Wind power plant global yawing method with net power improvement as target

Abstract

The invention relates to the technical field of wind power plant control, in particular to a wind power plant global yawing method with net power improvement as a target. The method comprises the following steps: collecting unit operation data and flow field monitoring data in a wind power plant to form a wind field state data set; constructing a parameterized wake flow simulation model fused with topographic features, and dynamically calibrating parameters through a gradient descent method to establish a dynamic digital twinborn body; taking a yaw angle of each unit as a decision variable, taking full-field net power maximization as a target, and setting constraint conditions to construct an optimization problem; solving an optimal yaw angle combination by adopting a collaborative optimization algorithm, converting the optimal yaw angle combination into a control instruction, and issuing and executing the control instruction; and performing feedback correction on the weight coefficient of the optimization problem according to the newly collected operation data. According to the method, dynamic digital twin bodies fused with topographic features are constructed, full-field collaborative optimization is achieved, wake flow interference and topographic effects among units are fully considered, and the problem that net power improvement of traditional single-machine yaw control is limited is solved.

Description

Technical Field

[0001] This invention relates to the field of wind farm control technology, and more specifically, to a global yaw method for wind farms aimed at improving net power output. Background Technology

[0002] Yaw control in wind farms is one of the core technologies for improving wind power utilization efficiency. By adjusting the yaw angle of the wind turbine, the turbine rotor is precisely aligned with the wind direction to maximize wind energy capture. Existing yaw control methods mostly adopt independent control logic for each turbine, failing to fully consider the wake interference between turbines within the wind farm and ignoring the influence of terrain features on the wind farm flow pattern, resulting in limited overall net power output. Furthermore, traditional methods lack dynamic balancing of load losses during turbine yaw, easily increasing the risk of turbine structural fatigue due to excessive pursuit of individual turbine power. Therefore, this invention provides a global yaw method for wind farms aimed at improving net power output. Summary of the Invention

[0003] The purpose of this invention is to provide a global yaw method for wind farms with the goal of improving net power, in order to solve the problem that traditional methods mentioned in the background art lack dynamic balancing of load loss during the yaw process of the unit.

[0004] To achieve the above objectives, the present invention aims to provide a global yaw method for wind farms with the goal of increasing net power output, comprising the following steps:

[0005] S1. Collect the unit operation data and flow field monitoring data in the wind farm, perform spatiotemporal alignment and normalization processing, and form a wind farm status dataset.

[0006] S2. Construct a parameterized wake simulation model that integrates terrain features, and combine it with the wind field state dataset to dynamically calibrate the parameters of the parameterized wake simulation model using the gradient descent method, thereby establishing a dynamic digital twin of the wind farm.

[0007] S3. Based on a dynamic digital twin, the yaw angle of each unit is used as the decision variable, the net power of the entire field is maximized as the optimization objective, and constraints are set to construct an optimization problem.

[0008] S4. Using a collaborative optimization algorithm, the optimization problem is solved in a dynamic digital twin to obtain the optimal yaw angle combination for each wind turbine.

[0009] S5. Convert the optimal yaw angle combination into control commands, send them to each wind turbine to perform yaw operations, and simultaneously collect new operating data;

[0010] S6. Based on the new operating data, the weight coefficients of the optimization problem are corrected.

[0011] As a further improvement to this technical solution, the specific steps involved in S1 for collecting wind farm unit operation data and flow field monitoring data, and performing spatiotemporal alignment and normalization processing are as follows:

[0012] Collect operating data of each unit in the wind farm, including wind speed, wind direction, power output, yaw angle, pitch angle, speed, and unit status signals;

[0013] Simultaneously collect flow field monitoring data, including three-dimensional wind field information provided by lidar, ultrasonic anemometer, and meteorological tower;

[0014] Clean the unit operation data and flow field monitoring data to remove outliers, missing values ​​and jump data;

[0015] The data is unified to the same time reference by using time interpolation, and the location information of the wind farm coordinate system and the monitoring equipment is utilized.

[0016] The data are subjected to min-max normalization, and the processed data is organized into a structured dataset according to the time series, forming a wind field status dataset.

