is designed to convert turbulence intensity into power allocation weighting coefficients. The specific steps are as follows: S31. Standardization of turbulence intensity: The specific turbulence intensity values ​​of each unit calculated in S1 are normalized. Specifically, the average turbulence level of all operating units in the field is calculated as a baseline, and then the turbulence value of a single unit is divided by this baseline to obtain a relative ratio. S32. Introduce a penalty mechanism for fatigue state: The health status of the unit assessed in step two is converted into a braking signal for weight calculation. An attenuation coefficient is designed. When the vibration value of the unit is within the safe range, the coefficient is close to one. When the vibration acceleration is detected to exceed the preset warning threshold, the coefficient will decay non-linearly and rapidly approach zero. This means that when calculating the final weight, the score of the unit will be forced to be significantly reduced. S33. Generate final priority weights: Combine the above two factors and use the reciprocal of the smoothed turbulence ratio as the base weight: the smaller the turbulence, the larger the reciprocal, and the higher the base score. Multiply by the fatigue penalty coefficient, and then normalize the calculated result across the entire field to ensure that the sum of the weights of all units equals 1. The final output weight factor represents the priority capability of each unit in the next stage of power allocation.

5. The method for dynamic active power allocation in a wind farm based on turbulence intensity as described in claim 1, characterized in that, In step S4, an optimization model is established with the dual objectives of minimizing the full-field power tracking error and minimizing the full-field fatigue load increment. The specific steps are as follows: S41. Determine the variables to be solved: The variables to be solved are how much more electricity each unit should generate and how much less electricity it should generate in the next control cycle. It is an allocation problem of adjustment quantity, not an allocation of absolute value. S42. Constructing a dual-core objective logic: The model simultaneously considers two seemingly contradictory metrics: The primary objective is to ensure accuracy, meaning that the total power output of all units after adjustments must be exactly equal to the command value issued by the power grid dispatch center, and the difference between the actual total calculated by the model and the command value must be minimized. The second objective is to reduce losses and avoid subjecting high-turbulence, high-vibration units to large power fluctuations. The objective function will penalize the power adjustments assigned to high-risk units, which are required to perform drastic pitch and torque adjustments, resulting in very poor performance and high penalties. Conversely, assigning power fluctuations to stable units will result in very low penalties. S43 sets a hard boundary condition: the sum of the total output must meet the target; this is a mandatory condition. Output range constraints: The power of each unit cannot be lower than the minimum grid-connected power, nor can it be higher than the maximum power that the unit can actually capture under the current turbulent wind conditions; Adjustment rate constraint: Considering the response inertia of the pitch mechanism and generator, the power change per second cannot exceed the specified upper limit of the grade rate to prevent mechanical shock; S44. Solution and Engineering Processing: The objective function includes an absolute value penalty for the power variation range, transforming it into a standard solvable form. Under the premise of satisfying the equality constraints, the solver will automatically find the optimal path that minimizes the total adjustment range of each unit and avoids high-risk units. The final output is the most reasonable power adjustment command for each unit at the current moment.

6. The method for dynamic active power allocation in a wind farm based on turbulence intensity as described in claim 1, characterized in that, In step S6, the physical feasibility of the calculated baseline power allocation is verified, as follows; The theoretically allocated baseline power is successively verified by multiple constraints, including the upper limit of rated capacity, the lower limit of minimum operating power, the pitch / torque ramp rate limit, the turbulence intensity derating boundary, and the vibration protection threshold. The intersection of all limiting conditions is taken as the final feasible target power command. When a power shortfall occurs due to the limiting, it is redistributed among the remaining healthy units according to priority weights. At the same time, the cutoff amount and the reason for the limiting are fed back to the next control cycle as optimization prior information.

Images