A wind farm power optimization distribution method and system based on fatigue load balancing

Abstract

This invention provides a wind farm power optimization allocation method and system based on fatigue load balancing, belonging to the field of wind farm operation and maintenance technology. By collecting wind farm operating parameters, fatigue loads, and wind condition data in real time to construct a dynamic dataset, sensitivity correlation features are extracted and combined with fatigue load balancing constraints to reverse-driven dynamic threshold ranges. This adapts to the complex and ever-changing wind conditions of wind farms, effectively balancing the fatigue loads of each unit, reducing the difference in fatigue damage rates between units, effectively extending the entire lifespan of units, and improving the overall operational stability and reliability of the wind farm. With the goal of minimizing the fatigue load balancing across the entire farm, the total power demand is decomposed based on real-time wind conditions, and an optimization function with sensitivity constraints is constructed. The optimal power command is adaptively solved and issued, accurately decomposing the total power demand, reducing power tracking deviation, optimizing wind resource utilization efficiency, and reducing equipment losses caused by sudden changes in operating conditions.

Description

Technical Field

[0001] This invention belongs to the field of wind farm operation and maintenance technology, specifically relating to a wind farm power optimization allocation method and system based on fatigue load balancing. Background Technology

[0002] Wind farm power optimization allocation is a technical means to rationally allocate the output power of each wind turbine in the farm by combining the operating characteristics of wind turbines, real-time wind speed forecast data and grid dispatch instructions. It can avoid wake interference between wind turbines, maximize the overall power generation of the wind farm, and at the same time meet the grid's constraints on power fluctuations, thereby improving the economic efficiency of wind farm operation and grid connection stability.

[0003] The invention patent application with application number 202211723945.7 discloses a wind farm frequency regulation power allocation method that takes into account the historical fatigue load of the units. This application aims to solve the problem that "the prior art considers the allocation of active power of the units from the perspective of fatigue load, which reduces the fatigue of the units in the station while meeting the active power demand. However, the historical fatigue state of the wind turbine units is not considered in the optimization process, and the coordinated distribution of fatigue of the units in the station is not considered, which increases the number of maintenance and operation times and costs of the wind farm".

[0004] However, in the process of optimizing the allocation of wind farm power based on fatigue load balancing, the sensitivity of the allocation is also a key point worth paying attention to. How to sensitively trigger the optimization of wind farm power allocation in real time to cope with the real-time operation scenarios and states of wind farms is also a goal that urgently needs to be improved.

[0005] To address this, we propose a wind farm power optimization allocation method based on fatigue load balancing. Summary of the Invention

[0006] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a wind farm power optimization allocation method and system based on fatigue load balancing.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a wind farm power optimization allocation method based on fatigue load balancing, comprising the following steps: Real-time data collection of operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine in the wind farm is used to construct a real-time operating status dataset of the wind farm based on the collected parameters. The real-time running status dataset is preprocessed to obtain a preprocessed dataset. Sensitivity correlation features characterizing sudden changes in running status are extracted from the preprocessed dataset. Based on the sensitivity correlation feature and fatigue load balance constraint, the allowable range of fatigue damage throughout the life cycle of the generator set is calculated. The threshold range of sensitivity correlation feature under different wind conditions is derived by reverse derivation of the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine set does not exceed a preset threshold. The sensitivity-related features are compared with the threshold ranges of sensitivity-related features under different wind conditions. If any sensitivity-related feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered; otherwise, the refresh step is executed. After triggering the power optimization allocation process, with the goal of minimizing the fatigue load balance of each wind turbine, the total power demand of the wind farm is decomposed based on real-time wind condition parameters, and a power allocation objective function with sensitivity correlation feature constraints is constructed. The optimal power command for each unit is solved through an adaptive weighting strategy. The optimal power command is issued to each wind turbine to perform adjustment. After all units have completed adjustment, the refresh procedure is executed.

[0008] In the step of collecting real-time operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine generator set in the wind farm, and constructing a real-time operating status dataset of the wind farm based on the collected parameters, the operating parameters include the output power of the wind turbine generator set, generator speed, yaw angle, pitch angle, gearbox oil temperature, and nacelle vibration acceleration; the fatigue load characteristic parameters include the blade root bending load, tower lateral vibration load, nacelle longitudinal impact load, gearbox input shaft torque, and hub axial load; and the wind condition parameters include real-time wind speed, wind direction, turbulence intensity, wind shear index, and atmospheric stability level.

[0009] The method for preprocessing the real-time running status dataset to obtain a preprocessed dataset, and then extracting sensitivity correlation features representing sudden changes in running status from the preprocessed dataset, is as follows: Outliers in the real-time operating status dataset are eliminated by combining the 3σ criterion with the load mutation judgment condition. The load mutation judgment condition is that the difference between the fatigue load characteristic parameters at adjacent sampling times exceeds the instantaneous load fluctuation range allowed by the unit design. Smoothing is performed on the runtime state dataset after outlier removal to obtain a preprocessed dataset; Sensitivity correlation features characterizing sudden changes in runtime state are extracted from the preprocessed dataset, specifically: By integrating the output power, fatigue load characteristic parameters, and wind speed parameters of wind turbine generators, a multi-dimensional sensitivity correlation feature is constructed:

[0010] In the formula: Let be the sensitivity correlation characteristic value of the i-th wind turbine generator at time t; , , These are the feature fusion weight coefficients; Let be the rate of change of the output power of the i-th wind turbine generator at time t; This refers to the rated power of the wind turbine generator set. Let be the fatigue load gradient of the i-th wind turbine generator at time t; The rated fatigue load of the i-th wind turbine generator unit; Let be the standard deviation of wind speed at time t; Let be the average wind speed at time t.

