Adaptive optimization control methods, devices, media, and equipment for the coupled operation state of wind turbine generator sets

Abstract

This disclosure relates to an adaptive optimization control method, device, medium, and equipment for the coupled operating state of a wind turbine generator set. The method includes acquiring wind turbine generator set operating state data within a sliding time window; updating the state of each coupled node in the operating coupled state map online based on the time-delayed response of upstream node state changes to downstream node load, vibration, and power fluctuations, as well as the downstream short-term damage increment. In the updated map, the dominant risk link is determined based on edge weights and downstream damage increments. According to the upstream and downstream nodes and risk types in the dominant risk link, load reduction targets are adaptively allocated through control weights. When a control action is predicted to amplify the risk of downstream nodes, the control constraint boundary is tightened. Under the control weights and constraint boundaries, the control parameters for the next control cycle are solved in a rolling manner and sent to the corresponding main control system for execution. This improves the operational stability, safety, and overall lifespan of the wind turbine generator set.

Description

Technical Field

[0001] This disclosure relates to the field of wind power generation control technology, specifically to an adaptive optimization control method, device, medium, and equipment for the overall operation coupling state of a wind turbine generator set. Background Technology

[0002] As wind turbines develop towards higher power, longer blades, taller towers, and more complex operating conditions, the coupling relationships between aerodynamic loads, structural loads, electrical power fluctuations, and actuator movements during overall turbine operation become more pronounced. Under conditions such as turbulent winds, yaw errors, gusts, wind shear, tower shadow effects, grid-friendly adjustments, and power-limited operation, the loads at the blade root, the forward and backward vibrations of the tower, the lateral vibrations of the tower, the torsional vibration of the drive train, the nacelle acceleration, the generator torque fluctuations, and the power output fluctuations are often not independent but exhibit significant time-delay transmission and coupling amplification relationships. For example, yaw errors not only reduce wind capture efficiency but may also cause asymmetrical loading on the rotor, further increasing the lateral load on the tower; while pitch control can reduce blade aerodynamic loads, excessively fast or frequent pitch control may increase fatigue in the pitch actuator and may affect tower vibration through changes in aerodynamic thrust; rapid adjustment of generator torque can improve power tracking but may also cause torsional vibration in the drive train and fluctuations in rotor speed. Therefore, controlling a single component or control objective is no longer sufficient to fully meet the comprehensive requirements of large wind turbine units for safety, stability, load reduction, and power generation efficiency.

[0003] In existing technologies, various control methods have been proposed for load control and power control of wind turbine generators. For example, the independent pitch control method for variable-speed, variable-pitch wind turbine generators obtains the coordinated pitch control quantity by detecting the generator output power and detecting the root load of the three blades and the rotor azimuth angle. The blade root load is converted into overturning load components and yaw load components through Park transformation, thereby achieving independent pitch control to reduce blade unbalanced loads. This scheme can effectively suppress blade cyclic loads and is a typical scheme in independent pitch load reduction control. However, it mainly focuses on the correspondence between blade root loads and pitch control, and the control object is still biased towards the blade load side. It lacks a unified modeling and identification mechanism for the dynamic coupling and transmission relationship between tower, drivetrain, yaw system, pitch actuator fatigue, and power fluctuations on the generator side. Especially under operating conditions where multiple risks exist simultaneously, this type of method struggles to determine whether to prioritize reducing blade load, tower load, drivetrain impact, or actuator action costs, and it also struggles to dynamically reallocate the control weights of different actuators according to the risk propagation direction.

[0004] For example, power control methods and systems based on yaw and pitch linkage control for load reduction determine the turbine power based on the current wind speed, turbine azimuth, pitch angle, torque, and rotational speed. They then determine the power state based on the relationship between turbine power and rated power, and generate corresponding control commands to improve the load issues of ultra-large wind turbine generators and extend their service life. This approach acknowledges the linkage between yaw control, pitch control, and power control, and can, to some extent, balance power control and load reduction control. However, this approach largely determines the control strategy based on power state, failing to further abstract the blades, tower, drive train, yaw system, pitch system, and generator side into online-updable coupled nodes. It also fails to construct an operational coupling state map based on the time-delayed response of upstream node state changes to downstream node load, vibration, or power fluctuations. Therefore, under complex wind conditions, when a control action might reduce the risk of one component but amplify the risk of another, existing yaw and pitch linkage control still lacks directional identification and constraint tightening mechanisms for the "risk link."

[0005] In addition, feedforward control and model predictive control are also used in wind turbine control. For example, US8025476B2 discloses a technical solution that uses an upstream wind condition measurement device to provide upstream wind condition measurements to the system controller and generates wind turbine operation control commands based on control algorithm parameters. The purpose is to enable the wind turbine to respond in advance to upcoming flow changes. Publicly available research also includes schemes that incorporate upstream wind speed information into model predictive control to achieve collective pitch and torque control, thereby considering state constraints such as speed during full-load operation. The above technologies can improve the control lag problem when wind speed disturbances arrive, but they focus on wind condition feedforward or constrained optimization and do not solve the problem of risk transmission path identification between multiple components of the turbine. They also do not disclose a whole-machine coupled control mechanism that uses the dominant risk link as the basis for adaptive adjustment of control weights and control constraint boundaries.

[0006] In summary, while existing wind turbine control technologies disclose solutions such as independent pitch control for load reduction, yaw-pitch linkage, feedforward disturbance compensation, and constraint optimization control, they still have the following shortcomings: First, most solutions focus on single or local targets such as blade load, power deviation, or yaw error, failing to describe the blades, tower, drive train, yaw system, pitch system, and generator side as a unified coupled whole-machine object. Second, existing solutions typically generate control commands based on thresholds, power states, or preset control laws, lacking dynamic risk link identification based on time-delay response, response intensity, and short-term damage increments. Third, in existing multi-actuator control, the division of labor among independent pitch control, collective pitch control, yaw correction, generator torque smoothing, and power limits largely relies on fixed strategies or empirical weights, making it difficult to achieve adaptive redistribution of control objectives under different dominant risk links. Fourth, when a certain control action may amplify the load or vibration of downstream components while reducing upstream risks, existing technologies lack predictive constraint tightening mechanisms for that control action. Summary of the Invention

[0007] The purpose of this disclosure is to provide an adaptive optimization control method, device, medium, and equipment for the coupled operating state of a wind turbine. By uniformly monitoring and coupling analysis of the operating states of the blades, tower, drive train, yaw system, pitch system, and generator side, the dominant risk links in the overall turbine operation process are identified online. Based on the risk propagation direction and risk type, independent pitch, collective pitch, yaw correction, generator torque smoothing, and power limit control strategies are adaptively adjusted. This solves the problem that existing wind turbine control methods cannot simultaneously take into account overall turbine stability, structural load reduction, power stability, and actuator lifespan.

[0008] To achieve the above objectives, in a first aspect, this disclosure provides an adaptive optimization control method for the coupled operating state of a wind turbine generator set, the method comprising: The operating status data of the wind turbine is acquired within a sliding time window; Based on the time-delay response of the upstream node's state change to the downstream node's load, vibration, and power fluctuations, as well as the downstream short-term damage increment, the state of each coupled node in the operational coupling state map is updated online. The map includes a node set, an edge set, and an edge weight set. The edge weights are calculated by weighting the response intensity, time delay, and damage increment. The coupled nodes are obtained by dividing the wind turbine's blades, tower, transmission chain, yaw system, pitch system, and power generation side into regions. In the running coupling state graph after the updated state of the coupled node, the dominant risk link is determined based on the edge weight and the downstream damage increment. The dominant risk link is the directed link in the running coupling state graph where the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. According to the upstream and downstream nodes and risk types in the dominant risk chain, the load reduction target is adaptively allocated to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control by controlling the weights. During the adaptive control process, when it is predicted that the adaptive control action will amplify the risk of the downstream node, the control constraint boundary is tightened. Under the control weights and the constraint boundaries, the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle are solved in a rolling manner to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact, and actuator action cost in the prediction time domain. The control parameters corresponding to each main control system are obtained, and the control parameters are sent to the corresponding main control system for execution.

[0009] Optionally, the weight of any edge in the running coupling state graph is calculated as follows: ; in, For the first The coupling node points to the first Edge weights of each coupled node For the first A state change at the first coupled node causes the... The normalized response intensity of the responses of each coupled node. For the first The coupling node to the first The time-delay response factor of each coupled node For the first Normalized value of short-time damage increment of each coupled node The actuator motion propagation factor is used to characterize the amplification or suppression effect of pitch, yaw, or generator torque actions on downstream coupled node loads, vibrations, or power fluctuations. , , and These are the corresponding weighting coefficients, and all are greater than 0.

[0010] Optionally, the time-delay response factor The maximum correlation response is determined by comparing the state change sequence of the upstream node with the response sequence of the downstream node within a preset time delay search interval, where the preset time delay search interval is from 0.1 seconds to 10 seconds. Specifically, when the time delay corresponding to the maximum correlation response is within the effective response range determined by the current speed, impeller azimuth angle, and control cycle of the wind turbine, the corresponding directed edge is retained; otherwise, the weight of the corresponding directed edge is decayed or set to zero.

[0011] Optionally, the determination of the dominant risk link includes: persistence determination and direction determination; The persistence determination is that the same directed link satisfies the condition that the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold within at least N consecutive control cycles. The directionality determination is that the state change of the upstream node occurs before the load, vibration or power fluctuation change of the downstream node, and the time delay between the two is within a preset time delay range, thereby eliminating pseudo-coupled links caused by synchronization disturbances or measurement noise.

[0012] Optionally, the risk types include at least: blade fatigue risk, tower vibration risk, transmission chain torsional vibration risk, yaw error amplification risk, pitch actuator fatigue risk, and power fluctuation risk; Specifically, the blade fatigue risk is determined based on the periodic and non-periodic components of the load at the root of the three blades; the tower vibration risk is determined based on the tower vibration amplitude, vibration frequency, and damping variation trend; the transmission chain torsional vibration risk is determined based on generator torque fluctuation and impeller speed fluctuation; and the pitch actuator fatigue risk is determined based on pitch rate, pitch acceleration, and number of actions per unit time.