[0017] As a further improvement to this technical solution, the specific steps involved in constructing the parameterized wake simulation model that incorporates terrain features in step S2 are as follows:

[0018] Based on the three-dimensional wind field information contained in the wind field state dataset, a wind field terrain feature matrix is ​​constructed using a gridded interpolation method. ;

[0019] A parameterized wake simulation model is constructed based on the Bastankhah model. The input of the model is the wind field state dataset, and the output is the wind speed distribution across the entire field.

[0020] Using a parametric wake simulation model, based on the unit thrust coefficient and upstream wind speed in the wind field state dataset, the wake wind speed distribution of a single wind turbine in flat terrain is calculated.

[0021] Based on wake wind speed distribution and wind field topographic feature matrix The wake of each wind turbine is corrected for terrain features, and the wakes of all wind turbines are superimposed to obtain the wind speed distribution across the entire field.

[0022] As a further improvement to this technical solution, the specific steps involved in S2, which involve dynamically calibrating the parameters of the parameterized wake simulation model using the gradient descent method to establish a dynamic digital twin of the wind farm, are as follows:

[0023] The wind speed prediction error is calculated by comparing the measured wind speed in the wind field state dataset with the predicted wind speed from the parametric wake simulation model at the same location. ;

[0024] The gradient descent method is used, combined with wind speed prediction error. For wake attenuation coefficient Perform iterative corrections;

[0025] Each time a new batch of wind field state data is received, a parameter update is performed, and the updated wake attenuation coefficient is... Substitute the parameterized wake simulation model and recalculate the overall wind speed distribution;

[0026] The updated parametric wake simulation model is combined with the real-time wind farm status dataset to form a dynamic digital twin of the wind farm.

[0027] As a further improvement to this technical solution, the specific steps involved in S3, with the yaw angle of each unit as the decision variable and the maximization of net power across the entire field as the optimization objective, are as follows:

[0028] Based on the full-field wind speed distribution output by the dynamic digital twin, combined with the wind energy utilization coefficient, air density and rotor swept area in the wind turbine characteristic table, the theoretical power generation of each wind turbine is calculated.

[0029] By introducing the yaw angle of each unit as an optimization variable, a cosine cubic function is used to reduce the theoretical power generation, and a yaw error compensation coefficient is added to calculate the corrected actual power generation. ;

[0030] Based on the difference between the yaw angle of the previous moment and the current yaw angle, as well as the absolute value of the yaw angle, and combined with the penalty weight, a load penalty term is constructed.

[0031] The net power index is calculated by subtracting the total load penalty cost from the sum of the actual power generation of all wind turbines. And the optimization objective is to maximize the net power of the entire field.

[0032] As a further improvement to this technical solution, in step S3, constraints are set, and the specific steps involved in constructing the optimization problem are as follows:

[0033] Set yaw angle range constraints to limit the yaw angle of each unit to the allowable range of the wind turbine design;

[0034] Set a yaw rate constraint to limit the yaw angle change within adjacent control cycles from not exceeding the allowable rate of change.

[0035] Set a lower limit constraint on power output to limit the actual power generation to not be lower than the lower limit;

[0036] Ultimately, an optimization problem is formed with the yaw angle of each unit as the decision variable, the goal of maximizing the net power of the entire field, and the constraints of yaw angle range, yaw rate of change, and power output lower limit.

[0037] As a further improvement to this technical solution, in step S4, a collaborative optimization algorithm is used to solve the optimization problem in the dynamic digital twin, involving the following specific steps:

[0038] S4.1 Initialize the yaw angle of each wind turbine to the current actual value;

[0039] S4.2, Decompose the overall yaw optimization problem into... There are several sub-problems, each corresponding to the yaw angle optimization of a single wind turbine. For the first sub-problem... For a typhoon generator, its local objective function is the net power increment;

[0040] S4.3 In each optimization iteration, based on the current yaw angle combination of all wind turbines, the dynamic digital twin is invoked to calculate the overall wind speed distribution. For each wind turbine... Calculate the local objective function Its yaw angle The gradient;

[0041] S4.4 Under the constraints of yaw angle range and yaw rate of change, the yaw angle is updated along the gradient direction using the gradient ascent method.

[0042] S4.5, if net power index If the change is less than the set threshold, then output the optimal yaw angle combination for all wind turbines; otherwise, return to step S4.3 to continue iterating.