[0011] Based on the aforementioned sensitivity correlation features and fatigue load balancing constraints, the allowable range of fatigue damage throughout the generator set's life cycle is calculated. The following method is used to deduce the sensitivity correlation feature threshold ranges under different wind conditions using these allowable fatigue damage ranges. Real-time wind conditions are classified into three categories based on turbulence intensity, wind speed fluctuation amplitude, and wind shear index: steady wind scenario, weak turbulence wind scenario, and strong turbulence wind scenario. Establish a quantitative correlation model for the relationship between fatigue damage rate and sensitivity of wind turbine generators under various wind conditions; Based on fatigue load balance constraints and quantitative correlation models, the allowable range of fatigue damage rate of wind turbine generators under each scenario is derived by using scenario proportion coefficients. Based on the allowable range of fatigue damage rate of the wind turbine generator, the threshold range of sensitivity correlation characteristics under different wind conditions is derived in reverse.

[0012] In the step of establishing a quantitative correlation model for the correlation characteristics between fatigue damage rate and sensitivity of wind turbine generators under various wind conditions, the quantitative correlation model is as follows: ; In the formula: Let be the fatigue damage rate of the i-th wind turbine generator at time t and wind condition scenario θ. The damage coefficient is related to the wind condition scenario; Let be the sensitivity correlation characteristic value of the i-th wind turbine generator at time t; This represents the base damage value for wind-related scenarios.

[0013] The formula for deriving the allowable range of fatigue damage rate of wind turbine generators under various scenarios, based on fatigue load balance constraints and a quantitative correlation model, using scenario proportion coefficients, is as follows:

[0014] in, This represents the average permissible fatigue damage rate for all units under scenario θ. This indicates the allowable fatigue damage value throughout the entire life cycle of the unit. Indicates the total lifespan of the unit design. This represents the damage allocation coefficient of unit i in scenario θ. This represents the maximum permissible difference in fatigue damage rates between units. The formula for deriving the threshold range of sensitivity correlation characteristics under different wind conditions based on the allowable range of fatigue damage rate of the wind turbine generator is expressed as follows:

[0015] in, , These are the upper and lower thresholds for the sensitivity-related features, respectively. Indicates the number of historical samples. , These represent the increments in damage rate and sensitivity characteristics at adjacent time points in historical data, respectively. , These represent the correlation characteristics between the average fatigue damage rate and the average sensitivity of unit i under scenario θ within the historical period.

[0016] After triggering the power optimization allocation process, with the goal of minimizing the fatigue load balance of each wind turbine, the total power demand of the wind farm is decomposed based on real-time wind condition parameters. A power allocation objective function with sensitivity correlation feature constraints is constructed, and the formula for solving the optimal power command of each unit through an adaptive weighting strategy is expressed as follows:

[0017]

[0018] In the formula: The objective function value; This represents the total number of wind turbine generators in the wind farm. Let be the fatigue damage rate of the i-th unit at time t; The average fatigue damage rate of all units at time t; The adaptive weights for the i-th unit; This is the penalty coefficient for sensitivity characteristics; Let be the sensitivity correlation characteristic value of the i-th unit at time t; Let be the optimal output power of the i-th unit at time t; This represents the initial power allocation value for the i-th unit based on real-time wind resource forecasting. This is the total power deviation penalty coefficient; Total power requirement; , These are the minimum and maximum allowable output power of the generator set; The maximum permissible rate of change of the unit's output power; This represents the upper threshold of the sensitivity correlation feature under wind condition scenario θ.

[0019] When decomposing the total power demand of a wind farm based on real-time wind condition parameters, the available power potential of each unit is calculated based on the wind profile prediction data of the location of each wind turbine in the wind farm. , in air density, The swept area of ​​the unit. The wind energy utilization coefficient, For the tip speed ratio, For variable pitch angle, Let be the real-time wind speed at the location of the i-th unit; The total power demand is initially decomposed based on the proportion of available power potential of each unit, resulting in an initial power allocation value: .

[0020] After issuing the optimal power command to each wind turbine for adjustment, the process also includes real-time monitoring of the actual output power of each turbine. and actual fatigue load Calculate the power tracking error and fatigue load equalization error: Calculate power tracking error: ; Calculate fatigue load equilibrium deviation ;in The fatigue damage rate is calculated based on the actual fatigue load. If the power tracking deviation of any unit exceeds the preset power deviation range, or the fatigue load balancing deviation exceeds the preset load balancing deviation range, the optimal output power command of that unit will be dynamically adjusted based on the deviation value. The adjustment amount is positively correlated with the deviation value, and the command will be reissued and executed until the deviation values ​​of all units are within the corresponding preset range.

[0021] Secondly, the present invention provides a wind farm power optimization allocation system based on fatigue load balancing, comprising: The data acquisition module is used to collect the operating parameters, fatigue load characteristic parameters and real-time wind condition parameters of each wind turbine in the wind farm in real time, and to build a real-time operating status dataset of the wind farm based on the collected parameters. The data preprocessing and feature extraction module is used to preprocess the real-time running status dataset to obtain the preprocessed dataset, and extract sensitivity correlation features that characterize sudden changes in running status from the preprocessed dataset. The calculation module is used to calculate the allowable range of fatigue damage throughout the entire life cycle of the generator set based on the sensitivity correlation feature and the fatigue load balance constraint. The threshold range of sensitivity correlation feature under different wind conditions is derived in reverse from the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine does not exceed a preset threshold. The judgment module is used to compare the sensitivity-related features with the threshold range of sensitivity-related features under different wind conditions. If any sensitivity-related feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered; if it does not exceed the threshold range, the refresh step is executed. The power allocation module is used to trigger the power optimization allocation process, aiming to minimize the fatigue load balance of each wind turbine generator set. Based on real-time wind condition parameters, it decomposes the total power demand of the wind farm, constructs a power allocation objective function with sensitivity correlation feature constraints, and solves the optimal power command for each unit through an adaptive weight strategy. The execution module is used to issue optimal power commands to each wind turbine generator set for adjustment. After all units have completed adjustment, the refresh step is executed.