[0013] Optionally, the step of adaptively allocating load reduction targets to independent pitch control, collective pitch control, yaw correction, generator torque smoothing, and power limit control by controlling weights according to the upstream and downstream nodes and risk types in the dominant risk chain includes: When the dominant risk link is from the blade node to the tower node, increase the control weight of independent pitch and tower damping, and limit the yaw correction amount that increases the tower lateral load. When the dominant risk link is a yaw system node pointing to a tower node, reduce the frequency of yaw actions and increase the control weight of generator torque smoothing and power limit. When the dominant risk link is a drive chain node pointing to a generator-side node, the constraint strength of the generator torque change rate is increased, and the abrupt change amplitude of the collective pitch control is reduced. When the dominant risk link is a pitch system node pointing to a blade node, the rate of change of independent pitch angle increment is limited, and part of the load reduction target is transferred to power limit control.

[0014] Optionally, the control constraint boundaries include: the range of independent pitch angle increments, the range of collective pitch angle increments, the upper limit of pitch rate, the upper limit of pitch acceleration, the range of yaw correction, the minimum interval of yaw action, the range of generator torque increments, the upper limit of generator torque change rate, and the reduction range of power limit. In the adaptive control process, when it is predicted that the adaptive control action may amplify the risk of downstream nodes, tightening the control constraint boundaries includes: During the adaptive control process, when it is predicted that the adaptive control action will increase the coupling risk value of the downstream node, the control constraint boundary corresponding to the adaptive control action is tightened, and the control constraint boundary corresponding to other adaptive control actions with risk suppression effect is relaxed.

[0015] Optionally, the rolling solution of the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle includes: A constraint optimization method combining prediction time domain and control time domain is adopted to solve the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment and power limit for the next control cycle in a rolling manner. The prediction time domain is from 0.5 seconds to 30 seconds, and the control time domain is one or more master control cycles. The comprehensive evaluation value is determined as follows: ; in, For comprehensive evaluation, As a power deviation evaluation metric, This is a metric for evaluating the incremental damage to the leaves. This is the evaluation metric for incremental tower damage. This is a quantity used for evaluating the impact on the transmission chain. This is a metric for evaluating the cost of yaw maneuvers. The cost evaluation metric for pitch and torque execution actions. , , , , and The weighting coefficients for the corresponding evaluation quantities are adjusted online according to the risk type of the dominant risk link.

[0016] Optionally, the method further includes: When a limited pitch rate, limited yaw action, lag in generator torque response, or deviation from the power limit command execution exceeds the corresponding threshold, the corresponding actuator will be marked as a capacity degradation actuator. Reduce the control weight corresponding to the degraded actuator and redistribute the load reduction target undertaken by the actuator to the non-degraded actuator.

[0017] Secondly, this disclosure provides an adaptive optimization control device for the overall operating coupling state of a wind turbine generator set, comprising: The data acquisition module is configured to acquire the operating status data of the wind turbine within a sliding time window; The coupling graph construction module is configured to update the state of each coupled node in the running coupling state graph online based on the time-delay response of the upstream node to the load, vibration and power fluctuation of the downstream node and the short-term damage increment of the downstream node, according to the state change of the upstream node within the sliding time window. The graph includes a set of nodes, a set of edges and a set of edge weights. The edge weights are calculated by weighting the response intensity, time delay and damage increment. The coupled nodes are obtained by dividing the wind turbine blades, tower, transmission chain, yaw system, pitch system and power generation side into regions. The risk link identification module is configured to determine the dominant risk link in the running coupling state graph after the updated state of the coupled node, based on the edge weight and the downstream damage increment. The dominant risk link is a directed link in the running coupling state graph in which the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. The control weight allocation module is configured to adaptively allocate load reduction targets to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control according to the upstream nodes, downstream nodes and risk types in the dominant risk link. During the adaptive control process, when it is predicted that the adaptive control action will amplify the risk of the downstream node, the control constraint boundary is tightened. The rolling optimization solution module is configured to, under the control weights and the constraint boundaries, continuously solve for the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle, so as to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact, and actuator action cost in the prediction time domain, obtain the control parameters corresponding to each main control system, and send the control parameters to the corresponding main control system for execution.

[0018] Thirdly, this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the first aspects.

[0019] Fourthly, this disclosure provides an electronic device, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method of any one of the first aspects.

[0020] By constructing an operational coupling state graph, the dynamic correlation between blade load, tower vibration, drivetrain torsional vibration, yaw error, pitch control status, and power fluctuations on the generator side of the wind turbine can be expressed online. This allows the main control system to determine the transmission path of risks between components at the whole-machine level, rather than relying solely on a single wind speed, power deviation, or local load threshold for control. By identifying the dominant risk link based on edge weights, time-delay response, and downstream short-term damage increments, it can distinguish between blade fatigue risk, tower vibration risk, drivetrain impact risk, yaw error amplification risk, and actuator fatigue risk, and adaptively allocate the priority of different control measures accordingly. When blade load is transmitted to tower vibration, the control weights related to independent pitch control and tower damping can be increased; when yaw action may amplify the lateral load on the tower, yaw correction constraints can be tightened and torque smoothing or power limits can be used for load reduction; when the pitch actuator operates frequently or its capacity degrades, some load reduction targets can be transferred to other actuators. This reduces control conflicts and excessive actuator movements caused by a single control strategy, ensuring stable power output while reducing short-term impacts and fatigue damage to blades, towers, and transmission chains. It also improves the operational stability, safety, and overall lifespan of wind turbines under turbulent, gusty, yaw error, power limitation, and complex coupling conditions.

[0021] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall structure of an adaptive optimization control system for the overall operation coupling state of a wind turbine generator set, according to an exemplary embodiment of this disclosure.

[0023] Figure 2 This is a flowchart illustrating an adaptive optimization control method for the overall operating coupling state of a wind turbine generator set according to an exemplary embodiment of this disclosure.

[0024] Figure 3 This is a schematic diagram illustrating the division of coupling nodes and the construction of the operational coupling state map according to an exemplary embodiment of this disclosure.

[0025] Figure 4 This is a schematic diagram illustrating a dominant risk link identification process according to an exemplary embodiment of this disclosure.

[0026] Figure 5 This is a schematic diagram illustrating the dynamic adjustment of control constraint boundaries according to an exemplary embodiment of the present disclosure.

[0027] Figure 6 This is a schematic diagram of a rolling optimization solution process according to an exemplary embodiment of the present disclosure.

[0028] Figure 7 This is a schematic diagram illustrating an actuator capability degradation identification and load reduction target reassignment process according to an exemplary embodiment of the present disclosure.

[0029] Figure 8 This is a block diagram of an adaptive optimization control device for the overall operating coupling state of a wind turbine generator set, as shown in an exemplary embodiment of the present disclosure.

[0030] Figure 9 This is a block diagram illustrating an electronic device according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0031] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0032] This disclosure provides an adaptive optimization control method, device, medium, and equipment for the coupled operating state of a wind turbine generator set. First, see... Figure 1 As shown, the adaptive optimization method for the coupled operating state of a wind turbine generator disclosed herein can be applied to an adaptive optimization control system for the coupled operating state of a wind turbine generator. This control system may include a data acquisition module, a coupling graph construction module, a risk link identification module, a control weight allocation module, a rolling optimization solution module, an actuator degradation identification module, and a command output module. Each module can be installed in the wind turbine generator's main control system, or it can be implemented collaboratively by the main control system, pitch controller, yaw controller, converter controller, and the wind farm monitoring system. For already operational turbine generators, this method can be implemented by adding software functional modules to the original main control system; for newly designed turbine generators, this method can be embedded into the main control strategy platform.

[0033] Figure 2 This is a flowchart illustrating an adaptive optimization control method for the overall operating coupling state of a wind turbine generator set, according to an exemplary embodiment of this disclosure. Figure 2 As shown, the method may include the following steps: Step S11: Obtain the operating status data of the wind turbine within the sliding time window; The operational status data of wind turbines refers to the comprehensive operational status formed by the interaction of factors such as aerodynamic loads, structural vibrations, transmission shocks, control actions, and power fluctuations between the blades, tower, drive train, yaw system, pitch system, and generator side during wind turbine operation. For example, increased yaw error may lead to uneven inflow to the rotor, further causing unbalanced loads at the blade roots and increased lateral vibration of the tower; while independent pitch control can reduce cyclic loads on the blades, excessively frequent operation may also increase the fatigue risk of the pitch actuator; sudden changes in generator torque may cause torsional vibration in the drive train, which in turn affects the aerodynamic loads on the blades through rotor speed fluctuations. Operational status data may include incoming wind speed data, wind direction deviation data, rotor speed data, power generation data, generator torque data, root load data of the three blades, tower vibration data, nacelle acceleration data, yaw angle data, pitch angle data, pitch rate data, and yaw action status data.

[0034] In this embodiment of the disclosure, combined with Figure 1 As shown, the data acquisition module is used to acquire the operating status data of the wind turbine within a sliding time window. The operating status data includes incoming wind speed, wind direction deviation, rotor speed, power generation, generator torque, root load of the three blades, tower vibration, nacelle acceleration, yaw angle, pitch angle, pitch rate, and yaw action status. Incoming wind speed can be obtained from the nacelle anemometer, lidar, or wind field estimation module; wind direction deviation can be calculated from the anemometer, nacelle direction, and yaw angle; rotor speed and generator torque can be read from the converter or main control system; blade root load can be obtained from blade root strain gauges, fiber optic sensors, load estimation models, or fusion estimation algorithms; tower vibration can be obtained from tower or nacelle accelerometers; pitch angle, pitch rate, and yaw status can be read from the pitch controller and yaw controller.

[0035] In one embodiment, the sliding time window can be configured to cover multiple master control cycles. Assuming the master control cycle is in the millisecond to hundreds of milliseconds range, the sliding time window can cover several control cycles to ensure that the time lag relationship between upstream node changes and downstream node responses can be observed. The sliding time window can be of a fixed length or adaptively adjusted based on wind speed change rate, turbulence intensity, or unit operating mode. When wind conditions are stable, the sliding time window can be appropriately extended to improve statistical stability; when gusts, yaw actions, or power limiting switching occur, the sliding time window can be appropriately shortened to improve response speed.

[0036] Within the sliding time window, the data acquisition module obtains the following data: Incoming wind speed data is used to characterize the external wind conditions acting on the rotor. The incoming wind speed can be measured by the nacelle anemometer, the equivalent inflow wind speed after filtering correction, or the forward incoming wind speed measured by lidar. For units using lidar, the incoming wind speed can also be used as a feedforward control input to predict blade load and power changes in advance.

[0037] Wind direction deviation data is used to characterize the angular deviation between the current impeller direction and the incoming flow direction. Wind direction deviation can be calculated based on the direction measured by the anemometer and the nacelle yaw angle, or it can be estimated through the imbalance of loads on the left and right blades, the nacelle lateral acceleration, or power deviation.