[0043] As a further improvement to this technical solution, the specific steps involved in converting the optimal yaw angle combination into control commands in step S5 are as follows:

[0044] Receive the optimal yaw angle combination for each wind turbine. Calculate its current actual yaw angle With optimal yaw angle The difference is used to obtain the change in yaw angle. ;

[0045] Based on the maximum yaw rate of the crew and the current control cycle Calculate the maximum permissible single change amount. ;

[0046] If the absolute value of the change in yaw angle does not exceed the maximum permissible single change, then the optimal yaw angle is directly set. Assign the value to the yaw command; otherwise, calculate the progressive command based on the maximum permissible single change amount and use it as the yaw command. ;

[0047] The yaw command is converted into a command format that the unit controller can recognize, and a timestamp and unit number are added to form a control command package, which is then sent to the corresponding wind turbine.

[0048] As a further improvement to this technical solution, the specific steps involved in S5, which involve sending the yaw operation to each wind turbine and simultaneously collecting new operating data, are as follows:

[0049] Monitor the actual yaw angle feedback value during control command execution. ;

[0050] When the actual yaw angle feedback value With yaw command When the absolute value of the difference is less than the set error threshold, the yaw is determined to be in place, and the unit enters a stable operating state.

[0051] After the yaw is in place, a new round of unit operation data and flow field monitoring data are collected to form a new wind field status dataset.

[0052] As a further improvement to this technical solution, the specific steps involved in the feedback correction of the weight coefficients of the optimization problem in step S6 are as follows:

[0053] Extract the actual load and yaw angle of each wind turbine from the fresh wind farm state dataset, and calculate the actual load variation. ;

[0054] Based on the actual load variations and yaw angles of all wind turbines, an adaptive threshold-based adjustment mechanism is used to penalize the yaw variation weights in the optimization problem. And the absolute value penalty weight of yaw Provide feedback and make corrections;

[0055] Corrected yaw change penalty weight And the corrected absolute yaw penalty weight This will be used to construct the next round of optimization problems.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0057] 1. In this wind farm global yaw method aimed at improving net power, a parameterized wake simulation model integrating terrain features is constructed and the parameters are dynamically calibrated by the gradient descent method to form a dynamic digital twin of the wind farm. Compared with the traditional single-unit yaw control that ignores the influence of terrain and wake interaction, it achieves high-fidelity simulation of the wind speed distribution across the entire field and solves the problem that the net power improvement is limited due to the failure to consider the wake interference between units and the terrain effect in the traditional method.

[0058] 2. In this wind farm global yaw method aimed at improving net power, the goal is to maximize the net power of the entire field. The method combines the actual power generation with a cosine cubic function and introduces a load penalty term. The optimal yaw angle combination is solved by a collaborative optimization algorithm. Compared with the traditional control logic that only pursues the power of a single unit, this method achieves a dynamic balance between power generation efficiency and unit load safety. At the same time, the weight coefficient is corrected by real-time data feedback, which solves the defects of traditional methods that are prone to unit structural fatigue and lack a closed-loop optimization mechanism. Attached Figure Description

[0059] Figure 1 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example:

[0062] Please see Figure 1 As shown, this embodiment provides a global yaw method for wind farms aimed at improving net power output, including the following steps:

[0063] S1. Collect the unit operation data and flow field monitoring data in the wind farm, perform spatiotemporal alignment and normalization processing, and form a wind farm status dataset.

[0064] In this embodiment, the operating data of each unit in the wind farm is collected, including wind speed, wind direction, power output, yaw angle, pitch angle, speed, and unit status signals.

[0065] Simultaneously collect flow field monitoring data, including three-dimensional wind field information provided by lidar, ultrasonic anemometer, and meteorological tower;

[0066] Clean the unit operation data and flow field monitoring data to remove outliers, missing values ​​and jump data;

[0067] The data is unified to the same time reference by using time interpolation method. The location information of the wind farm coordinate system unit and monitoring equipment is used to ensure that the data is consistent and comparable in time and space.

[0068] The data are subjected to maximum and minimum normalization, and the processed data is organized into a structured dataset according to the time series to form a wind field status dataset.

[0069] Each moment in the wind field state dataset corresponds to a state vector, which includes the operating data of each unit and the flow field monitoring data.