[0022] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a wind farm power optimization allocation method and system based on fatigue load balancing. It constructs a dynamic dataset by real-time collection of wind farm operating parameters, fatigue loads, and wind condition data. Sensitivity correlation features are extracted and combined with fatigue load balancing constraints to reverse-driven dynamic threshold ranges. This adapts to the complex and variable wind conditions of wind farms, effectively balancing the fatigue loads of each turbine, reducing differences in fatigue damage rates between turbines, effectively extending the turbine's lifespan, and improving the overall operational stability and reliability of the wind farm. Based on this, with the goal of minimizing overall fatigue load balancing, it decomposes the total power demand based on real-time wind conditions and constructs an optimization function with sensitivity constraints. It adaptively solves and issues the optimal power command, accurately decomposing the total power demand, reducing power tracking deviation, optimizing wind resource utilization efficiency, and reducing equipment losses caused by sudden changes in operating conditions. By dynamically adjusting the power command, it flexibly adapts to different turbulent scenarios, reducing operation and maintenance costs, while also considering the rationality and safety of turbine output power, and promptly avoiding overload and excessive power fluctuations. Attached Figure Description

[0023] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0024] To further understand the content of this invention, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely illustrative and not limiting of the invention.

[0025] Example 1 like Figure 1 As shown, a wind farm power optimization allocation method based on fatigue load balancing includes the following steps: S1: Real-time collection of operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine in the wind farm, and construction of a real-time operating status dataset of the wind farm based on the collected parameters. S2: Preprocess the real-time running status dataset to obtain a preprocessed dataset, and extract sensitivity correlation features that characterize sudden changes in running status from the preprocessed dataset; S3: Based on the sensitivity correlation feature and fatigue load balance constraint, calculate the allowable range of fatigue damage throughout the life cycle of the generator set. Use the allowable range of fatigue damage to deduce the threshold range of sensitivity correlation feature under different wind conditions. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine set does not exceed a preset threshold. S4: Compare the sensitivity-related features with the threshold ranges of sensitivity-related features under different wind conditions. If any sensitivity-related feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered; otherwise, the refresh step is executed. S5: After triggering the power optimization allocation process, with the goal of minimizing the fatigue load balance of each wind turbine, the total power demand of the wind farm is decomposed based on real-time wind condition parameters, and a power allocation objective function with sensitivity correlation feature constraints is constructed. The optimal power command for each unit is solved through an adaptive weighting strategy. S6: Issue the optimal power command to each unit to perform adjustment. After all units have completed adjustment, refresh the execution steps.

[0026] Specifically, in S1, the operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine generator in the wind farm are collected in real time, and a real-time operating status dataset of the wind farm is constructed based on the collected parameters.

[0027] Operating parameters include wind turbine output power, generator speed, yaw angle, pitch angle, gearbox oil temperature, and nacelle vibration acceleration; fatigue load characteristic parameters include blade root bending load, tower lateral vibration load, nacelle longitudinal impact load, gearbox input shaft torque, and hub axial load; wind condition parameters include real-time wind speed, wind direction, turbulence intensity, wind shear index, and atmospheric stability level.

[0028] Among them, the wind shear index is based on real-time wind speed synchronous measurement data at different heights within the wind farm, and is confirmed by fitting and calculating the logarithmic wind profile model. The atmospheric stability level is combined with the real-time near-surface temperature gradient, on-site wind speed and solar radiation intensity monitoring data, and is comprehensively confirmed according to the preset stability classification judgment criteria.

[0029] Specifically, in S2, the real-time running status dataset is preprocessed, including outlier removal and smoothing, to obtain a preprocessed dataset. Sensitivity correlation features that characterize sudden changes in running status are then extracted from the preprocessed dataset.

[0030] S21: The outlier removal and smoothing stage of the real-time operation status dataset. Outliers are removed by combining the 3σ criterion with the load mutation judgment condition. The load mutation judgment condition is that the difference of fatigue load characteristic parameters between adjacent sampling times exceeds the instantaneous load fluctuation range allowed by the unit design. S22: Perform smoothing on the runtime state dataset after removing outliers to obtain a preprocessed dataset; the formula for smoothing is as follows:

[0031] In the formula: For the first Taiwanese crew at all times Fatigue load characteristic parameter values ​​after smoothing; The length of the sliding window; Let be the dynamic weighting coefficient of the k-th sampling point within the sliding window at time t; For the first The original fatigue load characteristic parameter value of the unit at the sampling time corresponding to the kth sampling point within the sliding window; For fixed data sampling period.

[0032] The above formula is based on the original fatigue load characteristic parameters at multiple sampling times within the sliding window. By assigning corresponding weights to different sampling points through dynamic weighting coefficients, it fully integrates the effective information of historical sampling data, offsets the residual interference after instantaneous fluctuations and outlier removal, and makes the processed fatigue load characteristic parameters more consistent with the actual operating state of the unit, effectively improving the stability and reliability of the data, and providing support for subsequent feature extraction and model construction.

[0033] in, ; In the formula: This is the load change rate weighting coefficient. ∈[0.1,5], for units with high structural fatigue tolerance, the value is biased towards the upper limit of the interval, and for units with low structural fatigue tolerance, the value is biased towards the lower limit of the interval; This represents the maximum permissible load change value for the unit. This represents the load difference between the current sampling point and historical sampling points within the sliding window.