[0038] Rotor speed and power generation data are used to characterize the unit's operating conditions and power point tracking status. When the rotor speed deviates from the target value, it is usually necessary to adjust it through collective pitch control and generator torque control; when the power fluctuation is too large, torque smoothing, pitch smoothing, or power limit control need to be considered.

[0039] Generator torque data is used to characterize the drive train and generator-side control states. Rapid changes in generator torque can cause torsional vibration in the drive train, which in turn can affect the impeller speed and blade aerodynamic loads. Therefore, generator torque is both a control variable and an important input for coupled state analysis.

[0040] The root load data of the three blades is used to characterize the load distribution and periodic load state of the blades. The root loads of the three blades can be denoted as the root load of the first blade, the root load of the second blade, and the root load of the third blade, respectively. By comparing the root loads of the three blades, the impeller load imbalance and periodic load components can be identified.

[0041] Tower vibration and nacelle acceleration data are used to characterize the tower structure response. Tower vibration can include forward and backward tower vibration and lateral tower vibration. Forward and backward tower vibration is typically related to thrust fluctuations in the wind direction, rotor thrust variations, and collective pitch control; lateral tower vibration is typically related to yaw error, rotor unbalanced loads, yaw action, and lateral wind loads. Nacelle acceleration can be used as a direct measurement of the tower top response.

[0042] Yaw angle and yaw action status data are used to characterize the node status of the yaw system. The yaw action status can include the yaw motor start / stop status, yaw direction, yaw speed, yaw braking status, and yaw action duration. Since yaw action may change the inflow direction and affect the lateral load on the tower, this disclosure includes the yaw system as an independent coupled node in the diagram.

[0043] Pitch angle and pitch rate data are used to characterize the node status of the pitch system. The pitch angle may include the actual pitch angle of the three blades and the collective pitch angle; the pitch rate may include the pitch rate of each blade and its rate of change. If independent pitch actions are too frequent, it may cause fatigue risk to the pitch actuator. Therefore, this disclosure not only uses pitch as a control actuator, but also incorporates the pitch system status into risk link identification.

[0044] To ensure temporal consistency among data sources, the data acquisition module can align timestamps for data from different sources. For data with different sampling frequencies, resampling, interpolation, or maintaining the previous valid value can be used to convert them to a unified control period. For data with obvious anomalies, amplitude limiting, filtering, or anomaly removal can be performed. For load-related data, low-pass filtering can be used to extract trend components, or band-pass filtering can be used to extract periodic components. This data preprocessing does not change the core idea of ​​this disclosure; its purpose is to improve the stability of map construction and risk identification.

[0045] Step S12: Based on the time-delay response of the upstream node's state change to the downstream node's load, vibration, and power fluctuations, as well as the downstream short-term damage increment, update the state of each coupled node in the running coupled state map online. The map includes a node set, an edge set, and an edge weight set. The edge weights are calculated by weighting the response intensity, time delay, and damage increment. The coupled nodes are obtained by dividing the wind turbine's blades, tower, transmission chain, yaw system, pitch system, and power generation side into regions. In this context, a coupling node refers to a component or functional part of the generator set that is abstracted as a state unit in the overall coupled state analysis. In this embodiment, the coupling nodes include at least blade nodes, tower nodes, transmission chain nodes, yaw system nodes, pitch system nodes, and generator-side nodes. Blade nodes can be characterized by the root loads of the three blades, blade azimuth angle, and blade load imbalance; tower nodes can be characterized by tower forward / backward vibration, tower lateral vibration, tower top acceleration, and tower bending moment; transmission chain nodes can be characterized by generator torque, impeller speed fluctuation, and drive shaft torsional vibration amplitude; yaw system nodes can be characterized by wind direction deviation, yaw angle, yaw action state, and yaw braking state; pitch system nodes can be characterized by pitch angle, pitch rate, pitch acceleration, and number of pitch actions; and generator-side nodes can be characterized by generator power, power fluctuation, grid-connected active power command, and torque control state.

[0046] The operational coupling state graph is a data structure consisting of coupled nodes, directed edges between nodes, and edge weights. Nodes represent components of the unit's operating state, directed edges represent the dynamic impact of upstream node state changes on downstream node loads, vibrations, or power fluctuations, and edge weights represent the strength of this impact. The operational coupling state graph can be updated online in each or multiple control cycles to reflect changes in coupling relationships caused by wind conditions, unit operating conditions, and actuator actions.

[0047] Among them, time-delay response refers to the phenomenon that after the state of the upstream coupled node changes, the downstream coupled node does not respond immediately, but rather exhibits changes in load, vibration, or power fluctuations after a certain time delay. For example, after yaw angle adjustment, the impact of the change in impeller inflow direction on blade load and tower lateral vibration may be reflected after several control cycles; the impact of generator torque changes on transmission chain torsional vibration and impeller speed fluctuations also has a dynamic response time.

[0048] The short-time damage increment refers to the change in damage at a coupled node due to changes in load, vibration, or impact within a sliding time window. The short-time damage increment can be calculated using load amplitude, cycle count, vibration amplitude, impact intensity, or equivalent fatigue index, and is used to determine whether the risk at downstream nodes increases with changes at upstream nodes.

[0049] See Figure 3 As shown, the coupling map construction device is used to... Figure 3 As shown, the blades, tower, drive train, yaw system, pitch system, and generator side are divided into coupling nodes, and the directed influence relationships between nodes are identified within a sliding time window. This module does not employ fixed empirical rules, but instead calculates the intensity of the impact of upstream node state changes on downstream node responses online based on current wind conditions and unit status. Thus, the coupling map can be updated according to the operating status under conditions such as low wind speed, near rated wind speed, high wind speed power-limited operation, yaw operation, gust conditions, and enhanced turbulence conditions.

[0050] Combination Figure 3 As shown, the set of coupled nodes can be represented as: ; in, For the set of coupled nodes, For the leaf node, For tower nodes, For transmission chain nodes, As a yaw system node, For the pitch system node, This is the node on the power generation side.

[0051] The running coupling state diagram can be represented as: ; in, To run the coupled state map, For the set of coupled nodes, Let be a set of directed edges. It is the set of directed edge weights.

[0052] For any two coupled nodes and If detected within the sliding time window The state change precedes If load, vibration, or power fluctuations occur, and a time-delay response relationship exists between them that satisfies preset conditions, then a system is established from... point to The directed edges. The edge weight of the directed edge is used to characterize... right The intensity of the impact.

[0053] Step S13: In the running coupling state graph after the updated state of the coupled node, the dominant risk link is determined based on the edge weight and the downstream damage increment. The dominant risk link is the directed link in the running coupling state graph where the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. The dominant risk link refers to a directed link in the operational coupling state graph that meets preset judgment conditions. This directed link has edge weights continuously exceeding edge weight thresholds, and the short-term damage increment of downstream nodes exceeds the risk threshold. Simultaneously, changes in upstream nodes precede the responses of downstream nodes. The dominant risk link represents the risk propagation path that most needs to be monitored and suppressed by the control system under the current operating condition.

[0054] In this embodiment of the disclosure, combined with Figure 1 As shown, the risk link identification module is used to combine Figure 4 As shown, the dominant risk link is determined based on edge weights and downstream short-term damage increments. Risk link identification is not simply about whether a load exceeds a threshold, but rather whether a change in the state of an upstream node continuously increases the risk at downstream nodes within a preset time delay range. For example, if a change in the movement of the yaw system node continuously causes an increase in the lateral vibration of the tower, it can be identified as a dominant risk link "from the yaw system node to the tower node"; similarly, if increased torsional vibration in the drive train causes an increase in power generation fluctuations, it can be identified as a dominant risk link "from the drive train node to the power generation node".

[0055] Step S14: According to the upstream node, downstream node and risk type in the dominant risk link, the load reduction target is adaptively allocated to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control by controlling the weight. During the adaptive control process, when it is predicted that the adaptive control action will amplify the risk of the downstream node, the control constraint boundary is tightened. Here, the load reduction target refers to the control objective set to reduce the load on the blades, tower, drive train, or actuator. This disclosure does not assign the load reduction target to a specific actuator, but rather adaptively allocates the load reduction target to independent pitch control, collective pitch control, yaw correction, generator torque smoothing, and power limit control based on the upstream and downstream nodes and risk type of the dominant risk link.

[0056] Among them, the control constraint boundary refers to the boundary conditions that the control command must meet during the solution and execution process, including the range of independent pitch angle increment, the range of collective pitch angle increment, the upper limit of pitch rate, the upper limit of pitch acceleration, the range of yaw correction, the minimum interval of yaw action, the upper limit of generator torque change rate, the reduction range of power limit, the upper limit of speed, the load safety threshold, and the vibration safety threshold, etc.

[0057] In this embodiment of the disclosure, combined with Figure 1 As shown, the control weight allocation module is used to... Figure 5 As shown, the dominant risk links are mapped to the weights and constraint boundaries of different control actuators. When the risk link from the blade node to the tower node is dominant, the control system increases the weight of independent pitch control and the weight of tower damping control. When the risk link from the yaw system node to the tower node is dominant, the control system reduces the frequency of yaw actions to prevent yaw correction from further amplifying the lateral load on the tower. When the risk link from the drivetrain node to the generator node is dominant, the generator torque smoothing weight is increased and the torque change rate is limited. When the risk link from the pitch system node to the blade node is dominant, the rate of change of independent pitch angle increment is limited, and some load reduction targets are transferred to power limit control or torque control.

[0058] Step S15: Under the control weight and the constraint boundary, the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment and power limit of the next control cycle are solved in a rolling manner to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact and actuator action cost in the prediction time domain, so as to obtain the control parameters corresponding to each main control system, and the control parameters are sent to the corresponding main control system for execution.

[0059] In this embodiment of the disclosure, combined with Figure 1 As shown, the rolling optimization solution module is used to solve according to Figure 6The flowchart illustrates how to solve for the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle within both the prediction and control time domains. This module can be implemented using model predictive control, quadratic programming, sequence optimization, or fast optimization algorithms under rule constraints. Considering that the main control cycle of wind turbines is typically short, simplified prediction models and constraint lookup tables can be pre-established to reduce online computation.

[0060] Furthermore, corresponding control commands can be generated based on the control parameters. The command output module then distributes the control commands obtained from the rolling optimization solution to the actuators of the corresponding main control system. Specifically, this includes outputting independent pitch angle increments and collective pitch angle increments to the pitch controller, yaw corrections to the yaw controller, generator torque increments to the converter controller, and power limits to the power control unit. The command output module can also upload control weights, constraint boundaries, dominant risk links, and actuator degradation states to the wind farm monitoring system, facilitating maintenance personnel to view the unit's operating status.