[0070] S2. Construct a parameterized wake simulation model that integrates terrain features, and combine it with the wind field state dataset to dynamically calibrate the parameters of the parameterized wake simulation model using the gradient descent method, thereby establishing a dynamic digital twin of the wind farm.

[0071] In this embodiment, based on the three-dimensional wind field information (including wind farm terrain elevation and ground roughness distribution data) contained in the wind field state dataset, a wind field terrain feature matrix is ​​constructed using a gridded interpolation method. ;

[0072] Among them, the wind field topographic feature matrix is ​​a numerical grid, in which each grid cell stores the elevation and roughness values ​​at that location, which is used to quantitatively describe the impact of topographic relief and surface roughness on wind flow;

[0073] A parameterized wake simulation model is constructed based on the Bastankhah model. The input of the model is the wind field state dataset, and the output is the wind speed distribution across the entire field.

[0074] Using a parametric wake simulation model, based on the unit thrust coefficient and upstream wind speed in the wind field state dataset, the wake wind speed distribution of a single wind turbine in flat terrain is calculated.

[0075]

[0076] in, Indicates downstream spatial location The wake wind speed at the location; This represents the upstream wind speed, taken from the wind field state dataset. The thrust coefficient of the wind turbine is calculated from the pitch angle and tip speed ratio in the unit's operating data. This is the wake attenuation coefficient, which is related to atmospheric stability; The diameter of the fan rotor; These are the coordinates of the downstream location relative to the wind turbine; The wake width parameter describes the lateral spread of the wake. The Bastankhah model is a parameterized wake model based on a Gaussian function to describe the wake velocity distribution. It predicts the downstream wind speed distribution using the wake width parameter and the velocity deficit formula.

[0077] Based on wake wind speed distribution and wind field topographic feature matrix The wake of each wind turbine is corrected for terrain features, and the wakes of all wind turbines are superimposed to obtain the overall wind speed distribution:

[0078]

[0079] in, Spatial location The overall wind speed at the location is the actual wind speed after considering the wake effects of all wind turbines; For the first Typhoon generators in flat terrain The wake wind speed at the location; This indicates that all upwind fans are superimposed. This is a terrain correction function, with values ​​between 0.8 and 1.2. When the value is greater than 1, it reflects the enhancing effect of terrain on wake propagation; when the value is less than 1, it reflects the weakening effect of terrain on wake propagation. For the first Coordinates of the typhoon generator;

[0080] The terrain correction function takes the following form:

[0081]

[0082] in, downstream point With respect to the location of the fan The height difference; For reference height (wheel hub height can be used); For position Surface roughness; For reference roughness (typical value for flat terrain can be taken); As a high-impact coefficient, This is the roughness influence coefficient, which can be calibrated based on wind tunnel experiments or field observation data. Its value range is generally [range missing]. .

[0083] In this embodiment, the wind speed prediction error is calculated by comparing the measured wind speed in the wind field state dataset with the predicted wind speed from the parametric wake simulation model at the same location. ;

[0084]

[0085] in, To measure the wind speed, Predict wind speed using a parameterized wake simulation model at the same location point. This represents the wind speed prediction error.

[0086] In the parametric wake simulation model, the wake attenuation coefficient is selected. As a dynamic calibration parameter; The value reflects the diffusion and attenuation characteristics of the wake downstream, and is affected by factors such as atmospheric stability and topographic relief, and is time-varying;

[0087] The gradient descent method is used, combined with wind speed prediction error. For wake attenuation coefficient Perform iterative corrections;

[0088]

[0089] in, The wake attenuation coefficient at the current moment. This is the updated wake attenuation coefficient; The learning rate has a range of values. Control the magnitude of parameter updates; To predict wind speed The sensitivity can be assessed using a parameterized wake simulation model. Approximate calculation of nearby small perturbations;

[0090] Each time a new batch of wind field state data is received (e.g., updated every 5 minutes), a parameter update is performed, and the updated wake attenuation coefficient is applied. Substitute the parameterized wake simulation model and recalculate the overall wind speed distribution;

[0091] The updated parametric wake simulation model is combined with the real-time wind farm status dataset to form a dynamic digital twin of the wind farm.

[0092] Among them, the dynamic digital twin reflects the wind field flow pattern, wake superposition and terrain effect in real time, and provides a high-fidelity simulation environment for subsequent optimization.