[0034] S23: Extract sensitivity correlation features representing sudden changes in operating state from the preprocessed dataset, and integrate the output power, fatigue load characteristic parameters, and wind condition parameters of the wind turbine generator to construct multi-dimensional sensitivity correlation features: ; In the formula: Let be the sensitivity correlation characteristic value of the i-th wind turbine generator at time t; , , These are the feature fusion weight coefficients; Let be the rate of change of the output power of the i-th wind turbine generator at time t; This refers to the rated power of the wind turbine generator set. Let be the fatigue load gradient of the i-th wind turbine generator at time t; The rated fatigue load of the i-th wind turbine generator unit; Let be the standard deviation of wind speed at time t; Let be the average wind speed at time t.

[0035] in, , , All values ​​are positive and their sum is 1, and the values ​​are dynamically adjusted according to real-time wind conditions, especially in stable wind scenarios. The highest percentage of values ​​were found in turbulent wind scenarios. The value with the highest percentage.

[0036] , Indicates the data sampling time interval.

[0037] The above formula comprehensively considers three core dimensions: the dynamic trend of unit output power, the rate of change of fatigue load, and the degree of instability of wind conditions. By balancing the impact of each dimension on the operating status through feature fusion weight coefficients, a multi-dimensional sensitivity correlation feature is constructed, which can comprehensively and accurately capture sudden changes in the unit's operating status and provide a more comprehensive and reliable basis for subsequent optimization of the triggering mechanism.

[0038] Specifically, in S3, the allowable range of fatigue damage throughout the entire life cycle of the generator set is calculated based on the sensitivity correlation feature combined with the fatigue load balance constraint. The threshold range of sensitivity correlation feature under different wind conditions is derived in reverse through the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine generator set does not exceed a preset threshold.

[0039] S31: Real-time wind condition scenario parameters are divided into three categories—stable wind scenario, weakly turbulent wind scenario, and strongly turbulent wind scenario—based on turbulence intensity, wind speed fluctuation amplitude, and wind shear index. A wind condition scenario identification parameter θ is defined, where θ=0 corresponds to a stable wind scenario, θ=1 corresponds to a weakly turbulent wind scenario, and θ=2 corresponds to a strongly turbulent wind scenario. The cumulative proportion coefficient of each scenario is calculated based on historical wind condition statistics of the wind farm. ,in Indicates the historical statistical period The cumulative duration of the internal wind condition scenario θ.

[0040] S32: Establish a quantitative correlation model for the relationship between fatigue damage rate and sensitivity of wind turbine generators under various wind conditions: ; In the formula: Let be the fatigue damage rate of the i-th wind turbine generator at time t and wind condition scenario θ. The damage coefficient is related to the wind condition scenario; Let be the sensitivity correlation characteristic value of the i-th wind turbine generator at time t; This represents the base damage value for wind-related scenarios.

[0041] The above formula first divides the real-time wind conditions into three scenarios based on turbulence intensity, wind speed fluctuation amplitude, and wind shear index. It then obtains the cumulative proportion coefficient of each scenario by combining historical statistical data. Next, it establishes a linear relationship between fatigue damage rate and sensitivity characteristics. By using scenario-related damage coefficients and basic damage values, it quantifies the degree of correlation between the two under different wind conditions, ensuring that the calculation of fatigue damage rate is no longer divorced from the actual wind condition background, thereby improving its fit with actual operation and calculation accuracy.

[0042] S33: Based on the allowable fatigue damage value of wind turbine generator set throughout its entire life cycle (fatigue load balance constraint) and a quantitative correlation model, the allowable damage rate range under each scenario is derived through scenario proportion coefficients:

[0043] in, This represents the average permissible fatigue damage rate for all units under scenario θ. , This indicates the allowable fatigue damage value throughout the entire life cycle of the unit. Indicates the total lifespan of the unit design. This represents the damage allocation coefficient of unit i in scenario θ. , This indicates the maximum permissible difference in fatigue damage rates between units as preset.

[0044] The above formula combines the allowable fatigue damage value of the unit throughout its entire life cycle, the cumulative proportion of each wind condition scenario, and the damage distribution coefficient of a single unit in different scenarios. First, the average allowable fatigue damage rate in each scenario is derived. Then, by using the preset maximum allowable damage rate difference between units, the allowable range of damage rate of a single unit in the corresponding scenario is determined. This ensures that the damage of the unit throughout its entire life cycle does not exceed the allowable value, and provides a clear and executable constraint standard for achieving fatigue load balancing in the future. S34: Based on the allowable range of fatigue damage rate of the wind turbine generator set, the threshold range of sensitivity correlation characteristics under different wind conditions is derived in reverse: ;

[0045] in, , These are the upper and lower thresholds for the sensitivity-related features, respectively. Indicates the number of historical samples. , These represent the increments in damage rate and sensitivity characteristics at adjacent time points in historical data, respectively. , These represent the correlation characteristics between the average fatigue damage rate and the average sensitivity of unit i under scenario θ within the historical period.

[0046] The above formula is based on the established quantitative model of the allowable range of damage rate and the correlation between fatigue damage rate and sensitivity. The upper and lower limit thresholds of the sensitivity correlation feature are obtained by reverse calculation, so that the threshold setting is directly linked to the fatigue damage control target of the unit. This ensures that the optimization triggering conditions are more targeted, and the optimization process is only started when the damage rate corresponding to the sensitivity feature may exceed the equilibrium range, avoiding meaningless frequent adjustments, thereby improving the operating efficiency.

[0047] Specifically, in S4, the real-time extracted sensitivity-related features are compared with the threshold range of sensitivity-related features under different wind conditions. If any feature exceeds the corresponding threshold range, the power optimization allocation process is triggered; if it does not exceed the threshold range, the refresh step is executed.

[0048] Specifically, in S5, after triggering optimization, with the goal of minimizing the fatigue load balance of each unit, the total power demand of the wind farm is decomposed based on real-time wind condition parameters, a power allocation objective function with sensitivity feature constraints is constructed, and the optimal output power command of each unit is solved through an adaptive weighting strategy.