[0061] The aforementioned technical solution constructs an operational coupling state map to express the dynamic correlation between blade loads, tower vibration, drivetrain torsional vibration, yaw error, pitch control status, and power fluctuations on the generator side in a wind turbine unit online. This allows the main control system to determine the transmission path of risks between components at the whole-machine level, rather than relying solely on a single wind speed, power deviation, or local load threshold for control. By identifying the dominant risk link based on edge weights, time-delay response, and downstream short-term damage increments, it can distinguish between blade fatigue risk, tower vibration risk, drivetrain impact risk, yaw error amplification risk, and actuator fatigue risk, and adaptively allocate the priority of different control measures accordingly. When blade loads are transmitted to tower vibration, the control weights related to independent pitch control and tower damping can be increased; when yaw action may amplify the lateral load on the tower, yaw correction constraints can be tightened and torque smoothing or power limits can be used for load reduction; when the pitch actuator operates frequently or its capacity degrades, some load reduction targets can be transferred to other actuators. This reduces control conflicts and excessive actuator movements caused by a single control strategy, ensuring stable power output while reducing short-term impacts and fatigue damage to blades, towers, and transmission chains. It also improves the operational stability, safety, and overall lifespan of wind turbines under turbulent, gusty, yaw error, power limitation, and complex coupling conditions.

[0062] Optionally, the weight of any edge in the running coupling state graph is calculated as follows: ; in, For the first The coupling node points to the first Edge weights of each coupled node For the first A state change at the first coupled node causes the... The normalized response intensity of the responses of each coupled node. For the first The coupling node to the first The time-delay response factor of each coupled node For the first Normalized value of short-time damage increment of each coupled node The actuator motion propagation factor is used to characterize the amplification or suppression effect of pitch, yaw, or generator torque actions on downstream coupled node loads, vibrations, or power fluctuations. , , and These are the corresponding weighting coefficients, and all are greater than 0.

[0063] Among them, response strength The correlation between the state change sequence of the upstream node and the response sequence of the downstream node within the sliding time window can be determined by the normalized correlation, regression response gain, or perturbation response amplitude.

[0064] Time-delay response factor This is used to reflect the temporal relationship between changes in upstream nodes and responses in downstream nodes. If the time delay corresponding to the maximum response is within a preset effective response interval, the directed edge is considered to have directional significance. If the two change almost synchronously, or the downstream node changes before the upstream node, the directed edge may be a pseudo-coupling or a synchronous change caused by a common disturbance, and its edge weight should be reduced or the directed edge should not be established. This can avoid misidentifying the synchronous response caused by gusts acting on both the blades and the tower as a risk propagation from the blade node to the tower node.

[0065] Short-term damage increment This is used to reflect the actual risk changes at downstream nodes. Even if there is a strong response relationship between two nodes, if the damage at the downstream node does not increase, the link should not be identified as a dominant risk link. For example, a yaw correction may cause a short-term change in the lateral acceleration of the tower, but if this change does not cause the short-term damage increment of the tower to exceed the risk threshold, then there is no need to increase the risk level of the link. The short-term damage increment can be calculated based on the load amplitude, the root mean square value of vibration, the fatigue equivalent load, the cycle count results, or the equivalent damage model.

[0066] Actuator motion propagation factor This factor characterizes the amplification or suppression effect of pitch, yaw, or generator torque actions on downstream coupled node loads, vibrations, or power fluctuations. This factor is one of the key technical features that distinguishes this disclosure from ordinary coupled state identification. Traditional graphs or correlation analyses often only consider the correlation between state variables, while this disclosure includes actuator actions in the edge weights, enabling the system to identify whether "the control action itself is becoming a cause of risk propagation." For example, independent pitch actions may reduce blade cyclic loads, but if pitch actions are too rapid or frequent, they may also increase the fatigue risk of the pitch system; yaw correction may improve wind efficiency, but frequent yaw when tower lateral vibration is high may amplify tower lateral risk.

[0067] Optionally, the time-delay response factor The maximum correlation response is determined by comparing the state change sequence of the upstream node with the response sequence of the downstream node within a preset time delay search interval, where the preset time delay search interval is from 0.1 seconds to 10 seconds. Specifically, when the time delay corresponding to the maximum correlation response is within the effective response range determined by the current speed, impeller azimuth angle, and control cycle of the wind turbine, the corresponding directed edge is retained; otherwise, the weight of the corresponding directed edge is decayed or set to zero.

[0068] The preset time delay search range can be determined based on the unit speed, impeller azimuth angle, main control cycle, and component response characteristics. For example, a shorter time delay search range can be set for the response between blade load and tower vibration; a longer time delay search range can be set for the effect of yaw action on the lateral load of the tower; and the time delay search range can be set based on the natural frequency of the transmission chain and the control cycle for the effect of generator torque changes on the torsional vibration of the transmission chain.

[0069] In this embodiment, when the time delay corresponding to the maximum correlation response is within the effective response interval, the corresponding directed edge is retained; when the time delay corresponding to the maximum correlation response is not within the effective response interval, the weight of the corresponding directed edge is decayed or set to zero. In this way, the operational coupling state graph constructed in this disclosure can reflect the real dynamic response between nodes and suppress pseudo-coupling caused by measurement noise and common disturbances.

[0070] Optionally, the determination of the dominant risk link includes: persistence determination and direction determination; The persistence determination is that the same directed link satisfies the condition that the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold within at least N consecutive control cycles. In this embodiment of the disclosure, persistence determination refers to the condition that the same directed link satisfies the following conditions for at least three consecutive control cycles: the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. The purpose of persistence determination is to eliminate sporadic noise and transient disturbances. For example, the appearance of a spike in the lateral vibration of the tower during a certain control cycle does not necessarily indicate that the yaw system node forms a dominant risk link to the tower node; only when the yaw system state change corresponds to an increase in the lateral damage increment of the tower in multiple consecutive control cycles is the risk link considered to have control significance.

[0071] The directionality determination is that the state change of the upstream node occurs before the load, vibration or power fluctuation change of the downstream node, and the time delay between the two is within a preset time delay range, thereby eliminating pseudo-coupled links caused by synchronization disturbances or measurement noise.

[0072] In this embodiment, directionality determination refers to the situation where changes in the state of an upstream node occur before changes in the load, vibration, or power fluctuations of a downstream node, and the time lag between the two is within a preset time lag range. The purpose of directionality determination is to identify the direction of risk propagation. For example, changes in blade load and tower vibration may be simultaneously affected by gusts of wind. Without directionality determination, the system may mistakenly assume that changes in the blade nodes caused changes in the tower nodes. By determining whether the upstream change occurs first, and combining this with the time lag range, the direction of the directed link can be determined more accurately.

[0073] In this embodiment of the disclosure, the dominant risk chain can be represented as: ; in, For the first The coupling node points to the first The dominant risk link of each coupled node; For the first The coupling node points to the first Directed edges of a coupled node; Let the weight of the directed edge be denoted as . The edge weight threshold; For the first Short-term damage increment of each coupled node; Risk threshold; For the results of persistence and directionality determination, when both persistence and directionality determinations are satisfied, The value is 1, otherwise it is 0.

[0074] Optionally, the risk types include at least: blade fatigue risk, tower vibration risk, transmission chain torsional vibration risk, yaw error amplification risk, pitch actuator fatigue risk, and power fluctuation risk; Specifically, the blade fatigue risk is determined based on the periodic and non-periodic components of the load at the root of the three blades; the tower vibration risk is determined based on the tower vibration amplitude, vibration frequency, and damping variation trend; the transmission chain torsional vibration risk is determined based on generator torque fluctuation and impeller speed fluctuation; and the pitch actuator fatigue risk is determined based on pitch rate, pitch acceleration, and number of actions per unit time.

[0075] For example, under highly turbulent wind conditions, a significant imbalance in the load at the roots of the three blades occurs, while the lateral vibration of the tower continues to increase after several control cycles. If the operational coupling state diagram shows that the edge weights pointing from the blade nodes to the tower nodes continuously exceed the threshold, and the short-term damage increment of the tower nodes exceeds the risk threshold, then the directed link from the blade nodes to the tower nodes is identified as the dominant risk link. In this case, the control system should subsequently increase the weight of independent pitch control and limit yaw actions that may increase the lateral load on the tower.

[0076] For example, during frequent yaw system operations, although wind direction deviation may decrease, tower lateral vibration and nacelle lateral acceleration continue to increase. If the edge weight of the yaw system node pointing to the tower node continuously exceeds the threshold, and the short-term damage increment of the tower node exceeds the risk threshold, then this link is identified as the dominant risk link. In this case, the control system should not continue to blindly increase yaw correction, but should reduce the frequency of yaw actions and increase the weight of torque smoothing or power limit control to avoid further expansion of tower lateral risk.

[0077] Optionally, in step S14, the step of adaptively allocating the load reduction target to independent pitch control, collective pitch control, yaw correction, generator torque smoothing, and power limit control according to the upstream and downstream nodes and risk types in the dominant risk link by controlling the weights includes: When the dominant risk link is from the blade node to the tower node, increase the control weight of independent pitch and tower damping, and limit the yaw correction amount that increases the tower lateral load. When the dominant risk link is a yaw system node pointing to a tower node, reduce the frequency of yaw actions and increase the control weight of generator torque smoothing and power limit. When the dominant risk link is a drive chain node pointing to a generator-side node, the constraint strength of the generator torque change rate is increased, and the abrupt change amplitude of the collective pitch control is reduced. When the dominant risk link is a pitch system node pointing to a blade node, the rate of change of independent pitch angle increment is limited, and part of the load reduction target is transferred to power limit control.

[0078] In this embodiment of the disclosure, conventional control methods often use pitch control, yaw control, and torque control to serve different objectives. Pitch control is mainly used for speed and power regulation, with independent pitch control used for blade load reduction; yaw control is mainly used for wind control to improve wind energy capture; and torque control is mainly used for power point tracking or drivetrain control. Conflicts may exist between different control channels. For example, yaw correction can improve power generation efficiency, but may increase tower lateral load under certain operating conditions; independent pitch control can reduce blade load, but excessive movement can increase fatigue in the pitch mechanism; power limits can reduce load, but may result in power generation loss. This disclosure, by driving control weight allocation through a dominant risk link, enables dynamic coordination between different objectives.

[0079] In one embodiment, the set of control weights can be represented as: ; in, To control the set of weights; Independent pitch control weights; For collective pitch control weights; Adjust the control weights for yaw correction; Assign weights to generator torque smoothing control; Weights are used to control power limits.