[0093] S3. Based on a dynamic digital twin, the yaw angle of each unit is used as the decision variable, the net power of the entire field is maximized as the optimization objective, and constraints are set to construct an optimization problem.

[0094] In this embodiment, based on the full-field wind speed distribution output by the dynamic digital twin, and combined with the wind energy utilization coefficient, air density, and rotor swept area in the wind turbine characteristic table, the theoretical power generation of each wind turbine is calculated. ;

[0095]

[0096] in, For wind turbine Theoretical power generation capacity; For air density, the value is taken as the real-time density value in the wind field state dataset (e.g., calculated from temperature and pressure data). The rotor swept area, ; The wind energy utilization coefficient is the tip speed ratio. and pitch angle The function is retrieved from the fan characteristic table; For wind turbine The actual wind speed at the wheel hub height is determined by the overall wind speed distribution in the dynamic digital twin. Give; The tip speed ratio is calculated from the fan speed and the wind speed. This is the current pitch angle;

[0097] Introducing the yaw angle of each unit As an optimization variable, a cosine cubic function is used to reduce the theoretical power generation, while a yaw error compensation coefficient is added to calculate the corrected actual power generation. ;

[0098]

[0099] in, For wind turbine Corrected actual power generation; For wind turbine Yaw angle (optimization variable) ; The yaw error compensation coefficient (usually taken as...) ), used to correct the dynamic response delay of the yaw system; It is a cosine cubic function, reflecting the cubic effect of the reduction in effective wind speed caused by yaw on power;

[0100] Based on the difference between the yaw angle of the previous moment and the current yaw angle, as well as the absolute value of the yaw angle, and combined with the penalty weight, a load penalty term is constructed.

[0101] Defined as:

[0102]

[0103] in, For wind turbine The load penalty term, For wind turbine The yaw angle at the previous moment; The penalty weight for yaw change (e.g., 0.01~0.05) reflects the dynamic off-center load risk; The absolute value penalty weight for yaw (e.g., 0.001~0.01) reflects the static off-center load risk;

[0104] The net power index is calculated by subtracting the total load penalty cost from the sum of the actual power generation of all wind turbines. And the optimization objective is to maximize the net power of the entire field;

[0105] Optimization goal:

[0106]

[0107] in, This represents the total number of wind turbines in the wind farm. The load penalty factor (unit: kW / deg) is used to unify the dimensions of power and load, and is generally taken as 0.1% to 1% of the rated power of a single unit (e.g., 2 to 20 kW / deg). This is a net power rating, expressed in kW.

[0108] In this embodiment, a yaw angle range constraint is set to limit the yaw angle of each unit to the allowable range of the wind turbine design.

[0109] Yaw angle range constraints:

[0110]

[0111] Set a yaw rate constraint to limit the yaw angle change within adjacent control cycles from not exceeding the allowable rate of change.

[0112] Yaw rate constraint:

[0113]

[0114] Its existence, The maximum rate of change of yaw angle, usually To prevent excessive yaw rate from causing structural impact;

[0115] Set a lower limit constraint on power output to limit the actual power generation to not be lower than the lower limit;

[0116] Power output lower limit constraint:

[0117]

[0118] in, This is the lower limit of power output, typically The power should be 10% to 20% of the rated power to avoid a severe drop in power due to excessive yaw.

[0119] Ultimately, an optimization problem is formed with the yaw angle of each unit as the decision variable, the goal of maximizing the net power of the entire field, and the constraints of yaw angle range, yaw rate of change, and power output lower limit.

[0120] The optimization problem is expressed as:

[0121]

[0122]

[0123]

[0124]

[0125] This optimization problem aims to maximize power generation while also ensuring the safety of the unit load, making it suitable for closed-loop optimization control scenarios in actual wind farms.

[0126] S4. Using a collaborative optimization algorithm, the optimization problem is solved in a dynamic digital twin to obtain the optimal yaw angle combination for each wind turbine.