[0049] The objective function for power allocation with sensitivity feature constraints is expressed as:

[0050] ; In the formula: The objective function value; This represents the total number of wind turbine generators in the wind farm. Let be the fatigue damage rate of the i-th unit at time t; The average fatigue damage rate of all units at time t; The adaptive weights for the i-th unit; This is the penalty coefficient for sensitivity characteristics; Let be the sensitivity correlation characteristic value of the i-th unit at time t; Let be the optimal output power of the i-th unit at time t; This represents the initial power allocation value for the i-th unit based on real-time wind resource forecasting. This is the total power deviation penalty coefficient; Total power requirement; , These are the minimum and maximum allowable output power of the generator set; The maximum permissible rate of change of the unit's output power; This represents the upper threshold of the sensitivity correlation feature under wind condition scenario θ.

[0051] in, ∈(0,1], the smaller the cumulative fatigue damage of the i-th unit up to the present, the better the real-time wind resources at its location, and the closer the sensitivity correlation feature is to the stable threshold, the larger its value is, and vice versa; >0, the larger the sensitivity correlation characteristic value of the i-th unit, the stronger the constraint requirement for power allocation and the larger its value; conversely, the smaller the value. >0 indicates that the higher the priority of the total power demand of the wind farm and the lower the tolerance for the deviation between the total power output and the demand value, the larger the value will be, and vice versa.

[0052] The above formula takes minimizing the difference in fatigue damage rate among units as its core. It fully reflects the individual differences of different units in terms of damage accumulation, wind resource utilization, and operating status through adaptive weights. At the same time, it introduces a sensitivity feature penalty term to suppress possible sudden changes in operating status during power regulation and adds a total power deviation penalty term to ensure that the total power demand of the wind farm is met. Combined with constraints such as the upper and lower limits of unit power, the maximum allowable power change rate, and the sensitivity feature threshold, a multi-objective optimization framework is constructed that takes into account fatigue load balance, stable operating status, and total power compliance. This makes the power allocation scheme more comprehensive and reasonable, and fits the actual operating needs of the wind farm.

[0053] Adaptive weighting strategies include: ; Fatigue damage weighting coefficient ; Wind resource utilization weighting coefficient ; Operating status weight coefficient ; In the formula: The cumulative fatigue damage of the i-th unit at time t; This represents the allowable fatigue damage value throughout the entire life cycle of the unit. Let be the real-time wind speed at the location of the i-th unit; The average real-time wind speed of the wind farm; Let be the sensitivity correlation characteristic value of the i-th unit at time t; This represents the upper threshold of the sensitivity correlation feature under wind condition scenario θ.

[0054] Adaptive weights After normalization, the sum of all weights is 1.

[0055] In the above formula, the fatigue damage weight coefficient directly reflects the remaining space of accumulated damage and allowable damage of the unit, the wind resource utilization weight coefficient reflects the wind resource advantage of the location of a single unit, and the operating status weight coefficient is related to the fit between the current sensitivity characteristics of the unit and the scenario threshold. After multiplying the three and normalizing them, the final adaptive weight can accurately reflect the damage risk of the unit itself, while also taking into account the wind resource utilization efficiency and operating stability. This allows the power allocation to naturally tilt towards units with low damage risk, good wind resource conditions, and stable operating status, effectively improving the overall optimization effect.

[0056] Preferably, when decomposing the total power demand of a wind farm based on real-time wind condition parameters, the available power potential of each unit is calculated based on the wind profile prediction data of the location of each wind turbine in the wind farm: , in air density, The swept area of ​​the unit. The wind energy utilization coefficient, For the tip speed ratio, For variable pitch angle, Let represent the real-time wind speed at the location of the i-th wind turbine. The wind profile prediction data for the locations of each wind turbine in the wind farm are obtained through multi-source data fusion and microscale downscaling methods. Based on numerical weather prediction and ground observation, spatial downscaling is performed using computational fluid dynamics simulation or statistical models, and dynamic calibration is combined with measured data from the wind tower and nacelle to obtain a detailed prediction of the vertical wind speed distribution at each turbine location.

[0057] The total power demand is initially decomposed based on the proportion of available power potential of each unit, resulting in the initial power allocation:

[0058] The above formula combines the real-time wind speed at the unit's location, the unit's swept area, air density, and the wind energy utilization coefficient related to the tip speed ratio and pitch angle to accurately calculate the available power potential of a single unit. The total power demand is initially decomposed according to the potential ratio of each unit to ensure that the initial allocation is in line with the actual wind resources. Then, the initial power is fine-tuned based on the difference between the sensitivity characteristics and the scenario threshold, so that the correction amount is linked to the risk of sudden changes in the unit's operating state, thereby making full use of wind resources and further improving the balance of power distribution and operational stability.

[0059] Furthermore, combining real-time sensitivity correlation features The initial power allocation is corrected, and the correction amount is... ,in The correction coefficient is positively correlated with the redundancy of the unit's sensitivity characteristics. Finally, the corrected pre-allocated power (i.e., the optimal output power) is obtained as the input to the power allocation objective function.

[0060] Specifically, in S6, after issuing the optimal power command to each unit for adjustment, it also includes real-time monitoring of the actual output power of each unit. and actual fatigue load Calculate power tracking error and fatigue load balance deviation ,in This represents the fatigue damage rate calculated based on the actual fatigue load.

[0061] The above formula quantifies the tracking effect of power regulation by calculating the relative deviation between the actual output power and the optimal power command. At the same time, it calculates the relative deviation between the damage rate obtained based on the actual fatigue load and the average damage rate to evaluate the balancing effect of the fatigue load. The two deviation indicators provide a clear and quantifiable basis for whether the power command needs to be dynamically adjusted in the future, ensuring that the execution effect of the optimized command can achieve the expected goal, and further guaranteeing the balancing of fatigue load and the achievement of power output standards. If any unit Exceeding the preset power deviation range or If the load balance deviation exceeds the preset range, the optimal output power command for the unit will be dynamically adjusted based on the deviation value. The adjustment amount is positively correlated with the deviation value, and the command will be reissued and executed until the deviation values ​​of all units are within the corresponding preset range.