[0080] When the dominant risk link is from the blade node to the tower node, it indicates that blade load imbalance or periodic load is being transmitted to the tower vibration risk. In this case, the weight of independent pitch control is increased, generating independent pitch angle increments based on the root loads of the three blades and the impeller azimuth angle to reduce blade periodic loads. Simultaneously, the weight of tower damping-related control is increased to suppress tower forward / backward or lateral vibrations during collective pitch or torque control, and the yaw correction amount, which increases tower lateral loads, is limited. This prevents continued yaw when tower lateral risk is high, thereby reducing risk propagation.

[0081] When the dominant risk link is from the yaw system node to the tower node, it indicates that yaw actions or changes in yaw error have an adverse effect on the lateral vibration of the tower. In this case, the frequency of yaw actions should be reduced, and the weight of generator torque smoothing control and power limit control should be increased. Specifically, the minimum interval between yaw actions can be extended, the amount of single yaw correction can be reduced, or yaw can be allowed only when the yaw benefit significantly outweighs the load cost. At the same time, the load on the tower structure can be reduced by limiting torque mutations and appropriately lowering the power limit.

[0082] When the dominant risk link is a drivetrain node pointing to a generator-side node, it indicates that torsional vibration or torque fluctuations in the drivetrain are causing power fluctuations or grid-friendliness risks. In this case, increasing the generator torque change rate constraint strength and reducing the abrupt change amplitude of collective pitch control will make torque control smoother. If the drivetrain impact risk is high, the power setting can be appropriately reduced or the power ramp rate limited.

[0083] When the dominant risk link is from the pitch system node to the blade node, it indicates that the pitch control action itself may cause blade load disturbances or increase the fatigue risk of the pitch mechanism. In this case, limit the rate of change of independent pitch angle increments and shift some load reduction targets to power limit control or torque smoothing control. This can avoid over-reliance on independent pitch control for load reduction, which could lead to increased fatigue in the actuators.

[0084] Optionally, the control constraint boundaries include: the range of independent pitch angle increments, the range of collective pitch angle increments, the upper limit of pitch rate, the upper limit of pitch acceleration, the range of yaw correction, the minimum interval of yaw action, the range of generator torque increments, the upper limit of generator torque change rate, and the reduction range of power limit. In this embodiment of the disclosure, the control constraint boundary can be represented as: ; in, To control the set of constraint boundaries; This is the boundary for independent pitch angle increments; This is the boundary for the collective pitch angle increment; The pitch rate boundary; For the variable pitch acceleration boundary; For yaw correction and yaw action interval boundaries; This represents the boundary between the generator torque increment and the rate of torque change. This represents the lowering margin of the power limit.

[0085] In step S14, during the adaptive control process, when it is predicted that the adaptive control action may amplify the risk of downstream nodes, tightening the control constraint boundaries includes: During the adaptive control process, when it is predicted that the adaptive control action will increase the coupling risk value of the downstream node, the control constraint boundary corresponding to the adaptive control action is tightened, and the control constraint boundary corresponding to other adaptive control actions with risk suppression effect is relaxed.

[0086] In this embodiment of the disclosure, when it is predicted that a certain control action will increase the coupling risk value of downstream nodes, the constraint boundary corresponding to that control action is tightened, while the constraint boundaries corresponding to other control actions with risk suppression effects are appropriately relaxed. For example, if it is predicted that yaw correction will increase lateral vibration of the tower, the yaw correction amount boundary and the minimum yaw action interval boundary are tightened; at the same time, the available range of independent pitch or power limit control is relaxed, so that the system can achieve the load reduction target through other actuators. If it is predicted that frequent independent pitch actions will increase fatigue of the pitch actuator, the independent pitch angle increment and pitch rate boundaries are tightened, while the weight of torque smoothing or power limit is increased.

[0087] Optionally, the rolling solution of the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle includes: Combination Figure 6 As shown, a constraint optimization method combining prediction time domain and control time domain is adopted to solve the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment and power limit for the next control cycle in a rolling manner. The prediction time domain is from 0.5 seconds to 30 seconds, and the control time domain is one or more master control cycles. The comprehensive evaluation value is determined as follows: ; in, For comprehensive evaluation, As a power deviation evaluation metric, This is a metric for evaluating the incremental damage to the leaves. This is the evaluation metric for incremental tower damage. This is a quantity used for evaluating the impact on the transmission chain. This is a metric for evaluating the cost of yaw maneuvers. The cost evaluation metric for pitch and torque execution actions. , , , , and The weighting coefficients for the corresponding evaluation quantities are adjusted online according to the risk type of the dominant risk link.

[0088] It should be noted that the aforementioned weighting coefficients are not fixed, but are adjusted online according to the risk type of the dominant risk pathway. For example, when blade fatigue risk is dominant, the weighting coefficients are increased. When the risk of tower vibration is dominant, increase When the risk of impact on the transmission chain is dominant, improve When yaw action has a significant impact on tower risk, increase... When the risk of actuator fatigue is high, increase... In this way, the comprehensive evaluation function can reflect the control objective that most needs to be optimized under the current operating conditions.

[0089] The control variables for rolling optimization can be expressed as: ; in, For the set of control variables; For independent pitch angle increments; This represents the collective pitch angle increment; This is the yaw correction amount; This represents the generator torque increment. This is the adjustment amount for the power limit.

[0090] The objective is to achieve a comprehensive evaluation value that satisfies the boundary constraints. The constraints must be reduced to at least the following: impeller speed does not exceed the safety limit; power generation does not exceed the allowable power; blade root load does not exceed the set threshold; tower vibration does not exceed the safety threshold; nacelle acceleration does not exceed the safety threshold; pitch angle, pitch rate, and pitch acceleration do not exceed the actuator capability boundary; yaw action meets the minimum interval and single correction limits; generator torque change rate does not exceed the allowable boundary of the drivetrain; and the reduction in power limits does not exceed the range allowed by grid dispatching and operation strategies.

[0091] In one specific implementation, the rolling optimization solution can be obtained using the following process: First, read the current control cycle's operating status data, operating coupling status graph, dominant risk link, control weight, and constraint boundary.

[0092] Secondly, a simplified prediction model is used to predict power deviation, blade damage increment, tower damage increment, drivetrain impact, and actuator action cost in the prediction time domain under candidate control actions. The prediction model can be a linearized wind turbine model, a lookup table model, a data-driven model, or a combination of a physical model and a data-corrected model.

[0093] Next, candidate control actions are constrained and screened. If a candidate yaw action would increase the lateral risk of the tower, the yaw action is excluded or penalized; if an independent pitch action exceeds the pitch rate limit, the action is restricted; if a torque increment would increase the drive train impact, its selectable magnitude is reduced.

[0094] Then, select the control action that minimizes the comprehensive evaluation value or maximizes the reduction from the remaining candidate control actions. For main control systems with high real-time requirements, finite candidate set search, quadratic programming, or table lookup approximation methods can be used; for control platforms with strong computing power, model predictive control can be used.

[0095] Finally, the obtained independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit are sent to the corresponding controllers. After the command is issued, the system re-acquires data and updates the coupling graph in the next control cycle, forming a rolling closed loop.

[0096] Through the aforementioned rolling optimization, this disclosure can achieve dynamic balancing among different risk objectives. For example, when wind speed suddenly increases and causes blade load to rise, the system can prioritize increasing the independent pitch control weight; when tower lateral vibration increases, the system can suppress yaw action and reduce structural load through power limits; when drive train torsional vibration increases, the system can increase torque smoothing weight; when the fatigue risk of pitch actuators increases, the system can reduce the frequency of pitch control actions and transfer some load reduction objectives to other actuators.

[0097] Optionally, the method further includes: When a limited pitch rate, limited yaw action, lag in generator torque response, or deviation from the power limit command execution exceeds the corresponding threshold, the corresponding actuator will be marked as a capacity degradation actuator. Reduce the control weight corresponding to the degraded actuator and redistribute the load reduction target undertaken by the actuator to the non-degraded actuator.

[0098] Actuator capability degradation refers to the state in which an actuator cannot complete the control action as expected due to mechanical wear, decreased hydraulic or electrical response capability, control delay, limited action, or fault warning. For example, the pitch system may have insufficient pitch rate due to temperature, hydraulic pressure, or motor capacity limitations; the yaw system may have limited yaw action due to braking or rapid changes in wind direction; generator torque control may have response lag; and there may be deviations in power limit execution.

[0099] In this embodiment of the disclosure, combined with Figure 1 As shown, the actuator degradation identification module is used to... Figure 7 The logic shown involves an actuator degradation identification module monitoring the capability status of each actuator. If the deviation between the actual response and the expected response of an actuator exceeds a corresponding threshold, it is marked as a capability degradation actuator, and the control weight allocation module is notified to reduce the control weight of that actuator. For example, when the pitch rate cannot meet the command requirements, the main load reduction target is no longer allocated to independent pitch control; instead, some load reduction targets are transferred to power limits or generator torque smoothing control. When yaw action is limited, the yaw correction is reduced, and adverse loads are suppressed through independent pitch control or power limiting.

[0100] Combination Figure 7As shown, when the actuator degradation identification module detects limited pitch rate, limited yaw action, lag in generator torque response, or a deviation in power limit command execution exceeding the corresponding threshold, it marks the corresponding actuator as a capacity-degraded actuator. The actuator degradation identification module can determine this by the deviation between the command value and the actual response value. For example, if the main control system outputs a target pitch rate to the pitch controller, but the actual pitch rate is consistently lower than the target value and the deviation exceeds the threshold, it indicates that the pitch actuator's capability has decreased; if the yaw angle change does not reach the expected value after the yaw correction command is issued, or the yaw action is frequently interrupted, it indicates that the yaw actuator is limited; if there is a continuous lag between the generator torque command and the actual torque, it indicates that the torque control capability has decreased.

[0101] The actuator response deviation can be expressed as: ; in, For the first Response deviation of the actuator; For the first Control command values ​​for actuators; For the first The actual response value of the class executor.

[0102] when When the corresponding threshold is exceeded consecutively, the first... Actuators are marked as capacity-degraded actuators, and their control weights are reduced. For load reduction targets originally assigned to degraded actuators, the system reassigns them to non-degraded actuators. For example, if independent pitch capability degrades, the independent pitch load reduction weight is reduced, while the power limit and torque smoothing weights are increased; if yaw capability degrades, the yaw correction control weight is reduced, while the pitch control weight related to blade and tower load reduction is increased; if torque control capability degrades, the torque smoothing control weight is reduced, while the collective pitch or power limit control weights are appropriately increased.