[0127] In this embodiment, S4.1, initialize the yaw angle of each wind turbine to the current actual value;

[0128] S4.2, Decompose the overall yaw optimization problem into... Each sub-problem corresponds to the yaw angle optimization of a single wind turbine. Coordination between turbines is achieved through the overall wind speed distribution. For a typhoon generator, its local objective function is the net power increment, expressed as:

[0129]

[0130] in, For the first The local objective function of a typhoon generator. For wind turbine The actual power generation capacity; For load penalty terms; This is the load penalty factor;

[0131] S4.3 In each optimization iteration, based on the current yaw angle combination of all wind turbines, the dynamic digital twin is invoked to calculate the overall wind speed distribution. For each wind turbine... Calculate the local objective function Its yaw angle The gradient;

[0132]

[0133] in, This reflects the marginal impact of yaw on power; This reflects the marginal impact of yaw changes on the load. It is a symbolic function; For local objective function Its yaw angle The gradient;

[0134] S4.4 Under the constraints of yaw angle range and yaw rate of change, the yaw angle is updated by gradient ascent method along the gradient direction to ensure iteration in the direction of increasing net power.

[0135]

[0136] in, For the first Yaw angle in the next iteration; For the first Yaw angle in the next iteration; The learning step size is set to a value ranging from 0.1° to 1.0°.

[0137] S4.5, if net power index If the change is less than a set threshold (e.g., 0.1%), or the number of iterations reaches the upper limit (e.g., 50 times), then the optimal yaw angle combination for all wind turbines is output. Otherwise, return to step S4.3 and continue iterating.

[0138] S5. Convert the optimal yaw angle combination into control commands, send them to each wind turbine to perform yaw operations, and simultaneously collect new operating data;

[0139] In this embodiment, the optimal yaw angle combination is received for each wind turbine. Calculate its current actual yaw angle With optimal yaw angle The difference is used to obtain the change in yaw angle. ;

[0140]

[0141] in, Indicates the unit The amount of yaw angle change to be performed, in degrees (°). For wind turbines The optimal yaw angle, For wind turbines The actual yaw angle;

[0142] Based on the maximum yaw rate of the crew (like ) and the current control cycle (e.g., 30 seconds), calculate the maximum permissible single change. ;

[0143]

[0144] in, This is the maximum yaw rate of the unit. To control the cycle, This represents the maximum permissible single change.

[0145] If the absolute value of the change in yaw angle does not exceed the maximum permissible single change, then the optimal yaw angle is directly set. Assign the value to the yaw command; otherwise, calculate the progressive command based on the maximum permissible single change amount and use it as the yaw command. ;

[0146]

[0147] in, For the unit Yaw command;

[0148] The yaw command is converted into a command format that the unit controller can recognize (such as register write value in Modbus protocol or CAN message), and a timestamp and unit number are added to form a control command package, which is then sent to the corresponding wind turbine.

[0149] In this embodiment, the actual yaw angle feedback value is monitored during the execution of control commands. ;

[0150] When the actual yaw angle feedback value With yaw command When the absolute value of the difference is less than the set error threshold (e.g., 0.5°), the yaw is determined to be in place, and the unit enters a stable operating state.

[0151] After the yaw is in place, a new round of unit operation data and flow field monitoring data are collected to form a new wind field status dataset.

[0152] S6. Based on the new operational data, the weight coefficients of the optimization problem are corrected to achieve closed-loop optimization;

[0153] In this embodiment, the actual loads (such as tower bending moment and blade root bending moment) and yaw angles of each wind turbine are extracted from the fresh wind farm state dataset, and the actual load changes are calculated. ;

[0154]

[0155] in, Before yaw Actual load monitoring values ​​of typhoon turbines For the first time after yaw Actual load monitoring values ​​of typhoon turbines; For the first The actual load variation of the typhoon generator;

[0156] Based on the actual load variations and yaw angles of all wind turbines, an adaptive threshold-based adjustment mechanism is used to penalize the yaw variation weights in the optimization problem. And the absolute value penalty weight of yaw Provide feedback and make corrections;

[0157] The corrected formula is as follows:

[0158]

[0159]

[0160] in, The average actual load change over the entire field is given by... We get the summation and average. The preset safety load variation threshold (e.g., 5% of the rated load); This represents the absolute value of the average yaw angle for the entire field. The preset safe yaw angle threshold (e.g., 15°); This is a correction factor, ranging from 0.1 to 0.3, to prevent sudden changes in weights; The penalty weight for the corrected yaw change. The penalty weight for the current yaw change. The weight of the correction for the absolute value of the yaw. The penalty weight is the absolute value of the current yaw.