[0062] Furthermore, the difference in fatigue damage rate among the units is constrained to not exceed a preset threshold: Preset threshold For dynamic thresholds, ,in The basic damage difference threshold is set as a preset ratio based on the allowable fatigue damage value throughout the unit's entire life cycle. This is the normalized value of turbulence intensity for the current wind conditions, taken as the ratio of real-time turbulence intensity to the maximum turbulence intensity allowed by the unit design.

[0063] The above formula combines the basic damage difference threshold with the normalized value of turbulence intensity in the current wind condition scenario to construct a dynamically changing damage rate difference threshold. This allows the threshold to adaptively adjust with changes in turbulence intensity. When the turbulence intensity is high and the operating environment is more complex, the difference limit can be appropriately relaxed. Under stable wind conditions, the difference is strictly controlled to ensure the balancing effect. This makes the constraints more in line with the actual operation under different wind conditions, improving the rationality and flexibility of the constraints.

[0064] The methods described in the above embodiments can be adapted to different wind conditions, making the fatigue damage of each unit in the wind farm more balanced, effectively extending the service life of the units, reducing maintenance costs, and at the same time accurately allocating power, improving power generation efficiency and power tracking accuracy, reducing deviations, ensuring stable operation, making full use of wind resources, and bringing higher economic and operational benefits to the wind farm.

[0065] Example 2 The XX wind farm deploys 10 identical wind turbine generators and employs this power optimization allocation method to achieve efficient scheduling. First, it collects real-time operating parameters for each unit, including output power, generator speed, yaw angle, and gearbox oil temperature; fatigue load characteristic parameters such as blade root bending load, tower lateral vibration load, and gearbox input shaft torque; and wind condition parameters such as real-time wind speed, wind direction, turbulence intensity, and wind shear index. These data are then integrated to form a real-time operating status dataset for the wind farm.

[0066] The dataset was preprocessed: by combining the 3σ criterion with load mutation judgment (the load difference between adjacent sampling times does not exceed the instantaneous fluctuation range allowed by the unit design), 3 abnormal data points were removed, and then the dynamic weighted smoothing process was performed by sliding window to obtain the smoothed blade root bending load of Unit 2 at time t as 836kN, and the smoothed tower lateral vibration load of Unit 7 as 312kN.

[0067] Sensitivity correlation features were extracted: The output power change rate, fatigue load gradient and wind condition fluctuation coefficient of each unit were integrated to calculate the characteristic value of each unit at time t. The characteristic value of unit 1 was 0.29, unit 4 was 0.35, unit 8 was 0.48, and the characteristic value of the remaining units was between 0.24 and 0.43.

[0068] Calculate the sensitivity-related feature threshold range: Based on the real-time wind condition parameters, the current wind scenario is determined to be a weak turbulent wind scenario (θ=1). Based on the historical statistics of the wind farm, the cumulative proportion coefficient of this scenario is calculated to be 0.36. Combined with the allowable value of fatigue damage throughout the entire life cycle of the unit, the allowable range of fatigue damage rate under this scenario is derived to be [0.007, 0.015]. The sensitivity-related feature threshold range is obtained by reverse conversion [0.19, 0.44].

[0069] After comparing the eigenvalues ​​with the thresholds, it was found that the eigenvalue of Unit 8 (0.48) exceeded the upper limit, triggering the power optimization allocation process. Based on the wind profile prediction data of each unit's location, the available power potential was calculated. The total power demand of the wind farm, 14.5MW, was initially allocated according to the potential ratio, with Unit 1 initially allocated 1.5MW and Unit 8 initially allocated 1.6MW. Then, based on the sensitivity characteristics, a correction was made, with Unit 8 receiving a correction of -0.25MW, and a pre-allocation of 1.35MW after correction.

[0070] With the goal of minimizing fatigue load balance, a power allocation objective function with sensitivity characteristic constraints is constructed. An adaptive weighting strategy is used to calculate the weights of each unit: Unit 1 is 0.10, Unit 4 is 0.11, Unit 8 is 0.08, and the weights of the remaining units are between 0.09 and 0.12 (summing up to 1 after normalization). Solving the objective function yields the optimal output power for each unit: Unit 1 is 1.46 MW, Unit 4 is 1.52 MW, Unit 8 is 1.33 MW, and the optimal power for the remaining units is between 1.38 and 1.55 MW. Furthermore, the power variation of all units does not exceed the maximum allowable rate of change.

[0071] After issuing the optimal power command, the operating status of each unit was monitored in real time: Unit 8's actual output power was 1.34MW, with a power tracking deviation of 0.008; the fatigue damage rate of Unit 8, calculated based on the actual fatigue load, was 0.014, with a balance deviation of 0.027 from the average damage rate of 0.011 for all units, both within the preset allowable range. The power tracking deviation and load balance deviation of the other units also met the requirements, requiring no additional adjustments. The process was then refreshed to proceed to the next round of real-time monitoring and optimization.

[0072] In summary, the methods described in the above embodiments effectively balance the fatigue load of each turbine by adapting to the complex and ever-changing wind conditions of the wind farm in real time, reducing the difference in fatigue damage rates between turbines, effectively extending the entire life cycle of the turbines, and improving the overall operational stability and reliability of the wind farm. At the same time, they accurately decompose the total power demand, reduce power tracking deviation, optimize wind resource utilization efficiency, reduce equipment losses caused by sudden changes in operating conditions, flexibly adapt to different turbulent scenarios by dynamically adjusting power commands, reduce operation and maintenance costs, and take into account the rationality and safety of the turbine output power, thus avoiding overload and excessive power fluctuations in a timely manner.