[0103] This actuator degradation redistribution mechanism can maintain the overall load reduction capacity of the wind turbine when the actuator capacity changes or is locally limited, thus avoiding the control system from still allocating control tasks according to the ideal actuator capacity, thereby improving the long-term operational reliability of the wind turbine.

[0104] The present disclosure is compared with the prior art through specific embodiments and comparative examples below: 1. Experimental platform and testing methods.

[0105] To verify the technical effectiveness of this disclosure, a hardware-in-the-loop (HIL) simulation test platform for the wind turbine main control algorithm was built in a laboratory. This platform includes a wind turbine dynamics simulation module, a main control algorithm controller, a pitch actuator simulation module, a yaw actuator simulation module, a converter torque response simulation module, a load and vibration acquisition module, and a data recording module. The wind turbine dynamics simulation module simulates blade aerodynamic loads, tower vibration, drivetrain torsional vibration, power generation variations, and the impact of yaw error. The main control algorithm controller runs the adaptive optimization control method for the coupled operating state of the entire turbine and various proportional control strategies disclosed in this disclosure. The data recording module records indicators such as blade root loads, tower vibration, nacelle acceleration, torque impact, power fluctuations, number of yaw actions, and number of pitch actions.

[0106] The test subject was a three-bladed horizontal-axis variable-speed pitch wind turbine. To facilitate comparison of the differences between various control strategies, all embodiments and comparative examples used the same turbine model, the same wind condition input, the same sampling period, and the same safety constraints. Test wind conditions included enhanced turbulence, yaw disturbance, gust impact, and actuator capacity degradation. Each set of conditions was run continuously for 30 minutes; the first 5 minutes were considered the control stabilization phase and not included in the statistics, while the remaining 25 minutes were considered the valid statistical interval. Each control strategy was tested three times, and the average value was taken as the final test result.

[0107] The following evaluation indicators were used in the experiment: The equivalent fatigue load at the blade root is used to evaluate the blade's load reduction effect. The lower the value, the lower the risk of blade fatigue damage.

[0108] The equivalent fatigue load in the front and rear directions of the tower is used to evaluate the structural load of the tower in the prevailing wind direction.

[0109] The equivalent lateral fatigue load on the tower is used to evaluate the effects of yaw error, yaw action, and lateral inflow on the tower structure.

[0110] The peak torque impact of the transmission chain is used to evaluate the impact of generator torque changes and impeller speed fluctuations on the transmission chain.

[0111] The standard deviation of power fluctuation is used to evaluate the impact of control strategies on the smoothness of output power.

[0112] The number of pitch and yaw actions is used to evaluate the actuator's workload.

[0113] The comprehensive evaluation value is used to evaluate the overall control effect of power deviation, blade damage, tower damage, transmission chain impact, and actuator action cost.

[0114] The overall evaluation score is calculated as follows: ; in, This is a comprehensive evaluation value; This is a power deviation evaluation metric. This is a metric for evaluating incremental leaf damage. This is the evaluation metric for incremental tower damage; For evaluating the impact of the transmission chain; This is a metric for evaluating the cost of yaw maneuvers. The cost evaluation metric for pitch and torque execution actions; , , , , and These are the weighting coefficients for the corresponding evaluation quantities. In this embodiment, each evaluation quantity uses a dimensionless value normalized to the benchmark control strategy. The lower the comprehensive evaluation value, the better the comprehensive control effect.

[0115] The relative unload rate of the blades, tower, and drive train is calculated as follows: ; in, This refers to the relative load reduction rate; The corresponding load evaluation value in Comparative Example 1; The values ​​represent the corresponding load evaluation values ​​in the embodiments or other comparative examples. The load evaluation values ​​can be the equivalent fatigue load at the blade root, the equivalent fatigue load of the tower, or the peak torque impact value of the transmission chain.

[0116] 2. Example 1: Adaptive optimization control method based on the coupling state of the complete machine.

[0117] This embodiment employs the complete control method disclosed herein. Specifically, it involves: collecting incoming wind speed, wind direction deviation, impeller speed, power generation, generator torque, root load of the three blades, tower vibration, nacelle acceleration, yaw angle, pitch angle, pitch rate, and yaw action status within a sliding time window; classifying the blades, tower, drive train, yaw system, pitch system, and generator side as coupling nodes; updating the operational coupling state map online based on the time-delay response of upstream node state changes to downstream node load, vibration, and power fluctuations, as well as the downstream short-term damage increment; identifying the dominant risk link based on edge weights and downstream damage increments; adaptively allocating load reduction targets to independent pitch, collective pitch, yaw correction, generator torque smoothing, and power limit control according to the upstream and downstream nodes and risk type of the dominant risk link; tightening the constraint boundary of the control action when it is predicted that a certain control action will amplify the risk of the downstream node; and finally, rolling the solution for the control command of the next control cycle.

[0118] In this embodiment, the edge weights are calculated using the following formula: ; in, For the first The coupling node points to the first Edge weights of each coupled node; For the first A state change at the first coupled node causes the... The normalized response intensity of the response of each coupled node; For the first The coupling node to the first The time-delay response factor of each coupled node; For the first Normalized value of short-time damage increment of each coupled node; The actuator motion propagation factor; , , and These are the corresponding weighting coefficients. In this embodiment, Take 0.35, Take 0.20, Take 0.30, Set the value to 0.15. The sum of the above coefficients is 1, which is used to ensure that the edge weights are within the normalized evaluation range.

[0119] This embodiment uses continuity and directionality to determine the dominant risk link. Continuity determination requires that the same directed link satisfy the following conditions for at least three consecutive control cycles: the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. Directionality determination requires that changes in the state of the upstream node precede changes in the load, vibration, or power fluctuations of the downstream node, and that the time delay between the two is within a preset time delay range. This method can eliminate spurious coupling caused by the combined effects of gusts or measurement noise.

[0120] Under enhanced turbulence conditions, the system identifies the dominant risk link from blade nodes to tower nodes, thus increasing the weight of independent pitch control and limiting yaw corrections that could increase tower lateral loads. Under yaw disturbance conditions, the system identifies the dominant risk link from yaw system nodes to tower nodes, thus reducing yaw frequency and increasing the weight of generator torque smoothing control and power limit control. Under gust impact conditions, the system identifies the dominant risk link from drivetrain nodes to generator-side nodes, thus limiting the generator torque change rate and reducing drivetrain impact. Under actuator capacity degradation conditions, the system detects limited pitch rate and redistributes some blade load reduction targets to power limit control and generator torque smoothing control.

[0121] 3. Example 2: Coupled graph and dominant risk link identification are used, but actuator degradation reassignment is not enabled.

[0122] This embodiment is basically the same as Embodiment 1, except that: this embodiment uses online updating of the running coupling state map, identification of dominant risk links, adaptive allocation of multi-actuator control weights, and dynamic adjustment of constraint boundaries, but does not enable actuator capability degradation identification and load reduction target reallocation. When the pitch actuator capability degrades, the system still allocates load reduction targets according to the control weights under normal conditions.

[0123] Under normal actuator conditions, Embodiment 2 and Embodiment 1 have similar load reduction effects; however, when pitch rate or yaw action is limited, Embodiment 1 should exhibit a more stable overall control effect.

[0124] 4. Example 3: Using coupling graphs and risk link identification, but with fixed control weights.

[0125] This embodiment is basically the same as Embodiment 1, except that: this embodiment can construct an operational coupling state map and identify the dominant risk link, but the weights of independent pitch, collective pitch, yaw correction, generator torque smoothing and power limit control remain fixed and are not dynamically adjusted with changes in the dominant risk link.

[0126] If only risk links are identified without mapping them to control weights and constraint boundaries, the control system can perceive the source of risk, but cannot fully adjust the actuator strategy according to the risk propagation path. Therefore, the overall load reduction effect will be weaker than that of Example 1.

[0127] 5. Comparative Example 1: Traditional collective pitch control and generator torque control.

[0128] Comparative Example 1 employs a traditional wind turbine master control strategy, which only performs collective pitch and generator torque control based on rotor speed, power generation, and rated power deviation. This strategy does not collect or utilize the root loads of the three blades to form independent pitch load reduction, does not construct an operational coupling state map, does not identify the dominant risk link, and does not adjust the control weights between yaw, pitch, torque, and power limits based on the risk link.

[0129] Comparative Example 1 serves as a baseline control group to evaluate the technical effectiveness of this disclosure relative to traditional master control strategies.

[0130] 6. Comparative Example 2: Independent pitch control.

[0131] Comparative Example 2 adds independent pitch control to Comparative Example 1. This strategy generates independent pitch angle increments based on the root loads of the three blades and the impeller azimuth angle to reduce the cyclic load on the blades. However, this strategy does not construct a coupled state map of the entire unit's operation, does not identify the dominant risk links between the blades, tower, drive train, yaw system, pitch system, and generator side, and does not dynamically adjust the weights of yaw correction, generator torque smoothing, and power limit control based on the risk links.

[0132] 7. Comparative Example 3: Yaw and Pitch Linkage Load Reduction Control.

[0133] Comparative Example 3 employs a yaw and pitch linkage load reduction control strategy. This strategy generates yaw correction and pitch control commands based on wind direction deviation, wind speed, pitch angle, and power status to balance power capture and load control. However, this strategy does not establish an operational coupling state map based on time-delay response, short-term damage increment, and actuator motion propagation factor, nor does it tighten control constraint boundaries that may amplify downstream risks based on the dominant risk link.

[0134] 8. Comparative Example 4: Predictive Control Using a Common Multivariate Model.

[0135] Comparative Example 4 employs a conventional multivariate model predictive control, incorporating power deviation, blade load, tower load, and actuator action costs into a unified objective function, while simultaneously outputting pitch, torque, and power control commands. Although this strategy considers multi-objective optimization, its weight coefficients are fixed, it does not construct an operational coupling state graph, does not identify dominant risk links, and does not dynamically reallocate control weights and constraint boundaries based on upstream nodes, downstream nodes, and risk types.

[0136] Comparative Example 4 illustrates that even if conventional multivariate optimization control can compromise among multiple objectives, it is difficult to determine which component link the risk propagates along, and it cannot adjust control weights and constraint boundaries for specific risk links. Therefore, its overall effect is lower than that of this disclosure.

[0137] 9. Experimental data and results analysis.