[0161] Corrected yaw change penalty weight And the corrected absolute yaw penalty weight This will be used to construct the next round of optimization problems.

[0162] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A global yaw method for wind farms aimed at improving net power output, characterized in that: Includes the following steps: S1. Collect the unit operation data and flow field monitoring data in the wind farm, perform spatiotemporal alignment and normalization processing, and form a wind farm status dataset. S2. Construct a parameterized wake simulation model that integrates terrain features, and combine it with the wind field state dataset to dynamically calibrate the parameters of the parameterized wake simulation model using the gradient descent method, thereby establishing a dynamic digital twin of the wind farm. S3. Based on a dynamic digital twin, the yaw angle of each unit is used as the decision variable, the net power of the entire field is maximized as the optimization objective, and constraints are set to construct an optimization problem. S4. Using a collaborative optimization algorithm, the optimization problem is solved in a dynamic digital twin to obtain the optimal yaw angle combination for each wind turbine. S5. Convert the optimal yaw angle combination into control commands, send them to each wind turbine to perform yaw operations, and simultaneously collect new operating data; S6. Based on the new operating data, the weight coefficients of the optimization problem are corrected.

2. The global yaw method for wind farms aimed at improving net power output according to claim 1, characterized in that: In step S1, the specific steps involved in collecting wind farm turbine operation data and flow field monitoring data, and performing spatiotemporal alignment and normalization processing are as follows: Collect operating data of each unit in the wind farm, including wind speed, wind direction, power output, yaw angle, pitch angle, speed, and unit status signals; Simultaneously collect flow field monitoring data, including three-dimensional wind field information provided by lidar, ultrasonic anemometer, and meteorological tower; Clean the unit operation data and flow field monitoring data to remove outliers, missing values ​​and jump data; The data is unified to the same time reference by using time interpolation, and the location information of the wind farm coordinate system and the monitoring equipment is utilized. The data are subjected to min-max normalization, and the processed data is organized into a structured dataset according to the time series, forming a wind field status dataset.

3. The global yaw method for wind farms aimed at improving net power output according to claim 1, characterized in that: In step S2, the specific steps involved in constructing a parametric wake simulation model that incorporates terrain features are as follows: Based on the three-dimensional wind field information contained in the wind field state dataset, a wind field terrain feature matrix is ​​constructed using a gridded interpolation method. ; A parameterized wake simulation model is constructed based on the Bastankhah model. The input of the model is the wind field state dataset, and the output is the wind speed distribution across the entire field. Using a parametric wake simulation model, based on the unit thrust coefficient and upstream wind speed in the wind field state dataset, the wake wind speed distribution of a single wind turbine in flat terrain is calculated. Based on wake wind speed distribution and wind field topographic feature matrix The wake of each wind turbine is corrected for terrain features, and the wakes of all wind turbines are superimposed to obtain the wind speed distribution across the entire field.

4. The global yaw method for wind farms aimed at improving net power output according to claim 3, characterized in that: In step S2, the specific steps involved in dynamically calibrating the parameters of the parameterized wake simulation model using the gradient descent method to establish a dynamic digital twin of the wind farm are as follows: The wind speed prediction error is calculated by comparing the measured wind speed in the wind field state dataset with the predicted wind speed from the parametric wake simulation model at the same location. ; The gradient descent method is used, combined with wind speed prediction error. For wake attenuation coefficient Perform iterative corrections; Each time a new batch of wind field state data is received, a parameter update is performed, and the updated wake attenuation coefficient is... Substitute the parameterized wake simulation model and recalculate the overall wind speed distribution; The updated parametric wake simulation model is combined with the real-time wind farm status dataset to form a dynamic digital twin of the wind farm.