[0073] Example 3 A wind farm power optimization allocation system based on fatigue load balancing includes: The data acquisition module is used to collect the operating parameters, fatigue load characteristic parameters and real-time wind condition parameters of each wind turbine in the wind farm in real time, and to build a real-time operating status dataset of the wind farm based on the collected parameters. The data preprocessing and feature extraction module is used to preprocess the real-time running status dataset to obtain the preprocessed dataset, and extract sensitivity correlation features that characterize sudden changes in running status from the preprocessed dataset. The calculation module is used to calculate the allowable range of fatigue damage throughout the entire life cycle of the generator set based on the sensitivity correlation feature and the fatigue load balance constraint. The threshold range of sensitivity correlation feature under different wind conditions is derived in reverse from the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine does not exceed a preset threshold. The judgment module is used to compare the sensitivity-related features with the threshold range of sensitivity-related features under different wind conditions. If any sensitivity-related feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered; if it does not exceed the threshold range, the refresh step is executed. The power allocation module is used to trigger the power optimization allocation process, aiming to minimize the fatigue load balance of each wind turbine generator set. Based on real-time wind condition parameters, it decomposes the total power demand of the wind farm, constructs a power allocation objective function with sensitivity correlation feature constraints, and solves the optimal power command for each unit through an adaptive weight strategy. The execution module is used to issue optimal power commands to each wind turbine generator set for adjustment. After all units have completed adjustment, the refresh step is executed.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A wind farm power optimization distribution method based on fatigue load equalization, characterized in that, Includes the following steps: Real-time data collection of operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine in the wind farm is used to construct a real-time operating status dataset of the wind farm based on the collected parameters. The real-time running status dataset is preprocessed to obtain a preprocessed dataset, and sensitivity correlation features characterizing sudden changes in running status are extracted from the preprocessed dataset. Based on the sensitivity correlation feature and fatigue load balance constraint, the allowable range of fatigue damage throughout the life cycle of the generator set is calculated. The threshold range of sensitivity correlation feature under different wind conditions is derived by reverse derivation of the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine set does not exceed a preset threshold. The sensitivity correlation feature is compared with the sensitivity correlation feature threshold range under different wind conditions. If any sensitivity correlation feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered. If the steps are not exceeded, then refresh the process. After triggering the power optimization allocation process, with the goal of minimizing the fatigue load balance of each wind turbine, the total power demand of the wind farm is decomposed based on real-time wind condition parameters, and a power allocation objective function with sensitivity correlation feature constraints is constructed. The optimal power command for each unit is solved through an adaptive weighting strategy. The optimal power command is issued to each wind turbine generator set for adjustment. After all units have completed the adjustment, the refresh procedure is executed.

2. The wind farm power optimization distribution method based on fatigue load balance according to claim 1, characterized in that, In the step of collecting real-time operating parameters, fatigue load characteristic parameters, and real-time wind condition parameters of each wind turbine generator set in the wind farm, and constructing a real-time operating status dataset of the wind farm based on the collected parameters, the operating parameters include the output power of the wind turbine generator set, generator speed, yaw angle, pitch angle, gearbox oil temperature, and nacelle vibration acceleration; the fatigue load characteristic parameters include the blade root bending load, tower lateral vibration load, nacelle longitudinal impact load, gearbox input shaft torque, and hub axial load; and the wind condition parameters include real-time wind speed, wind direction, turbulence intensity, wind shear index, and atmospheric stability level.

3. The wind farm power optimization distribution method based on fatigue load equalization according to claim 2, characterized in that, The method for preprocessing the real-time running status dataset to obtain a preprocessed dataset, and then extracting sensitivity correlation features representing sudden changes in running status from the preprocessed dataset, is as follows: Outliers in the real-time operating status dataset are eliminated by combining the 3σ criterion with the load mutation judgment condition. The load mutation judgment condition is that the difference between the fatigue load characteristic parameters at adjacent sampling times exceeds the instantaneous load fluctuation range allowed by the unit design. Smoothing is performed on the runtime state dataset after outlier removal to obtain a preprocessed dataset; Sensitivity correlation features characterizing sudden changes in runtime state are extracted from the preprocessed dataset, specifically: By integrating the output power, fatigue load characteristic parameters, and wind speed parameters of wind turbine generators, a multi-dimensional sensitivity correlation feature is constructed: In the formula: is the sensitivity-related characteristic value of the ith wind turbine generator set at time t; , , is a characteristic fusion weight coefficient; is the output power change rate of the ith wind turbine generator set at time t; is the rated power of the wind turbine generator set; is the fatigue load gradient of the ith wind turbine generator set at time t; is the rated fatigue load of the ith wind turbine generator set; is the standard deviation of the wind speed at time t; is the average wind speed at time t.

4. The wind farm power optimization distribution method based on fatigue load equalization according to claim 3, characterized in that, Based on the aforementioned sensitivity correlation features and fatigue load balancing constraints, the allowable range of fatigue damage throughout the generator set's life cycle is calculated. The following method is used to deduce the sensitivity correlation feature threshold ranges under different wind conditions using these allowable fatigue damage ranges. Real-time wind conditions are classified into three categories based on turbulence intensity, wind speed fluctuation amplitude, and wind shear index: steady wind scenario, weak turbulence wind scenario, and strong turbulence wind scenario. Establish a quantitative correlation model for the relationship between fatigue damage rate and sensitivity of wind turbine generators under various wind conditions; Based on fatigue load balance constraints and quantitative correlation models, the allowable range of fatigue damage rate of wind turbine generators under each scenario is derived by using scenario proportion coefficients. Based on the allowable range of fatigue damage rate of the wind turbine generator, the threshold range of sensitivity correlation characteristics under different wind conditions is derived in reverse.