[0138] Table 1 shows the test results of blade and tower loads under turbulent enhanced conditions:

[0139] Table 1 As shown in Table 1, under enhanced turbulence conditions, Comparative Example 2, due to the use of independent pitch control, reduced the equivalent fatigue load at the blade root by 11.0% compared to Comparative Example 1, but the equivalent fatigue load on the tower side only decreased by 3.5%. This indicates that independent pitch control alone mainly affects the cyclic load on the blades and is insufficient to fully suppress the propagation of the blade load to the tower vibration. Comparative Example 3, due to the introduction of yaw pitch linkage control, reduced the equivalent fatigue load on the tower side by 6.4% compared to Comparative Example 1, but the load reduction rate at the blade root was lower than that of Comparative Example 2.

[0140] Example 1 achieved the lowest values ​​for equivalent fatigue load at the blade root, equivalent fatigue load in the forward and backward directions of the tower, and equivalent fatigue load in the lateral direction of the tower, reducing them by 19.9%, 14.8%, and 17.4% respectively compared to Comparative Example 1. This disclosure, after identifying the dominant risk link from the blade node to the tower node, does not add independent pitch control separately, but simultaneously adjusts the independent pitch control weight, the tower vibration suppression weight, and the yaw constraint boundary, thus enabling a simultaneous reduction in blade load and tower load.

[0141] Table 2 shows the test results of yaw action and tower lateral load under yaw disturbance conditions:

[0142] Table 2 As shown in Table 2, Comparative Example 3 reduced the average wind direction deviation to 5.1° and increased the power generation retention rate to 101.6% through more aggressive yaw and pitch control. However, the number of yaw actions reached 31, and the equivalent lateral fatigue load on the tower remained at 2860 kN·m. This indicates that under yaw disturbance conditions, simply pursuing a reduction in wind direction deviation may increase the burden of yaw actions and potentially lead to an increase in lateral tower risk.

[0143] The average wind direction deviation in Example 1 was 6.6°, slightly higher than that in Comparative Example 3, but the number of yaw maneuvers was reduced to 15, the equivalent lateral fatigue load on the tower was reduced to 2525 kN·m, and the nacelle lateral acceleration RMS was reduced to 0.169 m·s². - ², while the power generation rate still reaches 100.2%. This result shows that this disclosure does not simply suppress yaw, but after identifying the dominant risk link between the yaw system node and the tower node, it tightens the constraint boundaries of the yaw correction amount and the minimum interval of yaw action, and compensates for the load reduction target through torque smoothing and appropriate power limit, thereby reducing the lateral load on the tower while basically maintaining the power generation.

[0144] Table 3 shows the test results of the transmission chain and power fluctuation under gust impact conditions:

[0145] Table 3 As shown in Table 3, under gust impact conditions, Comparative Example 1, which only uses traditional collective pitch and torque control, experiences a peak torque impact of 815 kN·m in the drivetrain, a power fluctuation standard deviation of 286 kW, and a maximum power overshoot of 8.7%. Comparative Example 4, after adopting ordinary multivariate model predictive control, reduces the peak torque impact of the drivetrain to 720 kN·m, indicating that multi-objective optimization can reduce the impact to a certain extent.

[0146] In Example 1, the peak torque impact of the drivetrain was reduced to 655 kN·m, a 19.6% reduction compared to Comparative Example 1. The standard deviation of power fluctuation was reduced to 205 kW, and the maximum power overshoot was reduced to 4.8%. This disclosure, after identifying the dominant risk link between the drivetrain nodes and the generator-side nodes during the gust impact phase, increases the weight of generator torque smoothing control while limiting the amplitude of collective pitch abrupt changes, thereby reducing drivetrain impact and power fluctuation.

[0147] Table 4 shows the test results of control effect under actuator capability degradation conditions:

[0148] Table 4 In Table 4, the number of control mismatches refers to the number of times the controller's output control commands failed to be executed as expected due to actuator limitations. Table 4 shows that, under pitch rate limitation, although Comparative Example 2 has an independent pitch load reduction function, the control mismatch count reaches 27 times because the pitch actuator cannot fully execute the commands, resulting in a significant decrease in the load reduction effect on the blades and tower. Although Comparative Example 4 has ordinary multivariate optimization capabilities, the control mismatch count is still 18 times because it failed to identify actuator capability degradation and reallocate the load reduction target.

[0149] Example 1 reduces the number of control mismatches to 5 and the overall evaluation value to 0.756 by identifying actuator capability degradation and reallocating load reduction targets. When a pitch rate limitation is detected, this disclosure reduces the weight of independent pitch control and transfers part of the load reduction target to power limit control and generator torque smoothing control. Therefore, it can still maintain a good overall load reduction effect under actuator degradation conditions. Example 2 does not enable actuator degradation reallocation, and its overall evaluation value is 0.812, which is worse than Example 1. Table 5 presents the statistics of overall technical effectiveness:

[0150] Table 5 As shown in Table 5, Example 1 outperforms all comparative examples in terms of overall blade load reduction rate, overall tower load reduction rate, drivetrain impact reduction rate, power fluctuation reduction rate, and actuator operation cost reduction rate. While Comparative Example 2 reduces blade load, its actuator operation cost reduction rate is negative, indicating that frequent independent pitch control increases the actuator burden. Comparative Example 3 improves yaw wind response and some tower load, but the increased number of yaw actions also increases the actuator operation cost. Although Comparative Example 4 achieves some overall effect through multi-objective optimization, its overall evaluation value is still higher than Example 1 due to the lack of an operational coupling state map and identification of the dominant risk link.

[0151] Example 3 is an improvement over Comparative Example 4, demonstrating that coupling graphs and risk link identification can improve the control strategy. Example 2 is a further improvement over Example 3, demonstrating that adaptive allocation of control weights driven by dominant risk links and adjustment of constraint boundaries can significantly improve the load reduction effect. Example 1 is a further improvement over Example 2, demonstrating that actuator capability degradation identification and load reduction target reassignment can improve the robustness of the system under non-ideal execution conditions.

[0152] This disclosure also provides an adaptive optimization control device 800 for the overall operation coupling state of a wind turbine generator, see [link]. Figure 8 As shown, it includes: The data acquisition module 810 is configured to acquire the operating status data of the wind turbine within a sliding time window; The coupling graph construction module 820 is configured to update the state of each coupled node in the running coupling state graph online based on the time-delay response of the load, vibration and power fluctuation of the downstream node and the short-term damage increment of the downstream node to the state changes of the upstream node within the sliding time window. The graph includes a set of nodes, a set of edges and a set of edge weights. The edge weights are calculated by weighting the response intensity, time delay and damage increment. The coupled nodes are obtained by dividing the wind turbine blades, tower, transmission chain, yaw system, pitch system and power generation side into regions. The risk link identification module 830 is configured to determine the dominant risk link in the running coupling state graph after the updated state of the coupled node, based on the edge weight and the downstream damage increment. The dominant risk link is a directed link in the running coupling state graph in which the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. The control weight allocation module 840 is configured to adaptively allocate load reduction targets to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control according to the upstream nodes, downstream nodes and risk types in the dominant risk link, and tighten the control constraint boundary during the adaptive control process when it is predicted that the adaptive control action will amplify the risk of the downstream node. The rolling optimization solution module 850 is configured to, under the control weights and the constraint boundaries, continuously solve for the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle, so as to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact, and actuator action cost in the prediction time domain, obtain the control parameters corresponding to each main control system, and send the control parameters to the corresponding main control system for execution.

[0153] Optionally, the weight of any edge in the running coupling state graph is calculated as follows: ; in, For the first The coupling node points to the first Edge weights of each coupled node For the first A state change at the first coupled node causes the... The normalized response intensity of the responses of each coupled node. For the first The coupling node to the first The time-delay response factor of each coupled node For the first Normalized value of short-time damage increment of each coupled node The actuator motion propagation factor is used to characterize the amplification or suppression effect of pitch, yaw, or generator torque actions on downstream coupled node loads, vibrations, or power fluctuations. , , and These are the corresponding weighting coefficients, and all are greater than 0.

[0154] Optionally, the time-delay response factor The maximum correlation response is determined by comparing the state change sequence of the upstream node with the response sequence of the downstream node within a preset time delay search interval, where the preset time delay search interval is from 0.1 seconds to 10 seconds. Specifically, when the time delay corresponding to the maximum correlation response is within the effective response range determined by the current speed, impeller azimuth angle, and control cycle of the wind turbine, the corresponding directed edge is retained; otherwise, the weight of the corresponding directed edge is decayed or set to zero.

[0155] Optionally, the determination of the dominant risk link includes: persistence determination and direction determination; The persistence determination is that the same directed link satisfies the condition that the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold within at least N consecutive control cycles. The directionality determination is that the state change of the upstream node occurs before the load, vibration or power fluctuation change of the downstream node, and the time delay between the two is within a preset time delay range, thereby eliminating pseudo-coupled links caused by synchronization disturbances or measurement noise.

[0156] Optionally, the risk types include at least: blade fatigue risk, tower vibration risk, transmission chain torsional vibration risk, yaw error amplification risk, pitch actuator fatigue risk, and power fluctuation risk; Specifically, the blade fatigue risk is determined based on the periodic and non-periodic components of the load at the root of the three blades; the tower vibration risk is determined based on the tower vibration amplitude, vibration frequency, and damping variation trend; the transmission chain torsional vibration risk is determined based on generator torque fluctuation and impeller speed fluctuation; and the pitch actuator fatigue risk is determined based on pitch rate, pitch acceleration, and number of actions per unit time.

[0157] Optionally, the control weight allocation module 840 is configured to: When the dominant risk link is from the blade node to the tower node, increase the control weight of independent pitch and tower damping, and limit the yaw correction amount that increases the tower lateral load. When the dominant risk link is a yaw system node pointing to a tower node, reduce the frequency of yaw actions and increase the control weight of generator torque smoothing and power limit. When the dominant risk link is a drive chain node pointing to a generator-side node, the constraint strength of the generator torque change rate is increased, and the abrupt change amplitude of the collective pitch control is reduced. When the dominant risk link is a pitch system node pointing to a blade node, the rate of change of independent pitch angle increment is limited, and part of the load reduction target is transferred to power limit control.

[0158] Optionally, the control constraint boundaries include: the range of independent pitch angle increments, the range of collective pitch angle increments, the upper limit of pitch rate, the upper limit of pitch acceleration, the range of yaw correction, the minimum interval of yaw action, the range of generator torque increments, the upper limit of generator torque change rate, and the reduction range of power limit. The control weight allocation module 840 is configured as follows: During the adaptive control process, when it is predicted that the adaptive control action will increase the coupling risk value of the downstream node, the control constraint boundary corresponding to the adaptive control action is tightened, and the control constraint boundary corresponding to other adaptive control actions with risk suppression effect is relaxed.