5. The global yaw method for wind farms aimed at improving net power output according to claim 1, characterized in that: In S3, the specific steps involved, with the yaw angle of each unit as the decision variable and the maximization of net power across the entire field as the optimization objective, are as follows: Based on the full-field wind speed distribution output by the dynamic digital twin, combined with the wind energy utilization coefficient, air density and rotor swept area in the wind turbine characteristic table, the theoretical power generation of each wind turbine is calculated. By introducing the yaw angle of each unit as an optimization variable, a cosine cubic function is used to reduce the theoretical power generation, and a yaw error compensation coefficient is added to calculate the corrected actual power generation. ; Based on the difference between the yaw angle of the previous moment and the current yaw angle, as well as the absolute value of the yaw angle, and combined with the penalty weight, a load penalty term is constructed. The net power index is calculated by subtracting the total load penalty cost from the sum of the actual power generation of all wind turbines. And the optimization objective is to maximize the net power of the entire field.

6. The global yaw method for wind farms aimed at improving net power output according to claim 5, characterized in that: In step S3, the specific steps involved in setting constraints and constructing the optimization problem are as follows: Set yaw angle range constraints to limit the yaw angle of each unit to the allowable range of the wind turbine design; Set a yaw rate constraint to limit the yaw angle change within adjacent control cycles from not exceeding the allowable rate of change. Set a lower limit constraint on power output to limit the actual power generation to not be lower than the lower limit; Ultimately, an optimization problem is formed with the yaw angle of each unit as the decision variable, the goal of maximizing the net power of the entire field, and the constraints of yaw angle range, yaw rate of change, and power output lower limit.

7. The global yaw method for wind farms aimed at improving net power output according to claim 1, characterized in that: In step S4, the specific steps involved in solving the optimization problem in the dynamic digital twin using a collaborative optimization algorithm are as follows: S4.1 Initialize the yaw angle of each wind turbine to the current actual value; S4.2, Decompose the overall yaw optimization problem into... There are several sub-problems, each corresponding to the yaw angle optimization of a single wind turbine. For the first sub-problem... For a typhoon generator, its local objective function is the net power increment; S4.3 In each optimization iteration, based on the current yaw angle combination of all wind turbines, the dynamic digital twin is invoked to calculate the overall wind speed distribution. For each wind turbine... Calculate the local objective function Its yaw angle The gradient; S4.4 Under the constraints of yaw angle range and yaw rate of change, the yaw angle is updated along the gradient direction using the gradient ascent method. S4.5, if net power index If the change is less than the set threshold, then output the optimal yaw angle combination for all wind turbines; otherwise, return to step S4.3 to continue iterating.

8. The global yaw method for wind farms aimed at improving net power output according to claim 1, characterized in that: In step S5, the specific steps involved in converting the optimal yaw angle combination into control commands are as follows: Receive the optimal yaw angle combination for each wind turbine. Calculate its current actual yaw angle With optimal yaw angle The difference is used to obtain the change in yaw angle. ; Based on the maximum yaw rate of the crew and the current control cycle Calculate the maximum permissible single change amount. ; If the absolute value of the change in yaw angle does not exceed the maximum permissible single change, then the optimal yaw angle is directly set. Assign the value to the yaw command; otherwise, calculate the progressive command based on the maximum permissible single change amount and use it as the yaw command. ; The yaw command is converted into a command format that the unit controller can recognize, and a timestamp and unit number are added to form a control command package, which is then sent to the corresponding wind turbine.

9. A global yaw method for wind farms aimed at improving net power output, as described in claim 8, characterized in that: In step S5, the specific steps involved in issuing yaw control commands to each wind turbine and simultaneously collecting new operational data are as follows: Monitor the actual yaw angle feedback value during control command execution. ; When the actual yaw angle feedback value With yaw command When the absolute value of the difference is less than the set error threshold, the yaw is determined to be in place, and the unit enters a stable operating state. After the yaw is in place, a new round of unit operation data and flow field monitoring data are collected to form a new wind field status dataset.

10. A global yaw method for wind farms aimed at improving net power output, as described in claim 1, characterized in that: In step S6, the specific steps involved in the feedback correction of the weight coefficients of the optimization problem are as follows: Extract the actual load and yaw angle of each wind turbine from the fresh wind farm state dataset, and calculate the actual load variation. ; Based on the actual load variations and yaw angles of all wind turbines, an adaptive threshold-based adjustment mechanism is used to penalize the yaw variation weights in the optimization problem. And the absolute value penalty weight of yaw Provide feedback and make corrections; Corrected yaw change penalty weight And the corrected absolute yaw penalty weight This will be used to construct the next round of optimization problems.

Images