5. The wind farm power optimization allocation method based on fatigue load balancing according to claim 4, characterized in that, In the step of establishing a quantitative correlation model for the correlation characteristics between fatigue damage rate and sensitivity of wind turbine generators under various wind conditions, the quantitative correlation model is as follows: ； In the formula: Let be the fatigue damage rate of the i-th wind turbine generator at time t and wind condition scenario θ. The damage coefficient is related to the wind condition scenario; Let be the sensitivity correlation characteristic value of the i-th wind turbine generator at time t; This represents the base damage value for wind conditions.

6. The wind farm power optimization allocation method based on fatigue load balancing according to claim 5, characterized in that, The formula for deriving the allowable range of fatigue damage rate of wind turbine generators under various scenarios, based on fatigue load balance constraints and a quantitative correlation model, using scenario proportion coefficients, is as follows: in, This represents the average allowable fatigue damage rate for all units under scenario θ. This indicates the allowable fatigue damage value throughout the entire life cycle of the unit. Indicates the total lifespan of the unit design. This represents the damage allocation coefficient of unit i in scenario θ. This represents the maximum permissible difference in fatigue damage rates between units. The formula for deriving the threshold range of sensitivity correlation characteristics under different wind conditions based on the allowable range of fatigue damage rate of the wind turbine generator is expressed as follows: in, , These are the upper and lower thresholds for the sensitivity-related features, respectively. Indicates the number of historical samples. , These represent the increments in damage rate and sensitivity characteristics at adjacent time points in historical data, respectively. , These represent the correlation characteristics between the average fatigue damage rate and the average sensitivity of unit i under scenario θ within the historical period.

7. The wind farm power optimization allocation method based on fatigue load balancing according to claim 6, characterized in that, After triggering the power optimization allocation process, with the goal of minimizing the fatigue load balance of each wind turbine, the total power demand of the wind farm is decomposed based on real-time wind condition parameters. A power allocation objective function with sensitivity correlation feature constraints is constructed, and the formula for solving the optimal power command of each unit through an adaptive weighting strategy is expressed as follows: In the formula: The objective function value; This represents the total number of wind turbine generators in the wind farm. Let be the fatigue damage rate of the i-th unit at time t; The average fatigue damage rate of all units at time t; The adaptive weights for the i-th unit; This is the penalty coefficient for sensitivity characteristics; Let be the sensitivity correlation characteristic value of the i-th unit at time t; Let be the optimal output power of the i-th unit at time t; This represents the initial power allocation value for the i-th unit based on real-time wind resource forecasting. This is the total power deviation penalty coefficient; Total power requirement; , These are the minimum and maximum allowable output power of the unit; The maximum permissible rate of change of the unit's output power; This represents the upper threshold of the sensitivity correlation feature under wind condition scenario θ.

8. The wind farm power optimization allocation method based on fatigue load balancing according to claim 7, characterized in that, When decomposing the total power demand of a wind farm based on real-time wind condition parameters, the available power potential of each unit is calculated based on the wind profile prediction data of the location of each wind turbine in the wind farm. ， in air density, The swept area of ​​the unit. The wind energy utilization coefficient, For the tip speed ratio, For variable pitch angle, Let be the real-time wind speed at the location of the i-th unit; The total power demand is initially decomposed based on the proportion of available power potential of each unit, resulting in an initial power allocation value: 。 9. The wind farm power optimization allocation method based on fatigue load balancing according to claim 8, characterized in that, After issuing the optimal power command to each wind turbine for adjustment, the process also includes real-time monitoring of the actual output power of each turbine. and actual fatigue load Calculate the power tracking error and fatigue load equalization error: Calculate power tracking error: ; Calculate fatigue load equilibrium deviation ;in The fatigue damage rate is calculated based on the actual fatigue load. If the power tracking deviation of any unit exceeds the preset power deviation range, or the fatigue load balancing deviation exceeds the preset load balancing deviation range, the optimal output power command of that unit will be dynamically adjusted based on the deviation value. The adjustment amount is positively correlated with the deviation value, and the command will be reissued and executed until the deviation values ​​of all units are within the corresponding preset range.

10. A wind farm power optimization allocation system based on fatigue load balancing, based on the wind farm power optimization allocation method based on fatigue load balancing as described in any one of claims 1 to 9, characterized in that, include: The data acquisition module is used to collect the operating parameters, fatigue load characteristic parameters and real-time wind condition parameters of each wind turbine in the wind farm in real time, and to build a real-time operating status dataset of the wind farm based on the collected parameters. The data preprocessing and feature extraction module is used to preprocess the real-time running status dataset to obtain the preprocessed dataset, and extract sensitivity correlation features that characterize sudden changes in running status from the preprocessed dataset. The calculation module is used to calculate the allowable range of fatigue damage throughout the entire life cycle of the generator set based on the sensitivity correlation feature and the fatigue load balance constraint. The threshold range of sensitivity correlation feature under different wind conditions is derived in reverse from the allowable range of fatigue damage. The fatigue load constraint is that the difference in fatigue damage rate of each wind turbine does not exceed a preset threshold. The judgment module is used to compare the sensitivity-related features with the threshold range of sensitivity-related features under different wind conditions. If any sensitivity-related feature exceeds the corresponding dynamic feature threshold range, the power optimization allocation process is triggered. If the steps are not exceeded, then refresh the process. The power allocation module is used to trigger the power optimization allocation process, aiming to minimize the fatigue load balance of each wind turbine generator set. Based on real-time wind condition parameters, it decomposes the total power demand of the wind farm, constructs a power allocation objective function with sensitivity correlation feature constraints, and solves the optimal power command for each unit through an adaptive weight strategy. The execution module is used to issue the optimal power command to each wind turbine to perform adjustment. After all units have completed adjustment, the execution steps are refreshed.

Images