[0159] Optionally, the rolling optimization solution module 850 is configured as follows: A constraint optimization method combining prediction time domain and control time domain is adopted to solve the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment and power limit for the next control cycle in a rolling manner. The prediction time domain is from 0.5 seconds to 30 seconds, and the control time domain is one or more master control cycles. The comprehensive evaluation value is determined as follows: ; in, For comprehensive evaluation, As a power deviation evaluation metric, This is a metric for evaluating the incremental damage to the leaves. This is the evaluation metric for incremental tower damage. This is a quantity used for evaluating the impact on the transmission chain. This is a metric for evaluating the cost of yaw maneuvers. The cost evaluation metric for pitch and torque execution actions. , , , , and The weighting coefficients for the corresponding evaluation quantities are adjusted online according to the risk type of the dominant risk link.

[0160] Optionally, the device 800 further includes: an actuator degradation identification module, configured to: When a limited pitch rate, limited yaw action, lag in generator torque response, or deviation from the power limit command execution exceeds the corresponding threshold, the corresponding actuator will be marked as a capacity degradation actuator. Reduce the control weight corresponding to the degraded actuator and redistribute the load reduction target undertaken by the actuator to the non-degraded actuator.

[0161] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0162] This disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described in the foregoing embodiments.

[0163] This disclosure also provides an electronic device, including: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0164] Figure 9 This is a block diagram illustrating an electronic device 300 according to an exemplary embodiment. Figure 9 As shown, the electronic device 300 may include a processor 301 and a memory 302. The electronic device 300 may also include one or more of a multimedia component 303, an input / output (I / O) interface 304, and a communication component 305.

[0165] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps in the aforementioned adaptive optimization control method for the coupled operation state of the wind turbine generator set. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, and application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 303 may include a screen and audio components. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 302 or transmitted via communication component 305. The audio component also includes at least one speaker for outputting audio signals. I / O interface 304 provides an interface between processor 301 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 305 is used for wired or wireless communication between the electronic device 300 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or combinations thereof, is not limited here. Therefore, the corresponding communication component 305 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0166] In an exemplary embodiment, the electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the aforementioned adaptive optimization control method for the coupled operating state of the wind turbine generator set.

[0167] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the adaptive optimization control method for the overall operating coupling state of the wind turbine generator described above. For example, the computer-readable storage medium may be the memory 302 including the program instructions, which may be executed by the processor 301 of the electronic device 300 to complete the adaptive optimization control method for the overall operating coupling state of the wind turbine generator described above.

[0168] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0169] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0170] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for adaptive optimization control of coupling state of a wind turbine system, characterized in that, Includes the following steps: The operating status data of the wind turbine is acquired within a sliding time window; Based on the time-delay response of the upstream node's state change to the downstream node's load, vibration, and power fluctuations, as well as the downstream short-term damage increment, the state of each coupled node in the operational coupling state map is updated online. The map includes a node set, an edge set, and an edge weight set. The edge weights are calculated by weighting the response intensity, time delay, and damage increment. The coupled nodes are obtained by dividing the wind turbine's blades, tower, transmission chain, yaw system, pitch system, and power generation side into regions. In the running coupling state graph after the updated state of the coupled node, the dominant risk link is determined based on the edge weight and the downstream damage increment. The dominant risk link is the directed link in the running coupling state graph where the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. According to the upstream and downstream nodes and risk types in the dominant risk chain, the load reduction target is adaptively allocated to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control by controlling the weights. During the adaptive control process, when it is predicted that the adaptive control action will amplify the risk of the downstream node, the control constraint boundary is tightened. Under the control weights and the constraint boundaries, the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle are solved in a rolling manner to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact, and actuator action cost in the prediction time domain. The control parameters corresponding to each main control system are obtained, and the control parameters are sent to the corresponding main control system for execution.

2. The method of claim 1, wherein, The weight of any edge in the operational coupling state graph is calculated as follows: ； in, For the first The coupling node points to the first Edge weights of each coupled node For the first A state change at the first coupled node causes the... The normalized response intensity of the responses of each coupled node. For the first The coupling node to the first The time-delay response factor of each coupled node. For the first Normalized value of short-time damage increment of each coupled node The actuator motion propagation factor is used to characterize the amplification or suppression effect of pitch, yaw, or generator torque actions on downstream coupled node loads, vibrations, or power fluctuations. , , and These are the corresponding weighting coefficients, and all are greater than 0.

3. The method according to claim 2, characterized in that, The time-delay response factor The maximum correlation response is determined by comparing the state change sequence of the upstream node with the response sequence of the downstream node within a preset time delay search interval, where the preset time delay search interval is from 0.1 seconds to 10 seconds. Specifically, when the time delay corresponding to the maximum correlation response is within the effective response range determined by the current speed, impeller azimuth angle, and control cycle of the wind turbine, the corresponding directed edge is retained; otherwise, the weight of the corresponding directed edge is decayed or set to zero.

4. The method according to claim 1, characterized in that, The determination of the dominant risk link includes: persistence determination and direction determination; The persistence determination is that the same directed link satisfies the condition that the edge weight exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold within at least N consecutive control cycles. The directionality determination is that the state change of the upstream node occurs before the load, vibration or power fluctuation change of the downstream node, and the time delay between the two is within a preset time delay range, thereby eliminating pseudo-coupled links caused by synchronization disturbances or measurement noise.

5. The method according to claim 1, characterized in that, The risk types include at least: blade fatigue risk, tower vibration risk, transmission chain torsional vibration risk, yaw error amplification risk, pitch actuator fatigue risk, and power fluctuation risk. Specifically, the blade fatigue risk is determined based on the periodic and non-periodic components of the load at the root of the three blades; the tower vibration risk is determined based on the tower vibration amplitude, vibration frequency, and damping variation trend; the transmission chain torsional vibration risk is determined based on generator torque fluctuation and impeller speed fluctuation; and the pitch actuator fatigue risk is determined based on pitch rate, pitch acceleration, and number of actions per unit time.

6. The method according to claim 1, characterized in that, The method of adaptively allocating load reduction targets to independent pitch control, collective pitch control, yaw correction, generator torque smoothing, and power limit control based on upstream and downstream nodes and risk types in the dominant risk chain, through control weights, includes: When the dominant risk link is from the blade node to the tower node, increase the control weight of independent pitch and tower damping, and limit the yaw correction amount that increases the tower lateral load. When the dominant risk link is a yaw system node pointing to a tower node, reduce the frequency of yaw actions and increase the control weight of generator torque smoothing and power limit. When the dominant risk link is a drive chain node pointing to a generator-side node, the constraint strength of the generator torque change rate is increased, and the abrupt change amplitude of the collective pitch control is reduced. When the dominant risk link is a pitch system node pointing to a blade node, the rate of change of independent pitch angle increment is limited, and part of the load reduction target is transferred to power limit control.

7. The method according to claim 1, characterized in that, The control constraint boundaries include: the range of independent pitch angle increments, the range of collective pitch angle increments, the upper limit of pitch rate, the upper limit of pitch acceleration, the range of yaw correction, the minimum interval of yaw action, the range of generator torque increments, the upper limit of generator torque change rate, and the reduction range of power limit. In the adaptive control process, when it is predicted that the adaptive control action may amplify the risk of downstream nodes, tightening the control constraint boundaries includes: During the adaptive control process, when it is predicted that the adaptive control action will increase the coupling risk value of the downstream node, the control constraint boundary corresponding to the adaptive control action is tightened, and the control constraint boundary corresponding to other adaptive control actions with risk suppression effect is relaxed.

8. The method according to claim 1, characterized in that, The rolling solution for the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle includes: A constraint optimization method combining prediction time domain and control time domain is adopted to solve the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment and power limit for the next control cycle in a rolling manner. The prediction time domain is from 0.5 seconds to 30 seconds, and the control time domain is one or more master control cycles. The comprehensive evaluation value is determined as follows: ； in, For comprehensive evaluation, As a power deviation evaluation metric, This is a metric for evaluating the incremental damage to the leaves. This is the evaluation metric for incremental tower damage. This is a quantity used for evaluating the impact on the transmission chain. This is a metric for evaluating the cost of yaw maneuvers. The cost evaluation metric for pitch and torque execution actions. , , , , and The weighting coefficients for the corresponding evaluation quantities are adjusted online according to the risk type of the dominant risk link.

9. The method according to any one of claims 1-8, characterized in that, The method further includes: When a limited pitch rate, limited yaw action, lag in generator torque response, or deviation from the power limit command execution exceeds the corresponding threshold, the corresponding actuator will be marked as a capacity degradation actuator. Reduce the control weight corresponding to the degraded actuator and redistribute the load reduction target undertaken by the actuator to the non-degraded actuator.

10. A wind turbine generator set overall operation coupling state adaptive optimization control device, characterized in that, include: The data acquisition module is configured to acquire the operating status data of the wind turbine within a sliding time window; The coupling graph construction module is configured to update the state of each coupled node in the running coupling state graph online based on the time-delay response of the upstream node to the load, vibration and power fluctuation of the downstream node and the short-term damage increment of the downstream node, according to the state change of the upstream node within the sliding time window. The graph includes a set of nodes, a set of edges and a set of edge weights. The edge weights are calculated by weighting the response intensity, time delay and damage increment. The coupled nodes are obtained by dividing the wind turbine blades, tower, transmission chain, yaw system, pitch system and power generation side into regions. The risk link identification module is configured to determine the dominant risk link in the running coupling state graph after the updated state of the coupled node, based on the edge weight and the downstream damage increment. The dominant risk link is a directed link in the running coupling state graph in which the edge weight continuously exceeds the edge weight threshold and the downstream damage increment exceeds the risk threshold. The control weight allocation module is configured to adaptively allocate load reduction targets to independent pitch control, collective pitch control, yaw correction, generator torque smoothing and power limit control according to the upstream nodes, downstream nodes and risk types in the dominant risk link. During the adaptive control process, when it is predicted that the adaptive control action will amplify the risk of the downstream node, the control constraint boundary is tightened. The rolling optimization solution module is configured to, under the control weights and the constraint boundaries, continuously solve for the independent pitch angle increment, collective pitch angle increment, yaw correction, generator torque increment, and power limit for the next control cycle, so as to reduce the comprehensive evaluation value of power deviation, blade and tower damage increment, transmission chain impact, and actuator action cost in the prediction time domain, obtain the control parameters corresponding to each main control system, and send the control parameters to the corresponding main control system for execution.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1-9.

12. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1-9.

Images